checkin from eve

this relates to https://github.com/Minres/DBT-RISE-RISCV/issues/8 :
Debugger loses control when trap vector fetch fails

and https://github.com/Minres/DBT-RISE-RISCV/issues/7 : Two debugger
single-steps are required at reset vector

.clang-format

@@ -1,4 +1,3 @@
----
 Language:        Cpp
 # BasedOnStyle:  LLVM
 # should be in line with IndentWidth
@@ -13,8 +12,8 @@ AllowAllParametersOfDeclarationOnNextLine: true
 AllowShortBlocksOnASingleLine: false
 AllowShortCaseLabelsOnASingleLine: false
 AllowShortFunctionsOnASingleLine: All
-AllowShortIfStatementsOnASingleLine: true
+AllowShortIfStatementsOnASingleLine: false
-AllowShortLoopsOnASingleLine: true
+AllowShortLoopsOnASingleLine: false
 AlwaysBreakAfterDefinitionReturnType: None
 AlwaysBreakAfterReturnType: None
 AlwaysBreakBeforeMultilineStrings: false
@@ -39,8 +38,8 @@ BreakBeforeTernaryOperators: true
 BreakConstructorInitializersBeforeComma: true
 BreakAfterJavaFieldAnnotations: false
 BreakStringLiterals: true
-ColumnLimit:     120
+ColumnLimit:     140
-CommentPragmas:  '^ IWYU pragma:'
+CommentPragmas:  '^( IWYU pragma:| @suppress)'
 ConstructorInitializerAllOnOneLineOrOnePerLine: false
 ConstructorInitializerIndentWidth: 0
 ContinuationIndentWidth: 4
@@ -76,13 +75,13 @@ PenaltyBreakFirstLessLess: 120
 PenaltyBreakString: 1000
 PenaltyExcessCharacter: 1000000
 PenaltyReturnTypeOnItsOwnLine: 60
-PointerAlignment: Right
+PointerAlignment: Left
 ReflowComments:  true
 SortIncludes:    true
 SpaceAfterCStyleCast: false
 SpaceAfterTemplateKeyword: true
 SpaceBeforeAssignmentOperators: true
-SpaceBeforeParens: ControlStatements
+SpaceBeforeParens: Never
 SpaceInEmptyParentheses: false
 SpacesBeforeTrailingComments: 1
 SpacesInAngles:  false

.gitignore

@@ -1,5 +1,6 @@
 .DS_Store
 /*.il
+/.settings
 /avr-instr.html
 /blink.S
 /flash.*
@@ -14,7 +15,6 @@
 /*.ods
 /build*/
 /*.logs
-language.settings.xml
 /*.gtkw
 /Debug wo LLVM/
 /*.txdb

.settings/org.eclipse.cdt.codan.core.prefs

@@ -1,73 +0,0 @@
-eclipse.preferences.version=1
-org.eclipse.cdt.codan.checkers.errnoreturn=Warning
-org.eclipse.cdt.codan.checkers.errnoreturn.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"No return\\")",implicit\=>false}
-org.eclipse.cdt.codan.checkers.errreturnvalue=Error
-org.eclipse.cdt.codan.checkers.errreturnvalue.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"Unused return value\\")"}
-org.eclipse.cdt.codan.checkers.nocommentinside=-Error
-org.eclipse.cdt.codan.checkers.nocommentinside.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"Nesting comments\\")"}
-org.eclipse.cdt.codan.checkers.nolinecomment=-Error
-org.eclipse.cdt.codan.checkers.nolinecomment.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"Line comments\\")"}
-org.eclipse.cdt.codan.checkers.noreturn=Error
-org.eclipse.cdt.codan.checkers.noreturn.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"No return value\\")",implicit\=>false}
-org.eclipse.cdt.codan.internal.checkers.AbstractClassCreation=Error
-org.eclipse.cdt.codan.internal.checkers.AbstractClassCreation.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"Abstract class cannot be instantiated\\")"}
-org.eclipse.cdt.codan.internal.checkers.AmbiguousProblem=Error
-org.eclipse.cdt.codan.internal.checkers.AmbiguousProblem.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"Ambiguous problem\\")"}
-org.eclipse.cdt.codan.internal.checkers.AssignmentInConditionProblem=Warning
-org.eclipse.cdt.codan.internal.checkers.AssignmentInConditionProblem.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"Assignment in condition\\")"}
-org.eclipse.cdt.codan.internal.checkers.AssignmentToItselfProblem=Error
-org.eclipse.cdt.codan.internal.checkers.AssignmentToItselfProblem.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"Assignment to itself\\")"}
-org.eclipse.cdt.codan.internal.checkers.CaseBreakProblem=Warning
-org.eclipse.cdt.codan.internal.checkers.CaseBreakProblem.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"No break at end of case\\")",no_break_comment\=>"no break",last_case_param\=>false,empty_case_param\=>false,enable_fallthrough_quickfix_param\=>false}
-org.eclipse.cdt.codan.internal.checkers.CatchByReference=Warning
-org.eclipse.cdt.codan.internal.checkers.CatchByReference.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"Catching by reference is recommended\\")",unknown\=>false,exceptions\=>()}
-org.eclipse.cdt.codan.internal.checkers.CircularReferenceProblem=Error
-org.eclipse.cdt.codan.internal.checkers.CircularReferenceProblem.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"Circular inheritance\\")"}
-org.eclipse.cdt.codan.internal.checkers.ClassMembersInitialization=Warning
-org.eclipse.cdt.codan.internal.checkers.ClassMembersInitialization.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"Class members should be properly initialized\\")",skip\=>true}
-org.eclipse.cdt.codan.internal.checkers.DecltypeAutoProblem=Error
-org.eclipse.cdt.codan.internal.checkers.DecltypeAutoProblem.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"Invalid 'decltype(auto)' specifier\\")"}
-org.eclipse.cdt.codan.internal.checkers.FieldResolutionProblem=Error
-org.eclipse.cdt.codan.internal.checkers.FieldResolutionProblem.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"Field cannot be resolved\\")"}
-org.eclipse.cdt.codan.internal.checkers.FunctionResolutionProblem=Error
-org.eclipse.cdt.codan.internal.checkers.FunctionResolutionProblem.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"Function cannot be resolved\\")"}
-org.eclipse.cdt.codan.internal.checkers.InvalidArguments=Error
-org.eclipse.cdt.codan.internal.checkers.InvalidArguments.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"Invalid arguments\\")"}
-org.eclipse.cdt.codan.internal.checkers.InvalidTemplateArgumentsProblem=Error
-org.eclipse.cdt.codan.internal.checkers.InvalidTemplateArgumentsProblem.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"Invalid template argument\\")"}
-org.eclipse.cdt.codan.internal.checkers.LabelStatementNotFoundProblem=Error
-org.eclipse.cdt.codan.internal.checkers.LabelStatementNotFoundProblem.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"Label statement not found\\")"}
-org.eclipse.cdt.codan.internal.checkers.MemberDeclarationNotFoundProblem=Error
-org.eclipse.cdt.codan.internal.checkers.MemberDeclarationNotFoundProblem.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"Member declaration not found\\")"}
-org.eclipse.cdt.codan.internal.checkers.MethodResolutionProblem=Error
-org.eclipse.cdt.codan.internal.checkers.MethodResolutionProblem.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"Method cannot be resolved\\")"}
-org.eclipse.cdt.codan.internal.checkers.NamingConventionFunctionChecker=-Info
-org.eclipse.cdt.codan.internal.checkers.NamingConventionFunctionChecker.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"Name convention for function\\")",pattern\=>"^[a-z]",macro\=>true,exceptions\=>()}
-org.eclipse.cdt.codan.internal.checkers.NonVirtualDestructorProblem=Warning
-org.eclipse.cdt.codan.internal.checkers.NonVirtualDestructorProblem.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"Class has a virtual method and non-virtual destructor\\")"}
-org.eclipse.cdt.codan.internal.checkers.OverloadProblem=Error
-org.eclipse.cdt.codan.internal.checkers.OverloadProblem.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"Invalid overload\\")"}
-org.eclipse.cdt.codan.internal.checkers.RedeclarationProblem=Error
-org.eclipse.cdt.codan.internal.checkers.RedeclarationProblem.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"Invalid redeclaration\\")"}
-org.eclipse.cdt.codan.internal.checkers.RedefinitionProblem=Error
-org.eclipse.cdt.codan.internal.checkers.RedefinitionProblem.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"Invalid redefinition\\")"}
-org.eclipse.cdt.codan.internal.checkers.ReturnStyleProblem=-Warning
-org.eclipse.cdt.codan.internal.checkers.ReturnStyleProblem.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"Return with parenthesis\\")"}
-org.eclipse.cdt.codan.internal.checkers.ScanfFormatStringSecurityProblem=-Warning
-org.eclipse.cdt.codan.internal.checkers.ScanfFormatStringSecurityProblem.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"Format String Vulnerability\\")"}
-org.eclipse.cdt.codan.internal.checkers.StatementHasNoEffectProblem=Warning
-org.eclipse.cdt.codan.internal.checkers.StatementHasNoEffectProblem.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"Statement has no effect\\")",macro\=>true,exceptions\=>()}
-org.eclipse.cdt.codan.internal.checkers.SuggestedParenthesisProblem=Warning
-org.eclipse.cdt.codan.internal.checkers.SuggestedParenthesisProblem.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"Suggested parenthesis around expression\\")",paramNot\=>false}
-org.eclipse.cdt.codan.internal.checkers.SuspiciousSemicolonProblem=Warning
-org.eclipse.cdt.codan.internal.checkers.SuspiciousSemicolonProblem.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"Suspicious semicolon\\")",else\=>false,afterelse\=>false}
-org.eclipse.cdt.codan.internal.checkers.TypeResolutionProblem=Error
-org.eclipse.cdt.codan.internal.checkers.TypeResolutionProblem.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"Type cannot be resolved\\")"}
-org.eclipse.cdt.codan.internal.checkers.UnusedFunctionDeclarationProblem=Warning
-org.eclipse.cdt.codan.internal.checkers.UnusedFunctionDeclarationProblem.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"Unused function declaration\\")",macro\=>true}
-org.eclipse.cdt.codan.internal.checkers.UnusedStaticFunctionProblem=Warning
-org.eclipse.cdt.codan.internal.checkers.UnusedStaticFunctionProblem.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"Unused static function\\")",macro\=>true}
-org.eclipse.cdt.codan.internal.checkers.UnusedVariableDeclarationProblem=Warning
-org.eclipse.cdt.codan.internal.checkers.UnusedVariableDeclarationProblem.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"Unused variable declaration in file scope\\")",macro\=>true,exceptions\=>("@(\#)","$Id")}
-org.eclipse.cdt.codan.internal.checkers.VariableResolutionProblem=Error
-org.eclipse.cdt.codan.internal.checkers.VariableResolutionProblem.params={launchModes\=>{RUN_ON_FULL_BUILD\=>true,RUN_ON_INC_BUILD\=>true,RUN_ON_FILE_OPEN\=>false,RUN_ON_FILE_SAVE\=>false,RUN_AS_YOU_TYPE\=>true,RUN_ON_DEMAND\=>true},suppression_comment\=>"@suppress(\\"Symbol is not resolved\\")"}

.settings/org.eclipse.cdt.core.prefs

@@ -1,13 +0,0 @@
-eclipse.preferences.version=1
-environment/project/cdt.managedbuild.config.gnu.exe.debug.1751741082/append=true
-environment/project/cdt.managedbuild.config.gnu.exe.debug.1751741082/appendContributed=true
-environment/project/cdt.managedbuild.config.gnu.exe.release.1745230171.1259602404/LLVM_HOME/delimiter=\:
-environment/project/cdt.managedbuild.config.gnu.exe.release.1745230171.1259602404/LLVM_HOME/operation=append
-environment/project/cdt.managedbuild.config.gnu.exe.release.1745230171.1259602404/LLVM_HOME/value=/usr/lib/llvm-6.0
-environment/project/cdt.managedbuild.config.gnu.exe.release.1745230171.1259602404/append=true
-environment/project/cdt.managedbuild.config.gnu.exe.release.1745230171.1259602404/appendContributed=true
-environment/project/cdt.managedbuild.config.gnu.exe.release.1745230171/LLVM_HOME/delimiter=\:
-environment/project/cdt.managedbuild.config.gnu.exe.release.1745230171/LLVM_HOME/operation=append
-environment/project/cdt.managedbuild.config.gnu.exe.release.1745230171/LLVM_HOME/value=/usr/lib/llvm-6.0
-environment/project/cdt.managedbuild.config.gnu.exe.release.1745230171/append=true
-environment/project/cdt.managedbuild.config.gnu.exe.release.1745230171/appendContributed=true

.settings/org.eclipse.cdt.managedbuilder.core.prefs

@@ -1,37 +0,0 @@
-eclipse.preferences.version=1
-environment/buildEnvironmentInclude/cdt.managedbuild.config.gnu.exe.debug.1751741082/CPATH/delimiter=\:
-environment/buildEnvironmentInclude/cdt.managedbuild.config.gnu.exe.debug.1751741082/CPATH/operation=remove
-environment/buildEnvironmentInclude/cdt.managedbuild.config.gnu.exe.debug.1751741082/CPLUS_INCLUDE_PATH/delimiter=\:
-environment/buildEnvironmentInclude/cdt.managedbuild.config.gnu.exe.debug.1751741082/CPLUS_INCLUDE_PATH/operation=remove
-environment/buildEnvironmentInclude/cdt.managedbuild.config.gnu.exe.debug.1751741082/C_INCLUDE_PATH/delimiter=\:
-environment/buildEnvironmentInclude/cdt.managedbuild.config.gnu.exe.debug.1751741082/C_INCLUDE_PATH/operation=remove
-environment/buildEnvironmentInclude/cdt.managedbuild.config.gnu.exe.debug.1751741082/append=true
-environment/buildEnvironmentInclude/cdt.managedbuild.config.gnu.exe.debug.1751741082/appendContributed=true
-environment/buildEnvironmentInclude/cdt.managedbuild.config.gnu.exe.release.1745230171.1259602404/CPATH/delimiter=\:
-environment/buildEnvironmentInclude/cdt.managedbuild.config.gnu.exe.release.1745230171.1259602404/CPATH/operation=remove
-environment/buildEnvironmentInclude/cdt.managedbuild.config.gnu.exe.release.1745230171.1259602404/CPLUS_INCLUDE_PATH/delimiter=\:
-environment/buildEnvironmentInclude/cdt.managedbuild.config.gnu.exe.release.1745230171.1259602404/CPLUS_INCLUDE_PATH/operation=remove
-environment/buildEnvironmentInclude/cdt.managedbuild.config.gnu.exe.release.1745230171.1259602404/C_INCLUDE_PATH/delimiter=\:
-environment/buildEnvironmentInclude/cdt.managedbuild.config.gnu.exe.release.1745230171.1259602404/C_INCLUDE_PATH/operation=remove
-environment/buildEnvironmentInclude/cdt.managedbuild.config.gnu.exe.release.1745230171.1259602404/append=true
-environment/buildEnvironmentInclude/cdt.managedbuild.config.gnu.exe.release.1745230171.1259602404/appendContributed=true
-environment/buildEnvironmentInclude/cdt.managedbuild.config.gnu.exe.release.1745230171/CPATH/delimiter=\:
-environment/buildEnvironmentInclude/cdt.managedbuild.config.gnu.exe.release.1745230171/CPATH/operation=remove
-environment/buildEnvironmentInclude/cdt.managedbuild.config.gnu.exe.release.1745230171/CPLUS_INCLUDE_PATH/delimiter=\:
-environment/buildEnvironmentInclude/cdt.managedbuild.config.gnu.exe.release.1745230171/CPLUS_INCLUDE_PATH/operation=remove
-environment/buildEnvironmentInclude/cdt.managedbuild.config.gnu.exe.release.1745230171/C_INCLUDE_PATH/delimiter=\:
-environment/buildEnvironmentInclude/cdt.managedbuild.config.gnu.exe.release.1745230171/C_INCLUDE_PATH/operation=remove
-environment/buildEnvironmentInclude/cdt.managedbuild.config.gnu.exe.release.1745230171/append=true
-environment/buildEnvironmentInclude/cdt.managedbuild.config.gnu.exe.release.1745230171/appendContributed=true
-environment/buildEnvironmentLibrary/cdt.managedbuild.config.gnu.exe.debug.1751741082/LIBRARY_PATH/delimiter=\:
-environment/buildEnvironmentLibrary/cdt.managedbuild.config.gnu.exe.debug.1751741082/LIBRARY_PATH/operation=remove
-environment/buildEnvironmentLibrary/cdt.managedbuild.config.gnu.exe.debug.1751741082/append=true
-environment/buildEnvironmentLibrary/cdt.managedbuild.config.gnu.exe.debug.1751741082/appendContributed=true
-environment/buildEnvironmentLibrary/cdt.managedbuild.config.gnu.exe.release.1745230171.1259602404/LIBRARY_PATH/delimiter=\:
-environment/buildEnvironmentLibrary/cdt.managedbuild.config.gnu.exe.release.1745230171.1259602404/LIBRARY_PATH/operation=remove
-environment/buildEnvironmentLibrary/cdt.managedbuild.config.gnu.exe.release.1745230171.1259602404/append=true
-environment/buildEnvironmentLibrary/cdt.managedbuild.config.gnu.exe.release.1745230171.1259602404/appendContributed=true
-environment/buildEnvironmentLibrary/cdt.managedbuild.config.gnu.exe.release.1745230171/LIBRARY_PATH/delimiter=\:
-environment/buildEnvironmentLibrary/cdt.managedbuild.config.gnu.exe.release.1745230171/LIBRARY_PATH/operation=remove
-environment/buildEnvironmentLibrary/cdt.managedbuild.config.gnu.exe.release.1745230171/append=true
-environment/buildEnvironmentLibrary/cdt.managedbuild.config.gnu.exe.release.1745230171/appendContributed=true

CMakeLists.txt

@@ -1,8 +1,9 @@
-cmake_minimum_required(VERSION 3.12)
+cmake_minimum_required(VERSION 3.18)
 list(APPEND CMAKE_MODULE_PATH ${CMAKE_CURRENT_SOURCE_DIR}/cmake)
-###############################################################################
+
+# ##############################################################################
 #
-###############################################################################
+# ##############################################################################
 project(dbt-rise-tgc VERSION 1.0.0)
 
 include(GNUInstallDirs)
@@ -15,32 +16,45 @@ find_package(Boost COMPONENTS coroutine REQUIRED)
 add_subdirectory(softfloat)
 
 set(LIB_SOURCES
-	src/iss/arch/tgc_c.cpp
+    src/iss/plugin/instruction_count.cpp
-	src/vm/tcc/vm_tgc_c.cpp
+    src/iss/arch/tgc5c.cpp
-	src/vm/interp/vm_tgc_c.cpp
+    src/vm/interp/vm_tgc5c.cpp
     src/vm/fp_functions.cpp
+    src/iss/debugger/csr_names.cpp
+    src/iss/semihosting/semihosting.cpp
 )
+
 if(WITH_TCC)
     list(APPEND LIB_SOURCES
-		src/vm/tcc/vm_tgc_c.cpp
+        src/vm/tcc/vm_tgc5c.cpp
     )
 endif()
+
 if(WITH_LLVM)
     list(APPEND LIB_SOURCES
-		src/vm/llvm/vm_tgc_c.cpp
+        src/vm/llvm/vm_tgc5c.cpp
         src/vm/llvm/fp_impl.cpp
     )
 endif()
 
+if(WITH_ASMJIT)
+    list(APPEND LIB_SOURCES
+        src/vm/asmjit/vm_tgc5c.cpp
+    )
+endif()
+
 # library files
 FILE(GLOB GEN_ISS_SOURCES ${CMAKE_CURRENT_SOURCE_DIR}/src-gen/iss/arch/*.cpp)
 FILE(GLOB GEN_VM_SOURCES ${CMAKE_CURRENT_SOURCE_DIR}/src-gen/vm/interp/vm_*.cpp)
+FILE(GLOB GEN_YAML_SOURCES ${CMAKE_CURRENT_SOURCE_DIR}/contrib/instr/*.yaml)
 list(APPEND LIB_SOURCES ${GEN_ISS_SOURCES} ${GEN_VM_SOURCES})
+
 foreach(FILEPATH ${GEN_ISS_SOURCES})
     get_filename_component(CORE ${FILEPATH} NAME_WE)
     string(TOUPPER ${CORE} CORE)
     list(APPEND LIB_DEFINES CORE_${CORE})
 endforeach()
+
 message(STATUS "Core defines are ${LIB_DEFINES}")
 
 if(WITH_LLVM)
@@ -53,17 +67,18 @@ if(WITH_TCC)
     list(APPEND LIB_SOURCES ${TCC_GEN_SOURCES})
 endif()
 
-if(TARGET RapidJSON)
+if(WITH_ASMJIT)
+    FILE(GLOB TCC_GEN_SOURCES ${CMAKE_CURRENT_SOURCE_DIR}/src-gen/vm/asmjit/vm_*.cpp)
+    list(APPEND LIB_SOURCES ${TCC_GEN_SOURCES})
+endif()
+
+if(TARGET yaml-cpp::yaml-cpp)
     list(APPEND LIB_SOURCES
         src/iss/plugin/cycle_estimate.cpp
-    	src/iss/plugin/pctrace.cpp
-    )
-endif()
-if(TARGET jsoncpp::jsoncpp)
-    list(APPEND LIB_SOURCES 
         src/iss/plugin/instruction_count.cpp
     )
 endif()
+
 # Define the library
 add_library(${PROJECT_NAME} SHARED ${LIB_SOURCES})
 
@@ -72,28 +87,37 @@ if("${CMAKE_CXX_COMPILER_ID}" STREQUAL "GNU")
 elseif("${CMAKE_CXX_COMPILER_ID}" STREQUAL "MSVC")
     target_compile_options(${PROJECT_NAME} PRIVATE /wd4293)
 endif()
+
 target_include_directories(${PROJECT_NAME} PUBLIC src)
 target_include_directories(${PROJECT_NAME} PUBLIC src-gen)
 
 target_force_link_libraries(${PROJECT_NAME} PRIVATE dbt-rise-core)
+
 # only re-export the include paths
 get_target_property(DBT_CORE_INCL dbt-rise-core INTERFACE_INCLUDE_DIRECTORIES)
 target_include_directories(${PROJECT_NAME} INTERFACE ${DBT_CORE_INCL})
 get_target_property(DBT_CORE_DEFS dbt-rise-core INTERFACE_COMPILE_DEFINITIONS)
-target_compile_definitions(${PROJECT_NAME} INTERFACE ${DBT_CORE_DEFS})
+
+if(NOT(DBT_CORE_DEFS STREQUAL DBT_CORE_DEFS-NOTFOUND))
+    target_compile_definitions(${PROJECT_NAME} INTERFACE ${DBT_CORE_DEFS})
+endif()
 
 target_link_libraries(${PROJECT_NAME} PUBLIC elfio::elfio softfloat scc-util Boost::coroutine)
-if(TARGET jsoncpp::jsoncpp)
+
-	target_link_libraries(${PROJECT_NAME} PUBLIC jsoncpp::jsoncpp)
+if(TARGET yaml-cpp::yaml-cpp)
-endif()
+    target_compile_definitions(${PROJECT_NAME} PUBLIC WITH_PLUGINS)
-if(TARGET lz4::lz4)
+    target_link_libraries(${PROJECT_NAME} PUBLIC yaml-cpp::yaml-cpp)
-    target_compile_definitions(${PROJECT_NAME} PUBLIC WITH_LZ4)
-    target_link_libraries(${PROJECT_NAME} PUBLIC lz4::lz4)
-endif()
-if(TARGET RapidJSON)
-    target_link_libraries(${PROJECT_NAME} PUBLIC RapidJSON)
 endif()
 
+# if(WITH_LLVM)
+#     target_compile_definitions(${PROJECT_NAME} PUBLIC ${LLVM_DEFINITIONS})
+#     target_include_directories(${PROJECT_NAME} PUBLIC ${LLVM_INCLUDE_DIRS})
+
+#     if(BUILD_SHARED_LIBS)
+#         target_link_libraries(${PROJECT_NAME} PUBLIC ${LLVM_LIBRARIES})
+#     endif()
+# endif()
+
 set_target_properties(${PROJECT_NAME} PROPERTIES
     VERSION ${PROJECT_VERSION}
     FRAMEWORK FALSE
@@ -111,14 +135,18 @@ install(DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR}/incl/iss COMPONENT ${PROJECT_NAME}
     DESTINATION ${CMAKE_INSTALL_INCLUDEDIR} # target directory
     FILES_MATCHING # install only matched files
     PATTERN "*.h" # select header files
-        )
+)
-###############################################################################
+install(FILES ${GEN_YAML_SOURCES} DESTINATION share/tgc-vp)
+
+# ##############################################################################
 #
-###############################################################################
+# ##############################################################################
+set(CMAKE_INSTALL_RPATH $ORIGIN/../${CMAKE_INSTALL_LIBDIR})
 project(tgc-sim)
 find_package(Boost COMPONENTS program_options thread REQUIRED)
 
 add_executable(${PROJECT_NAME} src/main.cpp)
+
 if(TARGET ${CORE_NAME}_cpp)
     list(APPEND TGC_SOURCES ${${CORE_NAME}_OUTPUT_FILES})
 else()
@@ -130,21 +158,20 @@ else()
 endif()
 
 foreach(F IN LISTS TGC_SOURCES)
-    if (${F} MATCHES ".*/arch/([^/]*)\.cpp")
+    if(${F} MATCHES ".*/arch/([^/]*)\.cpp")
         string(REGEX REPLACE ".*/([^/]*)\.cpp" "\\1" CORE_NAME_LC ${F})
         string(TOUPPER ${CORE_NAME_LC} CORE_NAME)
         target_compile_definitions(${PROJECT_NAME} PRIVATE CORE_${CORE_NAME})
     endif()
 endforeach()
 
-#if(WITH_LLVM)
+# if(WITH_LLVM)
 # target_compile_definitions(${PROJECT_NAME} PRIVATE WITH_LLVM)
 # #target_link_libraries(${PROJECT_NAME} PUBLIC ${llvm_libs})
-#endif()
+# endif()
-#if(WITH_TCC)
+# if(WITH_TCC)
 # target_compile_definitions(${PROJECT_NAME} PRIVATE WITH_TCC)
-#endif()
+# endif()
-
 target_link_libraries(${PROJECT_NAME} PUBLIC dbt-rise-tgc fmt::fmt)
 
 if(TARGET Boost::program_options)
@@ -152,8 +179,10 @@ if(TARGET Boost::program_options)
 else()
     target_link_libraries(${PROJECT_NAME} PUBLIC ${BOOST_program_options_LIBRARY})
 endif()
+
 target_link_libraries(${PROJECT_NAME} PUBLIC ${CMAKE_DL_LIBS})
-if (Tcmalloc_FOUND)
+
+if(Tcmalloc_FOUND)
     target_link_libraries(${PROJECT_NAME} PUBLIC ${Tcmalloc_LIBRARIES})
 endif(Tcmalloc_FOUND)
 
@@ -171,18 +200,26 @@ if(BUILD_TESTING)
     # ... CMake code to create tests ...
     add_test(NAME tgc-sim-interp
         COMMAND tgc-sim -f ${CMAKE_BINARY_DIR}/../../Firmwares/hello-world/hello --backend interp)
+
     if(WITH_TCC)
         add_test(NAME tgc-sim-tcc
             COMMAND tgc-sim -f ${CMAKE_BINARY_DIR}/../../Firmwares/hello-world/hello --backend tcc)
     endif()
+
     if(WITH_LLVM)
         add_test(NAME tgc-sim-llvm
             COMMAND tgc-sim -f ${CMAKE_BINARY_DIR}/../../Firmwares/hello-world/hello --backend llvm)
     endif()
+
+    if(WITH_ASMJIT)
+        add_test(NAME tgc-sim-asmjit
+            COMMAND tgc-sim -f ${CMAKE_BINARY_DIR}/../../Firmwares/hello-world/hello --backend asmjit)
+    endif()
 endif()
-###############################################################################
+
+# ##############################################################################
 #
-###############################################################################
+# ##############################################################################
 if(TARGET scc-sysc)
     project(dbt-rise-tgc_sc VERSION 1.0.0)
     set(LIB_SOURCES
@@ -194,18 +231,20 @@ if(TARGET scc-sysc)
     add_library(${PROJECT_NAME} ${LIB_SOURCES})
     target_compile_definitions(${PROJECT_NAME} PUBLIC WITH_SYSTEMC)
     target_compile_definitions(${PROJECT_NAME} PRIVATE CORE_${CORE_NAME})
+
     foreach(F IN LISTS TGC_SOURCES)
-        if (${F} MATCHES ".*/arch/([^/]*)\.cpp")
+        if(${F} MATCHES ".*/arch/([^/]*)\.cpp")
             string(REGEX REPLACE ".*/([^/]*)\.cpp" "\\1" CORE_NAME_LC ${F})
             string(TOUPPER ${CORE_NAME_LC} CORE_NAME)
             target_compile_definitions(${PROJECT_NAME} PRIVATE CORE_${CORE_NAME})
         endif()
     endforeach()
-    target_link_libraries(${PROJECT_NAME} PUBLIC dbt-rise-tgc scc-sysc)
-#    if(WITH_LLVM)
-#        target_link_libraries(${PROJECT_NAME} PUBLIC ${llvm_libs})
-#    endif()
 
+    target_link_libraries(${PROJECT_NAME} PUBLIC dbt-rise-tgc scc-sysc)
+
+    # if(WITH_LLVM)
+    # target_link_libraries(${PROJECT_NAME} PUBLIC ${llvm_libs})
+    # endif()
     set(LIB_HEADERS ${CMAKE_CURRENT_SOURCE_DIR}/src/sysc/core_complex.h)
     set_target_properties(${PROJECT_NAME} PROPERTIES
         VERSION ${PROJECT_VERSION}
@@ -223,3 +262,8 @@ if(TARGET scc-sysc)
     )
 endif()
 
+project(elfio-test)
+find_package(Boost COMPONENTS program_options thread REQUIRED)
+
+add_executable(${PROJECT_NAME} src/elfio.cpp)
+target_link_libraries(${PROJECT_NAME} PUBLIC elfio::elfio)

contrib/instr/.gitignore

@@ -0,0 +1 @@
+/*.yaml

contrib/instr/TGC5C_instr.yaml

@@ -1,536 +1,623 @@
 
-RV32I: 
+RVI: 
-  - LUI:
+  LUI:
+    index: 0
     encoding: 0b00000000000000000000000000110111
     mask: 0b00000000000000000000000001111111
     size:   32
     branch:   false
     delay:   1
-  - AUIPC:
+  AUIPC:
+    index: 1
     encoding: 0b00000000000000000000000000010111
     mask: 0b00000000000000000000000001111111
     size:   32
     branch:   false
     delay:   1
-  - JAL:
+  JAL:
+    index: 2
     encoding: 0b00000000000000000000000001101111
     mask: 0b00000000000000000000000001111111
-    attributes: [[name:no_cont]]
     size:   32
     branch:   true
     delay:   1
-  - JALR:
+  JALR:
+    index: 3
     encoding: 0b00000000000000000000000001100111
     mask: 0b00000000000000000111000001111111
-    attributes: [[name:no_cont]]
     size:   32
     branch:   true
-    delay:   1
+    delay:   [1,1]
-  - BEQ:
+  BEQ:
+    index: 4
     encoding: 0b00000000000000000000000001100011
     mask: 0b00000000000000000111000001111111
-    attributes: [[name:no_cont], [name:cond]]
     size:   32
     branch:   true
     delay:   [1,1]
-  - BNE:
+  BNE:
+    index: 5
     encoding: 0b00000000000000000001000001100011
     mask: 0b00000000000000000111000001111111
-    attributes: [[name:no_cont], [name:cond]]
     size:   32
     branch:   true
     delay:   [1,1]
-  - BLT:
+  BLT:
+    index: 6
     encoding: 0b00000000000000000100000001100011
     mask: 0b00000000000000000111000001111111
-    attributes: [[name:no_cont], [name:cond]]
     size:   32
     branch:   true
     delay:   [1,1]
-  - BGE:
+  BGE:
+    index: 7
     encoding: 0b00000000000000000101000001100011
     mask: 0b00000000000000000111000001111111
-    attributes: [[name:no_cont], [name:cond]]
     size:   32
     branch:   true
     delay:   [1,1]
-  - BLTU:
+  BLTU:
+    index: 8
     encoding: 0b00000000000000000110000001100011
     mask: 0b00000000000000000111000001111111
-    attributes: [[name:no_cont], [name:cond]]
     size:   32
     branch:   true
     delay:   [1,1]
-  - BGEU:
+  BGEU:
+    index: 9
     encoding: 0b00000000000000000111000001100011
     mask: 0b00000000000000000111000001111111
-    attributes: [[name:no_cont], [name:cond]]
     size:   32
     branch:   true
     delay:   [1,1]
-  - LB:
+  LB:
+    index: 10
     encoding: 0b00000000000000000000000000000011
     mask: 0b00000000000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - LH:
+  LH:
+    index: 11
     encoding: 0b00000000000000000001000000000011
     mask: 0b00000000000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - LW:
+  LW:
+    index: 12
     encoding: 0b00000000000000000010000000000011
     mask: 0b00000000000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - LBU:
+  LBU:
+    index: 13
     encoding: 0b00000000000000000100000000000011
     mask: 0b00000000000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - LHU:
+  LHU:
+    index: 14
     encoding: 0b00000000000000000101000000000011
     mask: 0b00000000000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - SB:
+  SB:
+    index: 15
     encoding: 0b00000000000000000000000000100011
     mask: 0b00000000000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - SH:
+  SH:
+    index: 16
     encoding: 0b00000000000000000001000000100011
     mask: 0b00000000000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - SW:
+  SW:
+    index: 17
     encoding: 0b00000000000000000010000000100011
     mask: 0b00000000000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - ADDI:
+  ADDI:
+    index: 18
     encoding: 0b00000000000000000000000000010011
     mask: 0b00000000000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - SLTI:
+  SLTI:
+    index: 19
     encoding: 0b00000000000000000010000000010011
     mask: 0b00000000000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - SLTIU:
+  SLTIU:
+    index: 20
     encoding: 0b00000000000000000011000000010011
     mask: 0b00000000000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - XORI:
+  XORI:
+    index: 21
     encoding: 0b00000000000000000100000000010011
     mask: 0b00000000000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - ORI:
+  ORI:
+    index: 22
     encoding: 0b00000000000000000110000000010011
     mask: 0b00000000000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - ANDI:
+  ANDI:
+    index: 23
     encoding: 0b00000000000000000111000000010011
     mask: 0b00000000000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - SLLI:
+  SLLI:
+    index: 24
     encoding: 0b00000000000000000001000000010011
     mask: 0b11111110000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - SRLI:
+  SRLI:
+    index: 25
     encoding: 0b00000000000000000101000000010011
     mask: 0b11111110000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - SRAI:
+  SRAI:
+    index: 26
     encoding: 0b01000000000000000101000000010011
     mask: 0b11111110000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - ADD:
+  ADD:
+    index: 27
     encoding: 0b00000000000000000000000000110011
     mask: 0b11111110000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - SUB:
+  SUB:
+    index: 28
     encoding: 0b01000000000000000000000000110011
     mask: 0b11111110000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - SLL:
+  SLL:
+    index: 29
     encoding: 0b00000000000000000001000000110011
     mask: 0b11111110000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - SLT:
+  SLT:
+    index: 30
     encoding: 0b00000000000000000010000000110011
     mask: 0b11111110000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - SLTU:
+  SLTU:
+    index: 31
     encoding: 0b00000000000000000011000000110011
     mask: 0b11111110000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - XOR:
+  XOR:
+    index: 32
     encoding: 0b00000000000000000100000000110011
     mask: 0b11111110000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - SRL:
+  SRL:
+    index: 33
     encoding: 0b00000000000000000101000000110011
     mask: 0b11111110000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - SRA:
+  SRA:
+    index: 34
     encoding: 0b01000000000000000101000000110011
     mask: 0b11111110000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - OR:
+  OR:
+    index: 35
     encoding: 0b00000000000000000110000000110011
     mask: 0b11111110000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - AND:
+  AND:
+    index: 36
     encoding: 0b00000000000000000111000000110011
     mask: 0b11111110000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - FENCE:
+  FENCE:
+    index: 37
     encoding: 0b00000000000000000000000000001111
     mask: 0b00000000000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - ECALL:
+  ECALL:
+    index: 38
     encoding: 0b00000000000000000000000001110011
     mask: 0b11111111111111111111111111111111
     attributes: [[name:no_cont]]
     size:   32
     branch:   false
     delay:   1
-  - EBREAK:
+  EBREAK:
+    index: 39
     encoding: 0b00000000000100000000000001110011
     mask: 0b11111111111111111111111111111111
     attributes: [[name:no_cont]]
     size:   32
     branch:   false
     delay:   1
-  - MRET:
+  MRET:
+    index: 40
     encoding: 0b00110000001000000000000001110011
     mask: 0b11111111111111111111111111111111
     attributes: [[name:no_cont]]
     size:   32
     branch:   false
     delay:   1
-  - WFI:
+  WFI:
+    index: 41
     encoding: 0b00010000010100000000000001110011
     mask: 0b11111111111111111111111111111111
     size:   32
     branch:   false
     delay:   1
 Zicsr: 
-  - CSRRW:
+  CSRRW:
+    index: 42
     encoding: 0b00000000000000000001000001110011
     mask: 0b00000000000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - CSRRS:
+  CSRRS:
+    index: 43
     encoding: 0b00000000000000000010000001110011
     mask: 0b00000000000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - CSRRC:
+  CSRRC:
+    index: 44
     encoding: 0b00000000000000000011000001110011
     mask: 0b00000000000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - CSRRWI:
+  CSRRWI:
+    index: 45
     encoding: 0b00000000000000000101000001110011
     mask: 0b00000000000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - CSRRSI:
+  CSRRSI:
+    index: 46
     encoding: 0b00000000000000000110000001110011
     mask: 0b00000000000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - CSRRCI:
+  CSRRCI:
+    index: 47
     encoding: 0b00000000000000000111000001110011
     mask: 0b00000000000000000111000001111111
     size:   32
     branch:   false
     delay:   1
 Zifencei: 
-  - FENCE_I:
+  FENCE_I:
+    index: 48
     encoding: 0b00000000000000000001000000001111
     mask: 0b00000000000000000111000001111111
     attributes: [[name:flush]]
     size:   32
     branch:   false
     delay:   1
-RV32M: 
+RVM: 
-  - MUL:
+  MUL:
+    index: 49
     encoding: 0b00000010000000000000000000110011
     mask: 0b11111110000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - MULH:
+  MULH:
+    index: 50
     encoding: 0b00000010000000000001000000110011
     mask: 0b11111110000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - MULHSU:
+  MULHSU:
+    index: 51
     encoding: 0b00000010000000000010000000110011
     mask: 0b11111110000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - MULHU:
+  MULHU:
+    index: 52
     encoding: 0b00000010000000000011000000110011
     mask: 0b11111110000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - DIV:
+  DIV:
+    index: 53
     encoding: 0b00000010000000000100000000110011
     mask: 0b11111110000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - DIVU:
+  DIVU:
+    index: 54
     encoding: 0b00000010000000000101000000110011
     mask: 0b11111110000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - REM:
+  REM:
+    index: 55
     encoding: 0b00000010000000000110000000110011
     mask: 0b11111110000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-  - REMU:
+  REMU:
+    index: 56
     encoding: 0b00000010000000000111000000110011
     mask: 0b11111110000000000111000001111111
     size:   32
     branch:   false
     delay:   1
-RV32IC: 
+Zca: 
-  - CADDI4SPN:
+  C__ADDI4SPN:
+    index: 57
     encoding: 0b0000000000000000
     mask: 0b1110000000000011
     size:   16
     branch:   false
     delay:   1
-  - CLW:
+  C__LW:
+    index: 58
     encoding: 0b0100000000000000
     mask: 0b1110000000000011
     size:   16
     branch:   false
     delay:   1
-  - CSW:
+  C__SW:
+    index: 59
     encoding: 0b1100000000000000
     mask: 0b1110000000000011
     size:   16
     branch:   false
     delay:   1
-  - CADDI:
+  C__ADDI:
+    index: 60
     encoding: 0b0000000000000001
     mask: 0b1110000000000011
     size:   16
     branch:   false
     delay:   1
-  - CNOP:
+  C__NOP:
+    index: 61
     encoding: 0b0000000000000001
     mask: 0b1110111110000011
     size:   16
     branch:   false
     delay:   1
-  - CJAL:
+  C__JAL:
+    index: 62
     encoding: 0b0010000000000001
     mask: 0b1110000000000011
-    attributes: [[name:no_cont]]
+    attributes: [[name:enable, value:1]]
     size:   16
     branch:   true
     delay:   1
-  - CLI:
+  C__LI:
+    index: 63
     encoding: 0b0100000000000001
     mask: 0b1110000000000011
     size:   16
     branch:   false
     delay:   1
-  - CLUI:
+  C__LUI:
+    index: 64
     encoding: 0b0110000000000001
     mask: 0b1110000000000011
     size:   16
     branch:   false
     delay:   1
-  - CADDI16SP:
+  C__ADDI16SP:
+    index: 65
     encoding: 0b0110000100000001
     mask: 0b1110111110000011
     size:   16
     branch:   false
     delay:   1
-  - CSRLI:
+  __reserved_clui:
+    index: 66
+    encoding: 0b0110000000000001
+    mask: 0b1111000001111111
+    size:   16
+    branch:   false
+    delay:   1
+  C__SRLI:
+    index: 67
     encoding: 0b1000000000000001
     mask: 0b1111110000000011
+    attributes: [[name:enable, value:1]]
     size:   16
     branch:   false
     delay:   1
-  - CSRAI:
+  C__SRAI:
+    index: 68
     encoding: 0b1000010000000001
     mask: 0b1111110000000011
+    attributes: [[name:enable, value:1]]
     size:   16
     branch:   false
     delay:   1
-  - CANDI:
+  C__ANDI:
+    index: 69
     encoding: 0b1000100000000001
     mask: 0b1110110000000011
     size:   16
     branch:   false
     delay:   1
-  - CSUB:
+  C__SUB:
+    index: 70
     encoding: 0b1000110000000001
     mask: 0b1111110001100011
     size:   16
     branch:   false
     delay:   1
-  - CXOR:
+  C__XOR:
+    index: 71
     encoding: 0b1000110000100001
     mask: 0b1111110001100011
     size:   16
     branch:   false
     delay:   1
-  - COR:
+  C__OR:
+    index: 72
     encoding: 0b1000110001000001
     mask: 0b1111110001100011
     size:   16
     branch:   false
     delay:   1
-  - CAND:
+  C__AND:
+    index: 73
     encoding: 0b1000110001100001
     mask: 0b1111110001100011
     size:   16
     branch:   false
     delay:   1
-  - CJ:
+  C__J:
+    index: 74
     encoding: 0b1010000000000001
     mask: 0b1110000000000011
-    attributes: [[name:no_cont]]
     size:   16
     branch:   true
     delay:   1
-  - CBEQZ:
+  C__BEQZ:
+    index: 75
     encoding: 0b1100000000000001
     mask: 0b1110000000000011
-    attributes: [[name:no_cont], [name:cond]]
     size:   16
     branch:   true
     delay:   [1,1]
-  - CBNEZ:
+  C__BNEZ:
+    index: 76
     encoding: 0b1110000000000001
     mask: 0b1110000000000011
-    attributes: [[name:no_cont], [name:cond]]
     size:   16
     branch:   true
     delay:   [1,1]
-  - CSLLI:
+  C__SLLI:
+    index: 77
     encoding: 0b0000000000000010
     mask: 0b1111000000000011
+    attributes: [[name:enable, value:1]]
     size:   16
     branch:   false
     delay:   1
-  - CLWSP:
+  C__LWSP:
+    index: 78
     encoding: 0b0100000000000010
     mask: 0b1110000000000011
     size:   16
     branch:   false
     delay:   1
-  - CMV:
+  C__MV:
+    index: 79
     encoding: 0b1000000000000010
     mask: 0b1111000000000011
     size:   16
     branch:   false
     delay:   1
-  - CJR:
+  C__JR:
+    index: 80
     encoding: 0b1000000000000010
     mask: 0b1111000001111111
-    attributes: [[name:no_cont]]
     size:   16
     branch:   true
     delay:   1
-  - CADD:
+  __reserved_cmv:
+    index: 81
+    encoding: 0b1000000000000010
+    mask: 0b1111111111111111
+    size:   16
+    branch:   false
+    delay:   1
+  C__ADD:
+    index: 82
     encoding: 0b1001000000000010
     mask: 0b1111000000000011
     size:   16
     branch:   false
     delay:   1
-  - CJALR:
+  C__JALR:
+    index: 83
     encoding: 0b1001000000000010
     mask: 0b1111000001111111
-    attributes: [[name:no_cont]]
     size:   16
     branch:   true
     delay:   1
-  - CEBREAK:
+  C__EBREAK:
+    index: 84
     encoding: 0b1001000000000010
     mask: 0b1111111111111111
-    attributes: [[name:no_cont]]
     size:   16
     branch:   false
     delay:   1
-  - CSWSP:
+  C__SWSP:
+    index: 85
     encoding: 0b1100000000000010
     mask: 0b1110000000000011
     size:   16
     branch:   false
     delay:   1
-  - DII:
+  DII:
+    index: 86
     encoding: 0b0000000000000000
     mask: 0b1111111111111111
-    attributes: [[name:no_cont]]
     size:   16
     branch:   false
     delay:   1

contrib/instr/TGC5C_slow.yaml

@@ -0,0 +1,650 @@
+RV32I:
+  ADD:
+    branch: false
+    delay: 1
+    encoding: 51
+    index: 27
+    mask: 4261441663
+    size: 32
+  ADDI:
+    branch: false
+    delay: 1
+    encoding: 19
+    index: 18
+    mask: 28799
+    size: 32
+  AND:
+    branch: false
+    delay: 1
+    encoding: 28723
+    index: 36
+    mask: 4261441663
+    size: 32
+  ANDI:
+    branch: false
+    delay: 1
+    encoding: 28691
+    index: 23
+    mask: 28799
+    size: 32
+  AUIPC:
+    branch: false
+    delay: 1
+    encoding: 23
+    index: 1
+    mask: 127
+    size: 32
+  BEQ:
+    branch: true
+    delay:
+    - 1
+    - 2
+    encoding: 99
+    index: 4
+    mask: 28799
+    size: 32
+  BGE:
+    branch: true
+    delay:
+    - 1
+    - 2
+    encoding: 20579
+    index: 7
+    mask: 28799
+    size: 32
+  BGEU:
+    branch: true
+    delay:
+    - 1
+    - 2
+    encoding: 28771
+    index: 9
+    mask: 28799
+    size: 32
+  BLT:
+    branch: true
+    delay:
+    - 1
+    - 2
+    encoding: 16483
+    index: 6
+    mask: 28799
+    size: 32
+  BLTU:
+    branch: true
+    delay:
+    - 1
+    - 2
+    encoding: 24675
+    index: 8
+    mask: 28799
+    size: 32
+  BNE:
+    branch: true
+    delay:
+    - 1
+    - 2
+    encoding: 4195
+    index: 5
+    mask: 28799
+    size: 32
+  EBREAK:
+    attributes:
+    - - name:no_cont
+    branch: false
+    delay: 3
+    encoding: 1048691
+    index: 39
+    mask: 4294967295
+    size: 32
+  ECALL:
+    attributes:
+    - - name:no_cont
+    branch: false
+    delay: 1
+    encoding: 115
+    index: 38
+    mask: 4294967295
+    size: 32
+  FENCE:
+    branch: false
+    delay: 1
+    encoding: 15
+    index: 37
+    mask: 28799
+    size: 32
+  JAL:
+    branch: true
+    delay: 2
+    encoding: 111
+    index: 2
+    mask: 127
+    size: 32
+  JALR:
+    branch: true
+    delay: 2
+    encoding: 103
+    index: 3
+    mask: 28799
+    size: 32
+  LB:
+    branch: false
+    delay: 2
+    encoding: 3
+    index: 10
+    mask: 28799
+    size: 32
+  LBU:
+    branch: false
+    delay: 2
+    encoding: 16387
+    index: 13
+    mask: 28799
+    size: 32
+  LH:
+    branch: false
+    delay: 2
+    encoding: 4099
+    index: 11
+    mask: 28799
+    size: 32
+  LHU:
+    branch: false
+    delay: 2
+    encoding: 20483
+    index: 14
+    mask: 28799
+    size: 32
+  LUI:
+    branch: false
+    delay: 1
+    encoding: 55
+    index: 0
+    mask: 127
+    size: 32
+  LW:
+    branch: false
+    delay: 2
+    encoding: 8195
+    index: 12
+    mask: 28799
+    size: 32
+  MRET:
+    attributes:
+    - - name:no_cont
+    branch: false
+    delay: 2
+    encoding: 807403635
+    index: 40
+    mask: 4294967295
+    size: 32
+  OR:
+    branch: false
+    delay: 1
+    encoding: 24627
+    index: 35
+    mask: 4261441663
+    size: 32
+  ORI:
+    branch: false
+    delay: 1
+    encoding: 24595
+    index: 22
+    mask: 28799
+    size: 32
+  SB:
+    branch: false
+    delay: 1
+    encoding: 35
+    index: 15
+    mask: 28799
+    size: 32
+  SH:
+    branch: false
+    delay: 1
+    encoding: 4131
+    index: 16
+    mask: 28799
+    size: 32
+  SLL:
+    branch: false
+    delay: X_24:20
+    encoding: 4147
+    index: 29
+    mask: 4261441663
+    size: 32
+  SLLI:
+    branch: false
+    delay: u_24:20
+    encoding: 4115
+    index: 24
+    mask: 4261441663
+    size: 32
+  SLT:
+    branch: false
+    delay: 1
+    encoding: 8243
+    index: 30
+    mask: 4261441663
+    size: 32
+  SLTI:
+    branch: false
+    delay: 1
+    encoding: 8211
+    index: 19
+    mask: 28799
+    size: 32
+  SLTIU:
+    branch: false
+    delay: 1
+    encoding: 12307
+    index: 20
+    mask: 28799
+    size: 32
+  SLTU:
+    branch: false
+    delay: 1
+    encoding: 12339
+    index: 31
+    mask: 4261441663
+    size: 32
+  SRA:
+    branch: false
+    delay: X_24:20
+    encoding: 1073762355
+    index: 34
+    mask: 4261441663
+    size: 32
+  SRAI:
+    branch: false
+    delay: u_24:20
+    encoding: 1073762323
+    index: 26
+    mask: 4261441663
+    size: 32
+  SRL:
+    branch: false
+    delay: X_24:20
+    encoding: 20531
+    index: 33
+    mask: 4261441663
+    size: 32
+  SRLI:
+    branch: false
+    delay: u_24:20
+    encoding: 20499
+    index: 25
+    mask: 4261441663
+    size: 32
+  SUB:
+    branch: false
+    delay: 1
+    encoding: 1073741875
+    index: 28
+    mask: 4261441663
+    size: 32
+  SW:
+    branch: false
+    delay: 1
+    encoding: 8227
+    index: 17
+    mask: 28799
+    size: 32
+  WFI:
+    branch: false
+    delay: 1
+    encoding: 273678451
+    index: 41
+    mask: 4294967295
+    size: 32
+  XOR:
+    branch: false
+    delay: 1
+    encoding: 16435
+    index: 32
+    mask: 4261441663
+    size: 32
+  XORI:
+    branch: false
+    delay: 1
+    encoding: 16403
+    index: 21
+    mask: 28799
+    size: 32
+RV32M:
+  DIV:
+    branch: false
+    delay: 33
+    encoding: 33570867
+    index: 53
+    mask: 4261441663
+    size: 32
+  DIVU:
+    branch: false
+    delay: 33
+    encoding: 33574963
+    index: 54
+    mask: 4261441663
+    size: 32
+  MUL:
+    branch: false
+    delay: 32
+    encoding: 33554483
+    index: 49
+    mask: 4261441663
+    size: 32
+  MULH:
+    branch: false
+    delay: 32
+    encoding: 33558579
+    index: 50
+    mask: 4261441663
+    size: 32
+  MULHSU:
+    branch: false
+    delay: 32
+    encoding: 33562675
+    index: 51
+    mask: 4261441663
+    size: 32
+  MULHU:
+    branch: false
+    delay: 32
+    encoding: 33566771
+    index: 52
+    mask: 4261441663
+    size: 32
+  REM:
+    branch: false
+    delay: 33
+    encoding: 33579059
+    index: 55
+    mask: 4261441663
+    size: 32
+  REMU:
+    branch: false
+    delay: 33
+    encoding: 33583155
+    index: 56
+    mask: 4261441663
+    size: 32
+Zca:
+  C__ADD:
+    branch: false
+    delay: 1
+    encoding: 36866
+    index: 82
+    mask: 61443
+    size: 16
+  C__ADDI:
+    branch: false
+    delay: 1
+    encoding: 1
+    index: 60
+    mask: 57347
+    size: 16
+  C__ADDI16SP:
+    branch: false
+    delay: 1
+    encoding: 24833
+    index: 65
+    mask: 61315
+    size: 16
+  C__ADDI4SPN:
+    branch: false
+    delay: 1
+    encoding: 0
+    index: 57
+    mask: 57347
+    size: 16
+  C__AND:
+    branch: false
+    delay: 1
+    encoding: 35937
+    index: 73
+    mask: 64611
+    size: 16
+  C__ANDI:
+    branch: false
+    delay: 1
+    encoding: 34817
+    index: 69
+    mask: 60419
+    size: 16
+  C__BEQZ:
+    branch: true
+    delay:
+    - 1
+    - 2
+    encoding: 49153
+    index: 75
+    mask: 57347
+    size: 16
+  C__BNEZ:
+    branch: true
+    delay:
+    - 1
+    - 2
+    encoding: 57345
+    index: 76
+    mask: 57347
+    size: 16
+  C__EBREAK:
+    branch: false
+    delay: 3
+    encoding: 36866
+    index: 84
+    mask: 65535
+    size: 16
+  C__J:
+    branch: true
+    delay: 1
+    encoding: 40961
+    index: 74
+    mask: 57347
+    size: 16
+  C__JAL:
+    attributes:
+    - - name:enable
+      - value:1
+    branch: true
+    delay: 1
+    encoding: 8193
+    index: 62
+    mask: 57347
+    size: 16
+  C__JALR:
+    branch: true
+    delay: 1
+    encoding: 36866
+    index: 83
+    mask: 61567
+    size: 16
+  C__JR:
+    branch: true
+    delay: 1
+    encoding: 32770
+    index: 80
+    mask: 61567
+    size: 16
+  C__LI:
+    branch: false
+    delay: 1
+    encoding: 16385
+    index: 63
+    mask: 57347
+    size: 16
+  C__LUI:
+    branch: false
+    delay: 1
+    encoding: 24577
+    index: 64
+    mask: 57347
+    size: 16
+  C__LW:
+    branch: false
+    delay: 2
+    encoding: 16384
+    index: 58
+    mask: 57347
+    size: 16
+  C__LWSP:
+    branch: false
+    delay: 2
+    encoding: 16386
+    index: 78
+    mask: 57347
+    size: 16
+  C__MV:
+    branch: false
+    delay: 1
+    encoding: 32770
+    index: 79
+    mask: 61443
+    size: 16
+  C__NOP:
+    branch: false
+    delay: 1
+    encoding: 1
+    index: 61
+    mask: 61315
+    size: 16
+  C__OR:
+    branch: false
+    delay: 1
+    encoding: 35905
+    index: 72
+    mask: 64611
+    size: 16
+  C__SLLI:
+    attributes:
+    - - name:enable
+      - value:1
+    branch: false
+    delay: u_12:12*16+u_6:2
+    encoding: 2
+    index: 77
+    mask: 61443
+    size: 16
+  C__SRAI:
+    attributes:
+    - - name:enable
+      - value:1
+    branch: false
+    delay: u_12:12*16+u_6:2
+    encoding: 33793
+    index: 68
+    mask: 64515
+    size: 16
+  C__SRLI:
+    attributes:
+    - - name:enable
+      - value:1
+    branch: false
+    delay: u_12:12*16+u_6:2
+    encoding: 32769
+    index: 67
+    mask: 64515
+    size: 16
+  C__SUB:
+    branch: false
+    delay: 1
+    encoding: 35841
+    index: 70
+    mask: 64611
+    size: 16
+  C__SW:
+    branch: false
+    delay: 1
+    encoding: 49152
+    index: 59
+    mask: 57347
+    size: 16
+  C__SWSP:
+    branch: false
+    delay: 1
+    encoding: 49154
+    index: 85
+    mask: 57347
+    size: 16
+  C__XOR:
+    branch: false
+    delay: 1
+    encoding: 35873
+    index: 71
+    mask: 64611
+    size: 16
+  DII:
+    branch: false
+    delay: 1
+    encoding: 0
+    index: 86
+    mask: 65535
+    size: 16
+  __reserved_clui:
+    branch: false
+    delay: 1
+    encoding: 24577
+    index: 66
+    mask: 61567
+    size: 16
+  __reserved_cmv:
+    branch: false
+    delay: 1
+    encoding: 32770
+    index: 81
+    mask: 65535
+    size: 16
+Zicsr:
+  CSRRC:
+    branch: false
+    delay: 1
+    encoding: 12403
+    index: 44
+    mask: 28799
+    size: 32
+  CSRRCI:
+    branch: false
+    delay: 1
+    encoding: 28787
+    index: 47
+    mask: 28799
+    size: 32
+  CSRRS:
+    branch: false
+    delay: 1
+    encoding: 8307
+    index: 43
+    mask: 28799
+    size: 32
+  CSRRSI:
+    branch: false
+    delay: 1
+    encoding: 24691
+    index: 46
+    mask: 28799
+    size: 32
+  CSRRW:
+    branch: false
+    delay: 1
+    encoding: 4211
+    index: 42
+    mask: 28799
+    size: 32
+  CSRRWI:
+    branch: false
+    delay: 1
+    encoding: 20595
+    index: 45
+    mask: 28799
+    size: 32
+Zifencei:
+  FENCE_I:
+    attributes:
+    - - name:flush
+    branch: false
+    delay: 1
+    encoding: 4111
+    index: 48
+    mask: 28799
+    size: 32

contrib/pa/README.md

@@ -19,7 +19,7 @@ setenv CXX $COWAREHOME/SLS/linux/common/bin/g++
 cmake -S . -B build/PA -DCMAKE_BUILD_TYPE=Debug -DUSE_CWR_SYSTEMC=ON -DBUILD_SHARED_LIBS=ON \
     -DCODEGEN=OFF -DCMAKE_INSTALL_PREFIX=${TGFS_INSTALL_ROOT}
 cmake --build build/PA --target install -j16
-cd dbt-rise-tgc/contrib
+cd dbt-rise-tgc/contrib/pa
 # import the TGC core itself
 pct tgc_import_tb.tcl
 ```
@@ -37,7 +37,7 @@ export CXX=$COWAREHOME/SLS/linux/common/bin/g++
 cmake -S . -B build/PA -DCMAKE_BUILD_TYPE=Debug -DUSE_CWR_SYSTEMC=ON -DBUILD_SHARED_LIBS=ON \
     -DCODEGEN=OFF -DCMAKE_INSTALL_PREFIX=${TGFS_INSTALL_ROOT}
 cmake --build build/PA --target install -j16
-cd dbt-rise-tgc/contrib
+cd dbt-rise-tgc/contrib/pa
 # import the TGC core itself
 pct tgc_import_tb.tcl
 ```

gen_input/TGC5C.core_desc

@@ -1,8 +1,8 @@
-import "ISA/RV32I.core_desc"
+import "ISA/RVI.core_desc"
 import "ISA/RVM.core_desc"
 import "ISA/RVC.core_desc"
 
-Core TGC_C provides RV32I, Zicsr, Zifencei, RV32M, RV32IC {
+Core TGC5C provides RV32I, Zicsr, Zifencei, RV32M, RV32IC {
     architectural_state {
         XLEN=32;
         // definitions for the architecture wrapper

gen_input/templates/CORENAME.cpp.gtl

@@ -1,5 +1,5 @@
 /*******************************************************************************
- * Copyright (C) 2017 - 2020 MINRES Technologies GmbH
+ * Copyright (C) 2024 MINRES Technologies GmbH
  * All rights reserved.
  *
  * Redistribution and use in source and binary forms, with or without
@@ -37,6 +37,7 @@ def getRegisterSizes(){
     return regs
 }
 %>
+// clang-format off
 #include "${coreDef.name.toLowerCase()}.h"
 #include "util/ities.h"
 #include <util/logging.h>
@@ -46,10 +47,10 @@ def getRegisterSizes(){
 
 using namespace iss::arch;
 
-constexpr std::array<const char*, ${registers.size}>    iss::arch::traits<iss::arch::${coreDef.name.toLowerCase()}>::reg_names;
+constexpr std::array<const char*, ${registers.size()}>    iss::arch::traits<iss::arch::${coreDef.name.toLowerCase()}>::reg_names;
-constexpr std::array<const char*, ${registers.size}>    iss::arch::traits<iss::arch::${coreDef.name.toLowerCase()}>::reg_aliases;
+constexpr std::array<const char*, ${registers.size()}>    iss::arch::traits<iss::arch::${coreDef.name.toLowerCase()}>::reg_aliases;
-constexpr std::array<const uint32_t, ${getRegisterSizes().size}> iss::arch::traits<iss::arch::${coreDef.name.toLowerCase()}>::reg_bit_widths;
+constexpr std::array<const uint32_t, ${getRegisterSizes().size()}> iss::arch::traits<iss::arch::${coreDef.name.toLowerCase()}>::reg_bit_widths;
-constexpr std::array<const uint32_t, ${getRegisterSizes().size}> iss::arch::traits<iss::arch::${coreDef.name.toLowerCase()}>::reg_byte_offsets;
+constexpr std::array<const uint32_t, ${getRegisterSizes().size()}> iss::arch::traits<iss::arch::${coreDef.name.toLowerCase()}>::reg_byte_offsets;
 
 ${coreDef.name.toLowerCase()}::${coreDef.name.toLowerCase()}()  = default;
 
@@ -73,4 +74,4 @@ uint8_t *${coreDef.name.toLowerCase()}::get_regs_base_ptr() {
 ${coreDef.name.toLowerCase()}::phys_addr_t ${coreDef.name.toLowerCase()}::virt2phys(const iss::addr_t &addr) {
     return phys_addr_t(addr.access, addr.space, addr.val&traits<${coreDef.name.toLowerCase()}>::addr_mask);
 }
-
+// clang-format on

gen_input/templates/CORENAME.h.gtl

@@ -1,5 +1,5 @@
 /*******************************************************************************
- * Copyright (C) 2017 - 2021 MINRES Technologies GmbH
+ * Copyright (C) 2024 MINRES Technologies GmbH
  * All rights reserved.
  *
  * Redistribution and use in source and binary forms, with or without
@@ -60,7 +60,7 @@ def getCString(def val){
 %>
 #ifndef _${coreDef.name.toUpperCase()}_H_
 #define _${coreDef.name.toUpperCase()}_H_
-
+// clang-format off
 #include <array>
 #include <iss/arch/traits.h>
 #include <iss/arch_if.h>
@@ -75,11 +75,11 @@ template <> struct traits<${coreDef.name.toLowerCase()}> {
 
     constexpr static char const* const core_type = "${coreDef.name}";
     
-    static constexpr std::array<const char*, ${registers.size}> reg_names{
+    static constexpr std::array<const char*, ${registers.size()}> reg_names{
-        {"${registers.collect{it.name}.join('", "')}"}};
+        {"${registers.collect{it.name.toLowerCase()}.join('", "')}"}};
  
-    static constexpr std::array<const char*, ${registers.size}> reg_aliases{
+    static constexpr std::array<const char*, ${registers.size()}> reg_aliases{
-        {"${registers.collect{it.alias}.join('", "')}"}};
+        {"${registers.collect{it.alias.toLowerCase()}.join('", "')}"}};
 
     enum constants {${constants.collect{c -> c.name+"="+getCString(c.value)}.join(', ')}};
 
@@ -99,17 +99,17 @@ template <> struct traits<${coreDef.name.toLowerCase()}> {
 
     using phys_addr_t = iss::typed_addr_t<iss::address_type::PHYSICAL>;
 
-    static constexpr std::array<const uint32_t, ${getRegisterSizes().size}> reg_bit_widths{
+    static constexpr std::array<const uint32_t, ${getRegisterSizes().size()}> reg_bit_widths{
         {${getRegisterSizes().join(',')}}};
 
-    static constexpr std::array<const uint32_t, ${getRegisterOffsets().size}> reg_byte_offsets{
+    static constexpr std::array<const uint32_t, ${getRegisterOffsets().size()}> reg_byte_offsets{
         {${getRegisterOffsets().join(',')}}};
 
     static const uint64_t addr_mask = (reg_t(1) << (XLEN - 1)) | ((reg_t(1) << (XLEN - 1)) - 1);
 
     enum sreg_flag_e { FLAGS };
 
-    enum mem_type_e { ${spaces.collect{it.name}.join(', ')} };
+    enum mem_type_e { ${spaces.collect{it.name}.join(', ')}, IMEM = MEM };
     
     enum class opcode_e {<%instructions.eachWithIndex{instr, index -> %>
         ${instr.instruction.name} = ${index},<%}%>
@@ -131,8 +131,6 @@ struct ${coreDef.name.toLowerCase()}: public arch_if {
 
     uint8_t* get_regs_base_ptr() override;
 
-    inline uint64_t get_icount() { return reg.icount; }
-
     inline bool should_stop() { return interrupt_sim; }
 
     inline uint64_t stop_code() { return interrupt_sim; }
@@ -141,8 +139,6 @@ struct ${coreDef.name.toLowerCase()}: public arch_if {
 
     virtual iss::sync_type needed_sync() const { return iss::NO_SYNC; }
 
-    inline uint32_t get_last_branch() { return reg.last_branch; }
-
 
 #pragma pack(push, 1)
     struct ${coreDef.name}_regs {<%
@@ -174,3 +170,4 @@ if(fcsr != null) {%>
 }
 }            
 #endif /* _${coreDef.name.toUpperCase()}_H_ */
+// clang-format on

gen_input/templates/CORENAME_decoder.cpp.gtl

@@ -1,86 +0,0 @@
-#include "${coreDef.name.toLowerCase()}.h"
-#include <vector>
-#include <array>
-#include <cstdlib>
-#include <algorithm>
-
-namespace iss {
-namespace arch {
-namespace {
-// according to
-// https://stackoverflow.com/questions/8871204/count-number-of-1s-in-binary-representation
-#ifdef __GCC__
-constexpr size_t bit_count(uint32_t u) { return __builtin_popcount(u); }
-#elif __cplusplus < 201402L
-constexpr size_t uCount(uint32_t u) { return u - ((u >> 1) & 033333333333) - ((u >> 2) & 011111111111); }
-constexpr size_t bit_count(uint32_t u) { return ((uCount(u) + (uCount(u) >> 3)) & 030707070707) % 63; }
-#else
-constexpr size_t bit_count(uint32_t u) {
-    size_t uCount = u - ((u >> 1) & 033333333333) - ((u >> 2) & 011111111111);
-    return ((uCount + (uCount >> 3)) & 030707070707) % 63;
-}
-#endif
-
-using opcode_e = traits<${coreDef.name.toLowerCase()}>::opcode_e;
-
-/****************************************************************************
- * start opcode definitions
- ****************************************************************************/
-struct instruction_desriptor {
-    size_t length;
-    uint32_t value;
-    uint32_t mask;
-    opcode_e op;
-};
-
-const std::array<instruction_desriptor, ${instructions.size}> instr_descr = {{
-     /* entries are: size, valid value, valid mask, function ptr */<%instructions.each{instr -> %>
-    {${instr.length}, ${instr.encoding}, ${instr.mask}, opcode_e::${instr.instruction.name}},<%}%>
-}};
-
-}
-
-template<>
-struct instruction_decoder<${coreDef.name.toLowerCase()}> {
-    using opcode_e = traits<${coreDef.name.toLowerCase()}>::opcode_e;
-    using code_word_t=traits<${coreDef.name.toLowerCase()}>::code_word_t;
-
-    struct instruction_pattern {
-        uint32_t value;
-        uint32_t mask;
-        opcode_e id;
-    };
-
-    std::array<std::vector<instruction_pattern>, 4> qlut;
-
-    template<typename T>
-    unsigned decode_instruction(T);
-
-    instruction_decoder() {
-        for (auto instr : instr_descr) {
-            auto quadrant = instr.value & 0x3;
-            qlut[quadrant].push_back(instruction_pattern{instr.value, instr.mask, instr.op});
-        }
-        for(auto& lut: qlut){
-            std::sort(std::begin(lut), std::end(lut), [](instruction_pattern const& a, instruction_pattern const& b){
-                return bit_count(a.mask) < bit_count(b.mask);
-            });
-        }
-    }
-};
-
-template<>
-unsigned instruction_decoder<${coreDef.name.toLowerCase()}>::decode_instruction<traits<${coreDef.name.toLowerCase()}>::code_word_t>(traits<${coreDef.name.toLowerCase()}>::code_word_t instr){
-    auto res = std::find_if(std::begin(qlut[instr&0x3]), std::end(qlut[instr&0x3]), [instr](instruction_pattern const& e){
-        return !((instr&e.mask) ^ e.value );
-    });
-    return static_cast<unsigned>(res!=std::end(qlut[instr&0x3])? res->id : opcode_e::MAX_OPCODE);
-}
-
-
-std::unique_ptr<instruction_decoder<${coreDef.name.toLowerCase()}>> traits<${coreDef.name.toLowerCase()}>::get_decoder(){
-    return std::make_unique<instruction_decoder<${coreDef.name.toLowerCase()}>>();
-}
-
-}
-}

gen_input/templates/CORENAME_instr.yaml.gtl

@@ -8,9 +8,10 @@
         instrGroups[groupName]+=it;
     }
     instrGroups
-}%><%getInstructionGroups().each{name, instrList -> %>
+}%><%int index = 0; getInstructionGroups().each{name, instrList -> %>
-${name}: <% instrList.findAll{!it.instruction.name.startsWith("__")}.each { %>
+${name}: <% instrList.each { %>
-  - ${it.instruction.name}:
+  ${it.instruction.name}:
+    index: ${index++}
     encoding: ${it.encoding}
     mask: ${it.mask}<%if(it.attributes.size) {%>
     attributes: ${it.attributes}<%}%>

gen_input/templates/CORENAME_sysc.cpp.gtl

@@ -1,5 +1,5 @@
 /*******************************************************************************
- * Copyright (C) 2023 MINRES Technologies GmbH
+ * Copyright (C) 2024 MINRES Technologies GmbH
  * All rights reserved.
  *
  * Redistribution and use in source and binary forms, with or without
@@ -29,46 +29,103 @@
  * POSSIBILITY OF SUCH DAMAGE.
  *
  *******************************************************************************/
-
+// clang-format off
-#include "iss_factory.h"
+#include <sysc/iss_factory.h>
 #include <iss/arch/${coreDef.name.toLowerCase()}.h>
 #include <iss/arch/riscv_hart_m_p.h>
 #include <iss/arch/riscv_hart_mu_p.h>
-#include "sc_core_adapter.h"
+#include <sysc/sc_core_adapter.h>
-#include "core_complex.h"
+#include <sysc/core_complex.h>
 #include <array>
-
+<%
+def array_count = coreDef.name.toLowerCase()=="tgc5d" || coreDef.name.toLowerCase()=="tgc5e"? 3 : 2;
+%>
 namespace iss {
 namespace interp {
 using namespace sysc;
-volatile std::array<bool, 2> ${coreDef.name.toLowerCase()}_init = {
+volatile std::array<bool, ${array_count}> ${coreDef.name.toLowerCase()}_init = {
         iss_factory::instance().register_creator("${coreDef.name.toLowerCase()}|m_p|interp", [](unsigned gdb_port, void* data) -> iss_factory::base_t {
-            auto* cc = reinterpret_cast<sysc::tgfs::core_complex*>(data);
+            auto* cc = reinterpret_cast<sysc::tgfs::core_complex_if*>(data);
             auto* cpu = new sc_core_adapter<arch::riscv_hart_m_p<arch::${coreDef.name.toLowerCase()}>>(cc);
             return {sysc::sc_cpu_ptr{cpu}, vm_ptr{create(static_cast<arch::${coreDef.name.toLowerCase()}*>(cpu), gdb_port)}};
         }),
         iss_factory::instance().register_creator("${coreDef.name.toLowerCase()}|mu_p|interp", [](unsigned gdb_port, void* data) -> iss_factory::base_t {
-            auto* cc = reinterpret_cast<sysc::tgfs::core_complex*>(data);
+            auto* cc = reinterpret_cast<sysc::tgfs::core_complex_if*>(data);
             auto* cpu = new sc_core_adapter<arch::riscv_hart_mu_p<arch::${coreDef.name.toLowerCase()}>>(cc);
             return {sysc::sc_cpu_ptr{cpu}, vm_ptr{create(static_cast<arch::${coreDef.name.toLowerCase()}*>(cpu), gdb_port)}};
-        })
+        })<%if(coreDef.name.toLowerCase()=="tgc5d" || coreDef.name.toLowerCase()=="tgc5e") {%>,
+        iss_factory::instance().register_creator("${coreDef.name.toLowerCase()}|mu_p_clic_pmp|interp", [](unsigned gdb_port, void* data) -> iss_factory::base_t {
+            auto* cc = reinterpret_cast<sysc::tgfs::core_complex_if*>(data);
+            auto* cpu = new sc_core_adapter<arch::riscv_hart_mu_p<arch::${coreDef.name.toLowerCase()}, (iss::arch::features_e)(iss::arch::FEAT_PMP | iss::arch::FEAT_EXT_N | iss::arch::FEAT_CLIC)>>(cc);
+            return {sysc::sc_cpu_ptr{cpu}, vm_ptr{create(static_cast<arch::${coreDef.name.toLowerCase()}*>(cpu), gdb_port)}};
+        })<%}%>
 };
 }
+#if defined(WITH_LLVM)
+namespace llvm {
+using namespace sysc;
+volatile std::array<bool, ${array_count}> ${coreDef.name.toLowerCase()}_init = {
+        iss_factory::instance().register_creator("${coreDef.name.toLowerCase()}|m_p|llvm", [](unsigned gdb_port, void* data) -> iss_factory::base_t {
+            auto* cc = reinterpret_cast<sysc::tgfs::core_complex_if*>(data);
+            auto* cpu = new sc_core_adapter<arch::riscv_hart_m_p<arch::${coreDef.name.toLowerCase()}>>(cc);
+            return {sysc::sc_cpu_ptr{cpu}, vm_ptr{create(static_cast<arch::${coreDef.name.toLowerCase()}*>(cpu), gdb_port)}};
+        }),
+        iss_factory::instance().register_creator("${coreDef.name.toLowerCase()}|mu_p|llvm", [](unsigned gdb_port, void* data) -> iss_factory::base_t {
+            auto* cc = reinterpret_cast<sysc::tgfs::core_complex_if*>(data);
+            auto* cpu = new sc_core_adapter<arch::riscv_hart_mu_p<arch::${coreDef.name.toLowerCase()}>>(cc);
+            return {sysc::sc_cpu_ptr{cpu}, vm_ptr{create(static_cast<arch::${coreDef.name.toLowerCase()}*>(cpu), gdb_port)}};
+        })<%if(coreDef.name.toLowerCase()=="tgc5d" || coreDef.name.toLowerCase()=="tgc5e") {%>,
+        iss_factory::instance().register_creator("${coreDef.name.toLowerCase()}|mu_p_clic_pmp|llvm", [](unsigned gdb_port, void* data) -> iss_factory::base_t {
+            auto* cc = reinterpret_cast<sysc::tgfs::core_complex_if*>(data);
+            auto* cpu = new sc_core_adapter<arch::riscv_hart_mu_p<arch::${coreDef.name.toLowerCase()}, (iss::arch::features_e)(iss::arch::FEAT_PMP | iss::arch::FEAT_EXT_N | iss::arch::FEAT_CLIC)>>(cc);
+            return {sysc::sc_cpu_ptr{cpu}, vm_ptr{create(static_cast<arch::${coreDef.name.toLowerCase()}*>(cpu), gdb_port)}};
+        })<%}%>
+};
+}
+#endif
 #if defined(WITH_TCC)
 namespace tcc {
 using namespace sysc;
-volatile std::array<bool, 2> ${coreDef.name.toLowerCase()}_init = {
+volatile std::array<bool, ${array_count}> ${coreDef.name.toLowerCase()}_init = {
         iss_factory::instance().register_creator("${coreDef.name.toLowerCase()}|m_p|tcc", [](unsigned gdb_port, void* data) -> iss_factory::base_t {
-            auto* cc = reinterpret_cast<sysc::tgfs::core_complex*>(data);
+            auto* cc = reinterpret_cast<sysc::tgfs::core_complex_if*>(data);
             auto* cpu = new sc_core_adapter<arch::riscv_hart_m_p<arch::${coreDef.name.toLowerCase()}>>(cc);
             return {sysc::sc_cpu_ptr{cpu}, vm_ptr{create(static_cast<arch::${coreDef.name.toLowerCase()}*>(cpu), gdb_port)}};
         }),
         iss_factory::instance().register_creator("${coreDef.name.toLowerCase()}|mu_p|tcc", [](unsigned gdb_port, void* data) -> iss_factory::base_t {
-            auto* cc = reinterpret_cast<sysc::tgfs::core_complex*>(data);
+            auto* cc = reinterpret_cast<sysc::tgfs::core_complex_if*>(data);
             auto* cpu = new sc_core_adapter<arch::riscv_hart_mu_p<arch::${coreDef.name.toLowerCase()}>>(cc);
             return {sysc::sc_cpu_ptr{cpu}, vm_ptr{create(static_cast<arch::${coreDef.name.toLowerCase()}*>(cpu), gdb_port)}};
-        })
+        })<%if(coreDef.name.toLowerCase()=="tgc5d" || coreDef.name.toLowerCase()=="tgc5e") {%>,
+        iss_factory::instance().register_creator("${coreDef.name.toLowerCase()}|mu_p_clic_pmp|tcc", [](unsigned gdb_port, void* data) -> iss_factory::base_t {
+            auto* cc = reinterpret_cast<sysc::tgfs::core_complex_if*>(data);
+            auto* cpu = new sc_core_adapter<arch::riscv_hart_mu_p<arch::${coreDef.name.toLowerCase()}, (iss::arch::features_e)(iss::arch::FEAT_PMP | iss::arch::FEAT_EXT_N | iss::arch::FEAT_CLIC)>>(cc);
+            return {sysc::sc_cpu_ptr{cpu}, vm_ptr{create(static_cast<arch::${coreDef.name.toLowerCase()}*>(cpu), gdb_port)}};
+        })<%}%>
+};
+}
+#endif
+#if defined(WITH_ASMJIT)
+namespace asmjit {
+using namespace sysc;
+volatile std::array<bool, ${array_count}> ${coreDef.name.toLowerCase()}_init = {
+        iss_factory::instance().register_creator("${coreDef.name.toLowerCase()}|m_p|asmjit", [](unsigned gdb_port, void* data) -> iss_factory::base_t {
+            auto* cc = reinterpret_cast<sysc::tgfs::core_complex_if*>(data);
+            auto* cpu = new sc_core_adapter<arch::riscv_hart_m_p<arch::${coreDef.name.toLowerCase()}>>(cc);
+            return {sysc::sc_cpu_ptr{cpu}, vm_ptr{create(static_cast<arch::${coreDef.name.toLowerCase()}*>(cpu), gdb_port)}};
+        }),
+        iss_factory::instance().register_creator("${coreDef.name.toLowerCase()}|mu_p|asmjit", [](unsigned gdb_port, void* data) -> iss_factory::base_t {
+            auto* cc = reinterpret_cast<sysc::tgfs::core_complex_if*>(data);
+            auto* cpu = new sc_core_adapter<arch::riscv_hart_mu_p<arch::${coreDef.name.toLowerCase()}>>(cc);
+            return {sysc::sc_cpu_ptr{cpu}, vm_ptr{create(static_cast<arch::${coreDef.name.toLowerCase()}*>(cpu), gdb_port)}};
+        })<%if(coreDef.name.toLowerCase()=="tgc5d" || coreDef.name.toLowerCase()=="tgc5e") {%>,
+        iss_factory::instance().register_creator("${coreDef.name.toLowerCase()}|mu_p_clic_pmp|asmjit", [](unsigned gdb_port, void* data) -> iss_factory::base_t {
+            auto* cc = reinterpret_cast<sysc::tgfs::core_complex_if*>(data);
+            auto* cpu = new sc_core_adapter<arch::riscv_hart_mu_p<arch::${coreDef.name.toLowerCase()}, (iss::arch::features_e)(iss::arch::FEAT_PMP | iss::arch::FEAT_EXT_N | iss::arch::FEAT_CLIC)>>(cc);
+            return {sysc::sc_cpu_ptr{cpu}, vm_ptr{create(static_cast<arch::${coreDef.name.toLowerCase()}*>(cpu), gdb_port)}};
+        })<%}%>
 };
 }
 #endif
 }
+// clang-format on

gen_input/templates/asmjit/CORENAME.cpp.gtl

@@ -0,0 +1,371 @@
+/*******************************************************************************
+ * Copyright (C) 2017-2024 MINRES Technologies GmbH
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ *    this list of conditions and the following disclaimer.
+ *
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ *    this list of conditions and the following disclaimer in the documentation
+ *    and/or other materials provided with the distribution.
+ *
+ * 3. Neither the name of the copyright holder nor the names of its contributors
+ *    may be used to endorse or promote products derived from this software
+ *    without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ *
+ *******************************************************************************/
+// clang-format off
+#include <iss/arch/${coreDef.name.toLowerCase()}.h>
+#include <iss/debugger/gdb_session.h>
+#include <iss/debugger/server.h>
+#include <iss/iss.h>
+#include <iss/asmjit/vm_base.h>
+#include <asmjit/asmjit.h>
+#include <util/logging.h>
+#include <iss/instruction_decoder.h>
+<%def fcsr = registers.find {it.name=='FCSR'}
+if(fcsr != null) {%>
+#include <vm/fp_functions.h><%}%>
+#ifndef FMT_HEADER_ONLY
+#define FMT_HEADER_ONLY
+#endif
+#include <fmt/format.h>
+
+#include <array>
+#include <iss/debugger/riscv_target_adapter.h>
+
+namespace iss {
+namespace asmjit {
+
+
+namespace ${coreDef.name.toLowerCase()} {
+using namespace ::asmjit;
+using namespace iss::arch;
+using namespace iss::debugger;
+
+template <typename ARCH> class vm_impl : public iss::asmjit::vm_base<ARCH> {
+public:
+    using traits = arch::traits<ARCH>;
+    using super = typename iss::asmjit::vm_base<ARCH>;
+    using virt_addr_t = typename super::virt_addr_t;
+    using phys_addr_t = typename super::phys_addr_t;
+    using code_word_t = typename super::code_word_t;
+    using mem_type_e = typename super::mem_type_e;
+    using addr_t = typename super::addr_t;
+
+    vm_impl();
+
+    vm_impl(ARCH &core, unsigned core_id = 0, unsigned cluster_id = 0);
+
+    void enableDebug(bool enable) { super::sync_exec = super::ALL_SYNC; }
+
+    target_adapter_if *accquire_target_adapter(server_if *srv) override {
+        debugger_if::dbg_enabled = true;
+        if (vm_base<ARCH>::tgt_adapter == nullptr)
+            vm_base<ARCH>::tgt_adapter = new riscv_target_adapter<ARCH>(srv, this->get_arch());
+        return vm_base<ARCH>::tgt_adapter;
+    }
+
+protected:
+    using super::get_ptr_for;
+    using super::get_reg_for;
+    using super::get_reg_for_Gp;
+    using super::load_reg_from_mem;
+    using super::load_reg_from_mem_Gp;
+    using super::write_reg_to_mem;
+    using super::gen_read_mem;
+    using super::gen_write_mem;
+    using super::gen_leave;
+    using super::gen_sync;
+   
+    using this_class = vm_impl<ARCH>;
+    using compile_func = continuation_e (this_class::*)(virt_addr_t&, code_word_t, jit_holder&);
+
+    continuation_e gen_single_inst_behavior(virt_addr_t&, unsigned int &, jit_holder&) override;
+    enum globals_e {TVAL = 0, GLOBALS_SIZE};
+    void gen_block_prologue(jit_holder& jh) override;
+    void gen_block_epilogue(jit_holder& jh) override;
+    inline const char *name(size_t index){return traits::reg_aliases.at(index);}
+<%if(fcsr != null) {%>
+    inline const char *fname(size_t index){return index < 32?name(index+traits::F0):"illegal";}   
+<%}%>
+    void gen_instr_prologue(jit_holder& jh);
+    void gen_instr_epilogue(jit_holder& jh);
+    inline void gen_raise(jit_holder& jh, uint16_t trap_id, uint16_t cause);
+    template <typename T, typename = typename std::enable_if<std::is_integral<T>::value>::type> void gen_set_tval(jit_holder& jh, T new_tval) ;
+    void gen_set_tval(jit_holder& jh, x86_reg_t _new_tval) ;
+
+    template<unsigned W, typename U, typename S = typename std::make_signed<U>::type>
+    inline S sext(U from) {
+        auto mask = (1ULL<<W) - 1;
+        auto sign_mask = 1ULL<<(W-1);
+        return (from & mask) | ((from & sign_mask) ? ~mask : 0);
+    }
+<%functions.each{ it.eachLine { %>
+    ${it}<%}%>
+<%}%>
+private:
+    /****************************************************************************
+     * start opcode definitions
+     ****************************************************************************/
+    struct instruction_descriptor {
+        uint32_t length;
+        uint32_t value;
+        uint32_t mask;
+        compile_func op;
+    };
+
+    const std::array<instruction_descriptor, ${instructions.size()}> instr_descr = {{
+         /* entries are: size, valid value, valid mask, function ptr */<%instructions.each{instr -> %>
+        /* instruction ${instr.instruction.name}, encoding '${instr.encoding}' */
+        {${instr.length}, ${instr.encoding}, ${instr.mask}, &this_class::__${generator.functionName(instr.name)}},<%}%>
+    }};
+
+    //needs to be declared after instr_descr
+    decoder instr_decoder;
+
+    /* instruction definitions */<%instructions.eachWithIndex{instr, idx -> %>
+    /* instruction ${idx}: ${instr.name} */
+    continuation_e __${generator.functionName(instr.name)}(virt_addr_t& pc, code_word_t instr, jit_holder& jh){
+        uint64_t PC = pc.val;
+        <%instr.fields.eachLine{%>${it}
+        <%}%>if(this->disass_enabled){
+            /* generate disass */
+            <%instr.disass.eachLine{%>
+            ${it}<%}%>
+            InvokeNode* call_print_disass;
+            char* mnemonic_ptr = strdup(mnemonic.c_str());
+            jh.disass_collection.push_back(mnemonic_ptr);
+            jh.cc.invoke(&call_print_disass, &print_disass, FuncSignature::build<void, void *, uint64_t, char *>());
+            call_print_disass->setArg(0, jh.arch_if_ptr);
+            call_print_disass->setArg(1, pc.val);
+            call_print_disass->setArg(2, mnemonic_ptr);
+
+        }
+        x86::Compiler& cc = jh.cc;
+        cc.comment(fmt::format("${instr.name}_{:#x}:",pc.val).c_str());
+        gen_sync(jh, PRE_SYNC, ${idx});
+        mov(cc, jh.pc, pc.val);
+        gen_set_tval(jh, instr);
+        pc = pc+${instr.length/8};
+        mov(cc, jh.next_pc, pc.val);
+
+        gen_instr_prologue(jh);
+        cc.comment("//behavior:");
+        /*generate behavior*/
+        <%instr.behavior.eachLine{%>${it}
+        <%}%>
+        gen_sync(jh, POST_SYNC, ${idx});
+        gen_instr_epilogue(jh);
+    	return returnValue;        
+    }
+    <%}%>
+    /****************************************************************************
+     * end opcode definitions
+     ****************************************************************************/
+    continuation_e illegal_instruction(virt_addr_t &pc, code_word_t instr, jit_holder& jh ) {
+        x86::Compiler& cc = jh.cc;
+        if(this->disass_enabled){          
+            auto mnemonic = std::string("illegal_instruction");
+            InvokeNode* call_print_disass;
+            char* mnemonic_ptr = strdup(mnemonic.c_str());
+            jh.disass_collection.push_back(mnemonic_ptr);
+            jh.cc.invoke(&call_print_disass, &print_disass, FuncSignature::build<void, void *, uint64_t, char *>());
+            call_print_disass->setArg(0, jh.arch_if_ptr);
+            call_print_disass->setArg(1, pc.val);
+            call_print_disass->setArg(2, mnemonic_ptr);
+        }
+        cc.comment(fmt::format("illegal_instruction{:#x}:",pc.val).c_str());
+        gen_sync(jh, PRE_SYNC, instr_descr.size());
+        mov(cc, jh.pc, pc.val);
+        gen_set_tval(jh, instr);
+        pc = pc + ((instr & 3) == 3 ? 4 : 2);
+        mov(cc, jh.next_pc, pc.val);
+        gen_instr_prologue(jh);
+        cc.comment("//behavior:");
+        gen_raise(jh, 0, 2);
+        gen_sync(jh, POST_SYNC, instr_descr.size());
+        gen_instr_epilogue(jh);
+        return ILLEGAL_INSTR;
+    }
+};
+
+template <typename ARCH> vm_impl<ARCH>::vm_impl() { this(new ARCH()); }
+
+template <typename ARCH>
+vm_impl<ARCH>::vm_impl(ARCH &core, unsigned core_id, unsigned cluster_id)
+: vm_base<ARCH>(core, core_id, cluster_id)
+, instr_decoder([this]() {
+        std::vector<generic_instruction_descriptor> g_instr_descr;
+        g_instr_descr.reserve(instr_descr.size());
+        for (uint32_t i = 0; i < instr_descr.size(); ++i) {
+            generic_instruction_descriptor new_instr_descr {instr_descr[i].value, instr_descr[i].mask, i};
+            g_instr_descr.push_back(new_instr_descr);
+        }
+        return std::move(g_instr_descr);
+    }()) {}
+
+template <typename ARCH>
+continuation_e vm_impl<ARCH>::gen_single_inst_behavior(virt_addr_t &pc, unsigned int &inst_cnt, jit_holder& jh) {
+    enum {TRAP_ID=1<<16};
+    code_word_t instr = 0;
+    phys_addr_t paddr(pc);
+    auto *const data = (uint8_t *)&instr;
+    if(this->core.has_mmu())
+        paddr = this->core.virt2phys(pc);
+    auto res = this->core.read(paddr, 4, data);
+    if (res != iss::Ok)
+        return ILLEGAL_FETCH;
+    if (instr == 0x0000006f || (instr&0xffff)==0xa001)
+        return JUMP_TO_SELF;
+    ++inst_cnt;
+    uint32_t inst_index = instr_decoder.decode_instr(instr);
+    compile_func f = nullptr;
+    if(inst_index < instr_descr.size())
+        f = instr_descr[inst_index].op;
+    if (f == nullptr) 
+        f = &this_class::illegal_instruction;
+    return (this->*f)(pc, instr, jh);
+}
+template <typename ARCH>
+void vm_impl<ARCH>::gen_instr_prologue(jit_holder& jh) {
+    auto& cc = jh.cc;
+
+    cc.comment("//gen_instr_prologue");
+
+    x86_reg_t current_trap_state = get_reg_for(cc, traits::TRAP_STATE);
+    mov(cc, current_trap_state, get_ptr_for(jh, traits::TRAP_STATE));
+    mov(cc, get_ptr_for(jh, traits::PENDING_TRAP), current_trap_state);
+
+}
+template <typename ARCH>
+void vm_impl<ARCH>::gen_instr_epilogue(jit_holder& jh) {
+    auto& cc = jh.cc;
+
+    cc.comment("//gen_instr_epilogue");
+    x86_reg_t current_trap_state = get_reg_for(cc, traits::TRAP_STATE);
+    mov(cc, current_trap_state, get_ptr_for(jh, traits::TRAP_STATE));
+    cmp(cc, current_trap_state, 0);
+    cc.jne(jh.trap_entry);
+    cc.inc(get_ptr_for(jh, traits::ICOUNT));
+    cc.inc(get_ptr_for(jh, traits::CYCLE));
+}
+template <typename ARCH>
+void vm_impl<ARCH>::gen_block_prologue(jit_holder& jh){
+    jh.pc = load_reg_from_mem_Gp(jh, traits::PC);
+    jh.next_pc = load_reg_from_mem_Gp(jh, traits::NEXT_PC);
+    jh.globals.resize(GLOBALS_SIZE);
+    jh.globals[TVAL] = get_reg_Gp(jh.cc, 64, false);
+}
+template <typename ARCH>
+void vm_impl<ARCH>::gen_block_epilogue(jit_holder& jh){
+    x86::Compiler& cc = jh.cc;
+    cc.comment("//gen_block_epilogue");
+    cc.ret(jh.next_pc);
+
+    cc.bind(jh.trap_entry);
+    this->write_back(jh);
+
+    x86::Gp current_trap_state = get_reg_for_Gp(cc, traits::TRAP_STATE);
+    mov(cc, current_trap_state, get_ptr_for(jh, traits::TRAP_STATE));
+
+    x86::Gp current_pc = get_reg_for_Gp(cc, traits::PC);
+    mov(cc, current_pc, get_ptr_for(jh, traits::PC));
+
+    cc.comment("//enter trap call;");
+    InvokeNode* call_enter_trap;
+    cc.invoke(&call_enter_trap, &enter_trap, FuncSignature::build<uint64_t, void*, uint64_t, uint64_t, uint64_t>());
+    call_enter_trap->setArg(0, jh.arch_if_ptr);
+    call_enter_trap->setArg(1, current_trap_state);
+    call_enter_trap->setArg(2, current_pc);
+    call_enter_trap->setArg(3, jh.globals[TVAL]);
+
+    x86_reg_t current_next_pc = get_reg_for(cc, traits::NEXT_PC);
+    mov(cc, current_next_pc, get_ptr_for(jh, traits::NEXT_PC));
+    mov(cc, jh.next_pc, current_next_pc);
+
+    mov(cc, get_ptr_for(jh, traits::LAST_BRANCH), static_cast<int>(UNKNOWN_JUMP));
+    cc.ret(jh.next_pc);
+}
+template <typename ARCH>
+inline void vm_impl<ARCH>::gen_raise(jit_holder& jh, uint16_t trap_id, uint16_t cause) {
+    auto& cc = jh.cc;
+    cc.comment("//gen_raise");
+    auto tmp1 = get_reg_for(cc, traits::TRAP_STATE);
+    mov(cc, tmp1, 0x80ULL << 24 | (cause << 16) | trap_id);
+    mov(cc, get_ptr_for(jh, traits::TRAP_STATE), tmp1);
+    cc.jmp(jh.trap_entry);
+}
+template <typename ARCH>
+template <typename T, typename>
+void vm_impl<ARCH>::gen_set_tval(jit_holder& jh, T new_tval) {
+        mov(jh.cc, jh.globals[TVAL], new_tval);
+    }
+template <typename ARCH>
+void vm_impl<ARCH>::gen_set_tval(jit_holder& jh, x86_reg_t _new_tval) {
+    if(nonstd::holds_alternative<x86::Gp>(_new_tval)) {
+        x86::Gp new_tval = nonstd::get<x86::Gp>(_new_tval);
+        if(new_tval.size() < 8)
+            new_tval = gen_ext_Gp(jh.cc, new_tval, 64, false);
+        mov(jh.cc, jh.globals[TVAL], new_tval);
+    } else {
+        throw std::runtime_error("Variant not supported in gen_set_tval");
+    }
+}
+
+} // namespace tgc5c
+
+template <>
+std::unique_ptr<vm_if> create<arch::${coreDef.name.toLowerCase()}>(arch::${coreDef.name.toLowerCase()} *core, unsigned short port, bool dump) {
+    auto ret = new ${coreDef.name.toLowerCase()}::vm_impl<arch::${coreDef.name.toLowerCase()}>(*core, dump);
+    if (port != 0) debugger::server<debugger::gdb_session>::run_server(ret, port);
+    return std::unique_ptr<vm_if>(ret);
+}
+} // namespace asmjit
+} // namespace iss
+
+#include <iss/arch/riscv_hart_m_p.h>
+#include <iss/arch/riscv_hart_mu_p.h>
+#include <iss/factory.h>
+namespace iss {
+namespace {
+volatile std::array<bool, 2> dummy = {
+        core_factory::instance().register_creator("${coreDef.name.toLowerCase()}|m_p|asmjit", [](unsigned port, void* init_data) -> std::tuple<cpu_ptr, vm_ptr>{
+            auto* cpu = new iss::arch::riscv_hart_m_p<iss::arch::${coreDef.name.toLowerCase()}>();
+		    auto vm = new asmjit::${coreDef.name.toLowerCase()}::vm_impl<arch::${coreDef.name.toLowerCase()}>(*cpu, false);
+		    if (port != 0) debugger::server<debugger::gdb_session>::run_server(vm, port);
+            if(init_data){
+                auto* cb = reinterpret_cast<semihosting_cb_t<arch::traits<arch::${coreDef.name.toLowerCase()}>::reg_t>*>(init_data);
+                cpu->set_semihosting_callback(*cb);
+            }
+            return {cpu_ptr{cpu}, vm_ptr{vm}};
+        }),
+        core_factory::instance().register_creator("${coreDef.name.toLowerCase()}|mu_p|asmjit", [](unsigned port, void* init_data) -> std::tuple<cpu_ptr, vm_ptr>{
+            auto* cpu = new iss::arch::riscv_hart_mu_p<iss::arch::${coreDef.name.toLowerCase()}>();
+		    auto vm = new asmjit::${coreDef.name.toLowerCase()}::vm_impl<arch::${coreDef.name.toLowerCase()}>(*cpu, false);
+		    if (port != 0) debugger::server<debugger::gdb_session>::run_server(vm, port);
+            if(init_data){
+                auto* cb = reinterpret_cast<semihosting_cb_t<arch::traits<arch::${coreDef.name.toLowerCase()}>::reg_t>*>(init_data);
+                cpu->set_semihosting_callback(*cb);
+            }
+            return {cpu_ptr{cpu}, vm_ptr{vm}};
+        })
+};
+}
+}
+// clang-format on

gen_input/templates/interp/CORENAME.cpp.gtl

@@ -1,5 +1,5 @@
 /*******************************************************************************
- * Copyright (C) 2021 MINRES Technologies GmbH
+ * Copyright (C) 2017-2024 MINRES Technologies GmbH
  * All rights reserved.
  *
  * Redistribution and use in source and binary forms, with or without
@@ -34,15 +34,22 @@ def nativeTypeSize(int size){
     if(size<=8) return 8; else if(size<=16) return 16; else if(size<=32) return 32; else return 64;
 }
 %>
+// clang-format off
+#include <cstdint>
 #include <iss/arch/${coreDef.name.toLowerCase()}.h>
 #include <iss/debugger/gdb_session.h>
 #include <iss/debugger/server.h>
 #include <iss/iss.h>
 #include <iss/interp/vm_base.h>
+#include <vm/fp_functions.h>
 #include <util/logging.h>
-#include <sstream>
 #include <boost/coroutine2/all.hpp>
 #include <functional>
+#include <exception>
+#include <vector>
+#include <sstream>
+#include <iss/instruction_decoder.h>
+
 
 #ifndef FMT_HEADER_ONLY
 #define FMT_HEADER_ONLY
@@ -59,6 +66,10 @@ using namespace iss::arch;
 using namespace iss::debugger;
 using namespace std::placeholders;
 
+struct memory_access_exception : public std::exception{
+    memory_access_exception(){}
+};
+
 template <typename ARCH> class vm_impl : public iss::interp::vm_base<ARCH> {
 public:
     using traits = arch::traits<ARCH>;
@@ -89,37 +100,20 @@ protected:
     using compile_ret_t = virt_addr_t;
     using compile_func = compile_ret_t (this_class::*)(virt_addr_t &pc, code_word_t instr);
 
-    inline const char *name(size_t index){return index<traits::reg_aliases.size()?traits::reg_aliases[index]:"illegal";}
+    inline const char *name(size_t index){return traits::reg_aliases.at(index);}
+<%
+def fcsr = registers.find {it.name=='FCSR'}
+if(fcsr != null) {%>
+    inline const char *fname(size_t index){return index < 32?name(index+traits::F0):"illegal";}     
+<%}%>
 
-    typename arch::traits<ARCH>::opcode_e decode_inst_id(code_word_t instr);
     virt_addr_t execute_inst(finish_cond_e cond, virt_addr_t start, uint64_t icount_limit) override;
 
     // some compile time constants
-    // enum { MASK16 = 0b1111110001100011, MASK32 = 0b11111111111100000111000001111111 };
-    enum { MASK16 = 0b1111111111111111, MASK32 = 0b11111111111100000111000001111111 };
-    enum { EXTR_MASK16 = MASK16 >> 2, EXTR_MASK32 = MASK32 >> 2 };
-    enum {
-        LUT_SIZE = 1 << util::bit_count(static_cast<uint32_t>(EXTR_MASK32)),
-        LUT_SIZE_C = 1 << util::bit_count(static_cast<uint32_t>(EXTR_MASK16))
-    };
-
-    std::array<compile_func, LUT_SIZE> lut;
-
-    std::array<compile_func, LUT_SIZE_C> lut_00, lut_01, lut_10;
-    std::array<compile_func, LUT_SIZE> lut_11;
-
-    struct instruction_pattern {
-        uint32_t value;
-        uint32_t mask;
-        typename arch::traits<ARCH>::opcode_e id;
-    };
-
-    std::array<std::vector<instruction_pattern>, 4> qlut;
 
     inline void raise(uint16_t trap_id, uint16_t cause){
         auto trap_val =  0x80ULL << 24 | (cause << 16) | trap_id;
         this->core.reg.trap_state = trap_val;
-        this->template get_reg<uint${addrDataWidth}_t>(traits::NEXT_PC) = std::numeric_limits<uint${addrDataWidth}_t>::max();
     }
 
     inline void leave(unsigned lvl){
@@ -130,6 +124,13 @@ protected:
         this->core.wait_until(type);
     }
 
+    inline void set_tval(uint64_t new_tval){
+        tval = new_tval;
+    }
+
+    uint64_t fetch_count{0};
+    uint64_t tval{0};
+
     using yield_t = boost::coroutines2::coroutine<void>::push_type;
     using coro_t = boost::coroutines2::coroutine<void>::pull_type;
     std::vector<coro_t> spawn_blocks;
@@ -159,25 +160,20 @@ private:
      * start opcode definitions
      ****************************************************************************/
     struct instruction_descriptor {
-        size_t length;
+        uint32_t length;
         uint32_t value;
         uint32_t mask;
         typename arch::traits<ARCH>::opcode_e op;
     };
-    struct decoding_tree_node{
-        std::vector<instruction_descriptor> instrs;
-        std::vector<decoding_tree_node*> children;
-        uint32_t submask = std::numeric_limits<uint32_t>::max();
-        uint32_t value;
-        decoding_tree_node(uint32_t value) : value(value){}
-    };
 
-    decoding_tree_node* root {nullptr};
+    const std::array<instruction_descriptor, ${instructions.size()}> instr_descr = {{
-    const std::array<instruction_descriptor, ${instructions.size}> instr_descr = {{
          /* entries are: size, valid value, valid mask, function ptr */<%instructions.each{instr -> %>
         {${instr.length}, ${instr.encoding}, ${instr.mask}, arch::traits<ARCH>::opcode_e::${instr.instruction.name}},<%}%>
     }};
 
+    //needs to be declared after instr_descr
+    decoder instr_decoder;
+
     iss::status fetch_ins(virt_addr_t pc, uint8_t * data){
         if(this->core.has_mmu()) {
             auto phys_pc = this->core.virt2phys(pc);
@@ -197,57 +193,6 @@ private:
         }
         return iss::Ok;
     }
-    
-    void populate_decoding_tree(decoding_tree_node* root){
-        //create submask
-        for(auto instr: root->instrs){
-            root->submask &= instr.mask;
-        }
-        //put each instr according to submask&encoding into children
-        for(auto instr: root->instrs){
-            bool foundMatch = false;
-            for(auto child: root->children){
-                //use value as identifying trait
-                if(child->value == (instr.value&root->submask)){
-                    child->instrs.push_back(instr);
-                    foundMatch = true;
-                }
-            }
-            if(!foundMatch){
-                decoding_tree_node* child = new decoding_tree_node(instr.value&root->submask);
-                child->instrs.push_back(instr);
-                root->children.push_back(child);
-            }
-        }
-        root->instrs.clear();
-        //call populate_decoding_tree for all children
-        if(root->children.size() >1)
-            for(auto child: root->children){
-                populate_decoding_tree(child);      
-            }
-        else{
-            //sort instrs by value of the mask, this works bc we want to have the least restrictive one last
-            std::sort(root->children[0]->instrs.begin(), root->children[0]->instrs.end(), [](const instruction_descriptor& instr1, const instruction_descriptor& instr2) {
-            return instr1.mask > instr2.mask;
-            }); 
-        }
-    }
-    typename arch::traits<ARCH>::opcode_e  decode_instr(decoding_tree_node* node, code_word_t word){
-        if(!node->children.size()){
-            if(node->instrs.size() == 1) return node->instrs[0].op;
-            for(auto instr : node->instrs){
-                if((instr.mask&word) == instr.value) return instr.op;
-            }
-        }
-        else{
-            for(auto child : node->children){
-                if (child->value == (node->submask&word)){
-                    return decode_instr(child, word);
-                }  
-            }  
-        }
-        return arch::traits<ARCH>::opcode_e::MAX_OPCODE;
-    }
 };
 
 template <typename CODE_WORD> void debug_fn(CODE_WORD insn) {
@@ -273,16 +218,23 @@ constexpr size_t bit_count(uint32_t u) {
 
 template <typename ARCH>
 vm_impl<ARCH>::vm_impl(ARCH &core, unsigned core_id, unsigned cluster_id)
-: vm_base<ARCH>(core, core_id, cluster_id) {
+: vm_base<ARCH>(core, core_id, cluster_id)
-    root = new decoding_tree_node(std::numeric_limits<uint32_t>::max());
+, instr_decoder([this]() {
-    for(auto instr:instr_descr){
+        std::vector<generic_instruction_descriptor> g_instr_descr;
-        root->instrs.push_back(instr);
+        g_instr_descr.reserve(instr_descr.size());
+        for (uint32_t i = 0; i < instr_descr.size(); ++i) {
+            generic_instruction_descriptor new_instr_descr {instr_descr[i].value, instr_descr[i].mask, i};
+            g_instr_descr.push_back(new_instr_descr);
         }
-    populate_decoding_tree(root);
+        return std::move(g_instr_descr);
+    }()) {}
+
+inline bool is_icount_limit_enabled(finish_cond_e cond){
+    return (cond & finish_cond_e::ICOUNT_LIMIT) == finish_cond_e::ICOUNT_LIMIT;
 }
 
-inline bool is_count_limit_enabled(finish_cond_e cond){
+inline bool is_fcount_limit_enabled(finish_cond_e cond){
-    return (cond & finish_cond_e::COUNT_LIMIT) == finish_cond_e::COUNT_LIMIT;
+    return (cond & finish_cond_e::FCOUNT_LIMIT) == finish_cond_e::FCOUNT_LIMIT;
 }
 
 inline bool is_jump_to_self_enabled(finish_cond_e cond){
@@ -290,7 +242,7 @@ inline bool is_jump_to_self_enabled(finish_cond_e cond){
 }
 
 template <typename ARCH>
-typename vm_base<ARCH>::virt_addr_t vm_impl<ARCH>::execute_inst(finish_cond_e cond, virt_addr_t start, uint64_t icount_limit){
+typename vm_base<ARCH>::virt_addr_t vm_impl<ARCH>::execute_inst(finish_cond_e cond, virt_addr_t start, uint64_t count_limit){
     auto pc=start;
     auto* PC = reinterpret_cast<uint${addrDataWidth}_t*>(this->regs_base_ptr+arch::traits<ARCH>::reg_byte_offsets[arch::traits<ARCH>::PC]);
     auto* NEXT_PC = reinterpret_cast<uint${addrDataWidth}_t*>(this->regs_base_ptr+arch::traits<ARCH>::reg_byte_offsets[arch::traits<ARCH>::NEXT_PC]);
@@ -303,23 +255,35 @@ typename vm_base<ARCH>::virt_addr_t vm_impl<ARCH>::execute_inst(finish_cond_e co
     auto *const data = reinterpret_cast<uint8_t*>(&instr);
 
     while(!this->core.should_stop() &&
-            !(is_count_limit_enabled(cond) && icount >= icount_limit)){
+            !(is_icount_limit_enabled(cond) && icount >= count_limit) &&
+            !(is_fcount_limit_enabled(cond) && fetch_count >= count_limit)){
+        if(this->debugging_enabled())
+            this->tgt_adapter->check_continue(*PC);
+        pc.val=*PC;
         if(fetch_ins(pc, data)!=iss::Ok){
-            this->do_sync(POST_SYNC, std::numeric_limits<unsigned>::max());
+            if(this->sync_exec && PRE_SYNC) this->do_sync(PRE_SYNC, std::numeric_limits<unsigned>::max());
-            pc.val = super::core.enter_trap(std::numeric_limits<uint64_t>::max(), pc.val, 0);
+            process_spawn_blocks();
+            if(this->sync_exec && POST_SYNC) this->do_sync(PRE_SYNC, std::numeric_limits<unsigned>::max());
+            pc.val = super::core.enter_trap(arch::traits<ARCH>::RV_CAUSE_FETCH_ACCESS<<16, pc.val, 0);
         } else {
             if (is_jump_to_self_enabled(cond) &&
                     (instr == 0x0000006f || (instr&0xffff)==0xa001)) throw simulation_stopped(0); // 'J 0' or 'C.J 0'
-            auto inst_id = decode_instr(root, instr);
+            uint32_t inst_index = instr_decoder.decode_instr(instr);
+            opcode_e inst_id = arch::traits<ARCH>::opcode_e::MAX_OPCODE;;
+            if(inst_index <instr_descr.size())
+                inst_id = instr_descr[inst_index].op;
+
             // pre execution stuff
             this->core.reg.last_branch = 0;
             if(this->sync_exec && PRE_SYNC) this->do_sync(PRE_SYNC, static_cast<unsigned>(inst_id));
+            try{
                 switch(inst_id){<%instructions.eachWithIndex{instr, idx -> %>
                 case arch::traits<ARCH>::opcode_e::${instr.name}: {
                     <%instr.fields.eachLine{%>${it}
                     <%}%>if(this->disass_enabled){
                         /* generate console output when executing the command */<%instr.disass.eachLine{%>
                         ${it}<%}%>
+                        this->core.disass_output(pc.val, mnemonic);
                     }
                     // used registers<%instr.usedVariables.each{ k,v->
                     if(v.isArray) {%>
@@ -329,13 +293,14 @@ typename vm_base<ARCH>::virt_addr_t vm_impl<ARCH>::execute_inst(finish_cond_e co
                     *NEXT_PC = *PC + ${instr.length/8};
                     // execute instruction<%instr.behavior.eachLine{%>
                     ${it}<%}%>
-                TRAP_${instr.name}:break;
+                    break;
                 }// @suppress("No break at end of case")<%}%>
                 default: {
                     *NEXT_PC = *PC + ((instr & 3) == 3 ? 4 : 2);
                     raise(0,  2);
                 }
                 }
+            }catch(memory_access_exception& e){}
             // post execution stuff
             process_spawn_blocks();
             if(this->sync_exec && POST_SYNC) this->do_sync(POST_SYNC, static_cast<unsigned>(inst_id));
@@ -343,16 +308,18 @@ typename vm_base<ARCH>::virt_addr_t vm_impl<ARCH>::execute_inst(finish_cond_e co
             //    this->core.reg.trap_state =  this->core.reg.pending_trap;
             // trap check
             if(trap_state!=0){
-                super::core.enter_trap(trap_state, pc.val, instr);
+                //In case of Instruction address misaligned (cause = 0 and trapid = 0) need the targeted addr (in tval)
+                auto mcause = (trap_state>>16) & 0xff; 
+                super::core.enter_trap(trap_state, pc.val, mcause ? instr:tval);
             } else {
                 icount++;
                 instret++;
             }
-            cycle++;
+            *PC = *NEXT_PC;
-            pc.val=*NEXT_PC;
-            this->core.reg.PC = this->core.reg.NEXT_PC;
             this->core.reg.trap_state =  this->core.reg.pending_trap;
         }
+        fetch_count++;
+        cycle++;
     }
     return pc;
 }
@@ -368,24 +335,33 @@ std::unique_ptr<vm_if> create<arch::${coreDef.name.toLowerCase()}>(arch::${coreD
 } // namespace interp
 } // namespace iss
 
-#include <iss/factory.h>
 #include <iss/arch/riscv_hart_m_p.h>
 #include <iss/arch/riscv_hart_mu_p.h>
+#include <iss/factory.h>
 namespace iss {
 namespace {
 volatile std::array<bool, 2> dummy = {
-        core_factory::instance().register_creator("${coreDef.name.toLowerCase()}|m_p|interp", [](unsigned port, void*) -> std::tuple<cpu_ptr, vm_ptr>{
+        core_factory::instance().register_creator("${coreDef.name.toLowerCase()}|m_p|interp", [](unsigned port, void* init_data) -> std::tuple<cpu_ptr, vm_ptr>{
             auto* cpu = new iss::arch::riscv_hart_m_p<iss::arch::${coreDef.name.toLowerCase()}>();
 		    auto vm = new interp::${coreDef.name.toLowerCase()}::vm_impl<arch::${coreDef.name.toLowerCase()}>(*cpu, false);
 		    if (port != 0) debugger::server<debugger::gdb_session>::run_server(vm, port);
+            if(init_data){
+                auto* cb = reinterpret_cast<semihosting_cb_t<arch::traits<arch::${coreDef.name.toLowerCase()}>::reg_t>*>(init_data);
+                cpu->set_semihosting_callback(*cb);
+            }
             return {cpu_ptr{cpu}, vm_ptr{vm}};
         }),
-        core_factory::instance().register_creator("${coreDef.name.toLowerCase()}|mu_p|interp", [](unsigned port, void*) -> std::tuple<cpu_ptr, vm_ptr>{
+        core_factory::instance().register_creator("${coreDef.name.toLowerCase()}|mu_p|interp", [](unsigned port, void* init_data) -> std::tuple<cpu_ptr, vm_ptr>{
             auto* cpu = new iss::arch::riscv_hart_mu_p<iss::arch::${coreDef.name.toLowerCase()}>();
 		    auto vm = new interp::${coreDef.name.toLowerCase()}::vm_impl<arch::${coreDef.name.toLowerCase()}>(*cpu, false);
 		    if (port != 0) debugger::server<debugger::gdb_session>::run_server(vm, port);
+            if(init_data){
+                auto* cb = reinterpret_cast<semihosting_cb_t<arch::traits<arch::${coreDef.name.toLowerCase()}>::reg_t>*>(init_data);
+                cpu->set_semihosting_callback(*cb);
+            }
             return {cpu_ptr{cpu}, vm_ptr{vm}};
         })
 };
 }
 }
+// clang-format on

gen_input/templates/llvm/CORENAME.cpp.gtl

@@ -1,5 +1,5 @@
 /*******************************************************************************
- * Copyright (C) 2017, 2018 MINRES Technologies GmbH
+ * Copyright (C) 2017-2024 MINRES Technologies GmbH
  * All rights reserved.
  *
  * Redistribution and use in source and binary forms, with or without
@@ -29,14 +29,17 @@
  * POSSIBILITY OF SUCH DAMAGE.
  *
  *******************************************************************************/
-
+// clang-format off
 #include <iss/arch/${coreDef.name.toLowerCase()}.h>
 #include <iss/debugger/gdb_session.h>
 #include <iss/debugger/server.h>
 #include <iss/iss.h>
 #include <iss/llvm/vm_base.h>
 #include <util/logging.h>
-
+#include <iss/instruction_decoder.h>
+<%def fcsr = registers.find {it.name=='FCSR'}
+if(fcsr != null) {%>
+#include <vm/fp_functions.h><%}%>
 #ifndef FMT_HEADER_ONLY
 #define FMT_HEADER_ONLY
 #endif
@@ -58,6 +61,7 @@ using namespace iss::debugger;
 
 template <typename ARCH> class vm_impl : public iss::llvm::vm_base<ARCH> {
 public:
+    using traits = arch::traits<ARCH>;
     using super = typename iss::llvm::vm_base<ARCH>;
     using virt_addr_t = typename super::virt_addr_t;
     using phys_addr_t = typename super::phys_addr_t;
@@ -80,15 +84,17 @@ public:
 protected:
     using vm_base<ARCH>::get_reg_ptr;
 
-    inline const char *name(size_t index){return traits<ARCH>::reg_aliases.at(index);}
+    inline const char *name(size_t index){return traits::reg_aliases.at(index);}
-
+<%if(fcsr != null) {%>
+    inline const char *fname(size_t index){return index < 32?name(index+traits::F0):"illegal";}   
+<%}%>
     template <typename T> inline ConstantInt *size(T type) {
         return ConstantInt::get(getContext(), APInt(32, type->getType()->getScalarSizeInBits()));
     }
 
     void setup_module(Module* m) override {
         super::setup_module(m);
-        iss::llvm::fp_impl::add_fp_functions_2_module(m, traits<ARCH>::FP_REGS_SIZE, traits<ARCH>::XLEN);
+        iss::llvm::fp_impl::add_fp_functions_2_module(m, traits::FP_REGS_SIZE, traits::XLEN);
     }
 
     inline Value *gen_choose(Value *cond, Value *trueVal, Value *falseVal, unsigned size) {
@@ -98,219 +104,242 @@ protected:
     std::tuple<continuation_e, BasicBlock *> gen_single_inst_behavior(virt_addr_t &, unsigned int &, BasicBlock *) override;
 
     void gen_leave_behavior(BasicBlock *leave_blk) override;
-
     void gen_raise_trap(uint16_t trap_id, uint16_t cause);
-
     void gen_leave_trap(unsigned lvl);
-
     void gen_wait(unsigned type);
-
+    void set_tval(uint64_t new_tval);
+    void set_tval(Value* new_tval);
     void gen_trap_behavior(BasicBlock *) override;
-
+    void gen_instr_prologue();
-    void gen_trap_check(BasicBlock *bb);
+    void gen_instr_epilogue(BasicBlock *bb);
 
     inline Value *gen_reg_load(unsigned i, unsigned level = 0) {
         return this->builder.CreateLoad(this->get_typeptr(i), get_reg_ptr(i), false);
     }
 
     inline void gen_set_pc(virt_addr_t pc, unsigned reg_num) {
-        Value *next_pc_v = this->builder.CreateSExtOrTrunc(this->gen_const(traits<ARCH>::XLEN, pc.val),
+        Value *next_pc_v = this->builder.CreateSExtOrTrunc(this->gen_const(traits::XLEN, pc.val),
-                                                           this->get_type(traits<ARCH>::XLEN));
+                                                           this->get_type(traits::XLEN));
         this->builder.CreateStore(next_pc_v, get_reg_ptr(reg_num), true);
     }
 
     // some compile time constants
-    // enum { MASK16 = 0b1111110001100011, MASK32 = 0b11111111111100000111000001111111 };
-    enum { MASK16 = 0b1111111111111111, MASK32 = 0b11111111111100000111000001111111 };
-    enum { EXTR_MASK16 = MASK16 >> 2, EXTR_MASK32 = MASK32 >> 2 };
-    enum { LUT_SIZE = 1 << util::bit_count(static_cast<uint64_t>(EXTR_MASK32)), LUT_SIZE_C = 1 << util::bit_count(static_cast<uint64_t>(EXTR_MASK16)) };
 
     using this_class = vm_impl<ARCH>;
     using compile_func = std::tuple<continuation_e, BasicBlock *> (this_class::*)(virt_addr_t &pc,
                                                                                   code_word_t instr,
                                                                                   BasicBlock *bb);
-    std::array<compile_func, LUT_SIZE> lut;
+    template<unsigned W, typename U, typename S = typename std::make_signed<U>::type>
-
+    inline S sext(U from) {
-    std::array<compile_func, LUT_SIZE_C> lut_00, lut_01, lut_10;
+        auto mask = (1ULL<<W) - 1;
-    std::array<compile_func, LUT_SIZE> lut_11;
+        auto sign_mask = 1ULL<<(W-1);
-
+        return (from & mask) | ((from & sign_mask) ? ~mask : 0);
-	std::array<compile_func *, 4> qlut;
-
-	std::array<const uint32_t, 4> lutmasks = {{EXTR_MASK16, EXTR_MASK16, EXTR_MASK16, EXTR_MASK32}};
-
-    void expand_bit_mask(int pos, uint32_t mask, uint32_t value, uint32_t valid, uint32_t idx, compile_func lut[],
-                         compile_func f) {
-        if (pos < 0) {
-            lut[idx] = f;
-        } else {
-            auto bitmask = 1UL << pos;
-            if ((mask & bitmask) == 0) {
-                expand_bit_mask(pos - 1, mask, value, valid, idx, lut, f);
-            } else {
-                if ((valid & bitmask) == 0) {
-                    expand_bit_mask(pos - 1, mask, value, valid, (idx << 1), lut, f);
-                    expand_bit_mask(pos - 1, mask, value, valid, (idx << 1) + 1, lut, f);
-                } else {
-                    auto new_val = idx << 1;
-                    if ((value & bitmask) != 0) new_val++;
-                    expand_bit_mask(pos - 1, mask, value, valid, new_val, lut, f);
     }
-            }
+<%functions.each{ it.eachLine { %>
-        }
+    ${it}<%}%>
-    }
+<%}%>
-
-    inline uint32_t extract_fields(uint32_t val) { return extract_fields(29, val >> 2, lutmasks[val & 0x3], 0); }
-
-    uint32_t extract_fields(int pos, uint32_t val, uint32_t mask, uint32_t lut_val) {
-        if (pos >= 0) {
-            auto bitmask = 1UL << pos;
-            if ((mask & bitmask) == 0) {
-                lut_val = extract_fields(pos - 1, val, mask, lut_val);
-            } else {
-                auto new_val = lut_val << 1;
-                if ((val & bitmask) != 0) new_val++;
-                lut_val = extract_fields(pos - 1, val, mask, new_val);
-            }
-        }
-        return lut_val;
-    }
-
 private:
     /****************************************************************************
      * start opcode definitions
      ****************************************************************************/
-    struct InstructionDesriptor {
+    struct instruction_descriptor {
-        size_t length;
+        uint32_t length;
         uint32_t value;
         uint32_t mask;
         compile_func op;
     };
 
-    const std::array<InstructionDesriptor, ${instructions.size}> instr_descr = {{
+    const std::array<instruction_descriptor, ${instructions.size()}> instr_descr = {{
          /* entries are: size, valid value, valid mask, function ptr */<%instructions.each{instr -> %>
-        /* instruction ${instr.instruction.name} */
+        /* instruction ${instr.instruction.name}, encoding '${instr.encoding}' */
-        {${instr.length}, ${instr.value}, ${instr.mask}, &this_class::__${generator.functionName(instr.name)}},<%}%>
+        {${instr.length}, ${instr.encoding}, ${instr.mask}, &this_class::__${generator.functionName(instr.name)}},<%}%>
     }};
 
+    //needs to be declared after instr_descr
+    decoder instr_decoder;
+
     /* instruction definitions */<%instructions.eachWithIndex{instr, idx -> %>
     /* instruction ${idx}: ${instr.name} */
-    std::tuple<continuation_e, BasicBlock*> __${generator.functionName(instr.name)}(virt_addr_t& pc, code_word_t instr, BasicBlock* bb){<%instr.code.eachLine{%>
+    std::tuple<continuation_e, BasicBlock*> __${generator.functionName(instr.name)}(virt_addr_t& pc, code_word_t instr, BasicBlock* bb){
+        uint64_t PC = pc.val;
+        <%instr.fields.eachLine{%>${it}
+        <%}%>if(this->disass_enabled){
+            /* generate console output when executing the command */<%instr.disass.eachLine{%>
             ${it}<%}%>
+            std::vector<Value*> args {
+                this->core_ptr,
+                this->gen_const(64, pc.val),
+                this->builder.CreateGlobalStringPtr(mnemonic),
+            };
+            this->builder.CreateCall(this->mod->getFunction("print_disass"), args);
+        }
+        bb->setName(fmt::format("${instr.name}_0x{:X}",pc.val));
+        this->gen_sync(PRE_SYNC,${idx});
+        
+        this->gen_set_pc(pc, traits::PC);
+        this->set_tval(instr);
+        pc=pc+ ${instr.length/8};
+        this->gen_set_pc(pc, traits::NEXT_PC);
+        
+        this->gen_instr_prologue();
+        /*generate behavior*/
+        <%instr.behavior.eachLine{%>${it}
+        <%}%>
+        this->gen_sync(POST_SYNC, ${idx});
+        this->gen_instr_epilogue(bb);
+        this->builder.CreateBr(bb);
+    	return returnValue;        
     }
     <%}%>
     /****************************************************************************
      * end opcode definitions
      ****************************************************************************/
-    std::tuple<continuation_e, BasicBlock *> illegal_intruction(virt_addr_t &pc, code_word_t instr, BasicBlock *bb) {
+    std::tuple<continuation_e, BasicBlock *> illegal_instruction(virt_addr_t &pc, code_word_t instr, BasicBlock *bb) {
+        if(this->disass_enabled){
+            auto mnemonic = std::string("illegal_instruction");
+            std::vector<Value*> args {
+                this->core_ptr,
+                this->gen_const(64, pc.val),
+                this->builder.CreateGlobalStringPtr(mnemonic),
+            };
+            this->builder.CreateCall(this->mod->getFunction("print_disass"), args);
+        }
         this->gen_sync(iss::PRE_SYNC, instr_descr.size());
-        this->builder.CreateStore(this->builder.CreateLoad(this->get_typeptr(traits<ARCH>::NEXT_PC), get_reg_ptr(traits<ARCH>::NEXT_PC), true),
+        this->builder.CreateStore(this->builder.CreateLoad(this->get_typeptr(traits::NEXT_PC), get_reg_ptr(traits::NEXT_PC), true),
-                                   get_reg_ptr(traits<ARCH>::PC), true);
+                                   get_reg_ptr(traits::PC), true);
         this->builder.CreateStore(
-            this->builder.CreateAdd(this->builder.CreateLoad(this->get_typeptr(traits<ARCH>::ICOUNT), get_reg_ptr(traits<ARCH>::ICOUNT), true),
+            this->builder.CreateAdd(this->builder.CreateLoad(this->get_typeptr(traits::ICOUNT), get_reg_ptr(traits::ICOUNT), true),
                                      this->gen_const(64U, 1)),
-            get_reg_ptr(traits<ARCH>::ICOUNT), true);
+            get_reg_ptr(traits::ICOUNT), true);
         pc = pc + ((instr & 3) == 3 ? 4 : 2);
+        this->set_tval(instr);
         this->gen_raise_trap(0, 2);     // illegal instruction trap
 		this->gen_sync(iss::POST_SYNC, instr_descr.size());
-        this->gen_trap_check(this->leave_blk);
+        bb = this->leave_blk;
-        return std::make_tuple(BRANCH, nullptr);
+        this->gen_instr_epilogue(bb);
+        this->builder.CreateBr(bb);
+        return std::make_tuple(ILLEGAL_INSTR, nullptr);
     }    
 };
 
-template <typename CODE_WORD> void debug_fn(CODE_WORD insn) {
+template <typename CODE_WORD> void debug_fn(CODE_WORD instr) {
-    volatile CODE_WORD x = insn;
+    volatile CODE_WORD x = instr;
-    insn = 2 * x;
+    instr = 2 * x;
 }
 
 template <typename ARCH> vm_impl<ARCH>::vm_impl() { this(new ARCH()); }
 
 template <typename ARCH>
 vm_impl<ARCH>::vm_impl(ARCH &core, unsigned core_id, unsigned cluster_id)
-: vm_base<ARCH>(core, core_id, cluster_id) {
+: vm_base<ARCH>(core, core_id, cluster_id)
-    qlut[0] = lut_00.data();
+, instr_decoder([this]() {
-    qlut[1] = lut_01.data();
+        std::vector<generic_instruction_descriptor> g_instr_descr;
-    qlut[2] = lut_10.data();
+        g_instr_descr.reserve(instr_descr.size());
-    qlut[3] = lut_11.data();
+        for (uint32_t i = 0; i < instr_descr.size(); ++i) {
-    for (auto instr : instr_descr) {
+            generic_instruction_descriptor new_instr_descr {instr_descr[i].value, instr_descr[i].mask, i};
-        auto quantrant = instr.value & 0x3;
+            g_instr_descr.push_back(new_instr_descr);
-        expand_bit_mask(29, lutmasks[quantrant], instr.value >> 2, instr.mask >> 2, 0, qlut[quantrant], instr.op);
         }
-}
+        return std::move(g_instr_descr);
+    }()) {}
 
 template <typename ARCH>
 std::tuple<continuation_e, BasicBlock *>
 vm_impl<ARCH>::gen_single_inst_behavior(virt_addr_t &pc, unsigned int &inst_cnt, BasicBlock *this_block) {
     // we fetch at max 4 byte, alignment is 2
     enum {TRAP_ID=1<<16};
-    code_word_t insn = 0;
+    code_word_t instr = 0;
-    // const typename traits<ARCH>::addr_t upper_bits = ~traits<ARCH>::PGMASK;
+    // const typename traits::addr_t upper_bits = ~traits::PGMASK;
     phys_addr_t paddr(pc);
-    auto *const data = (uint8_t *)&insn;
+    auto *const data = (uint8_t *)&instr;
-    paddr = this->core.v2p(pc);
+    if(this->core.has_mmu())
-    //TODO: re-add page handling
+        paddr = this->core.virt2phys(pc);
-//    if ((pc.val & upper_bits) != ((pc.val + 2) & upper_bits)) { // we may cross a page boundary
-//        auto res = this->core.read(paddr, 2, data);
-//        if (res != iss::Ok) throw trap_access(TRAP_ID, pc.val);
-//        if ((insn & 0x3) == 0x3) { // this is a 32bit instruction
-//            res = this->core.read(this->core.v2p(pc + 2), 2, data + 2);
-//        }
-//    } else {
     auto res = this->core.read(paddr, 4, data);
-        if (res != iss::Ok) throw trap_access(TRAP_ID, pc.val);
+    if (res != iss::Ok) 
-//    }
+        return std::make_tuple(ILLEGAL_FETCH, nullptr);
-    if (insn == 0x0000006f || (insn&0xffff)==0xa001) throw simulation_stopped(0); // 'J 0' or 'C.J 0'
+    if (instr == 0x0000006f || (instr&0xffff)==0xa001)
-    // curr pc on stack
+        return std::make_tuple(JUMP_TO_SELF, nullptr);
     ++inst_cnt;
-    auto lut_val = extract_fields(insn);
+    uint32_t inst_index = instr_decoder.decode_instr(instr);
-    auto f = qlut[insn & 0x3][lut_val];
+    compile_func f = nullptr;
+    if(inst_index < instr_descr.size())
+        f = instr_descr[inst_index].op;
     if (f == nullptr) {
-        f = &this_class::illegal_intruction;
+        f = &this_class::illegal_instruction;
     }
-    return (this->*f)(pc, insn, this_block);
+    return (this->*f)(pc, instr, this_block);
 }
 
-template <typename ARCH> void vm_impl<ARCH>::gen_leave_behavior(BasicBlock *leave_blk) {
+template <typename ARCH>
+void vm_impl<ARCH>::gen_leave_behavior(BasicBlock *leave_blk) {
     this->builder.SetInsertPoint(leave_blk);
-    this->builder.CreateRet(this->builder.CreateLoad(this->get_typeptr(arch::traits<ARCH>::NEXT_PC), get_reg_ptr(arch::traits<ARCH>::NEXT_PC), false));
+    this->builder.CreateRet(this->builder.CreateLoad(this->get_typeptr(traits::NEXT_PC),get_reg_ptr(traits::NEXT_PC), false));
 }
 
-template <typename ARCH> void vm_impl<ARCH>::gen_raise_trap(uint16_t trap_id, uint16_t cause) {
+template <typename ARCH>
+void vm_impl<ARCH>::gen_raise_trap(uint16_t trap_id, uint16_t cause) {
     auto *TRAP_val = this->gen_const(32, 0x80 << 24 | (cause << 16) | trap_id);
-    this->builder.CreateStore(TRAP_val, get_reg_ptr(traits<ARCH>::TRAP_STATE), true);
+    this->builder.CreateStore(TRAP_val, get_reg_ptr(traits::TRAP_STATE), true);
-    this->builder.CreateStore(this->gen_const(32U, std::numeric_limits<uint32_t>::max()), get_reg_ptr(traits<ARCH>::LAST_BRANCH), false);
+    this->builder.CreateBr(this->trap_blk);
 }
 
-template <typename ARCH> void vm_impl<ARCH>::gen_leave_trap(unsigned lvl) {
+template <typename ARCH>
+void vm_impl<ARCH>::gen_leave_trap(unsigned lvl) {
     std::vector<Value *> args{ this->core_ptr, ConstantInt::get(getContext(), APInt(64, lvl)) };
     this->builder.CreateCall(this->mod->getFunction("leave_trap"), args);
-    auto *PC_val = this->gen_read_mem(traits<ARCH>::CSR, (lvl << 8) + 0x41, traits<ARCH>::XLEN / 8);
+    this->builder.CreateStore(this->gen_const(32U, static_cast<int>(UNKNOWN_JUMP)), get_reg_ptr(traits::LAST_BRANCH), false);
-    this->builder.CreateStore(PC_val, get_reg_ptr(traits<ARCH>::NEXT_PC), false);
-    this->builder.CreateStore(this->gen_const(32U, std::numeric_limits<uint32_t>::max()), get_reg_ptr(traits<ARCH>::LAST_BRANCH), false);
 }
 
-template <typename ARCH> void vm_impl<ARCH>::gen_wait(unsigned type) {
+template <typename ARCH>
+void vm_impl<ARCH>::gen_wait(unsigned type) {
     std::vector<Value *> args{ this->core_ptr, ConstantInt::get(getContext(), APInt(64, type)) };
     this->builder.CreateCall(this->mod->getFunction("wait"), args);
 }
 
-template <typename ARCH> void vm_impl<ARCH>::gen_trap_behavior(BasicBlock *trap_blk) {
+template <typename ARCH>
+inline void vm_impl<ARCH>::set_tval(uint64_t tval) {
+    auto tmp_tval = this->gen_const(64, tval);
+    this->set_tval(tmp_tval);
+}
+template <typename ARCH>
+inline void vm_impl<ARCH>::set_tval(Value* new_tval) {
+    this->builder.CreateStore(this->gen_ext(new_tval, 64, false), this->tval);
+}
+template <typename ARCH> 
+void vm_impl<ARCH>::gen_trap_behavior(BasicBlock *trap_blk) {
     this->builder.SetInsertPoint(trap_blk);
-    auto *trap_state_val = this->builder.CreateLoad(this->get_typeptr(arch::traits<ARCH>::TRAP_STATE), get_reg_ptr(traits<ARCH>::TRAP_STATE), true);
+    auto *trap_state_val = this->builder.CreateLoad(this->get_typeptr(traits::TRAP_STATE), get_reg_ptr(traits::TRAP_STATE), true);
-    this->builder.CreateStore(this->gen_const(32U, std::numeric_limits<uint32_t>::max()),
+    auto *cur_pc_val = this->builder.CreateLoad(this->get_typeptr(traits::PC), get_reg_ptr(traits::PC), true);
-                              get_reg_ptr(traits<ARCH>::LAST_BRANCH), false);
+    std::vector<Value *> args{this->core_ptr,
-    std::vector<Value *> args{this->core_ptr, this->adj_to64(trap_state_val),
+                                this->adj_to64(trap_state_val),
-                              this->adj_to64(this->builder.CreateLoad(this->get_typeptr(arch::traits<ARCH>::PC), get_reg_ptr(traits<ARCH>::PC), false))};
+                                this->adj_to64(cur_pc_val),
+                              this->adj_to64(this->builder.CreateLoad(this->get_type(64),this->tval))};
     this->builder.CreateCall(this->mod->getFunction("enter_trap"), args);
-    auto *trap_addr_val = this->builder.CreateLoad(this->get_typeptr(arch::traits<ARCH>::NEXT_PC), get_reg_ptr(traits<ARCH>::NEXT_PC), false);
+    this->builder.CreateStore(this->gen_const(32U, static_cast<int>(UNKNOWN_JUMP)), get_reg_ptr(traits::LAST_BRANCH), false);
+
+    auto *trap_addr_val = this->builder.CreateLoad(this->get_typeptr(traits::NEXT_PC), get_reg_ptr(traits::NEXT_PC), false);
     this->builder.CreateRet(trap_addr_val);
 }
+template <typename ARCH>
+void vm_impl<ARCH>::gen_instr_prologue() {
+    auto* trap_val =
+        this->builder.CreateLoad(this->get_typeptr(arch::traits<ARCH>::PENDING_TRAP), get_reg_ptr(arch::traits<ARCH>::PENDING_TRAP));
+    this->builder.CreateStore(trap_val, get_reg_ptr(arch::traits<ARCH>::TRAP_STATE), false);
+}
             
-template <typename ARCH> inline void vm_impl<ARCH>::gen_trap_check(BasicBlock *bb) {
+
-    auto *v = this->builder.CreateLoad(this->get_typeptr(arch::traits<ARCH>::TRAP_STATE), get_reg_ptr(arch::traits<ARCH>::TRAP_STATE), true);
+template <typename ARCH>
+void vm_impl<ARCH>::gen_instr_epilogue(BasicBlock *bb) {
+    auto* target_bb = BasicBlock::Create(this->mod->getContext(), "", this->func, bb);
+    auto *v = this->builder.CreateLoad(this->get_typeptr(traits::TRAP_STATE), get_reg_ptr(traits::TRAP_STATE), true);
     this->gen_cond_branch(this->builder.CreateICmp(
                               ICmpInst::ICMP_EQ, v,
                               ConstantInt::get(getContext(), APInt(v->getType()->getIntegerBitWidth(), 0))),
-                          bb, this->trap_blk, 1);
+                          target_bb, this->trap_blk, 1);
+    this->builder.SetInsertPoint(target_bb);
+    // update icount
+    auto* icount_val = this->builder.CreateAdd(
+        this->builder.CreateLoad(this->get_typeptr(arch::traits<ARCH>::ICOUNT), get_reg_ptr(arch::traits<ARCH>::ICOUNT)), this->gen_const(64U, 1));
+    this->builder.CreateStore(icount_val, get_reg_ptr(arch::traits<ARCH>::ICOUNT), false);
 }
 
 } // namespace ${coreDef.name.toLowerCase()}
@@ -324,24 +353,33 @@ std::unique_ptr<vm_if> create<arch::${coreDef.name.toLowerCase()}>(arch::${coreD
 } // namespace llvm
 } // namespace iss
 
-#include <iss/factory.h>
 #include <iss/arch/riscv_hart_m_p.h>
 #include <iss/arch/riscv_hart_mu_p.h>
+#include <iss/factory.h>
 namespace iss {
 namespace {
 volatile std::array<bool, 2> dummy = {
-        core_factory::instance().register_creator("${coreDef.name.toLowerCase()}|m_p|llvm", [](unsigned port, void*) -> std::tuple<cpu_ptr, vm_ptr>{
+        core_factory::instance().register_creator("${coreDef.name.toLowerCase()}|m_p|llvm", [](unsigned port, void* init_data) -> std::tuple<cpu_ptr, vm_ptr>{
             auto* cpu = new iss::arch::riscv_hart_m_p<iss::arch::${coreDef.name.toLowerCase()}>();
-            auto* vm = new llvm::${coreDef.name.toLowerCase()}::vm_impl<arch::${coreDef.name.toLowerCase()}>(*cpu, false);
+		    auto vm = new llvm::${coreDef.name.toLowerCase()}::vm_impl<arch::${coreDef.name.toLowerCase()}>(*cpu, false);
 		    if (port != 0) debugger::server<debugger::gdb_session>::run_server(vm, port);
+            if(init_data){
+                auto* cb = reinterpret_cast<std::function<void(arch_if*, arch::traits<arch::${coreDef.name.toLowerCase()}>::reg_t*, arch::traits<arch::${coreDef.name.toLowerCase()}>::reg_t*)>*>(init_data);
+                cpu->set_semihosting_callback(*cb);
+            }
             return {cpu_ptr{cpu}, vm_ptr{vm}};
         }),
-        core_factory::instance().register_creator("${coreDef.name.toLowerCase()}|mu_p|llvm", [](unsigned port, void*) -> std::tuple<cpu_ptr, vm_ptr>{
+        core_factory::instance().register_creator("${coreDef.name.toLowerCase()}|mu_p|llvm", [](unsigned port, void* init_data) -> std::tuple<cpu_ptr, vm_ptr>{
             auto* cpu = new iss::arch::riscv_hart_mu_p<iss::arch::${coreDef.name.toLowerCase()}>();
-            auto* vm = new llvm::${coreDef.name.toLowerCase()}::vm_impl<arch::${coreDef.name.toLowerCase()}>(*cpu, false);
+		    auto vm = new llvm::${coreDef.name.toLowerCase()}::vm_impl<arch::${coreDef.name.toLowerCase()}>(*cpu, false);
 		    if (port != 0) debugger::server<debugger::gdb_session>::run_server(vm, port);
+            if(init_data){
+                auto* cb = reinterpret_cast<std::function<void(arch_if*, arch::traits<arch::${coreDef.name.toLowerCase()}>::reg_t*, arch::traits<arch::${coreDef.name.toLowerCase()}>::reg_t*)>*>(init_data);
+                cpu->set_semihosting_callback(*cb);
+            }
             return {cpu_ptr{cpu}, vm_ptr{vm}};
         })
 };
 }
 }
+// clang-format on

gen_input/templates/tcc/CORENAME.cpp.gtl

@@ -1,5 +1,5 @@
 /*******************************************************************************
- * Copyright (C) 2020 MINRES Technologies GmbH
+ * Copyright (C) 2020-2024 MINRES Technologies GmbH
  * All rights reserved.
  *
  * Redistribution and use in source and binary forms, with or without
@@ -29,7 +29,7 @@
  * POSSIBILITY OF SUCH DAMAGE.
  *
  *******************************************************************************/
-
+// clang-format off
 #include <iss/arch/${coreDef.name.toLowerCase()}.h>
 #include <iss/debugger/gdb_session.h>
 #include <iss/debugger/server.h>
@@ -37,7 +37,10 @@
 #include <iss/tcc/vm_base.h>
 #include <util/logging.h>
 #include <sstream>
-
+#include <iss/instruction_decoder.h>
+<%def fcsr = registers.find {it.name=='FCSR'}
+if(fcsr != null) {%>
+#include <vm/fp_functions.h><%}%>
 #ifndef FMT_HEADER_ONLY
 #define FMT_HEADER_ONLY
 #endif
@@ -80,16 +83,21 @@ protected:
     using vm_base<ARCH>::get_reg_ptr;
 
     using this_class = vm_impl<ARCH>;
-    using compile_ret_t = std::tuple<continuation_e>;
+    using compile_ret_t = continuation_e;
     using compile_func = compile_ret_t (this_class::*)(virt_addr_t &pc, code_word_t instr, tu_builder&);
 
     inline const char *name(size_t index){return traits::reg_aliases.at(index);}
+<%
+if(fcsr != null) {%>
+    inline const char *fname(size_t index){return index < 32?name(index+traits::F0):"illegal";}   
+<%}%>
+    void add_prologue(tu_builder& tu) override;
 
     void setup_module(std::string m) override {
         super::setup_module(m);
     }
 
-    compile_ret_t gen_single_inst_behavior(virt_addr_t &, unsigned int &, tu_builder&) override;
+    compile_ret_t gen_single_inst_behavior(virt_addr_t &, tu_builder&) override;
 
     void gen_trap_behavior(tu_builder& tu) override;
 
@@ -97,7 +105,9 @@ protected:
 
     void gen_leave_trap(tu_builder& tu, unsigned lvl);
 
-    void gen_wait(tu_builder& tu, unsigned type);
+    inline void gen_set_tval(tu_builder& tu, uint64_t new_tval);
+
+    inline void gen_set_tval(tu_builder& tu, value new_tval);
 
     inline void gen_trap_check(tu_builder& tu) {
         tu("if(*trap_state!=0) goto trap_entry;");
@@ -128,47 +138,48 @@ protected:
         return (from & mask) | ((from & sign_mask) ? ~mask : 0);
     }
 
+<%functions.each{ it.eachLine { %>
+    ${it}<%}%>
+<%}%>
 private:
     /****************************************************************************
      * start opcode definitions
      ****************************************************************************/
     struct instruction_descriptor {
-        size_t length;
+        uint32_t length;
         uint32_t value;
         uint32_t mask;
         compile_func op;
     };
-    struct decoding_tree_node{
-        std::vector<instruction_descriptor> instrs;
-        std::vector<decoding_tree_node*> children;
-        uint32_t submask = std::numeric_limits<uint32_t>::max();
-        uint32_t value;
-        decoding_tree_node(uint32_t value) : value(value){}
-    };
 
-    decoding_tree_node* root {nullptr};
+    const std::array<instruction_descriptor, ${instructions.size()}> instr_descr = {{
-
-    const std::array<instruction_descriptor, ${instructions.size}> instr_descr = {{
          /* entries are: size, valid value, valid mask, function ptr */<%instructions.each{instr -> %>
         /* instruction ${instr.instruction.name}, encoding '${instr.encoding}' */
         {${instr.length}, ${instr.encoding}, ${instr.mask}, &this_class::__${generator.functionName(instr.name)}},<%}%>
     }};
 
+    //needs to be declared after instr_descr
+    decoder instr_decoder;
+
     /* instruction definitions */<%instructions.eachWithIndex{instr, idx -> %>
     /* instruction ${idx}: ${instr.name} */
     compile_ret_t __${generator.functionName(instr.name)}(virt_addr_t& pc, code_word_t instr, tu_builder& tu){
         tu("${instr.name}_{:#010x}:", pc.val);
         vm_base<ARCH>::gen_sync(tu, PRE_SYNC,${idx});
+        uint64_t PC = pc.val;
         <%instr.fields.eachLine{%>${it}
         <%}%>if(this->disass_enabled){
             /* generate console output when executing the command */<%instr.disass.eachLine{%>
             ${it}<%}%>
+            tu("print_disass(core_ptr, {:#x}, \"{}\");", pc.val, mnemonic);
         }
         auto cur_pc_val = tu.constant(pc.val, traits::reg_bit_widths[traits::PC]);
         pc=pc+ ${instr.length/8};
         gen_set_pc(tu, pc, traits::NEXT_PC);
-        tu.open_scope();<%instr.behavior.eachLine{%>
+        tu.open_scope();
-        ${it}<%}%>
+        this->gen_set_tval(tu, instr);
+        <%instr.behavior.eachLine{%>${it}
+        <%}%>
         tu.close_scope();
         vm_base<ARCH>::gen_sync(tu, POST_SYNC,${idx});
         gen_trap_check(tu);        
@@ -178,66 +189,18 @@ private:
     /****************************************************************************
      * end opcode definitions
      ****************************************************************************/
-    compile_ret_t illegal_intruction(virt_addr_t &pc, code_word_t instr, tu_builder& tu) {
+    compile_ret_t illegal_instruction(virt_addr_t &pc, code_word_t instr, tu_builder& tu) {
         vm_impl::gen_sync(tu, iss::PRE_SYNC, instr_descr.size());
+        if(this->disass_enabled){
+            /* generate console output when executing the command */
+            tu("print_disass(core_ptr, {:#x}, \"{}\");", pc.val, std::string("illegal_instruction"));
+        }
         pc = pc + ((instr & 3) == 3 ? 4 : 2);
-        gen_raise_trap(tu, 0, 2);     // illegal instruction trap
+        gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+        this->gen_set_tval(tu, instr);
         vm_impl::gen_sync(tu, iss::POST_SYNC, instr_descr.size());
         vm_impl::gen_trap_check(tu);
-        return BRANCH;
+        return ILLEGAL_INSTR;
-    }
-    
-    //decoding functionality
-
-    void populate_decoding_tree(decoding_tree_node* root){
-        //create submask
-        for(auto instr: root->instrs){
-            root->submask &= instr.mask;
-        }
-        //put each instr according to submask&encoding into children
-        for(auto instr: root->instrs){
-            bool foundMatch = false;
-            for(auto child: root->children){
-                //use value as identifying trait
-                if(child->value == (instr.value&root->submask)){
-                    child->instrs.push_back(instr);
-                    foundMatch = true;
-                }
-            }
-            if(!foundMatch){
-                decoding_tree_node* child = new decoding_tree_node(instr.value&root->submask);
-                child->instrs.push_back(instr);
-                root->children.push_back(child);
-            }
-        }
-        root->instrs.clear();
-        //call populate_decoding_tree for all children
-        if(root->children.size() >1)
-            for(auto child: root->children){
-                populate_decoding_tree(child);      
-            }
-        else{
-            //sort instrs by value of the mask, this works bc we want to have the least restrictive one last
-            std::sort(root->children[0]->instrs.begin(), root->children[0]->instrs.end(), [](const instruction_descriptor& instr1, const instruction_descriptor& instr2) {
-            return instr1.mask > instr2.mask;
-            }); 
-        }
-    }
-    compile_func decode_instr(decoding_tree_node* node, code_word_t word){
-        if(!node->children.size()){
-            if(node->instrs.size() == 1) return node->instrs[0].op;
-            for(auto instr : node->instrs){
-                if((instr.mask&word) == instr.value) return instr.op;
-            }
-        }
-        else{
-            for(auto child : node->children){
-                if (child->value == (node->submask&word)){
-                    return decode_instr(child, word);
-                }  
-            }  
-        }
-        return nullptr;
     }
 };
 
@@ -250,64 +213,102 @@ template <typename ARCH> vm_impl<ARCH>::vm_impl() { this(new ARCH()); }
 
 template <typename ARCH>
 vm_impl<ARCH>::vm_impl(ARCH &core, unsigned core_id, unsigned cluster_id)
-: vm_base<ARCH>(core, core_id, cluster_id) {
+: vm_base<ARCH>(core, core_id, cluster_id)
-    root = new decoding_tree_node(std::numeric_limits<uint32_t>::max());
+, instr_decoder([this]() {
-    for(auto instr:instr_descr){
+        std::vector<generic_instruction_descriptor> g_instr_descr;
-        root->instrs.push_back(instr);
+        g_instr_descr.reserve(instr_descr.size());
+        for (uint32_t i = 0; i < instr_descr.size(); ++i) {
+            generic_instruction_descriptor new_instr_descr {instr_descr[i].value, instr_descr[i].mask, i};
+            g_instr_descr.push_back(new_instr_descr);
         }
-    populate_decoding_tree(root);
+        return std::move(g_instr_descr);
-}
+    }()) {}
 
 template <typename ARCH>
-std::tuple<continuation_e>
+continuation_e
-vm_impl<ARCH>::gen_single_inst_behavior(virt_addr_t &pc, unsigned int &inst_cnt, tu_builder& tu) {
+vm_impl<ARCH>::gen_single_inst_behavior(virt_addr_t &pc, tu_builder& tu) {
     // we fetch at max 4 byte, alignment is 2
     enum {TRAP_ID=1<<16};
     code_word_t instr = 0;
     phys_addr_t paddr(pc);
     if(this->core.has_mmu())
         paddr = this->core.virt2phys(pc);
-    //TODO: re-add page handling
-//    if ((pc.val & upper_bits) != ((pc.val + 2) & upper_bits)) { // we may cross a page boundary
-//        auto res = this->core.read(paddr, 2, data);
-//        if (res != iss::Ok) throw trap_access(TRAP_ID, pc.val);
-//        if ((insn & 0x3) == 0x3) { // this is a 32bit instruction
-//            res = this->core.read(this->core.v2p(pc + 2), 2, data + 2);
-//        }
-//    } else {
     auto res = this->core.read(paddr, 4, reinterpret_cast<uint8_t*>(&instr));
-        if (res != iss::Ok) throw trap_access(TRAP_ID, pc.val);
+    if (res != iss::Ok)
-//    }
+        return ILLEGAL_FETCH;
-    if (instr == 0x0000006f || (instr&0xffff)==0xa001) throw simulation_stopped(0); // 'J 0' or 'C.J 0'
+    if (instr == 0x0000006f || (instr&0xffff)==0xa001) 
-    // curr pc on stack
+        return JUMP_TO_SELF;
-    ++inst_cnt;
+    uint32_t inst_index = instr_decoder.decode_instr(instr);
-    auto f = decode_instr(root, instr);
+    compile_func f = nullptr;
+    if(inst_index < instr_descr.size())
+        f = instr_descr[inst_index].op;
     if (f == nullptr) {
-        f = &this_class::illegal_intruction;
+        f = &this_class::illegal_instruction;
     }
     return (this->*f)(pc, instr, tu);
 }
 
 template <typename ARCH> void vm_impl<ARCH>::gen_raise_trap(tu_builder& tu, uint16_t trap_id, uint16_t cause) {
     tu("  *trap_state = {:#x};", 0x80 << 24 | (cause << 16) | trap_id);
-    tu.store(traits::LAST_BRANCH, tu.constant(std::numeric_limits<uint32_t>::max(), 32));
 }
 
 template <typename ARCH> void vm_impl<ARCH>::gen_leave_trap(tu_builder& tu, unsigned lvl) {
     tu("leave_trap(core_ptr, {});", lvl);
     tu.store(traits::NEXT_PC, tu.read_mem(traits::CSR, (lvl << 8) + 0x41, traits::XLEN));
-    tu.store(traits::LAST_BRANCH, tu.constant(std::numeric_limits<uint32_t>::max(), 32));
+    tu.store(traits::LAST_BRANCH, tu.constant(static_cast<int>(UNKNOWN_JUMP), 32));
 }
 
-template <typename ARCH> void vm_impl<ARCH>::gen_wait(tu_builder& tu, unsigned type) {
+template <typename ARCH> void vm_impl<ARCH>::gen_set_tval(tu_builder& tu, uint64_t new_tval) {
+    tu(fmt::format("tval = {};", new_tval));
+}
+template <typename ARCH> void vm_impl<ARCH>::gen_set_tval(tu_builder& tu, value new_tval) {
+    tu(fmt::format("tval = {};", new_tval.str));
 }
 
 template <typename ARCH> void vm_impl<ARCH>::gen_trap_behavior(tu_builder& tu) {
     tu("trap_entry:");
-    tu("enter_trap(core_ptr, *trap_state, *pc, 0);");
+    this->gen_sync(tu, POST_SYNC, -1);    
-    tu.store(traits::LAST_BRANCH, tu.constant(std::numeric_limits<uint32_t>::max(),32));
+    tu("enter_trap(core_ptr, *trap_state, *pc, tval);");
+    tu.store(traits::LAST_BRANCH, tu.constant(static_cast<int>(UNKNOWN_JUMP),32));
     tu("return *next_pc;");
 }
+<%
+
+template <typename ARCH> void vm_impl<ARCH>::add_prologue(tu_builder& tu){
+    std::ostringstream os;
+    os << add_reg_ptr("trap_state", arch::traits<ARCH>::TRAP_STATE);
+    os << add_reg_ptr("pending_trap", arch::traits<ARCH>::PENDING_TRAP);
+if(fcsr != null) {%>
+    os << "uint32_t (*fget_flags)()=" << (uintptr_t)&fget_flags << ";\\n";
+    os << "uint32_t (*fadd_s)(uint32_t v1, uint32_t v2, uint8_t mode)=" << (uintptr_t)&fadd_s << ";\\n";
+    os << "uint32_t (*fsub_s)(uint32_t v1, uint32_t v2, uint8_t mode)=" << (uintptr_t)&fsub_s << ";\\n";
+    os << "uint32_t (*fmul_s)(uint32_t v1, uint32_t v2, uint8_t mode)=" << (uintptr_t)&fmul_s << ";\\n";
+    os << "uint32_t (*fdiv_s)(uint32_t v1, uint32_t v2, uint8_t mode)=" << (uintptr_t)&fdiv_s << ";\\n";
+    os << "uint32_t (*fsqrt_s)(uint32_t v1, uint8_t mode)=" << (uintptr_t)&fsqrt_s << ";\\n";
+    os << "uint32_t (*fcmp_s)(uint32_t v1, uint32_t v2, uint32_t op)=" << (uintptr_t)&fcmp_s << ";\\n";
+    os << "uint32_t (*fcvt_s)(uint32_t v1, uint32_t op, uint8_t mode)=" << (uintptr_t)&fcvt_s << ";\\n";
+    os << "uint32_t (*fmadd_s)(uint32_t v1, uint32_t v2, uint32_t v3, uint32_t op, uint8_t mode)=" << (uintptr_t)&fmadd_s << ";\\n";
+    os << "uint32_t (*fsel_s)(uint32_t v1, uint32_t v2, uint32_t op)=" << (uintptr_t)&fsel_s << ";\\n";
+    os << "uint32_t (*fclass_s)( uint32_t v1 )=" << (uintptr_t)&fclass_s << ";\\n";
+    os << "uint32_t (*fconv_d2f)(uint64_t v1, uint8_t mode)=" << (uintptr_t)&fconv_d2f << ";\\n";
+    os << "uint64_t (*fconv_f2d)(uint32_t v1, uint8_t mode)=" << (uintptr_t)&fconv_f2d << ";\\n";
+    os << "uint64_t (*fadd_d)(uint64_t v1, uint64_t v2, uint8_t mode)=" << (uintptr_t)&fadd_d << ";\\n";
+    os << "uint64_t (*fsub_d)(uint64_t v1, uint64_t v2, uint8_t mode)=" << (uintptr_t)&fsub_d << ";\\n";
+    os << "uint64_t (*fmul_d)(uint64_t v1, uint64_t v2, uint8_t mode)=" << (uintptr_t)&fmul_d << ";\\n";
+    os << "uint64_t (*fdiv_d)(uint64_t v1, uint64_t v2, uint8_t mode)=" << (uintptr_t)&fdiv_d << ";\\n";
+    os << "uint64_t (*fsqrt_d)(uint64_t v1, uint8_t mode)=" << (uintptr_t)&fsqrt_d << ";\\n";
+    os << "uint64_t (*fcmp_d)(uint64_t v1, uint64_t v2, uint32_t op)=" << (uintptr_t)&fcmp_d << ";\\n";
+    os << "uint64_t (*fcvt_d)(uint64_t v1, uint32_t op, uint8_t mode)=" << (uintptr_t)&fcvt_d << ";\\n";
+    os << "uint64_t (*fmadd_d)(uint64_t v1, uint64_t v2, uint64_t v3, uint32_t op, uint8_t mode)=" << (uintptr_t)&fmadd_d << ";\\n";
+    os << "uint64_t (*fsel_d)(uint64_t v1, uint64_t v2, uint32_t op)=" << (uintptr_t)&fsel_d << ";\\n";
+    os << "uint64_t (*fclass_d)(uint64_t v1  )=" << (uintptr_t)&fclass_d << ";\\n";
+    os << "uint64_t (*fcvt_32_64)(uint32_t v1, uint32_t op, uint8_t mode)=" << (uintptr_t)&fcvt_32_64 << ";\\n";
+    os << "uint32_t (*fcvt_64_32)(uint64_t v1, uint32_t op, uint8_t mode)=" << (uintptr_t)&fcvt_64_32 << ";\\n";
+    os << "uint32_t (*unbox_s)(uint64_t v)=" << (uintptr_t)&unbox_s << ";\\n";
+    <%}%>
+    tu.add_prologue(os.str());
+}
+
 
 } // namespace ${coreDef.name.toLowerCase()}
 
@@ -320,24 +321,33 @@ std::unique_ptr<vm_if> create<arch::${coreDef.name.toLowerCase()}>(arch::${coreD
 } // namesapce tcc
 } // namespace iss
 
-#include <iss/factory.h>
 #include <iss/arch/riscv_hart_m_p.h>
 #include <iss/arch/riscv_hart_mu_p.h>
+#include <iss/factory.h>
 namespace iss {
 namespace {
 volatile std::array<bool, 2> dummy = {
-        core_factory::instance().register_creator("${coreDef.name.toLowerCase()}|m_p|tcc", [](unsigned port, void*) -> std::tuple<cpu_ptr, vm_ptr>{
+        core_factory::instance().register_creator("${coreDef.name.toLowerCase()}|m_p|tcc", [](unsigned port, void* init_data) -> std::tuple<cpu_ptr, vm_ptr>{
             auto* cpu = new iss::arch::riscv_hart_m_p<iss::arch::${coreDef.name.toLowerCase()}>();
 		    auto vm = new tcc::${coreDef.name.toLowerCase()}::vm_impl<arch::${coreDef.name.toLowerCase()}>(*cpu, false);
 		    if (port != 0) debugger::server<debugger::gdb_session>::run_server(vm, port);
+            if(init_data){
+                auto* cb = reinterpret_cast<semihosting_cb_t<arch::traits<arch::${coreDef.name.toLowerCase()}>::reg_t>*>(init_data);
+                cpu->set_semihosting_callback(*cb);
+            }
             return {cpu_ptr{cpu}, vm_ptr{vm}};
         }),
-        core_factory::instance().register_creator("${coreDef.name.toLowerCase()}|mu_p|tcc", [](unsigned port, void*) -> std::tuple<cpu_ptr, vm_ptr>{
+        core_factory::instance().register_creator("${coreDef.name.toLowerCase()}|mu_p|tcc", [](unsigned port, void* init_data) -> std::tuple<cpu_ptr, vm_ptr>{
             auto* cpu = new iss::arch::riscv_hart_mu_p<iss::arch::${coreDef.name.toLowerCase()}>();
 		    auto vm = new tcc::${coreDef.name.toLowerCase()}::vm_impl<arch::${coreDef.name.toLowerCase()}>(*cpu, false);
 		    if (port != 0) debugger::server<debugger::gdb_session>::run_server(vm, port);
+            if(init_data){
+                auto* cb = reinterpret_cast<semihosting_cb_t<arch::traits<arch::${coreDef.name.toLowerCase()}>::reg_t>*>(init_data);
+                cpu->set_semihosting_callback(*cb);
+            }
             return {cpu_ptr{cpu}, vm_ptr{vm}};
         })
 };
 }
 }
+// clang-format on

softfloat/.gitignore

@@ -0,0 +1,2 @@
+build/*/*.o
+build/*/*.a

softfloat/CMakeLists.txt

@@ -327,7 +327,7 @@ set(OTHERS
 
 set(LIB_SOURCES ${PRIMITIVES} ${SPECIALIZE} ${OTHERS})
 
-add_library(softfloat ${LIB_SOURCES})
+add_library(softfloat STATIC ${LIB_SOURCES})
 set_property(TARGET softfloat PROPERTY C_STANDARD 99)
 target_compile_definitions(softfloat PRIVATE 
 	SOFTFLOAT_ROUND_ODD 
@@ -347,7 +347,7 @@ set_target_properties(softfloat PROPERTIES
 
 install(TARGETS softfloat
   EXPORT ${PROJECT_NAME}Targets            # for downstream dependencies
-  ARCHIVE DESTINATION ${CMAKE_INSTALL_LIBDIR} COMPONENT libs   # static lib
+  ARCHIVE DESTINATION ${CMAKE_INSTALL_LIBDIR}/static COMPONENT libs   # static lib
   LIBRARY DESTINATION ${CMAKE_INSTALL_LIBDIR} COMPONENT libs   # shared lib
   FRAMEWORK DESTINATION ${CMAKE_INSTALL_LIBDIR} COMPONENT libs # for mac
   PUBLIC_HEADER DESTINATION ${CMAKE_INSTALL_INCLUDEDIR} COMPONENT devel   # headers for mac (note the different component -> different package)

softfloat/README.md

@@ -0,0 +1,24 @@
+
+Package Overview for Berkeley SoftFloat Release 3e
+==================================================
+
+John R. Hauser<br>
+2018 January 20
+
+
+Berkeley SoftFloat is a software implementation of binary floating-point
+that conforms to the IEEE Standard for Floating-Point Arithmetic.  SoftFloat
+is distributed in the form of C source code.  Building the SoftFloat sources
+generates a library file (typically `softfloat.a` or `libsoftfloat.a`)
+containing the floating-point subroutines.
+
+
+The SoftFloat package is documented in the following files in the `doc`
+subdirectory:
+
+* [SoftFloat.html](http://www.jhauser.us/arithmetic/SoftFloat-3/doc/SoftFloat.html) Documentation for using the SoftFloat functions.
+* [SoftFloat-source.html](http://www.jhauser.us/arithmetic/SoftFloat-3/doc/SoftFloat-source.html) Documentation for building SoftFloat.
+* [SoftFloat-history.html](http://www.jhauser.us/arithmetic/SoftFloat-3/doc/SoftFloat-history.html) History of the major changes to SoftFloat.
+
+Other files in the package comprise the source code for SoftFloat.
+

softfloat/build/Linux-RISCV64-GCC/Makefile

@@ -0,0 +1,399 @@
+
+#=============================================================================
+#
+# This Makefile is part of the SoftFloat IEEE Floating-Point Arithmetic
+# Package, Release 3e, by John R. Hauser.
+#
+# Copyright 2011, 2012, 2013, 2014, 2015, 2016, 2017 The Regents of the
+# University of California.  All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions are met:
+#
+#  1. Redistributions of source code must retain the above copyright notice,
+#     this list of conditions, and the following disclaimer.
+#
+#  2. Redistributions in binary form must reproduce the above copyright
+#     notice, this list of conditions, and the following disclaimer in the
+#     documentation and/or other materials provided with the distribution.
+#
+#  3. Neither the name of the University nor the names of its contributors
+#     may be used to endorse or promote products derived from this software
+#     without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS "AS IS", AND ANY
+# EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+# WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, ARE
+# DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY
+# DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+# (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+# LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
+# ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
+# THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+#
+#=============================================================================
+
+SOURCE_DIR ?= ../../source
+SPECIALIZE_TYPE ?= RISCV
+MARCH ?= rv64gcv_zfh_zfhmin
+MABI ?= lp64d
+
+SOFTFLOAT_OPTS ?= \
+  -DSOFTFLOAT_ROUND_ODD -DINLINE_LEVEL=5 -DSOFTFLOAT_FAST_DIV32TO16 \
+  -DSOFTFLOAT_FAST_DIV64TO32
+
+DELETE = rm -f
+C_INCLUDES = -I. -I$(SOURCE_DIR)/$(SPECIALIZE_TYPE) -I$(SOURCE_DIR)/include
+COMPILE_C = \
+  riscv64-unknown-linux-gnu-gcc -c -march=$(MARCH) -mabi=$(MABI) -Werror-implicit-function-declaration -DSOFTFLOAT_FAST_INT64 \
+    $(SOFTFLOAT_OPTS) $(C_INCLUDES) -O2 -o $@
+MAKELIB = ar crs $@
+
+OBJ = .o
+LIB = .a
+
+OTHER_HEADERS = $(SOURCE_DIR)/include/opts-GCC.h
+
+.PHONY: all
+all: softfloat$(LIB)
+
+OBJS_PRIMITIVES = \
+  s_eq128$(OBJ) \
+  s_le128$(OBJ) \
+  s_lt128$(OBJ) \
+  s_shortShiftLeft128$(OBJ) \
+  s_shortShiftRight128$(OBJ) \
+  s_shortShiftRightJam64$(OBJ) \
+  s_shortShiftRightJam64Extra$(OBJ) \
+  s_shortShiftRightJam128$(OBJ) \
+  s_shortShiftRightJam128Extra$(OBJ) \
+  s_shiftRightJam32$(OBJ) \
+  s_shiftRightJam64$(OBJ) \
+  s_shiftRightJam64Extra$(OBJ) \
+  s_shiftRightJam128$(OBJ) \
+  s_shiftRightJam128Extra$(OBJ) \
+  s_shiftRightJam256M$(OBJ) \
+  s_countLeadingZeros8$(OBJ) \
+  s_countLeadingZeros16$(OBJ) \
+  s_countLeadingZeros32$(OBJ) \
+  s_countLeadingZeros64$(OBJ) \
+  s_add128$(OBJ) \
+  s_add256M$(OBJ) \
+  s_sub128$(OBJ) \
+  s_sub256M$(OBJ) \
+  s_mul64ByShifted32To128$(OBJ) \
+  s_mul64To128$(OBJ) \
+  s_mul128By32$(OBJ) \
+  s_mul128To256M$(OBJ) \
+  s_approxRecip_1Ks$(OBJ) \
+  s_approxRecip32_1$(OBJ) \
+  s_approxRecipSqrt_1Ks$(OBJ) \
+  s_approxRecipSqrt32_1$(OBJ) \
+
+OBJS_SPECIALIZE = \
+  softfloat_raiseFlags$(OBJ) \
+  s_f16UIToCommonNaN$(OBJ) \
+  s_commonNaNToF16UI$(OBJ) \
+  s_propagateNaNF16UI$(OBJ) \
+  s_bf16UIToCommonNaN$(OBJ) \
+  s_commonNaNToBF16UI$(OBJ) \
+  s_f32UIToCommonNaN$(OBJ) \
+  s_commonNaNToF32UI$(OBJ) \
+  s_propagateNaNF32UI$(OBJ) \
+  s_f64UIToCommonNaN$(OBJ) \
+  s_commonNaNToF64UI$(OBJ) \
+  s_propagateNaNF64UI$(OBJ) \
+  extF80M_isSignalingNaN$(OBJ) \
+  s_extF80UIToCommonNaN$(OBJ) \
+  s_commonNaNToExtF80UI$(OBJ) \
+  s_propagateNaNExtF80UI$(OBJ) \
+  f128M_isSignalingNaN$(OBJ) \
+  s_f128UIToCommonNaN$(OBJ) \
+  s_commonNaNToF128UI$(OBJ) \
+  s_propagateNaNF128UI$(OBJ) \
+
+OBJS_OTHERS = \
+  s_roundToUI32$(OBJ) \
+  s_roundToUI64$(OBJ) \
+  s_roundToI32$(OBJ) \
+  s_roundToI64$(OBJ) \
+  s_normSubnormalBF16Sig$(OBJ) \
+  s_roundPackToBF16$(OBJ) \
+  s_normSubnormalF16Sig$(OBJ) \
+  s_roundPackToF16$(OBJ) \
+  s_normRoundPackToF16$(OBJ) \
+  s_addMagsF16$(OBJ) \
+  s_subMagsF16$(OBJ) \
+  s_mulAddF16$(OBJ) \
+  s_normSubnormalF32Sig$(OBJ) \
+  s_roundPackToF32$(OBJ) \
+  s_normRoundPackToF32$(OBJ) \
+  s_addMagsF32$(OBJ) \
+  s_subMagsF32$(OBJ) \
+  s_mulAddF32$(OBJ) \
+  s_normSubnormalF64Sig$(OBJ) \
+  s_roundPackToF64$(OBJ) \
+  s_normRoundPackToF64$(OBJ) \
+  s_addMagsF64$(OBJ) \
+  s_subMagsF64$(OBJ) \
+  s_mulAddF64$(OBJ) \
+  s_normSubnormalExtF80Sig$(OBJ) \
+  s_roundPackToExtF80$(OBJ) \
+  s_normRoundPackToExtF80$(OBJ) \
+  s_addMagsExtF80$(OBJ) \
+  s_subMagsExtF80$(OBJ) \
+  s_normSubnormalF128Sig$(OBJ) \
+  s_roundPackToF128$(OBJ) \
+  s_normRoundPackToF128$(OBJ) \
+  s_addMagsF128$(OBJ) \
+  s_subMagsF128$(OBJ) \
+  s_mulAddF128$(OBJ) \
+  softfloat_state$(OBJ) \
+  ui32_to_f16$(OBJ) \
+  ui32_to_f32$(OBJ) \
+  ui32_to_f64$(OBJ) \
+  ui32_to_extF80$(OBJ) \
+  ui32_to_extF80M$(OBJ) \
+  ui32_to_f128$(OBJ) \
+  ui32_to_f128M$(OBJ) \
+  ui64_to_f16$(OBJ) \
+  ui64_to_f32$(OBJ) \
+  ui64_to_f64$(OBJ) \
+  ui64_to_extF80$(OBJ) \
+  ui64_to_extF80M$(OBJ) \
+  ui64_to_f128$(OBJ) \
+  ui64_to_f128M$(OBJ) \
+  i32_to_f16$(OBJ) \
+  i32_to_f32$(OBJ) \
+  i32_to_f64$(OBJ) \
+  i32_to_extF80$(OBJ) \
+  i32_to_extF80M$(OBJ) \
+  i32_to_f128$(OBJ) \
+  i32_to_f128M$(OBJ) \
+  i64_to_f16$(OBJ) \
+  i64_to_f32$(OBJ) \
+  i64_to_f64$(OBJ) \
+  i64_to_extF80$(OBJ) \
+  i64_to_extF80M$(OBJ) \
+  i64_to_f128$(OBJ) \
+  i64_to_f128M$(OBJ) \
+  bf16_isSignalingNaN$(OBJ) \
+  bf16_to_f32$(OBJ) \
+  f16_to_ui32$(OBJ) \
+  f16_to_ui64$(OBJ) \
+  f16_to_i32$(OBJ) \
+  f16_to_i64$(OBJ) \
+  f16_to_ui32_r_minMag$(OBJ) \
+  f16_to_ui64_r_minMag$(OBJ) \
+  f16_to_i32_r_minMag$(OBJ) \
+  f16_to_i64_r_minMag$(OBJ) \
+  f16_to_f32$(OBJ) \
+  f16_to_f64$(OBJ) \
+  f16_to_extF80$(OBJ) \
+  f16_to_extF80M$(OBJ) \
+  f16_to_f128$(OBJ) \
+  f16_to_f128M$(OBJ) \
+  f16_roundToInt$(OBJ) \
+  f16_add$(OBJ) \
+  f16_sub$(OBJ) \
+  f16_mul$(OBJ) \
+  f16_mulAdd$(OBJ) \
+  f16_div$(OBJ) \
+  f16_rem$(OBJ) \
+  f16_sqrt$(OBJ) \
+  f16_eq$(OBJ) \
+  f16_le$(OBJ) \
+  f16_lt$(OBJ) \
+  f16_eq_signaling$(OBJ) \
+  f16_le_quiet$(OBJ) \
+  f16_lt_quiet$(OBJ) \
+  f16_isSignalingNaN$(OBJ) \
+  f32_to_ui32$(OBJ) \
+  f32_to_ui64$(OBJ) \
+  f32_to_i32$(OBJ) \
+  f32_to_i64$(OBJ) \
+  f32_to_ui32_r_minMag$(OBJ) \
+  f32_to_ui64_r_minMag$(OBJ) \
+  f32_to_i32_r_minMag$(OBJ) \
+  f32_to_i64_r_minMag$(OBJ) \
+  f32_to_bf16$(OBJ) \
+  f32_to_f16$(OBJ) \
+  f32_to_f64$(OBJ) \
+  f32_to_extF80$(OBJ) \
+  f32_to_extF80M$(OBJ) \
+  f32_to_f128$(OBJ) \
+  f32_to_f128M$(OBJ) \
+  f32_roundToInt$(OBJ) \
+  f32_add$(OBJ) \
+  f32_sub$(OBJ) \
+  f32_mul$(OBJ) \
+  f32_mulAdd$(OBJ) \
+  f32_div$(OBJ) \
+  f32_rem$(OBJ) \
+  f32_sqrt$(OBJ) \
+  f32_eq$(OBJ) \
+  f32_le$(OBJ) \
+  f32_lt$(OBJ) \
+  f32_eq_signaling$(OBJ) \
+  f32_le_quiet$(OBJ) \
+  f32_lt_quiet$(OBJ) \
+  f32_isSignalingNaN$(OBJ) \
+  f64_to_ui32$(OBJ) \
+  f64_to_ui64$(OBJ) \
+  f64_to_i32$(OBJ) \
+  f64_to_i64$(OBJ) \
+  f64_to_ui32_r_minMag$(OBJ) \
+  f64_to_ui64_r_minMag$(OBJ) \
+  f64_to_i32_r_minMag$(OBJ) \
+  f64_to_i64_r_minMag$(OBJ) \
+  f64_to_f16$(OBJ) \
+  f64_to_f32$(OBJ) \
+  f64_to_extF80$(OBJ) \
+  f64_to_extF80M$(OBJ) \
+  f64_to_f128$(OBJ) \
+  f64_to_f128M$(OBJ) \
+  f64_roundToInt$(OBJ) \
+  f64_add$(OBJ) \
+  f64_sub$(OBJ) \
+  f64_mul$(OBJ) \
+  f64_mulAdd$(OBJ) \
+  f64_div$(OBJ) \
+  f64_rem$(OBJ) \
+  f64_sqrt$(OBJ) \
+  f64_eq$(OBJ) \
+  f64_le$(OBJ) \
+  f64_lt$(OBJ) \
+  f64_eq_signaling$(OBJ) \
+  f64_le_quiet$(OBJ) \
+  f64_lt_quiet$(OBJ) \
+  f64_isSignalingNaN$(OBJ) \
+  extF80_to_ui32$(OBJ) \
+  extF80_to_ui64$(OBJ) \
+  extF80_to_i32$(OBJ) \
+  extF80_to_i64$(OBJ) \
+  extF80_to_ui32_r_minMag$(OBJ) \
+  extF80_to_ui64_r_minMag$(OBJ) \
+  extF80_to_i32_r_minMag$(OBJ) \
+  extF80_to_i64_r_minMag$(OBJ) \
+  extF80_to_f16$(OBJ) \
+  extF80_to_f32$(OBJ) \
+  extF80_to_f64$(OBJ) \
+  extF80_to_f128$(OBJ) \
+  extF80_roundToInt$(OBJ) \
+  extF80_add$(OBJ) \
+  extF80_sub$(OBJ) \
+  extF80_mul$(OBJ) \
+  extF80_div$(OBJ) \
+  extF80_rem$(OBJ) \
+  extF80_sqrt$(OBJ) \
+  extF80_eq$(OBJ) \
+  extF80_le$(OBJ) \
+  extF80_lt$(OBJ) \
+  extF80_eq_signaling$(OBJ) \
+  extF80_le_quiet$(OBJ) \
+  extF80_lt_quiet$(OBJ) \
+  extF80_isSignalingNaN$(OBJ) \
+  extF80M_to_ui32$(OBJ) \
+  extF80M_to_ui64$(OBJ) \
+  extF80M_to_i32$(OBJ) \
+  extF80M_to_i64$(OBJ) \
+  extF80M_to_ui32_r_minMag$(OBJ) \
+  extF80M_to_ui64_r_minMag$(OBJ) \
+  extF80M_to_i32_r_minMag$(OBJ) \
+  extF80M_to_i64_r_minMag$(OBJ) \
+  extF80M_to_f16$(OBJ) \
+  extF80M_to_f32$(OBJ) \
+  extF80M_to_f64$(OBJ) \
+  extF80M_to_f128M$(OBJ) \
+  extF80M_roundToInt$(OBJ) \
+  extF80M_add$(OBJ) \
+  extF80M_sub$(OBJ) \
+  extF80M_mul$(OBJ) \
+  extF80M_div$(OBJ) \
+  extF80M_rem$(OBJ) \
+  extF80M_sqrt$(OBJ) \
+  extF80M_eq$(OBJ) \
+  extF80M_le$(OBJ) \
+  extF80M_lt$(OBJ) \
+  extF80M_eq_signaling$(OBJ) \
+  extF80M_le_quiet$(OBJ) \
+  extF80M_lt_quiet$(OBJ) \
+  f128_to_ui32$(OBJ) \
+  f128_to_ui64$(OBJ) \
+  f128_to_i32$(OBJ) \
+  f128_to_i64$(OBJ) \
+  f128_to_ui32_r_minMag$(OBJ) \
+  f128_to_ui64_r_minMag$(OBJ) \
+  f128_to_i32_r_minMag$(OBJ) \
+  f128_to_i64_r_minMag$(OBJ) \
+  f128_to_f16$(OBJ) \
+  f128_to_f32$(OBJ) \
+  f128_to_extF80$(OBJ) \
+  f128_to_f64$(OBJ) \
+  f128_roundToInt$(OBJ) \
+  f128_add$(OBJ) \
+  f128_sub$(OBJ) \
+  f128_mul$(OBJ) \
+  f128_mulAdd$(OBJ) \
+  f128_div$(OBJ) \
+  f128_rem$(OBJ) \
+  f128_sqrt$(OBJ) \
+  f128_eq$(OBJ) \
+  f128_le$(OBJ) \
+  f128_lt$(OBJ) \
+  f128_eq_signaling$(OBJ) \
+  f128_le_quiet$(OBJ) \
+  f128_lt_quiet$(OBJ) \
+  f128_isSignalingNaN$(OBJ) \
+  f128M_to_ui32$(OBJ) \
+  f128M_to_ui64$(OBJ) \
+  f128M_to_i32$(OBJ) \
+  f128M_to_i64$(OBJ) \
+  f128M_to_ui32_r_minMag$(OBJ) \
+  f128M_to_ui64_r_minMag$(OBJ) \
+  f128M_to_i32_r_minMag$(OBJ) \
+  f128M_to_i64_r_minMag$(OBJ) \
+  f128M_to_f16$(OBJ) \
+  f128M_to_f32$(OBJ) \
+  f128M_to_extF80M$(OBJ) \
+  f128M_to_f64$(OBJ) \
+  f128M_roundToInt$(OBJ) \
+  f128M_add$(OBJ) \
+  f128M_sub$(OBJ) \
+  f128M_mul$(OBJ) \
+  f128M_mulAdd$(OBJ) \
+  f128M_div$(OBJ) \
+  f128M_rem$(OBJ) \
+  f128M_sqrt$(OBJ) \
+  f128M_eq$(OBJ) \
+  f128M_le$(OBJ) \
+  f128M_lt$(OBJ) \
+  f128M_eq_signaling$(OBJ) \
+  f128M_le_quiet$(OBJ) \
+  f128M_lt_quiet$(OBJ) \
+
+OBJS_ALL = $(OBJS_PRIMITIVES) $(OBJS_SPECIALIZE) $(OBJS_OTHERS)
+
+$(OBJS_ALL): \
+  $(OTHER_HEADERS) platform.h $(SOURCE_DIR)/include/primitiveTypes.h \
+  $(SOURCE_DIR)/include/primitives.h
+$(OBJS_SPECIALIZE) $(OBJS_OTHERS): \
+  $(SOURCE_DIR)/include/softfloat_types.h $(SOURCE_DIR)/include/internals.h \
+  $(SOURCE_DIR)/$(SPECIALIZE_TYPE)/specialize.h \
+  $(SOURCE_DIR)/include/softfloat.h
+
+$(OBJS_PRIMITIVES) $(OBJS_OTHERS): %$(OBJ): $(SOURCE_DIR)/%.c
+	$(COMPILE_C) $(SOURCE_DIR)/$*.c
+
+$(OBJS_SPECIALIZE): %$(OBJ): $(SOURCE_DIR)/$(SPECIALIZE_TYPE)/%.c
+	$(COMPILE_C) $(SOURCE_DIR)/$(SPECIALIZE_TYPE)/$*.c
+
+softfloat$(LIB): $(OBJS_ALL)
+	$(DELETE) $@
+	$(MAKELIB) $^
+
+.PHONY: clean
+clean:
+	$(DELETE) $(OBJS_ALL) softfloat$(LIB)
+

softfloat/build/Linux-RISCV64-GCC/platform.h

@@ -0,0 +1,54 @@
+
+/*============================================================================
+
+This C header file is part of the SoftFloat IEEE Floating-Point Arithmetic
+Package, Release 3e, by John R. Hauser.
+
+Copyright 2011, 2012, 2013, 2014, 2015, 2016, 2017 The Regents of the
+University of California.  All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+
+ 1. Redistributions of source code must retain the above copyright notice,
+    this list of conditions, and the following disclaimer.
+
+ 2. Redistributions in binary form must reproduce the above copyright notice,
+    this list of conditions, and the following disclaimer in the documentation
+    and/or other materials provided with the distribution.
+
+ 3. Neither the name of the University nor the names of its contributors may
+    be used to endorse or promote products derived from this software without
+    specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS "AS IS", AND ANY
+EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, ARE
+DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY
+DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
+ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+=============================================================================*/
+
+/*----------------------------------------------------------------------------
+*----------------------------------------------------------------------------*/
+#define LITTLEENDIAN 1
+
+/*----------------------------------------------------------------------------
+*----------------------------------------------------------------------------*/
+#ifdef __GNUC_STDC_INLINE__
+#define INLINE inline
+#else
+#define INLINE extern inline
+#endif
+
+/*----------------------------------------------------------------------------
+*----------------------------------------------------------------------------*/
+#define SOFTFLOAT_BUILTIN_CLZ 1
+#define SOFTFLOAT_INTRINSIC_INT128 1
+#include "opts-GCC.h"
+

softfloat/build/Linux-x86_64-GCC/Makefile

@@ -94,6 +94,8 @@ OBJS_SPECIALIZE = \
   s_f16UIToCommonNaN$(OBJ) \
   s_commonNaNToF16UI$(OBJ) \
   s_propagateNaNF16UI$(OBJ) \
+  s_bf16UIToCommonNaN$(OBJ) \
+  s_commonNaNToBF16UI$(OBJ) \
   s_f32UIToCommonNaN$(OBJ) \
   s_commonNaNToF32UI$(OBJ) \
   s_propagateNaNF32UI$(OBJ) \
@@ -114,6 +116,8 @@ OBJS_OTHERS = \
   s_roundToUI64$(OBJ) \
   s_roundToI32$(OBJ) \
   s_roundToI64$(OBJ) \
+  s_normSubnormalBF16Sig$(OBJ) \
+  s_roundPackToBF16$(OBJ) \
   s_normSubnormalF16Sig$(OBJ) \
   s_roundPackToF16$(OBJ) \
   s_normRoundPackToF16$(OBJ) \
@@ -172,6 +176,8 @@ OBJS_OTHERS = \
   i64_to_extF80M$(OBJ) \
   i64_to_f128$(OBJ) \
   i64_to_f128M$(OBJ) \
+  bf16_isSignalingNaN$(OBJ) \
+  bf16_to_f32$(OBJ) \
   f16_to_ui32$(OBJ) \
   f16_to_ui64$(OBJ) \
   f16_to_i32$(OBJ) \
@@ -209,6 +215,7 @@ OBJS_OTHERS = \
   f32_to_ui64_r_minMag$(OBJ) \
   f32_to_i32_r_minMag$(OBJ) \
   f32_to_i64_r_minMag$(OBJ) \
+  f32_to_bf16$(OBJ) \
   f32_to_f16$(OBJ) \
   f32_to_f64$(OBJ) \
   f32_to_extF80$(OBJ) \

softfloat/build/Linux-x86_64-GCC/platform.h

@@ -35,11 +35,11 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 =============================================================================*/
 
 /*----------------------------------------------------------------------------
-*----------------------------------------------------------------------------*/
+ *----------------------------------------------------------------------------*/
 #define LITTLEENDIAN 1
 
 /*----------------------------------------------------------------------------
-*----------------------------------------------------------------------------*/
+ *----------------------------------------------------------------------------*/
 #ifdef __GNUC_STDC_INLINE__
 //#define INLINE inline
 #define INLINE static
@@ -48,10 +48,9 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #endif
 
 /*----------------------------------------------------------------------------
-*----------------------------------------------------------------------------*/
+ *----------------------------------------------------------------------------*/
 #ifdef __GNUC__
 #define SOFTFLOAT_BUILTIN_CLZ 1
 #define SOFTFLOAT_INTRINSIC_INT128 1
 #endif
 #include "opts-GCC.h"
-

softfloat/build/template-FAST_INT64/Makefile

@@ -115,6 +115,8 @@ OBJS_OTHERS = \
   s_roundToUI64$(OBJ) \
   s_roundToI32$(OBJ) \
   s_roundToI64$(OBJ) \
+  s_normSubnormalBF16Sig$(OBJ) \
+  s_roundPackToBF16$(OBJ) \
   s_normSubnormalF16Sig$(OBJ) \
   s_roundPackToF16$(OBJ) \
   s_normRoundPackToF16$(OBJ) \
@@ -173,6 +175,8 @@ OBJS_OTHERS = \
   i64_to_extF80M$(OBJ) \
   i64_to_f128$(OBJ) \
   i64_to_f128M$(OBJ) \
+  bf16_isSignalingNaN$(OBJ) \
+  bf16_to_f32$(OBJ) \
   f16_to_ui32$(OBJ) \
   f16_to_ui64$(OBJ) \
   f16_to_i32$(OBJ) \
@@ -210,6 +214,7 @@ OBJS_OTHERS = \
   f32_to_ui64_r_minMag$(OBJ) \
   f32_to_i32_r_minMag$(OBJ) \
   f32_to_i64_r_minMag$(OBJ) \
+  f32_to_bf16$(OBJ) \
   f32_to_f16$(OBJ) \
   f32_to_f64$(OBJ) \
   f32_to_extF80$(OBJ) \

softfloat/doc/SoftFloat.html

@@ -508,7 +508,7 @@ significant extra cost.
 On computers where the word size is <NOBR>64 bits</NOBR> or larger, both
 function versions (<CODE>f128M_add</CODE> and <CODE>f128_add</CODE>) are
 provided, because the cost of passing by value is then more reasonable.
-Applications that must be portable accross both classes of computers must use
+Applications that must be portable across both classes of computers must use
 the pointer-based functions, as these are always implemented.
 However, if it is known that SoftFloat includes the by-value functions for all
 platforms of interest, programmers can use whichever version they prefer.

softfloat/source/8086-SSE/s_bf16UIToCommonNaN.c

@@ -0,0 +1,59 @@
+
+/*============================================================================
+
+This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
+Package, Release 3e, by John R. Hauser.
+
+Copyright 2011, 2012, 2013, 2014 The Regents of the University of California.
+All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+
+ 1. Redistributions of source code must retain the above copyright notice,
+    this list of conditions, and the following disclaimer.
+
+ 2. Redistributions in binary form must reproduce the above copyright notice,
+    this list of conditions, and the following disclaimer in the documentation
+    and/or other materials provided with the distribution.
+
+ 3. Neither the name of the University nor the names of its contributors may
+    be used to endorse or promote products derived from this software without
+    specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS "AS IS", AND ANY
+EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, ARE
+DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY
+DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
+ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+=============================================================================*/
+
+#include <stdint.h>
+#include "platform.h"
+#include "specialize.h"
+#include "softfloat.h"
+
+/*----------------------------------------------------------------------------
+| Assuming `uiA' has the bit pattern of a BF16 NaN, converts
+| this NaN to the common NaN form, and stores the resulting common NaN at the
+| location pointed to by `zPtr'.  If the NaN is a signaling NaN, the invalid
+| exception is raised.
+*----------------------------------------------------------------------------*/
+void softfloat_bf16UIToCommonNaN( uint_fast16_t uiA, struct commonNaN *zPtr )
+{
+
+    if ( softfloat_isSigNaNBF16UI( uiA ) ) {
+        softfloat_raiseFlags( softfloat_flag_invalid );
+    }
+    zPtr->sign = uiA>>15;
+    zPtr->v64  = (uint_fast64_t) uiA<<56;
+    zPtr->v0   = 0;
+
+}
+

softfloat/source/8086-SSE/s_commonNaNToBF16UI.c

@@ -0,0 +1,51 @@
+
+/*============================================================================
+
+This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
+Package, Release 3e, by John R. Hauser.
+
+Copyright 2011, 2012, 2013, 2014 The Regents of the University of California.
+All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+
+ 1. Redistributions of source code must retain the above copyright notice,
+    this list of conditions, and the following disclaimer.
+
+ 2. Redistributions in binary form must reproduce the above copyright notice,
+    this list of conditions, and the following disclaimer in the documentation
+    and/or other materials provided with the distribution.
+
+ 3. Neither the name of the University nor the names of its contributors may
+    be used to endorse or promote products derived from this software without
+    specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS "AS IS", AND ANY
+EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, ARE
+DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY
+DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
+ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+=============================================================================*/
+
+#include <stdint.h>
+#include "platform.h"
+#include "specialize.h"
+
+/*----------------------------------------------------------------------------
+| Converts the common NaN pointed to by `aPtr' into a BF16 NaN, and 
+| returns the bit pattern of this value as an unsigned integer.
+*----------------------------------------------------------------------------*/
+uint_fast16_t softfloat_commonNaNToBF16UI( const struct commonNaN *aPtr )
+{
+
+    return (uint_fast16_t) aPtr->sign<<15 | 0x7FC0 | aPtr->v64>>56;
+
+}
+

softfloat/source/8086-SSE/specialize.h

@@ -37,10 +37,10 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #ifndef specialize_h
 #define specialize_h 1
 
-#include <stdbool.h>
-#include <stdint.h>
 #include "primitiveTypes.h"
 #include "softfloat.h"
+#include <stdbool.h>
+#include <stdint.h>
 
 /*----------------------------------------------------------------------------
 | Default value for 'softfloat_detectTininess'.
@@ -62,12 +62,12 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 | The values to return on conversions to 64-bit integer formats that raise an
 | invalid exception.
 *----------------------------------------------------------------------------*/
-#define ui64_fromPosOverflow UINT64_C( 0xFFFFFFFFFFFFFFFF )
+#define ui64_fromPosOverflow UINT64_C(0xFFFFFFFFFFFFFFFF)
-#define ui64_fromNegOverflow UINT64_C( 0xFFFFFFFFFFFFFFFF )
+#define ui64_fromNegOverflow UINT64_C(0xFFFFFFFFFFFFFFFF)
-#define ui64_fromNaN         UINT64_C( 0xFFFFFFFFFFFFFFFF )
+#define ui64_fromNaN UINT64_C(0xFFFFFFFFFFFFFFFF)
-#define i64_fromPosOverflow  (-INT64_C( 0x7FFFFFFFFFFFFFFF ) - 1)
+#define i64_fromPosOverflow (-INT64_C(0x7FFFFFFFFFFFFFFF) - 1)
-#define i64_fromNegOverflow  (-INT64_C( 0x7FFFFFFFFFFFFFFF ) - 1)
+#define i64_fromNegOverflow (-INT64_C(0x7FFFFFFFFFFFFFFF) - 1)
-#define i64_fromNaN          (-INT64_C( 0x7FFFFFFFFFFFFFFF ) - 1)
+#define i64_fromNaN (-INT64_C(0x7FFFFFFFFFFFFFFF) - 1)
 
 /*----------------------------------------------------------------------------
 | "Common NaN" structure, used to transfer NaN representations from one format
@@ -92,7 +92,7 @@ struct commonNaN {
 | 16-bit floating-point signaling NaN.
 | Note:  This macro evaluates its argument more than once.
 *----------------------------------------------------------------------------*/
-#define softfloat_isSigNaNF16UI( uiA ) ((((uiA) & 0x7E00) == 0x7C00) && ((uiA) & 0x01FF))
+#define softfloat_isSigNaNF16UI(uiA) ((((uiA)&0x7E00) == 0x7C00) && ((uiA)&0x01FF))
 
 /*----------------------------------------------------------------------------
 | Assuming 'uiA' has the bit pattern of a 16-bit floating-point NaN, converts
@@ -100,13 +100,13 @@ struct commonNaN {
 | location pointed to by 'zPtr'.  If the NaN is a signaling NaN, the invalid
 | exception is raised.
 *----------------------------------------------------------------------------*/
-void softfloat_f16UIToCommonNaN( uint_fast16_t uiA, struct commonNaN *zPtr );
+void softfloat_f16UIToCommonNaN(uint_fast16_t uiA, struct commonNaN* zPtr);
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into a 16-bit floating-point
 | NaN, and returns the bit pattern of this value as an unsigned integer.
 *----------------------------------------------------------------------------*/
-uint_fast16_t softfloat_commonNaNToF16UI( const struct commonNaN *aPtr );
+uint_fast16_t softfloat_commonNaNToF16UI(const struct commonNaN* aPtr);
 
 /*----------------------------------------------------------------------------
 | Interpreting 'uiA' and 'uiB' as the bit patterns of two 16-bit floating-
@@ -114,8 +114,28 @@ uint_fast16_t softfloat_commonNaNToF16UI( const struct commonNaN *aPtr );
 | the combined NaN result.  If either 'uiA' or 'uiB' has the pattern of a
 | signaling NaN, the invalid exception is raised.
 *----------------------------------------------------------------------------*/
-uint_fast16_t
+uint_fast16_t softfloat_propagateNaNF16UI(uint_fast16_t uiA, uint_fast16_t uiB);
- softfloat_propagateNaNF16UI( uint_fast16_t uiA, uint_fast16_t uiB );
+
+/*----------------------------------------------------------------------------
+| Returns true when 16-bit unsigned integer 'uiA' has the bit pattern of a
+| 16-bit brain floating-point (BF16) signaling NaN.
+| Note:  This macro evaluates its argument more than once.
+*----------------------------------------------------------------------------*/
+#define softfloat_isSigNaNBF16UI(uiA) ((((uiA)&0x7FC0) == 0x7F80) && ((uiA)&0x003F))
+
+/*----------------------------------------------------------------------------
+| Assuming 'uiA' has the bit pattern of a 16-bit BF16 floating-point NaN, converts
+| this NaN to the common NaN form, and stores the resulting common NaN at the
+| location pointed to by 'zPtr'.  If the NaN is a signaling NaN, the invalid
+| exception is raised.
+*----------------------------------------------------------------------------*/
+void softfloat_bf16UIToCommonNaN(uint_fast16_t uiA, struct commonNaN* zPtr);
+
+/*----------------------------------------------------------------------------
+| Converts the common NaN pointed to by 'aPtr' into a 16-bit floating-point
+| NaN, and returns the bit pattern of this value as an unsigned integer.
+*----------------------------------------------------------------------------*/
+uint_fast16_t softfloat_commonNaNToBF16UI(const struct commonNaN* aPtr);
 
 /*----------------------------------------------------------------------------
 | The bit pattern for a default generated 32-bit floating-point NaN.
@@ -127,7 +147,7 @@ uint_fast16_t
 | 32-bit floating-point signaling NaN.
 | Note:  This macro evaluates its argument more than once.
 *----------------------------------------------------------------------------*/
-#define softfloat_isSigNaNF32UI( uiA ) ((((uiA) & 0x7FC00000) == 0x7F800000) && ((uiA) & 0x003FFFFF))
+#define softfloat_isSigNaNF32UI(uiA) ((((uiA)&0x7FC00000) == 0x7F800000) && ((uiA)&0x003FFFFF))
 
 /*----------------------------------------------------------------------------
 | Assuming 'uiA' has the bit pattern of a 32-bit floating-point NaN, converts
@@ -135,13 +155,13 @@ uint_fast16_t
 | location pointed to by 'zPtr'.  If the NaN is a signaling NaN, the invalid
 | exception is raised.
 *----------------------------------------------------------------------------*/
-void softfloat_f32UIToCommonNaN( uint_fast32_t uiA, struct commonNaN *zPtr );
+void softfloat_f32UIToCommonNaN(uint_fast32_t uiA, struct commonNaN* zPtr);
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into a 32-bit floating-point
 | NaN, and returns the bit pattern of this value as an unsigned integer.
 *----------------------------------------------------------------------------*/
-uint_fast32_t softfloat_commonNaNToF32UI( const struct commonNaN *aPtr );
+uint_fast32_t softfloat_commonNaNToF32UI(const struct commonNaN* aPtr);
 
 /*----------------------------------------------------------------------------
 | Interpreting 'uiA' and 'uiB' as the bit patterns of two 32-bit floating-
@@ -149,20 +169,20 @@ uint_fast32_t softfloat_commonNaNToF32UI( const struct commonNaN *aPtr );
 | the combined NaN result.  If either 'uiA' or 'uiB' has the pattern of a
 | signaling NaN, the invalid exception is raised.
 *----------------------------------------------------------------------------*/
-uint_fast32_t
+uint_fast32_t softfloat_propagateNaNF32UI(uint_fast32_t uiA, uint_fast32_t uiB);
- softfloat_propagateNaNF32UI( uint_fast32_t uiA, uint_fast32_t uiB );
 
 /*----------------------------------------------------------------------------
 | The bit pattern for a default generated 64-bit floating-point NaN.
 *----------------------------------------------------------------------------*/
-#define defaultNaNF64UI UINT64_C( 0xFFF8000000000000 )
+#define defaultNaNF64UI UINT64_C(0xFFF8000000000000)
 
 /*----------------------------------------------------------------------------
 | Returns true when 64-bit unsigned integer 'uiA' has the bit pattern of a
 | 64-bit floating-point signaling NaN.
 | Note:  This macro evaluates its argument more than once.
 *----------------------------------------------------------------------------*/
-#define softfloat_isSigNaNF64UI( uiA ) ((((uiA) & UINT64_C( 0x7FF8000000000000 )) == UINT64_C( 0x7FF0000000000000 )) && ((uiA) & UINT64_C( 0x0007FFFFFFFFFFFF )))
+#define softfloat_isSigNaNF64UI(uiA)                                                                                                       \
+    ((((uiA)&UINT64_C(0x7FF8000000000000)) == UINT64_C(0x7FF0000000000000)) && ((uiA)&UINT64_C(0x0007FFFFFFFFFFFF)))
 
 /*----------------------------------------------------------------------------
 | Assuming 'uiA' has the bit pattern of a 64-bit floating-point NaN, converts
@@ -170,13 +190,13 @@ uint_fast32_t
 | location pointed to by 'zPtr'.  If the NaN is a signaling NaN, the invalid
 | exception is raised.
 *----------------------------------------------------------------------------*/
-void softfloat_f64UIToCommonNaN( uint_fast64_t uiA, struct commonNaN *zPtr );
+void softfloat_f64UIToCommonNaN(uint_fast64_t uiA, struct commonNaN* zPtr);
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into a 64-bit floating-point
 | NaN, and returns the bit pattern of this value as an unsigned integer.
 *----------------------------------------------------------------------------*/
-uint_fast64_t softfloat_commonNaNToF64UI( const struct commonNaN *aPtr );
+uint_fast64_t softfloat_commonNaNToF64UI(const struct commonNaN* aPtr);
 
 /*----------------------------------------------------------------------------
 | Interpreting 'uiA' and 'uiB' as the bit patterns of two 64-bit floating-
@@ -184,14 +204,13 @@ uint_fast64_t softfloat_commonNaNToF64UI( const struct commonNaN *aPtr );
 | the combined NaN result.  If either 'uiA' or 'uiB' has the pattern of a
 | signaling NaN, the invalid exception is raised.
 *----------------------------------------------------------------------------*/
-uint_fast64_t
+uint_fast64_t softfloat_propagateNaNF64UI(uint_fast64_t uiA, uint_fast64_t uiB);
- softfloat_propagateNaNF64UI( uint_fast64_t uiA, uint_fast64_t uiB );
 
 /*----------------------------------------------------------------------------
 | The bit pattern for a default generated 80-bit extended floating-point NaN.
 *----------------------------------------------------------------------------*/
 #define defaultNaNExtF80UI64 0xFFFF
-#define defaultNaNExtF80UI0  UINT64_C( 0xC000000000000000 )
+#define defaultNaNExtF80UI0 UINT64_C(0xC000000000000000)
 
 /*----------------------------------------------------------------------------
 | Returns true when the 80-bit unsigned integer formed from concatenating
@@ -199,7 +218,8 @@ uint_fast64_t
 | floating-point signaling NaN.
 | Note:  This macro evaluates its arguments more than once.
 *----------------------------------------------------------------------------*/
-#define softfloat_isSigNaNExtF80UI( uiA64, uiA0 ) ((((uiA64) & 0x7FFF) == 0x7FFF) && ! ((uiA0) & UINT64_C( 0x4000000000000000 )) && ((uiA0) & UINT64_C( 0x3FFFFFFFFFFFFFFF )))
+#define softfloat_isSigNaNExtF80UI(uiA64, uiA0)                                                                                            \
+    ((((uiA64)&0x7FFF) == 0x7FFF) && !((uiA0)&UINT64_C(0x4000000000000000)) && ((uiA0)&UINT64_C(0x3FFFFFFFFFFFFFFF)))
 
 #ifdef SOFTFLOAT_FAST_INT64
 
@@ -215,16 +235,14 @@ uint_fast64_t
 | location pointed to by 'zPtr'.  If the NaN is a signaling NaN, the invalid
 | exception is raised.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_extF80UIToCommonNaN(uint_fast16_t uiA64, uint_fast64_t uiA0, struct commonNaN* zPtr);
- softfloat_extF80UIToCommonNaN(
-     uint_fast16_t uiA64, uint_fast64_t uiA0, struct commonNaN *zPtr );
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into an 80-bit extended
 | floating-point NaN, and returns the bit pattern of this value as an unsigned
 | integer.
 *----------------------------------------------------------------------------*/
-struct uint128 softfloat_commonNaNToExtF80UI( const struct commonNaN *aPtr );
+struct uint128 softfloat_commonNaNToExtF80UI(const struct commonNaN* aPtr);
 
 /*----------------------------------------------------------------------------
 | Interpreting the unsigned integer formed from concatenating 'uiA64' and
@@ -235,19 +253,13 @@ struct uint128 softfloat_commonNaNToExtF80UI( const struct commonNaN *aPtr );
 | result.  If either original floating-point value is a signaling NaN, the
 | invalid exception is raised.
 *----------------------------------------------------------------------------*/
-struct uint128
+struct uint128 softfloat_propagateNaNExtF80UI(uint_fast16_t uiA64, uint_fast64_t uiA0, uint_fast16_t uiB64, uint_fast64_t uiB0);
- softfloat_propagateNaNExtF80UI(
-     uint_fast16_t uiA64,
-     uint_fast64_t uiA0,
-     uint_fast16_t uiB64,
-     uint_fast64_t uiB0
- );
 
 /*----------------------------------------------------------------------------
 | The bit pattern for a default generated 128-bit floating-point NaN.
 *----------------------------------------------------------------------------*/
-#define defaultNaNF128UI64 UINT64_C( 0xFFFF800000000000 )
+#define defaultNaNF128UI64 UINT64_C(0xFFFF800000000000)
-#define defaultNaNF128UI0  UINT64_C( 0 )
+#define defaultNaNF128UI0 UINT64_C(0)
 
 /*----------------------------------------------------------------------------
 | Returns true when the 128-bit unsigned integer formed from concatenating
@@ -255,7 +267,8 @@ struct uint128
 | point signaling NaN.
 | Note:  This macro evaluates its arguments more than once.
 *----------------------------------------------------------------------------*/
-#define softfloat_isSigNaNF128UI( uiA64, uiA0 ) ((((uiA64) & UINT64_C( 0x7FFF800000000000 )) == UINT64_C( 0x7FFF000000000000 )) && ((uiA0) || ((uiA64) & UINT64_C( 0x00007FFFFFFFFFFF ))))
+#define softfloat_isSigNaNF128UI(uiA64, uiA0)                                                                                              \
+    ((((uiA64)&UINT64_C(0x7FFF800000000000)) == UINT64_C(0x7FFF000000000000)) && ((uiA0) || ((uiA64)&UINT64_C(0x00007FFFFFFFFFFF))))
 
 /*----------------------------------------------------------------------------
 | Assuming the unsigned integer formed from concatenating 'uiA64' and 'uiA0'
@@ -264,15 +277,13 @@ struct uint128
 | pointed to by 'zPtr'.  If the NaN is a signaling NaN, the invalid exception
 | is raised.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_f128UIToCommonNaN(uint_fast64_t uiA64, uint_fast64_t uiA0, struct commonNaN* zPtr);
- softfloat_f128UIToCommonNaN(
-     uint_fast64_t uiA64, uint_fast64_t uiA0, struct commonNaN *zPtr );
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into a 128-bit floating-point
 | NaN, and returns the bit pattern of this value as an unsigned integer.
 *----------------------------------------------------------------------------*/
-struct uint128 softfloat_commonNaNToF128UI( const struct commonNaN * );
+struct uint128 softfloat_commonNaNToF128UI(const struct commonNaN*);
 
 /*----------------------------------------------------------------------------
 | Interpreting the unsigned integer formed from concatenating 'uiA64' and
@@ -283,13 +294,7 @@ struct uint128 softfloat_commonNaNToF128UI( const struct commonNaN * );
 | If either original floating-point value is a signaling NaN, the invalid
 | exception is raised.
 *----------------------------------------------------------------------------*/
-struct uint128
+struct uint128 softfloat_propagateNaNF128UI(uint_fast64_t uiA64, uint_fast64_t uiA0, uint_fast64_t uiB64, uint_fast64_t uiB0);
- softfloat_propagateNaNF128UI(
-     uint_fast64_t uiA64,
-     uint_fast64_t uiA0,
-     uint_fast64_t uiB64,
-     uint_fast64_t uiB0
- );
 
 #else
 
@@ -304,18 +309,14 @@ struct uint128
 | common NaN at the location pointed to by 'zPtr'.  If the NaN is a signaling
 | NaN, the invalid exception is raised.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_extF80MToCommonNaN(const struct extFloat80M* aSPtr, struct commonNaN* zPtr);
- softfloat_extF80MToCommonNaN(
-     const struct extFloat80M *aSPtr, struct commonNaN *zPtr );
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into an 80-bit extended
 | floating-point NaN, and stores this NaN at the location pointed to by
 | 'zSPtr'.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_commonNaNToExtF80M(const struct commonNaN* aPtr, struct extFloat80M* zSPtr);
- softfloat_commonNaNToExtF80M(
-     const struct commonNaN *aPtr, struct extFloat80M *zSPtr );
 
 /*----------------------------------------------------------------------------
 | Assuming at least one of the two 80-bit extended floating-point values
@@ -323,12 +324,7 @@ void
 | at the location pointed to by 'zSPtr'.  If either original floating-point
 | value is a signaling NaN, the invalid exception is raised.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_propagateNaNExtF80M(const struct extFloat80M* aSPtr, const struct extFloat80M* bSPtr, struct extFloat80M* zSPtr);
- softfloat_propagateNaNExtF80M(
-     const struct extFloat80M *aSPtr,
-     const struct extFloat80M *bSPtr,
-     struct extFloat80M *zSPtr
- );
 
 /*----------------------------------------------------------------------------
 | The bit pattern for a default generated 128-bit floating-point NaN.
@@ -346,8 +342,7 @@ void
 | four 32-bit elements that concatenate in the platform's normal endian order
 | to form a 128-bit floating-point value.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_f128MToCommonNaN(const uint32_t* aWPtr, struct commonNaN* zPtr);
- softfloat_f128MToCommonNaN( const uint32_t *aWPtr, struct commonNaN *zPtr );
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into a 128-bit floating-point
@@ -355,8 +350,7 @@ void
 | 'zWPtr' points to an array of four 32-bit elements that concatenate in the
 | platform's normal endian order to form a 128-bit floating-point value.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_commonNaNToF128M(const struct commonNaN* aPtr, uint32_t* zWPtr);
- softfloat_commonNaNToF128M( const struct commonNaN *aPtr, uint32_t *zWPtr );
 
 /*----------------------------------------------------------------------------
 | Assuming at least one of the two 128-bit floating-point values pointed to by
@@ -366,11 +360,8 @@ void
 | and 'zWPtr' points to an array of four 32-bit elements that concatenate in
 | the platform's normal endian order to form a 128-bit floating-point value.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_propagateNaNF128M(const uint32_t* aWPtr, const uint32_t* bWPtr, uint32_t* zWPtr);
- softfloat_propagateNaNF128M(
-     const uint32_t *aWPtr, const uint32_t *bWPtr, uint32_t *zWPtr );
 
 #endif
 
 #endif
-

softfloat/source/8086/specialize.h

@@ -37,10 +37,10 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #ifndef specialize_h
 #define specialize_h 1
 
-#include <stdbool.h>
-#include <stdint.h>
 #include "primitiveTypes.h"
 #include "softfloat.h"
+#include <stdbool.h>
+#include <stdint.h>
 
 /*----------------------------------------------------------------------------
 | Default value for 'softfloat_detectTininess'.
@@ -62,12 +62,12 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 | The values to return on conversions to 64-bit integer formats that raise an
 | invalid exception.
 *----------------------------------------------------------------------------*/
-#define ui64_fromPosOverflow UINT64_C( 0xFFFFFFFFFFFFFFFF )
+#define ui64_fromPosOverflow UINT64_C(0xFFFFFFFFFFFFFFFF)
-#define ui64_fromNegOverflow UINT64_C( 0xFFFFFFFFFFFFFFFF )
+#define ui64_fromNegOverflow UINT64_C(0xFFFFFFFFFFFFFFFF)
-#define ui64_fromNaN         UINT64_C( 0xFFFFFFFFFFFFFFFF )
+#define ui64_fromNaN UINT64_C(0xFFFFFFFFFFFFFFFF)
-#define i64_fromPosOverflow  (-INT64_C( 0x7FFFFFFFFFFFFFFF ) - 1)
+#define i64_fromPosOverflow (-INT64_C(0x7FFFFFFFFFFFFFFF) - 1)
-#define i64_fromNegOverflow  (-INT64_C( 0x7FFFFFFFFFFFFFFF ) - 1)
+#define i64_fromNegOverflow (-INT64_C(0x7FFFFFFFFFFFFFFF) - 1)
-#define i64_fromNaN          (-INT64_C( 0x7FFFFFFFFFFFFFFF ) - 1)
+#define i64_fromNaN (-INT64_C(0x7FFFFFFFFFFFFFFF) - 1)
 
 /*----------------------------------------------------------------------------
 | "Common NaN" structure, used to transfer NaN representations from one format
@@ -92,7 +92,7 @@ struct commonNaN {
 | 16-bit floating-point signaling NaN.
 | Note:  This macro evaluates its argument more than once.
 *----------------------------------------------------------------------------*/
-#define softfloat_isSigNaNF16UI( uiA ) ((((uiA) & 0x7E00) == 0x7C00) && ((uiA) & 0x01FF))
+#define softfloat_isSigNaNF16UI(uiA) ((((uiA)&0x7E00) == 0x7C00) && ((uiA)&0x01FF))
 
 /*----------------------------------------------------------------------------
 | Assuming 'uiA' has the bit pattern of a 16-bit floating-point NaN, converts
@@ -100,13 +100,13 @@ struct commonNaN {
 | location pointed to by 'zPtr'.  If the NaN is a signaling NaN, the invalid
 | exception is raised.
 *----------------------------------------------------------------------------*/
-void softfloat_f16UIToCommonNaN( uint_fast16_t uiA, struct commonNaN *zPtr );
+void softfloat_f16UIToCommonNaN(uint_fast16_t uiA, struct commonNaN* zPtr);
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into a 16-bit floating-point
 | NaN, and returns the bit pattern of this value as an unsigned integer.
 *----------------------------------------------------------------------------*/
-uint_fast16_t softfloat_commonNaNToF16UI( const struct commonNaN *aPtr );
+uint_fast16_t softfloat_commonNaNToF16UI(const struct commonNaN* aPtr);
 
 /*----------------------------------------------------------------------------
 | Interpreting 'uiA' and 'uiB' as the bit patterns of two 16-bit floating-
@@ -114,8 +114,7 @@ uint_fast16_t softfloat_commonNaNToF16UI( const struct commonNaN *aPtr );
 | the combined NaN result.  If either 'uiA' or 'uiB' has the pattern of a
 | signaling NaN, the invalid exception is raised.
 *----------------------------------------------------------------------------*/
-uint_fast16_t
+uint_fast16_t softfloat_propagateNaNF16UI(uint_fast16_t uiA, uint_fast16_t uiB);
- softfloat_propagateNaNF16UI( uint_fast16_t uiA, uint_fast16_t uiB );
 
 /*----------------------------------------------------------------------------
 | The bit pattern for a default generated 32-bit floating-point NaN.
@@ -127,7 +126,7 @@ uint_fast16_t
 | 32-bit floating-point signaling NaN.
 | Note:  This macro evaluates its argument more than once.
 *----------------------------------------------------------------------------*/
-#define softfloat_isSigNaNF32UI( uiA ) ((((uiA) & 0x7FC00000) == 0x7F800000) && ((uiA) & 0x003FFFFF))
+#define softfloat_isSigNaNF32UI(uiA) ((((uiA)&0x7FC00000) == 0x7F800000) && ((uiA)&0x003FFFFF))
 
 /*----------------------------------------------------------------------------
 | Assuming 'uiA' has the bit pattern of a 32-bit floating-point NaN, converts
@@ -135,13 +134,13 @@ uint_fast16_t
 | location pointed to by 'zPtr'.  If the NaN is a signaling NaN, the invalid
 | exception is raised.
 *----------------------------------------------------------------------------*/
-void softfloat_f32UIToCommonNaN( uint_fast32_t uiA, struct commonNaN *zPtr );
+void softfloat_f32UIToCommonNaN(uint_fast32_t uiA, struct commonNaN* zPtr);
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into a 32-bit floating-point
 | NaN, and returns the bit pattern of this value as an unsigned integer.
 *----------------------------------------------------------------------------*/
-uint_fast32_t softfloat_commonNaNToF32UI( const struct commonNaN *aPtr );
+uint_fast32_t softfloat_commonNaNToF32UI(const struct commonNaN* aPtr);
 
 /*----------------------------------------------------------------------------
 | Interpreting 'uiA' and 'uiB' as the bit patterns of two 32-bit floating-
@@ -149,20 +148,20 @@ uint_fast32_t softfloat_commonNaNToF32UI( const struct commonNaN *aPtr );
 | the combined NaN result.  If either 'uiA' or 'uiB' has the pattern of a
 | signaling NaN, the invalid exception is raised.
 *----------------------------------------------------------------------------*/
-uint_fast32_t
+uint_fast32_t softfloat_propagateNaNF32UI(uint_fast32_t uiA, uint_fast32_t uiB);
- softfloat_propagateNaNF32UI( uint_fast32_t uiA, uint_fast32_t uiB );
 
 /*----------------------------------------------------------------------------
 | The bit pattern for a default generated 64-bit floating-point NaN.
 *----------------------------------------------------------------------------*/
-#define defaultNaNF64UI UINT64_C( 0xFFF8000000000000 )
+#define defaultNaNF64UI UINT64_C(0xFFF8000000000000)
 
 /*----------------------------------------------------------------------------
 | Returns true when 64-bit unsigned integer 'uiA' has the bit pattern of a
 | 64-bit floating-point signaling NaN.
 | Note:  This macro evaluates its argument more than once.
 *----------------------------------------------------------------------------*/
-#define softfloat_isSigNaNF64UI( uiA ) ((((uiA) & UINT64_C( 0x7FF8000000000000 )) == UINT64_C( 0x7FF0000000000000 )) && ((uiA) & UINT64_C( 0x0007FFFFFFFFFFFF )))
+#define softfloat_isSigNaNF64UI(uiA)                                                                                                       \
+    ((((uiA)&UINT64_C(0x7FF8000000000000)) == UINT64_C(0x7FF0000000000000)) && ((uiA)&UINT64_C(0x0007FFFFFFFFFFFF)))
 
 /*----------------------------------------------------------------------------
 | Assuming 'uiA' has the bit pattern of a 64-bit floating-point NaN, converts
@@ -170,13 +169,13 @@ uint_fast32_t
 | location pointed to by 'zPtr'.  If the NaN is a signaling NaN, the invalid
 | exception is raised.
 *----------------------------------------------------------------------------*/
-void softfloat_f64UIToCommonNaN( uint_fast64_t uiA, struct commonNaN *zPtr );
+void softfloat_f64UIToCommonNaN(uint_fast64_t uiA, struct commonNaN* zPtr);
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into a 64-bit floating-point
 | NaN, and returns the bit pattern of this value as an unsigned integer.
 *----------------------------------------------------------------------------*/
-uint_fast64_t softfloat_commonNaNToF64UI( const struct commonNaN *aPtr );
+uint_fast64_t softfloat_commonNaNToF64UI(const struct commonNaN* aPtr);
 
 /*----------------------------------------------------------------------------
 | Interpreting 'uiA' and 'uiB' as the bit patterns of two 64-bit floating-
@@ -184,14 +183,13 @@ uint_fast64_t softfloat_commonNaNToF64UI( const struct commonNaN *aPtr );
 | the combined NaN result.  If either 'uiA' or 'uiB' has the pattern of a
 | signaling NaN, the invalid exception is raised.
 *----------------------------------------------------------------------------*/
-uint_fast64_t
+uint_fast64_t softfloat_propagateNaNF64UI(uint_fast64_t uiA, uint_fast64_t uiB);
- softfloat_propagateNaNF64UI( uint_fast64_t uiA, uint_fast64_t uiB );
 
 /*----------------------------------------------------------------------------
 | The bit pattern for a default generated 80-bit extended floating-point NaN.
 *----------------------------------------------------------------------------*/
 #define defaultNaNExtF80UI64 0xFFFF
-#define defaultNaNExtF80UI0  UINT64_C( 0xC000000000000000 )
+#define defaultNaNExtF80UI0 UINT64_C(0xC000000000000000)
 
 /*----------------------------------------------------------------------------
 | Returns true when the 80-bit unsigned integer formed from concatenating
@@ -199,7 +197,8 @@ uint_fast64_t
 | floating-point signaling NaN.
 | Note:  This macro evaluates its arguments more than once.
 *----------------------------------------------------------------------------*/
-#define softfloat_isSigNaNExtF80UI( uiA64, uiA0 ) ((((uiA64) & 0x7FFF) == 0x7FFF) && ! ((uiA0) & UINT64_C( 0x4000000000000000 )) && ((uiA0) & UINT64_C( 0x3FFFFFFFFFFFFFFF )))
+#define softfloat_isSigNaNExtF80UI(uiA64, uiA0)                                                                                            \
+    ((((uiA64)&0x7FFF) == 0x7FFF) && !((uiA0)&UINT64_C(0x4000000000000000)) && ((uiA0)&UINT64_C(0x3FFFFFFFFFFFFFFF)))
 
 #ifdef SOFTFLOAT_FAST_INT64
 
@@ -215,16 +214,14 @@ uint_fast64_t
 | location pointed to by 'zPtr'.  If the NaN is a signaling NaN, the invalid
 | exception is raised.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_extF80UIToCommonNaN(uint_fast16_t uiA64, uint_fast64_t uiA0, struct commonNaN* zPtr);
- softfloat_extF80UIToCommonNaN(
-     uint_fast16_t uiA64, uint_fast64_t uiA0, struct commonNaN *zPtr );
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into an 80-bit extended
 | floating-point NaN, and returns the bit pattern of this value as an unsigned
 | integer.
 *----------------------------------------------------------------------------*/
-struct uint128 softfloat_commonNaNToExtF80UI( const struct commonNaN *aPtr );
+struct uint128 softfloat_commonNaNToExtF80UI(const struct commonNaN* aPtr);
 
 /*----------------------------------------------------------------------------
 | Interpreting the unsigned integer formed from concatenating 'uiA64' and
@@ -235,19 +232,13 @@ struct uint128 softfloat_commonNaNToExtF80UI( const struct commonNaN *aPtr );
 | result.  If either original floating-point value is a signaling NaN, the
 | invalid exception is raised.
 *----------------------------------------------------------------------------*/
-struct uint128
+struct uint128 softfloat_propagateNaNExtF80UI(uint_fast16_t uiA64, uint_fast64_t uiA0, uint_fast16_t uiB64, uint_fast64_t uiB0);
- softfloat_propagateNaNExtF80UI(
-     uint_fast16_t uiA64,
-     uint_fast64_t uiA0,
-     uint_fast16_t uiB64,
-     uint_fast64_t uiB0
- );
 
 /*----------------------------------------------------------------------------
 | The bit pattern for a default generated 128-bit floating-point NaN.
 *----------------------------------------------------------------------------*/
-#define defaultNaNF128UI64 UINT64_C( 0xFFFF800000000000 )
+#define defaultNaNF128UI64 UINT64_C(0xFFFF800000000000)
-#define defaultNaNF128UI0  UINT64_C( 0 )
+#define defaultNaNF128UI0 UINT64_C(0)
 
 /*----------------------------------------------------------------------------
 | Returns true when the 128-bit unsigned integer formed from concatenating
@@ -255,7 +246,8 @@ struct uint128
 | point signaling NaN.
 | Note:  This macro evaluates its arguments more than once.
 *----------------------------------------------------------------------------*/
-#define softfloat_isSigNaNF128UI( uiA64, uiA0 ) ((((uiA64) & UINT64_C( 0x7FFF800000000000 )) == UINT64_C( 0x7FFF000000000000 )) && ((uiA0) || ((uiA64) & UINT64_C( 0x00007FFFFFFFFFFF ))))
+#define softfloat_isSigNaNF128UI(uiA64, uiA0)                                                                                              \
+    ((((uiA64)&UINT64_C(0x7FFF800000000000)) == UINT64_C(0x7FFF000000000000)) && ((uiA0) || ((uiA64)&UINT64_C(0x00007FFFFFFFFFFF))))
 
 /*----------------------------------------------------------------------------
 | Assuming the unsigned integer formed from concatenating 'uiA64' and 'uiA0'
@@ -264,15 +256,13 @@ struct uint128
 | pointed to by 'zPtr'.  If the NaN is a signaling NaN, the invalid exception
 | is raised.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_f128UIToCommonNaN(uint_fast64_t uiA64, uint_fast64_t uiA0, struct commonNaN* zPtr);
- softfloat_f128UIToCommonNaN(
-     uint_fast64_t uiA64, uint_fast64_t uiA0, struct commonNaN *zPtr );
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into a 128-bit floating-point
 | NaN, and returns the bit pattern of this value as an unsigned integer.
 *----------------------------------------------------------------------------*/
-struct uint128 softfloat_commonNaNToF128UI( const struct commonNaN * );
+struct uint128 softfloat_commonNaNToF128UI(const struct commonNaN*);
 
 /*----------------------------------------------------------------------------
 | Interpreting the unsigned integer formed from concatenating 'uiA64' and
@@ -283,13 +273,7 @@ struct uint128 softfloat_commonNaNToF128UI( const struct commonNaN * );
 | If either original floating-point value is a signaling NaN, the invalid
 | exception is raised.
 *----------------------------------------------------------------------------*/
-struct uint128
+struct uint128 softfloat_propagateNaNF128UI(uint_fast64_t uiA64, uint_fast64_t uiA0, uint_fast64_t uiB64, uint_fast64_t uiB0);
- softfloat_propagateNaNF128UI(
-     uint_fast64_t uiA64,
-     uint_fast64_t uiA0,
-     uint_fast64_t uiB64,
-     uint_fast64_t uiB0
- );
 
 #else
 
@@ -304,18 +288,14 @@ struct uint128
 | common NaN at the location pointed to by 'zPtr'.  If the NaN is a signaling
 | NaN, the invalid exception is raised.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_extF80MToCommonNaN(const struct extFloat80M* aSPtr, struct commonNaN* zPtr);
- softfloat_extF80MToCommonNaN(
-     const struct extFloat80M *aSPtr, struct commonNaN *zPtr );
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into an 80-bit extended
 | floating-point NaN, and stores this NaN at the location pointed to by
 | 'zSPtr'.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_commonNaNToExtF80M(const struct commonNaN* aPtr, struct extFloat80M* zSPtr);
- softfloat_commonNaNToExtF80M(
-     const struct commonNaN *aPtr, struct extFloat80M *zSPtr );
 
 /*----------------------------------------------------------------------------
 | Assuming at least one of the two 80-bit extended floating-point values
@@ -323,12 +303,7 @@ void
 | at the location pointed to by 'zSPtr'.  If either original floating-point
 | value is a signaling NaN, the invalid exception is raised.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_propagateNaNExtF80M(const struct extFloat80M* aSPtr, const struct extFloat80M* bSPtr, struct extFloat80M* zSPtr);
- softfloat_propagateNaNExtF80M(
-     const struct extFloat80M *aSPtr,
-     const struct extFloat80M *bSPtr,
-     struct extFloat80M *zSPtr
- );
 
 /*----------------------------------------------------------------------------
 | The bit pattern for a default generated 128-bit floating-point NaN.
@@ -346,8 +321,7 @@ void
 | four 32-bit elements that concatenate in the platform's normal endian order
 | to form a 128-bit floating-point value.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_f128MToCommonNaN(const uint32_t* aWPtr, struct commonNaN* zPtr);
- softfloat_f128MToCommonNaN( const uint32_t *aWPtr, struct commonNaN *zPtr );
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into a 128-bit floating-point
@@ -355,8 +329,7 @@ void
 | 'zWPtr' points to an array of four 32-bit elements that concatenate in the
 | platform's normal endian order to form a 128-bit floating-point value.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_commonNaNToF128M(const struct commonNaN* aPtr, uint32_t* zWPtr);
- softfloat_commonNaNToF128M( const struct commonNaN *aPtr, uint32_t *zWPtr );
 
 /*----------------------------------------------------------------------------
 | Assuming at least one of the two 128-bit floating-point values pointed to by
@@ -366,11 +339,8 @@ void
 | and 'zWPtr' points to an array of four 32-bit elements that concatenate in
 | the platform's normal endian order to form a 128-bit floating-point value.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_propagateNaNF128M(const uint32_t* aWPtr, const uint32_t* bWPtr, uint32_t* zWPtr);
- softfloat_propagateNaNF128M(
-     const uint32_t *aWPtr, const uint32_t *bWPtr, uint32_t *zWPtr );
 
 #endif
 
 #endif
-

softfloat/source/ARM-VFPv2-defaultNaN/specialize.h

@@ -37,10 +37,10 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #ifndef specialize_h
 #define specialize_h 1
 
-#include <stdbool.h>
-#include <stdint.h>
 #include "primitiveTypes.h"
 #include "softfloat.h"
+#include <stdbool.h>
+#include <stdint.h>
 
 /*----------------------------------------------------------------------------
 | Default value for 'softfloat_detectTininess'.
@@ -62,18 +62,20 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 | The values to return on conversions to 64-bit integer formats that raise an
 | invalid exception.
 *----------------------------------------------------------------------------*/
-#define ui64_fromPosOverflow UINT64_C( 0xFFFFFFFFFFFFFFFF )
+#define ui64_fromPosOverflow UINT64_C(0xFFFFFFFFFFFFFFFF)
 #define ui64_fromNegOverflow 0
 #define ui64_fromNaN 0
-#define i64_fromPosOverflow  INT64_C( 0x7FFFFFFFFFFFFFFF )
+#define i64_fromPosOverflow INT64_C(0x7FFFFFFFFFFFFFFF)
-#define i64_fromNegOverflow  (-INT64_C( 0x7FFFFFFFFFFFFFFF ) - 1)
+#define i64_fromNegOverflow (-INT64_C(0x7FFFFFFFFFFFFFFF) - 1)
 #define i64_fromNaN 0
 
 /*----------------------------------------------------------------------------
 | "Common NaN" structure, used to transfer NaN representations from one format
 | to another.
 *----------------------------------------------------------------------------*/
-struct commonNaN { char _unused; };
+struct commonNaN {
+    char _unused;
+};
 
 /*----------------------------------------------------------------------------
 | The bit pattern for a default generated 16-bit floating-point NaN.
@@ -85,7 +87,7 @@ struct commonNaN { char _unused; };
 | 16-bit floating-point signaling NaN.
 | Note:  This macro evaluates its argument more than once.
 *----------------------------------------------------------------------------*/
-#define softfloat_isSigNaNF16UI( uiA ) ((((uiA) & 0x7E00) == 0x7C00) && ((uiA) & 0x01FF))
+#define softfloat_isSigNaNF16UI(uiA) ((((uiA)&0x7E00) == 0x7C00) && ((uiA)&0x01FF))
 
 /*----------------------------------------------------------------------------
 | Assuming 'uiA' has the bit pattern of a 16-bit floating-point NaN, converts
@@ -93,13 +95,15 @@ struct commonNaN { char _unused; };
 | location pointed to by 'zPtr'.  If the NaN is a signaling NaN, the invalid
 | exception is raised.
 *----------------------------------------------------------------------------*/
-#define softfloat_f16UIToCommonNaN( uiA, zPtr ) if ( ! ((uiA) & 0x0200) ) softfloat_raiseFlags( softfloat_flag_invalid )
+#define softfloat_f16UIToCommonNaN(uiA, zPtr)                                                                                              \
+    if(!((uiA)&0x0200))                                                                                                                    \
+    softfloat_raiseFlags(softfloat_flag_invalid)
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into a 16-bit floating-point
 | NaN, and returns the bit pattern of this value as an unsigned integer.
 *----------------------------------------------------------------------------*/
-#define softfloat_commonNaNToF16UI( aPtr ) ((uint_fast16_t) defaultNaNF16UI)
+#define softfloat_commonNaNToF16UI(aPtr) ((uint_fast16_t)defaultNaNF16UI)
 
 /*----------------------------------------------------------------------------
 | Interpreting 'uiA' and 'uiB' as the bit patterns of two 16-bit floating-
@@ -107,8 +111,7 @@ struct commonNaN { char _unused; };
 | the combined NaN result.  If either 'uiA' or 'uiB' has the pattern of a
 | signaling NaN, the invalid exception is raised.
 *----------------------------------------------------------------------------*/
-uint_fast16_t
+uint_fast16_t softfloat_propagateNaNF16UI(uint_fast16_t uiA, uint_fast16_t uiB);
- softfloat_propagateNaNF16UI( uint_fast16_t uiA, uint_fast16_t uiB );
 
 /*----------------------------------------------------------------------------
 | The bit pattern for a default generated 32-bit floating-point NaN.
@@ -120,7 +123,7 @@ uint_fast16_t
 | 32-bit floating-point signaling NaN.
 | Note:  This macro evaluates its argument more than once.
 *----------------------------------------------------------------------------*/
-#define softfloat_isSigNaNF32UI( uiA ) ((((uiA) & 0x7FC00000) == 0x7F800000) && ((uiA) & 0x003FFFFF))
+#define softfloat_isSigNaNF32UI(uiA) ((((uiA)&0x7FC00000) == 0x7F800000) && ((uiA)&0x003FFFFF))
 
 /*----------------------------------------------------------------------------
 | Assuming 'uiA' has the bit pattern of a 32-bit floating-point NaN, converts
@@ -128,13 +131,15 @@ uint_fast16_t
 | location pointed to by 'zPtr'.  If the NaN is a signaling NaN, the invalid
 | exception is raised.
 *----------------------------------------------------------------------------*/
-#define softfloat_f32UIToCommonNaN( uiA, zPtr ) if ( ! ((uiA) & 0x00400000) ) softfloat_raiseFlags( softfloat_flag_invalid )
+#define softfloat_f32UIToCommonNaN(uiA, zPtr)                                                                                              \
+    if(!((uiA)&0x00400000))                                                                                                                \
+    softfloat_raiseFlags(softfloat_flag_invalid)
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into a 32-bit floating-point
 | NaN, and returns the bit pattern of this value as an unsigned integer.
 *----------------------------------------------------------------------------*/
-#define softfloat_commonNaNToF32UI( aPtr ) ((uint_fast32_t) defaultNaNF32UI)
+#define softfloat_commonNaNToF32UI(aPtr) ((uint_fast32_t)defaultNaNF32UI)
 
 /*----------------------------------------------------------------------------
 | Interpreting 'uiA' and 'uiB' as the bit patterns of two 32-bit floating-
@@ -142,20 +147,20 @@ uint_fast16_t
 | the combined NaN result.  If either 'uiA' or 'uiB' has the pattern of a
 | signaling NaN, the invalid exception is raised.
 *----------------------------------------------------------------------------*/
-uint_fast32_t
+uint_fast32_t softfloat_propagateNaNF32UI(uint_fast32_t uiA, uint_fast32_t uiB);
- softfloat_propagateNaNF32UI( uint_fast32_t uiA, uint_fast32_t uiB );
 
 /*----------------------------------------------------------------------------
 | The bit pattern for a default generated 64-bit floating-point NaN.
 *----------------------------------------------------------------------------*/
-#define defaultNaNF64UI UINT64_C( 0x7FF8000000000000 )
+#define defaultNaNF64UI UINT64_C(0x7FF8000000000000)
 
 /*----------------------------------------------------------------------------
 | Returns true when 64-bit unsigned integer 'uiA' has the bit pattern of a
 | 64-bit floating-point signaling NaN.
 | Note:  This macro evaluates its argument more than once.
 *----------------------------------------------------------------------------*/
-#define softfloat_isSigNaNF64UI( uiA ) ((((uiA) & UINT64_C( 0x7FF8000000000000 )) == UINT64_C( 0x7FF0000000000000 )) && ((uiA) & UINT64_C( 0x0007FFFFFFFFFFFF )))
+#define softfloat_isSigNaNF64UI(uiA)                                                                                                       \
+    ((((uiA)&UINT64_C(0x7FF8000000000000)) == UINT64_C(0x7FF0000000000000)) && ((uiA)&UINT64_C(0x0007FFFFFFFFFFFF)))
 
 /*----------------------------------------------------------------------------
 | Assuming 'uiA' has the bit pattern of a 64-bit floating-point NaN, converts
@@ -163,13 +168,15 @@ uint_fast32_t
 | location pointed to by 'zPtr'.  If the NaN is a signaling NaN, the invalid
 | exception is raised.
 *----------------------------------------------------------------------------*/
-#define softfloat_f64UIToCommonNaN( uiA, zPtr ) if ( ! ((uiA) & UINT64_C( 0x0008000000000000 )) ) softfloat_raiseFlags( softfloat_flag_invalid )
+#define softfloat_f64UIToCommonNaN(uiA, zPtr)                                                                                              \
+    if(!((uiA)&UINT64_C(0x0008000000000000)))                                                                                              \
+    softfloat_raiseFlags(softfloat_flag_invalid)
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into a 64-bit floating-point
 | NaN, and returns the bit pattern of this value as an unsigned integer.
 *----------------------------------------------------------------------------*/
-#define softfloat_commonNaNToF64UI( aPtr ) ((uint_fast64_t) defaultNaNF64UI)
+#define softfloat_commonNaNToF64UI(aPtr) ((uint_fast64_t)defaultNaNF64UI)
 
 /*----------------------------------------------------------------------------
 | Interpreting 'uiA' and 'uiB' as the bit patterns of two 64-bit floating-
@@ -177,14 +184,13 @@ uint_fast32_t
 | the combined NaN result.  If either 'uiA' or 'uiB' has the pattern of a
 | signaling NaN, the invalid exception is raised.
 *----------------------------------------------------------------------------*/
-uint_fast64_t
+uint_fast64_t softfloat_propagateNaNF64UI(uint_fast64_t uiA, uint_fast64_t uiB);
- softfloat_propagateNaNF64UI( uint_fast64_t uiA, uint_fast64_t uiB );
 
 /*----------------------------------------------------------------------------
 | The bit pattern for a default generated 80-bit extended floating-point NaN.
 *----------------------------------------------------------------------------*/
 #define defaultNaNExtF80UI64 0x7FFF
-#define defaultNaNExtF80UI0  UINT64_C( 0xC000000000000000 )
+#define defaultNaNExtF80UI0 UINT64_C(0xC000000000000000)
 
 /*----------------------------------------------------------------------------
 | Returns true when the 80-bit unsigned integer formed from concatenating
@@ -192,7 +198,8 @@ uint_fast64_t
 | floating-point signaling NaN.
 | Note:  This macro evaluates its arguments more than once.
 *----------------------------------------------------------------------------*/
-#define softfloat_isSigNaNExtF80UI( uiA64, uiA0 ) ((((uiA64) & 0x7FFF) == 0x7FFF) && ! ((uiA0) & UINT64_C( 0x4000000000000000 )) && ((uiA0) & UINT64_C( 0x3FFFFFFFFFFFFFFF )))
+#define softfloat_isSigNaNExtF80UI(uiA64, uiA0)                                                                                            \
+    ((((uiA64)&0x7FFF) == 0x7FFF) && !((uiA0)&UINT64_C(0x4000000000000000)) && ((uiA0)&UINT64_C(0x3FFFFFFFFFFFFFFF)))
 
 #ifdef SOFTFLOAT_FAST_INT64
 
@@ -208,24 +215,25 @@ uint_fast64_t
 | location pointed to by 'zPtr'.  If the NaN is a signaling NaN, the invalid
 | exception is raised.
 *----------------------------------------------------------------------------*/
-#define softfloat_extF80UIToCommonNaN( uiA64, uiA0, zPtr ) if ( ! ((uiA0) & UINT64_C( 0x4000000000000000 )) ) softfloat_raiseFlags( softfloat_flag_invalid )
+#define softfloat_extF80UIToCommonNaN(uiA64, uiA0, zPtr)                                                                                   \
+    if(!((uiA0)&UINT64_C(0x4000000000000000)))                                                                                             \
+    softfloat_raiseFlags(softfloat_flag_invalid)
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into an 80-bit extended
 | floating-point NaN, and returns the bit pattern of this value as an unsigned
 | integer.
 *----------------------------------------------------------------------------*/
-#if defined INLINE && ! defined softfloat_commonNaNToExtF80UI
+#if defined INLINE && !defined softfloat_commonNaNToExtF80UI
 INLINE
-struct uint128 softfloat_commonNaNToExtF80UI( const struct commonNaN *aPtr )
+struct uint128 softfloat_commonNaNToExtF80UI(const struct commonNaN* aPtr) {
-{
     struct uint128 uiZ;
     uiZ.v64 = defaultNaNExtF80UI64;
     uiZ.v0 = defaultNaNExtF80UI0;
     return uiZ;
 }
 #else
-struct uint128 softfloat_commonNaNToExtF80UI( const struct commonNaN *aPtr );
+struct uint128 softfloat_commonNaNToExtF80UI(const struct commonNaN* aPtr);
 #endif
 
 /*----------------------------------------------------------------------------
@@ -237,19 +245,13 @@ struct uint128 softfloat_commonNaNToExtF80UI( const struct commonNaN *aPtr );
 | result.  If either original floating-point value is a signaling NaN, the
 | invalid exception is raised.
 *----------------------------------------------------------------------------*/
-struct uint128
+struct uint128 softfloat_propagateNaNExtF80UI(uint_fast16_t uiA64, uint_fast64_t uiA0, uint_fast16_t uiB64, uint_fast64_t uiB0);
- softfloat_propagateNaNExtF80UI(
-     uint_fast16_t uiA64,
-     uint_fast64_t uiA0,
-     uint_fast16_t uiB64,
-     uint_fast64_t uiB0
- );
 
 /*----------------------------------------------------------------------------
 | The bit pattern for a default generated 128-bit floating-point NaN.
 *----------------------------------------------------------------------------*/
-#define defaultNaNF128UI64 UINT64_C( 0x7FFF800000000000 )
+#define defaultNaNF128UI64 UINT64_C(0x7FFF800000000000)
-#define defaultNaNF128UI0  UINT64_C( 0 )
+#define defaultNaNF128UI0 UINT64_C(0)
 
 /*----------------------------------------------------------------------------
 | Returns true when the 128-bit unsigned integer formed from concatenating
@@ -257,7 +259,8 @@ struct uint128
 | point signaling NaN.
 | Note:  This macro evaluates its arguments more than once.
 *----------------------------------------------------------------------------*/
-#define softfloat_isSigNaNF128UI( uiA64, uiA0 ) ((((uiA64) & UINT64_C( 0x7FFF800000000000 )) == UINT64_C( 0x7FFF000000000000 )) && ((uiA0) || ((uiA64) & UINT64_C( 0x00007FFFFFFFFFFF ))))
+#define softfloat_isSigNaNF128UI(uiA64, uiA0)                                                                                              \
+    ((((uiA64)&UINT64_C(0x7FFF800000000000)) == UINT64_C(0x7FFF000000000000)) && ((uiA0) || ((uiA64)&UINT64_C(0x00007FFFFFFFFFFF))))
 
 /*----------------------------------------------------------------------------
 | Assuming the unsigned integer formed from concatenating 'uiA64' and 'uiA0'
@@ -266,23 +269,24 @@ struct uint128
 | pointed to by 'zPtr'.  If the NaN is a signaling NaN, the invalid exception
 | is raised.
 *----------------------------------------------------------------------------*/
-#define softfloat_f128UIToCommonNaN( uiA64, uiA0, zPtr ) if ( ! ((uiA64) & UINT64_C( 0x0000800000000000 )) ) softfloat_raiseFlags( softfloat_flag_invalid )
+#define softfloat_f128UIToCommonNaN(uiA64, uiA0, zPtr)                                                                                     \
+    if(!((uiA64)&UINT64_C(0x0000800000000000)))                                                                                            \
+    softfloat_raiseFlags(softfloat_flag_invalid)
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into a 128-bit floating-point
 | NaN, and returns the bit pattern of this value as an unsigned integer.
 *----------------------------------------------------------------------------*/
-#if defined INLINE && ! defined softfloat_commonNaNToF128UI
+#if defined INLINE && !defined softfloat_commonNaNToF128UI
 INLINE
-struct uint128 softfloat_commonNaNToF128UI( const struct commonNaN *aPtr )
+struct uint128 softfloat_commonNaNToF128UI(const struct commonNaN* aPtr) {
-{
     struct uint128 uiZ;
     uiZ.v64 = defaultNaNF128UI64;
     uiZ.v0 = defaultNaNF128UI0;
     return uiZ;
 }
 #else
-struct uint128 softfloat_commonNaNToF128UI( const struct commonNaN * );
+struct uint128 softfloat_commonNaNToF128UI(const struct commonNaN*);
 #endif
 
 /*----------------------------------------------------------------------------
@@ -294,13 +298,7 @@ struct uint128 softfloat_commonNaNToF128UI( const struct commonNaN * );
 | If either original floating-point value is a signaling NaN, the invalid
 | exception is raised.
 *----------------------------------------------------------------------------*/
-struct uint128
+struct uint128 softfloat_propagateNaNF128UI(uint_fast64_t uiA64, uint_fast64_t uiA0, uint_fast64_t uiB64, uint_fast64_t uiB0);
- softfloat_propagateNaNF128UI(
-     uint_fast64_t uiA64,
-     uint_fast64_t uiA0,
-     uint_fast64_t uiB64,
-     uint_fast64_t uiB0
- );
 
 #else
 
@@ -315,26 +313,23 @@ struct uint128
 | common NaN at the location pointed to by 'zPtr'.  If the NaN is a signaling
 | NaN, the invalid exception is raised.
 *----------------------------------------------------------------------------*/
-#define softfloat_extF80MToCommonNaN( aSPtr, zPtr ) if ( ! ((aSPtr)->signif & UINT64_C( 0x4000000000000000 )) ) softfloat_raiseFlags( softfloat_flag_invalid )
+#define softfloat_extF80MToCommonNaN(aSPtr, zPtr)                                                                                          \
+    if(!((aSPtr)->signif & UINT64_C(0x4000000000000000)))                                                                                  \
+    softfloat_raiseFlags(softfloat_flag_invalid)
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into an 80-bit extended
 | floating-point NaN, and stores this NaN at the location pointed to by
 | 'zSPtr'.
 *----------------------------------------------------------------------------*/
-#if defined INLINE && ! defined softfloat_commonNaNToExtF80M
+#if defined INLINE && !defined softfloat_commonNaNToExtF80M
 INLINE
-void
+void softfloat_commonNaNToExtF80M(const struct commonNaN* aPtr, struct extFloat80M* zSPtr) {
- softfloat_commonNaNToExtF80M(
-     const struct commonNaN *aPtr, struct extFloat80M *zSPtr )
-{
     zSPtr->signExp = defaultNaNExtF80UI64;
     zSPtr->signif = defaultNaNExtF80UI0;
 }
 #else
-void
+void softfloat_commonNaNToExtF80M(const struct commonNaN* aPtr, struct extFloat80M* zSPtr);
- softfloat_commonNaNToExtF80M(
-     const struct commonNaN *aPtr, struct extFloat80M *zSPtr );
 #endif
 
 /*----------------------------------------------------------------------------
@@ -343,12 +338,7 @@ void
 | at the location pointed to by 'zSPtr'.  If either original floating-point
 | value is a signaling NaN, the invalid exception is raised.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_propagateNaNExtF80M(const struct extFloat80M* aSPtr, const struct extFloat80M* bSPtr, struct extFloat80M* zSPtr);
- softfloat_propagateNaNExtF80M(
-     const struct extFloat80M *aSPtr,
-     const struct extFloat80M *bSPtr,
-     struct extFloat80M *zSPtr
- );
 
 /*----------------------------------------------------------------------------
 | The bit pattern for a default generated 128-bit floating-point NaN.
@@ -366,7 +356,9 @@ void
 | four 32-bit elements that concatenate in the platform's normal endian order
 | to form a 128-bit floating-point value.
 *----------------------------------------------------------------------------*/
-#define softfloat_f128MToCommonNaN( aWPtr, zPtr ) if ( ! ((aWPtr)[indexWordHi( 4 )] & UINT64_C( 0x0000800000000000 )) ) softfloat_raiseFlags( softfloat_flag_invalid )
+#define softfloat_f128MToCommonNaN(aWPtr, zPtr)                                                                                            \
+    if(!((aWPtr)[indexWordHi(4)] & UINT64_C(0x0000800000000000)))                                                                          \
+    softfloat_raiseFlags(softfloat_flag_invalid)
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into a 128-bit floating-point
@@ -374,19 +366,16 @@ void
 | 'zWPtr' points to an array of four 32-bit elements that concatenate in the
 | platform's normal endian order to form a 128-bit floating-point value.
 *----------------------------------------------------------------------------*/
-#if defined INLINE && ! defined softfloat_commonNaNToF128M
+#if defined INLINE && !defined softfloat_commonNaNToF128M
 INLINE
-void
+void softfloat_commonNaNToF128M(const struct commonNaN* aPtr, uint32_t* zWPtr) {
- softfloat_commonNaNToF128M( const struct commonNaN *aPtr, uint32_t *zWPtr )
+    zWPtr[indexWord(4, 3)] = defaultNaNF128UI96;
-{
+    zWPtr[indexWord(4, 2)] = defaultNaNF128UI64;
-    zWPtr[indexWord( 4, 3 )] = defaultNaNF128UI96;
+    zWPtr[indexWord(4, 1)] = defaultNaNF128UI32;
-    zWPtr[indexWord( 4, 2 )] = defaultNaNF128UI64;
+    zWPtr[indexWord(4, 0)] = defaultNaNF128UI0;
-    zWPtr[indexWord( 4, 1 )] = defaultNaNF128UI32;
-    zWPtr[indexWord( 4, 0 )] = defaultNaNF128UI0;
 }
 #else
-void
+void softfloat_commonNaNToF128M(const struct commonNaN* aPtr, uint32_t* zWPtr);
- softfloat_commonNaNToF128M( const struct commonNaN *aPtr, uint32_t *zWPtr );
 #endif
 
 /*----------------------------------------------------------------------------
@@ -397,11 +386,8 @@ void
 | and 'zWPtr' points to an array of four 32-bit elements that concatenate in
 | the platform's normal endian order to form a 128-bit floating-point value.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_propagateNaNF128M(const uint32_t* aWPtr, const uint32_t* bWPtr, uint32_t* zWPtr);
- softfloat_propagateNaNF128M(
-     const uint32_t *aWPtr, const uint32_t *bWPtr, uint32_t *zWPtr );
 
 #endif
 
 #endif
-

softfloat/source/ARM-VFPv2/specialize.h

@@ -37,10 +37,10 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #ifndef specialize_h
 #define specialize_h 1
 
-#include <stdbool.h>
-#include <stdint.h>
 #include "primitiveTypes.h"
 #include "softfloat.h"
+#include <stdbool.h>
+#include <stdint.h>
 
 /*----------------------------------------------------------------------------
 | Default value for 'softfloat_detectTininess'.
@@ -62,11 +62,11 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 | The values to return on conversions to 64-bit integer formats that raise an
 | invalid exception.
 *----------------------------------------------------------------------------*/
-#define ui64_fromPosOverflow UINT64_C( 0xFFFFFFFFFFFFFFFF )
+#define ui64_fromPosOverflow UINT64_C(0xFFFFFFFFFFFFFFFF)
 #define ui64_fromNegOverflow 0
 #define ui64_fromNaN 0
-#define i64_fromPosOverflow  INT64_C( 0x7FFFFFFFFFFFFFFF )
+#define i64_fromPosOverflow INT64_C(0x7FFFFFFFFFFFFFFF)
-#define i64_fromNegOverflow  (-INT64_C( 0x7FFFFFFFFFFFFFFF ) - 1)
+#define i64_fromNegOverflow (-INT64_C(0x7FFFFFFFFFFFFFFF) - 1)
 #define i64_fromNaN 0
 
 /*----------------------------------------------------------------------------
@@ -92,7 +92,7 @@ struct commonNaN {
 | 16-bit floating-point signaling NaN.
 | Note:  This macro evaluates its argument more than once.
 *----------------------------------------------------------------------------*/
-#define softfloat_isSigNaNF16UI( uiA ) ((((uiA) & 0x7E00) == 0x7C00) && ((uiA) & 0x01FF))
+#define softfloat_isSigNaNF16UI(uiA) ((((uiA)&0x7E00) == 0x7C00) && ((uiA)&0x01FF))
 
 /*----------------------------------------------------------------------------
 | Assuming 'uiA' has the bit pattern of a 16-bit floating-point NaN, converts
@@ -100,13 +100,13 @@ struct commonNaN {
 | location pointed to by 'zPtr'.  If the NaN is a signaling NaN, the invalid
 | exception is raised.
 *----------------------------------------------------------------------------*/
-void softfloat_f16UIToCommonNaN( uint_fast16_t uiA, struct commonNaN *zPtr );
+void softfloat_f16UIToCommonNaN(uint_fast16_t uiA, struct commonNaN* zPtr);
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into a 16-bit floating-point
 | NaN, and returns the bit pattern of this value as an unsigned integer.
 *----------------------------------------------------------------------------*/
-uint_fast16_t softfloat_commonNaNToF16UI( const struct commonNaN *aPtr );
+uint_fast16_t softfloat_commonNaNToF16UI(const struct commonNaN* aPtr);
 
 /*----------------------------------------------------------------------------
 | Interpreting 'uiA' and 'uiB' as the bit patterns of two 16-bit floating-
@@ -114,8 +114,7 @@ uint_fast16_t softfloat_commonNaNToF16UI( const struct commonNaN *aPtr );
 | the combined NaN result.  If either 'uiA' or 'uiB' has the pattern of a
 | signaling NaN, the invalid exception is raised.
 *----------------------------------------------------------------------------*/
-uint_fast16_t
+uint_fast16_t softfloat_propagateNaNF16UI(uint_fast16_t uiA, uint_fast16_t uiB);
- softfloat_propagateNaNF16UI( uint_fast16_t uiA, uint_fast16_t uiB );
 
 /*----------------------------------------------------------------------------
 | The bit pattern for a default generated 32-bit floating-point NaN.
@@ -127,7 +126,7 @@ uint_fast16_t
 | 32-bit floating-point signaling NaN.
 | Note:  This macro evaluates its argument more than once.
 *----------------------------------------------------------------------------*/
-#define softfloat_isSigNaNF32UI( uiA ) ((((uiA) & 0x7FC00000) == 0x7F800000) && ((uiA) & 0x003FFFFF))
+#define softfloat_isSigNaNF32UI(uiA) ((((uiA)&0x7FC00000) == 0x7F800000) && ((uiA)&0x003FFFFF))
 
 /*----------------------------------------------------------------------------
 | Assuming 'uiA' has the bit pattern of a 32-bit floating-point NaN, converts
@@ -135,13 +134,13 @@ uint_fast16_t
 | location pointed to by 'zPtr'.  If the NaN is a signaling NaN, the invalid
 | exception is raised.
 *----------------------------------------------------------------------------*/
-void softfloat_f32UIToCommonNaN( uint_fast32_t uiA, struct commonNaN *zPtr );
+void softfloat_f32UIToCommonNaN(uint_fast32_t uiA, struct commonNaN* zPtr);
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into a 32-bit floating-point
 | NaN, and returns the bit pattern of this value as an unsigned integer.
 *----------------------------------------------------------------------------*/
-uint_fast32_t softfloat_commonNaNToF32UI( const struct commonNaN *aPtr );
+uint_fast32_t softfloat_commonNaNToF32UI(const struct commonNaN* aPtr);
 
 /*----------------------------------------------------------------------------
 | Interpreting 'uiA' and 'uiB' as the bit patterns of two 32-bit floating-
@@ -149,20 +148,20 @@ uint_fast32_t softfloat_commonNaNToF32UI( const struct commonNaN *aPtr );
 | the combined NaN result.  If either 'uiA' or 'uiB' has the pattern of a
 | signaling NaN, the invalid exception is raised.
 *----------------------------------------------------------------------------*/
-uint_fast32_t
+uint_fast32_t softfloat_propagateNaNF32UI(uint_fast32_t uiA, uint_fast32_t uiB);
- softfloat_propagateNaNF32UI( uint_fast32_t uiA, uint_fast32_t uiB );
 
 /*----------------------------------------------------------------------------
 | The bit pattern for a default generated 64-bit floating-point NaN.
 *----------------------------------------------------------------------------*/
-#define defaultNaNF64UI UINT64_C( 0x7FF8000000000000 )
+#define defaultNaNF64UI UINT64_C(0x7FF8000000000000)
 
 /*----------------------------------------------------------------------------
 | Returns true when 64-bit unsigned integer 'uiA' has the bit pattern of a
 | 64-bit floating-point signaling NaN.
 | Note:  This macro evaluates its argument more than once.
 *----------------------------------------------------------------------------*/
-#define softfloat_isSigNaNF64UI( uiA ) ((((uiA) & UINT64_C( 0x7FF8000000000000 )) == UINT64_C( 0x7FF0000000000000 )) && ((uiA) & UINT64_C( 0x0007FFFFFFFFFFFF )))
+#define softfloat_isSigNaNF64UI(uiA)                                                                                                       \
+    ((((uiA)&UINT64_C(0x7FF8000000000000)) == UINT64_C(0x7FF0000000000000)) && ((uiA)&UINT64_C(0x0007FFFFFFFFFFFF)))
 
 /*----------------------------------------------------------------------------
 | Assuming 'uiA' has the bit pattern of a 64-bit floating-point NaN, converts
@@ -170,13 +169,13 @@ uint_fast32_t
 | location pointed to by 'zPtr'.  If the NaN is a signaling NaN, the invalid
 | exception is raised.
 *----------------------------------------------------------------------------*/
-void softfloat_f64UIToCommonNaN( uint_fast64_t uiA, struct commonNaN *zPtr );
+void softfloat_f64UIToCommonNaN(uint_fast64_t uiA, struct commonNaN* zPtr);
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into a 64-bit floating-point
 | NaN, and returns the bit pattern of this value as an unsigned integer.
 *----------------------------------------------------------------------------*/
-uint_fast64_t softfloat_commonNaNToF64UI( const struct commonNaN *aPtr );
+uint_fast64_t softfloat_commonNaNToF64UI(const struct commonNaN* aPtr);
 
 /*----------------------------------------------------------------------------
 | Interpreting 'uiA' and 'uiB' as the bit patterns of two 64-bit floating-
@@ -184,14 +183,13 @@ uint_fast64_t softfloat_commonNaNToF64UI( const struct commonNaN *aPtr );
 | the combined NaN result.  If either 'uiA' or 'uiB' has the pattern of a
 | signaling NaN, the invalid exception is raised.
 *----------------------------------------------------------------------------*/
-uint_fast64_t
+uint_fast64_t softfloat_propagateNaNF64UI(uint_fast64_t uiA, uint_fast64_t uiB);
- softfloat_propagateNaNF64UI( uint_fast64_t uiA, uint_fast64_t uiB );
 
 /*----------------------------------------------------------------------------
 | The bit pattern for a default generated 80-bit extended floating-point NaN.
 *----------------------------------------------------------------------------*/
 #define defaultNaNExtF80UI64 0x7FFF
-#define defaultNaNExtF80UI0  UINT64_C( 0xC000000000000000 )
+#define defaultNaNExtF80UI0 UINT64_C(0xC000000000000000)
 
 /*----------------------------------------------------------------------------
 | Returns true when the 80-bit unsigned integer formed from concatenating
@@ -199,7 +197,8 @@ uint_fast64_t
 | floating-point signaling NaN.
 | Note:  This macro evaluates its arguments more than once.
 *----------------------------------------------------------------------------*/
-#define softfloat_isSigNaNExtF80UI( uiA64, uiA0 ) ((((uiA64) & 0x7FFF) == 0x7FFF) && ! ((uiA0) & UINT64_C( 0x4000000000000000 )) && ((uiA0) & UINT64_C( 0x3FFFFFFFFFFFFFFF )))
+#define softfloat_isSigNaNExtF80UI(uiA64, uiA0)                                                                                            \
+    ((((uiA64)&0x7FFF) == 0x7FFF) && !((uiA0)&UINT64_C(0x4000000000000000)) && ((uiA0)&UINT64_C(0x3FFFFFFFFFFFFFFF)))
 
 #ifdef SOFTFLOAT_FAST_INT64
 
@@ -215,16 +214,14 @@ uint_fast64_t
 | location pointed to by 'zPtr'.  If the NaN is a signaling NaN, the invalid
 | exception is raised.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_extF80UIToCommonNaN(uint_fast16_t uiA64, uint_fast64_t uiA0, struct commonNaN* zPtr);
- softfloat_extF80UIToCommonNaN(
-     uint_fast16_t uiA64, uint_fast64_t uiA0, struct commonNaN *zPtr );
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into an 80-bit extended
 | floating-point NaN, and returns the bit pattern of this value as an unsigned
 | integer.
 *----------------------------------------------------------------------------*/
-struct uint128 softfloat_commonNaNToExtF80UI( const struct commonNaN *aPtr );
+struct uint128 softfloat_commonNaNToExtF80UI(const struct commonNaN* aPtr);
 
 /*----------------------------------------------------------------------------
 | Interpreting the unsigned integer formed from concatenating 'uiA64' and
@@ -235,19 +232,13 @@ struct uint128 softfloat_commonNaNToExtF80UI( const struct commonNaN *aPtr );
 | result.  If either original floating-point value is a signaling NaN, the
 | invalid exception is raised.
 *----------------------------------------------------------------------------*/
-struct uint128
+struct uint128 softfloat_propagateNaNExtF80UI(uint_fast16_t uiA64, uint_fast64_t uiA0, uint_fast16_t uiB64, uint_fast64_t uiB0);
- softfloat_propagateNaNExtF80UI(
-     uint_fast16_t uiA64,
-     uint_fast64_t uiA0,
-     uint_fast16_t uiB64,
-     uint_fast64_t uiB0
- );
 
 /*----------------------------------------------------------------------------
 | The bit pattern for a default generated 128-bit floating-point NaN.
 *----------------------------------------------------------------------------*/
-#define defaultNaNF128UI64 UINT64_C( 0x7FFF800000000000 )
+#define defaultNaNF128UI64 UINT64_C(0x7FFF800000000000)
-#define defaultNaNF128UI0  UINT64_C( 0 )
+#define defaultNaNF128UI0 UINT64_C(0)
 
 /*----------------------------------------------------------------------------
 | Returns true when the 128-bit unsigned integer formed from concatenating
@@ -255,7 +246,8 @@ struct uint128
 | point signaling NaN.
 | Note:  This macro evaluates its arguments more than once.
 *----------------------------------------------------------------------------*/
-#define softfloat_isSigNaNF128UI( uiA64, uiA0 ) ((((uiA64) & UINT64_C( 0x7FFF800000000000 )) == UINT64_C( 0x7FFF000000000000 )) && ((uiA0) || ((uiA64) & UINT64_C( 0x00007FFFFFFFFFFF ))))
+#define softfloat_isSigNaNF128UI(uiA64, uiA0)                                                                                              \
+    ((((uiA64)&UINT64_C(0x7FFF800000000000)) == UINT64_C(0x7FFF000000000000)) && ((uiA0) || ((uiA64)&UINT64_C(0x00007FFFFFFFFFFF))))
 
 /*----------------------------------------------------------------------------
 | Assuming the unsigned integer formed from concatenating 'uiA64' and 'uiA0'
@@ -264,15 +256,13 @@ struct uint128
 | pointed to by 'zPtr'.  If the NaN is a signaling NaN, the invalid exception
 | is raised.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_f128UIToCommonNaN(uint_fast64_t uiA64, uint_fast64_t uiA0, struct commonNaN* zPtr);
- softfloat_f128UIToCommonNaN(
-     uint_fast64_t uiA64, uint_fast64_t uiA0, struct commonNaN *zPtr );
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into a 128-bit floating-point
 | NaN, and returns the bit pattern of this value as an unsigned integer.
 *----------------------------------------------------------------------------*/
-struct uint128 softfloat_commonNaNToF128UI( const struct commonNaN * );
+struct uint128 softfloat_commonNaNToF128UI(const struct commonNaN*);
 
 /*----------------------------------------------------------------------------
 | Interpreting the unsigned integer formed from concatenating 'uiA64' and
@@ -283,13 +273,7 @@ struct uint128 softfloat_commonNaNToF128UI( const struct commonNaN * );
 | If either original floating-point value is a signaling NaN, the invalid
 | exception is raised.
 *----------------------------------------------------------------------------*/
-struct uint128
+struct uint128 softfloat_propagateNaNF128UI(uint_fast64_t uiA64, uint_fast64_t uiA0, uint_fast64_t uiB64, uint_fast64_t uiB0);
- softfloat_propagateNaNF128UI(
-     uint_fast64_t uiA64,
-     uint_fast64_t uiA0,
-     uint_fast64_t uiB64,
-     uint_fast64_t uiB0
- );
 
 #else
 
@@ -304,18 +288,14 @@ struct uint128
 | common NaN at the location pointed to by 'zPtr'.  If the NaN is a signaling
 | NaN, the invalid exception is raised.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_extF80MToCommonNaN(const struct extFloat80M* aSPtr, struct commonNaN* zPtr);
- softfloat_extF80MToCommonNaN(
-     const struct extFloat80M *aSPtr, struct commonNaN *zPtr );
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into an 80-bit extended
 | floating-point NaN, and stores this NaN at the location pointed to by
 | 'zSPtr'.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_commonNaNToExtF80M(const struct commonNaN* aPtr, struct extFloat80M* zSPtr);
- softfloat_commonNaNToExtF80M(
-     const struct commonNaN *aPtr, struct extFloat80M *zSPtr );
 
 /*----------------------------------------------------------------------------
 | Assuming at least one of the two 80-bit extended floating-point values
@@ -323,12 +303,7 @@ void
 | at the location pointed to by 'zSPtr'.  If either original floating-point
 | value is a signaling NaN, the invalid exception is raised.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_propagateNaNExtF80M(const struct extFloat80M* aSPtr, const struct extFloat80M* bSPtr, struct extFloat80M* zSPtr);
- softfloat_propagateNaNExtF80M(
-     const struct extFloat80M *aSPtr,
-     const struct extFloat80M *bSPtr,
-     struct extFloat80M *zSPtr
- );
 
 /*----------------------------------------------------------------------------
 | The bit pattern for a default generated 128-bit floating-point NaN.
@@ -346,8 +321,7 @@ void
 | four 32-bit elements that concatenate in the platform's normal endian order
 | to form a 128-bit floating-point value.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_f128MToCommonNaN(const uint32_t* aWPtr, struct commonNaN* zPtr);
- softfloat_f128MToCommonNaN( const uint32_t *aWPtr, struct commonNaN *zPtr );
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into a 128-bit floating-point
@@ -355,8 +329,7 @@ void
 | 'zWPtr' points to an array of four 32-bit elements that concatenate in the
 | platform's normal endian order to form a 128-bit floating-point value.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_commonNaNToF128M(const struct commonNaN* aPtr, uint32_t* zWPtr);
- softfloat_commonNaNToF128M( const struct commonNaN *aPtr, uint32_t *zWPtr );
 
 /*----------------------------------------------------------------------------
 | Assuming at least one of the two 128-bit floating-point values pointed to by
@@ -366,11 +339,8 @@ void
 | and 'zWPtr' points to an array of four 32-bit elements that concatenate in
 | the platform's normal endian order to form a 128-bit floating-point value.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_propagateNaNF128M(const uint32_t* aWPtr, const uint32_t* bWPtr, uint32_t* zWPtr);
- softfloat_propagateNaNF128M(
-     const uint32_t *aWPtr, const uint32_t *bWPtr, uint32_t *zWPtr );
 
 #endif
 
 #endif
-

softfloat/source/RISCV/s_bf16UIToCommonNaN.c

@@ -0,0 +1,5 @@
+
+/*----------------------------------------------------------------------------
+| This file intentionally contains no code.
+*----------------------------------------------------------------------------*/
+

softfloat/source/RISCV/s_commonNaNToBF16UI.c

@@ -0,0 +1,5 @@
+
+/*----------------------------------------------------------------------------
+| This file intentionally contains no code.
+*----------------------------------------------------------------------------*/
+

softfloat/source/RISCV/s_commonNaNToExtF80M.c

@@ -4,8 +4,8 @@
 This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
 Package, Release 3e, by John R. Hauser.
 
-Copyright 2011, 2012, 2013, 2014 The Regents of the University of California.
+Copyright 2011, 2012, 2013, 2014, 2015 The Regents of the University of
-All rights reserved.
+California.  All rights reserved.
 
 Redistribution and use in source and binary forms, with or without
 modification, are permitted provided that the following conditions are met:
@@ -34,9 +34,10 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 =============================================================================*/
 
-#include <stdint.h>
 #include "platform.h"
-#include "internals.h"
+#include "softfloat_types.h"
+
+#define softfloat_commonNaNToExtF80M softfloat_commonNaNToExtF80M
 #include "specialize.h"
 
 /*----------------------------------------------------------------------------
@@ -49,8 +50,8 @@ void
      const struct commonNaN *aPtr, struct extFloat80M *zSPtr )
 {
 
-    zSPtr->signExp = packToExtF80UI64( aPtr->sign, 0x7FFF );
+    zSPtr->signExp = defaultNaNExtF80UI64;
-    zSPtr->signif = UINT64_C( 0xC000000000000000 ) | aPtr->v64>>1;
+    zSPtr->signif  = defaultNaNExtF80UI0;
 
 }
 

softfloat/source/RISCV/s_commonNaNToExtF80UI.c

@@ -4,8 +4,8 @@
 This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
 Package, Release 3e, by John R. Hauser.
 
-Copyright 2011, 2012, 2013, 2014 The Regents of the University of California.
+Copyright 2011, 2012, 2013, 2014, 2015 The Regents of the University of
-All rights reserved.
+California.  All rights reserved.
 
 Redistribution and use in source and binary forms, with or without
 modification, are permitted provided that the following conditions are met:
@@ -34,9 +34,10 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 =============================================================================*/
 
-#include <stdint.h>
 #include "platform.h"
-#include "primitives.h"
+#include "primitiveTypes.h"
+
+#define softfloat_commonNaNToExtF80UI softfloat_commonNaNToExtF80UI
 #include "specialize.h"
 
 /*----------------------------------------------------------------------------
@@ -48,8 +49,8 @@ struct uint128 softfloat_commonNaNToExtF80UI( const struct commonNaN *aPtr )
 {
     struct uint128 uiZ;
 
-    uiZ.v64 = (uint_fast16_t) aPtr->sign<<15 | 0x7FFF;
+    uiZ.v64 = defaultNaNExtF80UI64;
-    uiZ.v0 = UINT64_C( 0xC000000000000000 ) | aPtr->v64>>1;
+    uiZ.v0  = defaultNaNExtF80UI0;
     return uiZ;
 
 }

softfloat/source/RISCV/s_commonNaNToF128M.c

@@ -4,8 +4,8 @@
 This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
 Package, Release 3e, by John R. Hauser.
 
-Copyright 2011, 2012, 2013, 2014 The Regents of the University of California.
+Copyright 2011, 2012, 2013, 2014, 2015 The Regents of the University of
-All rights reserved.
+California.  All rights reserved.
 
 Redistribution and use in source and binary forms, with or without
 modification, are permitted provided that the following conditions are met:
@@ -36,7 +36,9 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 #include <stdint.h>
 #include "platform.h"
-#include "primitives.h"
+#include "primitiveTypes.h"
+
+#define softfloat_commonNaNToF128M softfloat_commonNaNToF128M
 #include "specialize.h"
 
 /*----------------------------------------------------------------------------
@@ -49,8 +51,10 @@ void
  softfloat_commonNaNToF128M( const struct commonNaN *aPtr, uint32_t *zWPtr )
 {
 
-    softfloat_shortShiftRight128M( (const uint32_t *) &aPtr->v0, 16, zWPtr );
+    zWPtr[indexWord( 4, 3 )] = defaultNaNF128UI96;
-    zWPtr[indexWordHi( 4 )] |= (uint32_t) aPtr->sign<<31 | 0x7FFF8000;
+    zWPtr[indexWord( 4, 2 )] = defaultNaNF128UI64;
+    zWPtr[indexWord( 4, 1 )] = defaultNaNF128UI32;
+    zWPtr[indexWord( 4, 0 )] = defaultNaNF128UI0;
 
 }
 

softfloat/source/RISCV/s_commonNaNToF128UI.c

@@ -4,8 +4,8 @@
 This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
 Package, Release 3e, by John R. Hauser.
 
-Copyright 2011, 2012, 2013, 2014 The Regents of the University of California.
+Copyright 2011, 2012, 2013, 2014, 2015 The Regents of the University of
-All rights reserved.
+California.  All rights reserved.
 
 Redistribution and use in source and binary forms, with or without
 modification, are permitted provided that the following conditions are met:
@@ -34,9 +34,10 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 =============================================================================*/
 
-#include <stdint.h>
 #include "platform.h"
-#include "primitives.h"
+#include "primitiveTypes.h"
+
+#define softfloat_commonNaNToF128UI softfloat_commonNaNToF128UI
 #include "specialize.h"
 
 /*----------------------------------------------------------------------------
@@ -47,8 +48,8 @@ struct uint128 softfloat_commonNaNToF128UI( const struct commonNaN *aPtr )
 {
     struct uint128 uiZ;
 
-    uiZ = softfloat_shortShiftRight128( aPtr->v64, aPtr->v0, 16 );
+    uiZ.v64 = defaultNaNF128UI64;
-    uiZ.v64 |= (uint_fast64_t) aPtr->sign<<63 | UINT64_C( 0x7FFF800000000000 );
+    uiZ.v0  = defaultNaNF128UI0;
     return uiZ;
 
 }

softfloat/source/RISCV/s_commonNaNToF16UI.c

@@ -1,51 +1,5 @@
 
-/*============================================================================
-
-This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
-Package, Release 3e, by John R. Hauser.
-
-Copyright 2011, 2012, 2013, 2014, 2015 The Regents of the University of
-California.  All rights reserved.
-
-Redistribution and use in source and binary forms, with or without
-modification, are permitted provided that the following conditions are met:
-
- 1. Redistributions of source code must retain the above copyright notice,
-    this list of conditions, and the following disclaimer.
-
- 2. Redistributions in binary form must reproduce the above copyright notice,
-    this list of conditions, and the following disclaimer in the documentation
-    and/or other materials provided with the distribution.
-
- 3. Neither the name of the University nor the names of its contributors may
-    be used to endorse or promote products derived from this software without
-    specific prior written permission.
-
-THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS "AS IS", AND ANY
-EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
-WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, ARE
-DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY
-DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
-(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
-LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
-ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
-SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-=============================================================================*/
-
-#include <stdint.h>
-#include "platform.h"
-#include "specialize.h"
-
 /*----------------------------------------------------------------------------
-| Converts the common NaN pointed to by `aPtr' into a 16-bit floating-point
+| This file intentionally contains no code.
-| NaN, and returns the bit pattern of this value as an unsigned integer.
 *----------------------------------------------------------------------------*/
-uint_fast16_t softfloat_commonNaNToF16UI( const struct commonNaN *aPtr )
-{
-
-    return (uint_fast16_t) aPtr->sign<<15 | 0x7E00 | aPtr->v64>>54;
-
-}
 

softfloat/source/RISCV/s_commonNaNToF32UI.c

@@ -1,51 +1,5 @@
 
-/*============================================================================
-
-This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
-Package, Release 3e, by John R. Hauser.
-
-Copyright 2011, 2012, 2013, 2014 The Regents of the University of California.
-All rights reserved.
-
-Redistribution and use in source and binary forms, with or without
-modification, are permitted provided that the following conditions are met:
-
- 1. Redistributions of source code must retain the above copyright notice,
-    this list of conditions, and the following disclaimer.
-
- 2. Redistributions in binary form must reproduce the above copyright notice,
-    this list of conditions, and the following disclaimer in the documentation
-    and/or other materials provided with the distribution.
-
- 3. Neither the name of the University nor the names of its contributors may
-    be used to endorse or promote products derived from this software without
-    specific prior written permission.
-
-THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS "AS IS", AND ANY
-EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
-WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, ARE
-DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY
-DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
-(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
-LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
-ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
-SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-=============================================================================*/
-
-#include <stdint.h>
-#include "platform.h"
-#include "specialize.h"
-
 /*----------------------------------------------------------------------------
-| Converts the common NaN pointed to by `aPtr' into a 32-bit floating-point
+| This file intentionally contains no code.
-| NaN, and returns the bit pattern of this value as an unsigned integer.
 *----------------------------------------------------------------------------*/
-uint_fast32_t softfloat_commonNaNToF32UI( const struct commonNaN *aPtr )
-{
-
-    return (uint_fast32_t) aPtr->sign<<31 | 0x7FC00000 | aPtr->v64>>41;
-
-}
 

softfloat/source/RISCV/s_commonNaNToF64UI.c

@@ -1,53 +1,5 @@
 
-/*============================================================================
-
-This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
-Package, Release 3e, by John R. Hauser.
-
-Copyright 2011, 2012, 2013, 2014 The Regents of the University of California.
-All rights reserved.
-
-Redistribution and use in source and binary forms, with or without
-modification, are permitted provided that the following conditions are met:
-
- 1. Redistributions of source code must retain the above copyright notice,
-    this list of conditions, and the following disclaimer.
-
- 2. Redistributions in binary form must reproduce the above copyright notice,
-    this list of conditions, and the following disclaimer in the documentation
-    and/or other materials provided with the distribution.
-
- 3. Neither the name of the University nor the names of its contributors may
-    be used to endorse or promote products derived from this software without
-    specific prior written permission.
-
-THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS "AS IS", AND ANY
-EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
-WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, ARE
-DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY
-DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
-(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
-LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
-ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
-SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-=============================================================================*/
-
-#include <stdint.h>
-#include "platform.h"
-#include "specialize.h"
-
 /*----------------------------------------------------------------------------
-| Converts the common NaN pointed to by `aPtr' into a 64-bit floating-point
+| This file intentionally contains no code.
-| NaN, and returns the bit pattern of this value as an unsigned integer.
 *----------------------------------------------------------------------------*/
-uint_fast64_t softfloat_commonNaNToF64UI( const struct commonNaN *aPtr )
-{
-
-    return
-        (uint_fast64_t) aPtr->sign<<63 | UINT64_C( 0x7FF8000000000000 )
-            | aPtr->v64>>12;
-
-}
 

softfloat/source/RISCV/s_extF80MToCommonNaN.c

@@ -1,62 +1,5 @@
 
-/*============================================================================
-
-This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
-Package, Release 3e, by John R. Hauser.
-
-Copyright 2011, 2012, 2013, 2014 The Regents of the University of California.
-All rights reserved.
-
-Redistribution and use in source and binary forms, with or without
-modification, are permitted provided that the following conditions are met:
-
- 1. Redistributions of source code must retain the above copyright notice,
-    this list of conditions, and the following disclaimer.
-
- 2. Redistributions in binary form must reproduce the above copyright notice,
-    this list of conditions, and the following disclaimer in the documentation
-    and/or other materials provided with the distribution.
-
- 3. Neither the name of the University nor the names of its contributors may
-    be used to endorse or promote products derived from this software without
-    specific prior written permission.
-
-THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS "AS IS", AND ANY
-EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
-WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, ARE
-DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY
-DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
-(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
-LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
-ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
-SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-=============================================================================*/
-
-#include <stdint.h>
-#include "platform.h"
-#include "internals.h"
-#include "specialize.h"
-#include "softfloat.h"
-
 /*----------------------------------------------------------------------------
-| Assuming the 80-bit extended floating-point value pointed to by `aSPtr' is
+| This file intentionally contains no code.
-| a NaN, converts this NaN to the common NaN form, and stores the resulting
-| common NaN at the location pointed to by `zPtr'.  If the NaN is a signaling
-| NaN, the invalid exception is raised.
 *----------------------------------------------------------------------------*/
-void
- softfloat_extF80MToCommonNaN(
-     const struct extFloat80M *aSPtr, struct commonNaN *zPtr )
-{
-
-    if ( extF80M_isSignalingNaN( (const extFloat80_t *) aSPtr ) ) {
-        softfloat_raiseFlags( softfloat_flag_invalid );
-    }
-    zPtr->sign = signExtF80UI64( aSPtr->signExp );
-    zPtr->v64 = aSPtr->signif<<1;
-    zPtr->v0  = 0;
-
-}
 

softfloat/source/RISCV/s_extF80UIToCommonNaN.c

@@ -1,62 +1,5 @@
 
-/*============================================================================
-
-This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
-Package, Release 3e, by John R. Hauser.
-
-Copyright 2011, 2012, 2013, 2014 The Regents of the University of California.
-All rights reserved.
-
-Redistribution and use in source and binary forms, with or without
-modification, are permitted provided that the following conditions are met:
-
- 1. Redistributions of source code must retain the above copyright notice,
-    this list of conditions, and the following disclaimer.
-
- 2. Redistributions in binary form must reproduce the above copyright notice,
-    this list of conditions, and the following disclaimer in the documentation
-    and/or other materials provided with the distribution.
-
- 3. Neither the name of the University nor the names of its contributors may
-    be used to endorse or promote products derived from this software without
-    specific prior written permission.
-
-THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS "AS IS", AND ANY
-EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
-WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, ARE
-DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY
-DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
-(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
-LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
-ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
-SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-=============================================================================*/
-
-#include <stdint.h>
-#include "platform.h"
-#include "specialize.h"
-#include "softfloat.h"
-
 /*----------------------------------------------------------------------------
-| Assuming the unsigned integer formed from concatenating `uiA64' and `uiA0'
+| This file intentionally contains no code.
-| has the bit pattern of an 80-bit extended floating-point NaN, converts
-| this NaN to the common NaN form, and stores the resulting common NaN at the
-| location pointed to by `zPtr'.  If the NaN is a signaling NaN, the invalid
-| exception is raised.
 *----------------------------------------------------------------------------*/
-void
- softfloat_extF80UIToCommonNaN(
-     uint_fast16_t uiA64, uint_fast64_t uiA0, struct commonNaN *zPtr )
-{
-
-    if ( softfloat_isSigNaNExtF80UI( uiA64, uiA0 ) ) {
-        softfloat_raiseFlags( softfloat_flag_invalid );
-    }
-    zPtr->sign = uiA64>>15;
-    zPtr->v64  = uiA0<<1;
-    zPtr->v0   = 0;
-
-}
 

softfloat/source/RISCV/s_f128MToCommonNaN.c

@@ -1,62 +1,5 @@
 
-/*============================================================================
-
-This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
-Package, Release 3e, by John R. Hauser.
-
-Copyright 2011, 2012, 2013, 2014 The Regents of the University of California.
-All rights reserved.
-
-Redistribution and use in source and binary forms, with or without
-modification, are permitted provided that the following conditions are met:
-
- 1. Redistributions of source code must retain the above copyright notice,
-    this list of conditions, and the following disclaimer.
-
- 2. Redistributions in binary form must reproduce the above copyright notice,
-    this list of conditions, and the following disclaimer in the documentation
-    and/or other materials provided with the distribution.
-
- 3. Neither the name of the University nor the names of its contributors may
-    be used to endorse or promote products derived from this software without
-    specific prior written permission.
-
-THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS "AS IS", AND ANY
-EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
-WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, ARE
-DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY
-DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
-(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
-LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
-ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
-SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-=============================================================================*/
-
-#include <stdint.h>
-#include "platform.h"
-#include "primitives.h"
-#include "specialize.h"
-#include "softfloat.h"
-
 /*----------------------------------------------------------------------------
-| Assuming the 128-bit floating-point value pointed to by `aWPtr' is a NaN,
+| This file intentionally contains no code.
-| converts this NaN to the common NaN form, and stores the resulting common
-| NaN at the location pointed to by `zPtr'.  If the NaN is a signaling NaN,
-| the invalid exception is raised.  Argument `aWPtr' points to an array of
-| four 32-bit elements that concatenate in the platform's normal endian order
-| to form a 128-bit floating-point value.
 *----------------------------------------------------------------------------*/
-void
- softfloat_f128MToCommonNaN( const uint32_t *aWPtr, struct commonNaN *zPtr )
-{
-
-    if ( f128M_isSignalingNaN( (const float128_t *) aWPtr ) ) {
-        softfloat_raiseFlags( softfloat_flag_invalid );
-    }
-    zPtr->sign = aWPtr[indexWordHi( 4 )]>>31;
-    softfloat_shortShiftLeft128M( aWPtr, 16, (uint32_t *) &zPtr->v0 );
-
-}
 

softfloat/source/RISCV/s_f128UIToCommonNaN.c

@@ -1,65 +1,5 @@
 
-/*============================================================================
-
-This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
-Package, Release 3e, by John R. Hauser.
-
-Copyright 2011, 2012, 2013, 2014 The Regents of the University of California.
-All rights reserved.
-
-Redistribution and use in source and binary forms, with or without
-modification, are permitted provided that the following conditions are met:
-
- 1. Redistributions of source code must retain the above copyright notice,
-    this list of conditions, and the following disclaimer.
-
- 2. Redistributions in binary form must reproduce the above copyright notice,
-    this list of conditions, and the following disclaimer in the documentation
-    and/or other materials provided with the distribution.
-
- 3. Neither the name of the University nor the names of its contributors may
-    be used to endorse or promote products derived from this software without
-    specific prior written permission.
-
-THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS "AS IS", AND ANY
-EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
-WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, ARE
-DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY
-DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
-(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
-LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
-ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
-SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-=============================================================================*/
-
-#include <stdint.h>
-#include "platform.h"
-#include "primitives.h"
-#include "specialize.h"
-#include "softfloat.h"
-
 /*----------------------------------------------------------------------------
-| Assuming the unsigned integer formed from concatenating `uiA64' and `uiA0'
+| This file intentionally contains no code.
-| has the bit pattern of a 128-bit floating-point NaN, converts this NaN to
-| the common NaN form, and stores the resulting common NaN at the location
-| pointed to by `zPtr'.  If the NaN is a signaling NaN, the invalid exception
-| is raised.
 *----------------------------------------------------------------------------*/
-void
- softfloat_f128UIToCommonNaN(
-     uint_fast64_t uiA64, uint_fast64_t uiA0, struct commonNaN *zPtr )
-{
-    struct uint128 NaNSig;
-
-    if ( softfloat_isSigNaNF128UI( uiA64, uiA0 ) ) {
-        softfloat_raiseFlags( softfloat_flag_invalid );
-    }
-    NaNSig = softfloat_shortShiftLeft128( uiA64, uiA0, 16 );
-    zPtr->sign = uiA64>>63;
-    zPtr->v64  = NaNSig.v64;
-    zPtr->v0   = NaNSig.v0;
-
-}
 

softfloat/source/RISCV/s_f16UIToCommonNaN.c

@@ -1,59 +1,5 @@
 
-/*============================================================================
-
-This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
-Package, Release 3e, by John R. Hauser.
-
-Copyright 2011, 2012, 2013, 2014, 2015 The Regents of the University of
-California.  All rights reserved.
-
-Redistribution and use in source and binary forms, with or without
-modification, are permitted provided that the following conditions are met:
-
- 1. Redistributions of source code must retain the above copyright notice,
-    this list of conditions, and the following disclaimer.
-
- 2. Redistributions in binary form must reproduce the above copyright notice,
-    this list of conditions, and the following disclaimer in the documentation
-    and/or other materials provided with the distribution.
-
- 3. Neither the name of the University nor the names of its contributors may
-    be used to endorse or promote products derived from this software without
-    specific prior written permission.
-
-THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS "AS IS", AND ANY
-EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
-WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, ARE
-DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY
-DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
-(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
-LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
-ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
-SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-=============================================================================*/
-
-#include <stdint.h>
-#include "platform.h"
-#include "specialize.h"
-#include "softfloat.h"
-
 /*----------------------------------------------------------------------------
-| Assuming `uiA' has the bit pattern of a 16-bit floating-point NaN, converts
+| This file intentionally contains no code.
-| this NaN to the common NaN form, and stores the resulting common NaN at the
-| location pointed to by `zPtr'.  If the NaN is a signaling NaN, the invalid
-| exception is raised.
 *----------------------------------------------------------------------------*/
-void softfloat_f16UIToCommonNaN( uint_fast16_t uiA, struct commonNaN *zPtr )
-{
-
-    if ( softfloat_isSigNaNF16UI( uiA ) ) {
-        softfloat_raiseFlags( softfloat_flag_invalid );
-    }
-    zPtr->sign = uiA>>15;
-    zPtr->v64  = (uint_fast64_t) uiA<<54;
-    zPtr->v0   = 0;
-
-}
 

softfloat/source/RISCV/s_f32UIToCommonNaN.c

@@ -1,59 +1,5 @@
 
-/*============================================================================
-
-This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
-Package, Release 3e, by John R. Hauser.
-
-Copyright 2011, 2012, 2013, 2014 The Regents of the University of California.
-All rights reserved.
-
-Redistribution and use in source and binary forms, with or without
-modification, are permitted provided that the following conditions are met:
-
- 1. Redistributions of source code must retain the above copyright notice,
-    this list of conditions, and the following disclaimer.
-
- 2. Redistributions in binary form must reproduce the above copyright notice,
-    this list of conditions, and the following disclaimer in the documentation
-    and/or other materials provided with the distribution.
-
- 3. Neither the name of the University nor the names of its contributors may
-    be used to endorse or promote products derived from this software without
-    specific prior written permission.
-
-THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS "AS IS", AND ANY
-EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
-WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, ARE
-DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY
-DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
-(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
-LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
-ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
-SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-=============================================================================*/
-
-#include <stdint.h>
-#include "platform.h"
-#include "specialize.h"
-#include "softfloat.h"
-
 /*----------------------------------------------------------------------------
-| Assuming `uiA' has the bit pattern of a 32-bit floating-point NaN, converts
+| This file intentionally contains no code.
-| this NaN to the common NaN form, and stores the resulting common NaN at the
-| location pointed to by `zPtr'.  If the NaN is a signaling NaN, the invalid
-| exception is raised.
 *----------------------------------------------------------------------------*/
-void softfloat_f32UIToCommonNaN( uint_fast32_t uiA, struct commonNaN *zPtr )
-{
-
-    if ( softfloat_isSigNaNF32UI( uiA ) ) {
-        softfloat_raiseFlags( softfloat_flag_invalid );
-    }
-    zPtr->sign = uiA>>31;
-    zPtr->v64  = (uint_fast64_t) uiA<<41;
-    zPtr->v0   = 0;
-
-}
 

softfloat/source/RISCV/s_f64UIToCommonNaN.c

@@ -1,59 +1,5 @@
 
-/*============================================================================
-
-This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
-Package, Release 3e, by John R. Hauser.
-
-Copyright 2011, 2012, 2013, 2014 The Regents of the University of California.
-All rights reserved.
-
-Redistribution and use in source and binary forms, with or without
-modification, are permitted provided that the following conditions are met:
-
- 1. Redistributions of source code must retain the above copyright notice,
-    this list of conditions, and the following disclaimer.
-
- 2. Redistributions in binary form must reproduce the above copyright notice,
-    this list of conditions, and the following disclaimer in the documentation
-    and/or other materials provided with the distribution.
-
- 3. Neither the name of the University nor the names of its contributors may
-    be used to endorse or promote products derived from this software without
-    specific prior written permission.
-
-THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS "AS IS", AND ANY
-EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
-WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, ARE
-DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY
-DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
-(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
-LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
-ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
-SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-=============================================================================*/
-
-#include <stdint.h>
-#include "platform.h"
-#include "specialize.h"
-#include "softfloat.h"
-
 /*----------------------------------------------------------------------------
-| Assuming `uiA' has the bit pattern of a 64-bit floating-point NaN, converts
+| This file intentionally contains no code.
-| this NaN to the common NaN form, and stores the resulting common NaN at the
-| location pointed to by `zPtr'.  If the NaN is a signaling NaN, the invalid
-| exception is raised.
 *----------------------------------------------------------------------------*/
-void softfloat_f64UIToCommonNaN( uint_fast64_t uiA, struct commonNaN *zPtr )
-{
-
-    if ( softfloat_isSigNaNF64UI( uiA ) ) {
-        softfloat_raiseFlags( softfloat_flag_invalid );
-    }
-    zPtr->sign = uiA>>63;
-    zPtr->v64  = uiA<<12;
-    zPtr->v0   = 0;
-
-}
 

softfloat/source/RISCV/s_propagateNaNExtF80M.c

@@ -4,8 +4,8 @@
 This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
 Package, Release 3e, by John R. Hauser.
 
-Copyright 2011, 2012, 2013, 2014 The Regents of the University of California.
+Copyright 2011, 2012, 2013, 2014, 2015 The Regents of the University of
-All rights reserved.
+California.  All rights reserved.
 
 Redistribution and use in source and binary forms, with or without
 modification, are permitted provided that the following conditions are met:
@@ -34,10 +34,9 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 =============================================================================*/
 
-#include <stdbool.h>
 #include <stdint.h>
 #include "platform.h"
-#include "internals.h"
+#include "primitiveTypes.h"
 #include "specialize.h"
 #include "softfloat.h"
 
@@ -54,54 +53,22 @@ void
      struct extFloat80M *zSPtr
  )
 {
-    bool isSigNaNA;
+    uint_fast16_t ui64;
-    const struct extFloat80M *sPtr;
+    uint_fast64_t ui0;
-    bool isSigNaNB;
-    uint_fast16_t uiB64;
-    uint64_t uiB0;
-    uint_fast16_t uiA64;
-    uint64_t uiA0;
-    uint_fast16_t uiMagA64, uiMagB64;
 
-    isSigNaNA = extF80M_isSignalingNaN( (const extFloat80_t *) aSPtr );
+    ui64 = aSPtr->signExp;
-    sPtr = aSPtr;
+    ui0  = aSPtr->signif;
-    if ( ! bSPtr ) {
+    if (
-        if ( isSigNaNA ) softfloat_raiseFlags( softfloat_flag_invalid );
+        softfloat_isSigNaNExtF80UI( ui64, ui0 )
-        goto copy;
+            || (bSPtr
-    }
+                    && (ui64 = bSPtr->signExp,
-    isSigNaNB = extF80M_isSignalingNaN( (const extFloat80_t *) bSPtr );
+                        ui0  = bSPtr->signif,
-    if ( isSigNaNA | isSigNaNB ) {
+                        softfloat_isSigNaNExtF80UI( ui64, ui0 )))
+    ) {
         softfloat_raiseFlags( softfloat_flag_invalid );
-        if ( isSigNaNA ) {
-            uiB64 = bSPtr->signExp;
-            if ( isSigNaNB ) goto returnLargerUIMag;
-            uiB0 = bSPtr->signif;
-            if ( isNaNExtF80UI( uiB64, uiB0 ) ) goto copyB;
-            goto copy;
-        } else {
-            uiA64 = aSPtr->signExp;
-            uiA0 = aSPtr->signif;
-            if ( isNaNExtF80UI( uiA64, uiA0 ) ) goto copy;
-            goto copyB;
     }
-    }
+    zSPtr->signExp = defaultNaNExtF80UI64;
-    uiB64 = bSPtr->signExp;
+    zSPtr->signif  = defaultNaNExtF80UI0;
- returnLargerUIMag:
-    uiA64 = aSPtr->signExp;
-    uiMagA64 = uiA64 & 0x7FFF;
-    uiMagB64 = uiB64 & 0x7FFF;
-    if ( uiMagA64 < uiMagB64 ) goto copyB;
-    if ( uiMagB64 < uiMagA64 ) goto copy;
-    uiA0 = aSPtr->signif;
-    uiB0 = bSPtr->signif;
-    if ( uiA0 < uiB0 ) goto copyB;
-    if ( uiB0 < uiA0 ) goto copy;
-    if ( uiA64 < uiB64 ) goto copy;
- copyB:
-    sPtr = bSPtr;
- copy:
-    zSPtr->signExp = sPtr->signExp;
-    zSPtr->signif = sPtr->signif | UINT64_C( 0xC000000000000000 );
 
 }
 

softfloat/source/RISCV/s_propagateNaNExtF80UI.c

@@ -4,7 +4,7 @@
 This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
 Package, Release 3e, by John R. Hauser.
 
-Copyright 2011, 2012, 2013, 2014, 2018 The Regents of the University of
+Copyright 2011, 2012, 2013, 2014, 2015 The Regents of the University of
 California.  All rights reserved.
 
 Redistribution and use in source and binary forms, with or without
@@ -34,17 +34,16 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 =============================================================================*/
 
-#include <stdbool.h>
 #include <stdint.h>
 #include "platform.h"
-#include "internals.h"
+#include "primitiveTypes.h"
 #include "specialize.h"
 #include "softfloat.h"
 
 /*----------------------------------------------------------------------------
-| Interpreting the unsigned integer formed from concatenating 'uiA64' and
+| Interpreting the unsigned integer formed from concatenating `uiA64' and
-| 'uiA0' as an 80-bit extended floating-point value, and likewise interpreting
+| `uiA0' as an 80-bit extended floating-point value, and likewise interpreting
-| the unsigned integer formed from concatenating 'uiB64' and 'uiB0' as another
+| the unsigned integer formed from concatenating `uiB64' and `uiB0' as another
 | 80-bit extended floating-point value, and assuming at least on of these
 | floating-point values is a NaN, returns the bit pattern of the combined NaN
 | result.  If either original floating-point value is a signaling NaN, the
@@ -58,48 +57,16 @@ struct uint128
      uint_fast64_t uiB0
  )
 {
-    bool isSigNaNA, isSigNaNB;
-    uint_fast64_t uiNonsigA0, uiNonsigB0;
-    uint_fast16_t uiMagA64, uiMagB64;
     struct uint128 uiZ;
 
-    /*------------------------------------------------------------------------
+    if (
-    *------------------------------------------------------------------------*/
+           softfloat_isSigNaNExtF80UI( uiA64, uiA0 )
-    isSigNaNA = softfloat_isSigNaNExtF80UI( uiA64, uiA0 );
+        || softfloat_isSigNaNExtF80UI( uiB64, uiB0 )
-    isSigNaNB = softfloat_isSigNaNExtF80UI( uiB64, uiB0 );
+    ) {
-    /*------------------------------------------------------------------------
-    | Make NaNs non-signaling.
-    *------------------------------------------------------------------------*/
-    uiNonsigA0 = uiA0 | UINT64_C( 0xC000000000000000 );
-    uiNonsigB0 = uiB0 | UINT64_C( 0xC000000000000000 );
-    /*------------------------------------------------------------------------
-    *------------------------------------------------------------------------*/
-    if ( isSigNaNA | isSigNaNB ) {
         softfloat_raiseFlags( softfloat_flag_invalid );
-        if ( isSigNaNA ) {
-            if ( isSigNaNB ) goto returnLargerMag;
-            if ( isNaNExtF80UI( uiB64, uiB0 ) ) goto returnB;
-            goto returnA;
-        } else {
-            if ( isNaNExtF80UI( uiA64, uiA0 ) ) goto returnA;
-            goto returnB;
     }
-    }
+    uiZ.v64 = defaultNaNExtF80UI64;
- returnLargerMag:
+    uiZ.v0  = defaultNaNExtF80UI0;
-    uiMagA64 = uiA64 & 0x7FFF;
-    uiMagB64 = uiB64 & 0x7FFF;
-    if ( uiMagA64 < uiMagB64 ) goto returnB;
-    if ( uiMagB64 < uiMagA64 ) goto returnA;
-    if ( uiA0 < uiB0 ) goto returnB;
-    if ( uiB0 < uiA0 ) goto returnA;
-    if ( uiA64 < uiB64 ) goto returnA;
- returnB:
-    uiZ.v64 = uiB64;
-    uiZ.v0  = uiNonsigB0;
-    return uiZ;
- returnA:
-    uiZ.v64 = uiA64;
-    uiZ.v0  = uiNonsigA0;
     return uiZ;
 
 }

softfloat/source/RISCV/s_propagateNaNF128M.c

@@ -4,8 +4,8 @@
 This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
 Package, Release 3e, by John R. Hauser.
 
-Copyright 2011, 2012, 2013, 2014 The Regents of the University of California.
+Copyright 2011, 2012, 2013, 2014, 2015, 2018 The Regents of the University of
-All rights reserved.
+California.  All rights reserved.
 
 Redistribution and use in source and binary forms, with or without
 modification, are permitted provided that the following conditions are met:
@@ -34,43 +34,35 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 =============================================================================*/
 
-#include <stdbool.h>
 #include <stdint.h>
 #include "platform.h"
-#include "internals.h"
+#include "primitiveTypes.h"
 #include "specialize.h"
 #include "softfloat.h"
 
 /*----------------------------------------------------------------------------
 | Assuming at least one of the two 128-bit floating-point values pointed to by
-| `aWPtr' and `bWPtr' is a NaN, stores the combined NaN result at the location
+| 'aWPtr' and 'bWPtr' is a NaN, stores the combined NaN result at the location
-| pointed to by `zWPtr'.  If either original floating-point value is a
+| pointed to by 'zWPtr'.  If either original floating-point value is a
-| signaling NaN, the invalid exception is raised.  Each of `aWPtr', `bWPtr',
+| signaling NaN, the invalid exception is raised.  Each of 'aWPtr', 'bWPtr',
-| and `zWPtr' points to an array of four 32-bit elements that concatenate in
+| and 'zWPtr' points to an array of four 32-bit elements that concatenate in
 | the platform's normal endian order to form a 128-bit floating-point value.
 *----------------------------------------------------------------------------*/
 void
  softfloat_propagateNaNF128M(
      const uint32_t *aWPtr, const uint32_t *bWPtr, uint32_t *zWPtr )
 {
-    bool isSigNaNA;
-    const uint32_t *ptr;
 
-    ptr = aWPtr;
-    isSigNaNA = f128M_isSignalingNaN( (const float128_t *) aWPtr );
     if (
-        isSigNaNA
+        f128M_isSignalingNaN( (const float128_t *) aWPtr )
             || (bWPtr && f128M_isSignalingNaN( (const float128_t *) bWPtr ))
     ) {
         softfloat_raiseFlags( softfloat_flag_invalid );
-        if ( isSigNaNA ) goto copy;
     }
-    if ( ! softfloat_isNaNF128M( aWPtr ) ) ptr = bWPtr;
+    zWPtr[indexWord( 4, 3 )] = defaultNaNF128UI96;
- copy:
+    zWPtr[indexWord( 4, 2 )] = defaultNaNF128UI64;
-    zWPtr[indexWordHi( 4 )] = ptr[indexWordHi( 4 )] | 0x00008000;
+    zWPtr[indexWord( 4, 1 )] = defaultNaNF128UI32;
-    zWPtr[indexWord( 4, 2 )] = ptr[indexWord( 4, 2 )];
+    zWPtr[indexWord( 4, 0 )] = defaultNaNF128UI0;
-    zWPtr[indexWord( 4, 1 )] = ptr[indexWord( 4, 1 )];
-    zWPtr[indexWord( 4, 0 )] = ptr[indexWord( 4, 0 )];
 
 }
 

softfloat/source/RISCV/s_propagateNaNF128UI.c

@@ -4,8 +4,8 @@
 This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
 Package, Release 3e, by John R. Hauser.
 
-Copyright 2011, 2012, 2013, 2014 The Regents of the University of California.
+Copyright 2011, 2012, 2013, 2014, 2015 The Regents of the University of
-All rights reserved.
+California.  All rights reserved.
 
 Redistribution and use in source and binary forms, with or without
 modification, are permitted provided that the following conditions are met:
@@ -34,10 +34,9 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 =============================================================================*/
 
-#include <stdbool.h>
 #include <stdint.h>
 #include "platform.h"
-#include "internals.h"
+#include "primitiveTypes.h"
 #include "specialize.h"
 #include "softfloat.h"
 
@@ -58,23 +57,16 @@ struct uint128
      uint_fast64_t uiB0
  )
 {
-    bool isSigNaNA;
     struct uint128 uiZ;
 
-    isSigNaNA = softfloat_isSigNaNF128UI( uiA64, uiA0 );
+    if (
-    if ( isSigNaNA || softfloat_isSigNaNF128UI( uiB64, uiB0 ) ) {
+           softfloat_isSigNaNF128UI( uiA64, uiA0 )
+        || softfloat_isSigNaNF128UI( uiB64, uiB0 )
+    ) {
         softfloat_raiseFlags( softfloat_flag_invalid );
-        if ( isSigNaNA ) goto returnNonsigA;
     }
-    if ( isNaNF128UI( uiA64, uiA0 ) ) {
+    uiZ.v64 = defaultNaNF128UI64;
- returnNonsigA:
+    uiZ.v0  = defaultNaNF128UI0;
-        uiZ.v64 = uiA64;
-        uiZ.v0  = uiA0;
-    } else {
-        uiZ.v64 = uiB64;
-        uiZ.v0  = uiB0;
-    }
-    uiZ.v64 |= UINT64_C( 0x0000800000000000 );
     return uiZ;
 
 }

softfloat/source/RISCV/s_propagateNaNF16UI.c

@@ -4,7 +4,7 @@
 This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
 Package, Release 3e, by John R. Hauser.
 
-Copyright 2011, 2012, 2013, 2014, 2015 The Regents of the University of
+Copyright 2011, 2012, 2013, 2014, 2015, 2016 The Regents of the University of
 California.  All rights reserved.
 
 Redistribution and use in source and binary forms, with or without
@@ -34,10 +34,8 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 =============================================================================*/
 
-#include <stdbool.h>
 #include <stdint.h>
 #include "platform.h"
-#include "internals.h"
 #include "specialize.h"
 #include "softfloat.h"
 
@@ -50,14 +48,11 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 uint_fast16_t
  softfloat_propagateNaNF16UI( uint_fast16_t uiA, uint_fast16_t uiB )
 {
-    bool isSigNaNA;
 
-    isSigNaNA = softfloat_isSigNaNF16UI( uiA );
+    if ( softfloat_isSigNaNF16UI( uiA ) || softfloat_isSigNaNF16UI( uiB ) ) {
-    if ( isSigNaNA || softfloat_isSigNaNF16UI( uiB ) ) {
         softfloat_raiseFlags( softfloat_flag_invalid );
-        if ( isSigNaNA ) return uiA | 0x0200;
     }
-    return (isNaNF16UI( uiA ) ? uiA : uiB) | 0x0200;
+    return defaultNaNF16UI;
 
 }
 

softfloat/source/RISCV/s_propagateNaNF32UI.c

@@ -4,8 +4,8 @@
 This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
 Package, Release 3e, by John R. Hauser.
 
-Copyright 2011, 2012, 2013, 2014 The Regents of the University of California.
+Copyright 2011, 2012, 2013, 2014, 2015 The Regents of the University of
-All rights reserved.
+California.  All rights reserved.
 
 Redistribution and use in source and binary forms, with or without
 modification, are permitted provided that the following conditions are met:
@@ -34,10 +34,8 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 =============================================================================*/
 
-#include <stdbool.h>
 #include <stdint.h>
 #include "platform.h"
-#include "internals.h"
 #include "specialize.h"
 #include "softfloat.h"
 
@@ -50,14 +48,11 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 uint_fast32_t
  softfloat_propagateNaNF32UI( uint_fast32_t uiA, uint_fast32_t uiB )
 {
-    bool isSigNaNA;
 
-    isSigNaNA = softfloat_isSigNaNF32UI( uiA );
+    if ( softfloat_isSigNaNF32UI( uiA ) || softfloat_isSigNaNF32UI( uiB ) ) {
-    if ( isSigNaNA || softfloat_isSigNaNF32UI( uiB ) ) {
         softfloat_raiseFlags( softfloat_flag_invalid );
-        if ( isSigNaNA ) return uiA | 0x00400000;
     }
-    return (isNaNF32UI( uiA ) ? uiA : uiB) | 0x00400000;
+    return defaultNaNF32UI;
 
 }
 

softfloat/source/RISCV/s_propagateNaNF64UI.c

@@ -4,8 +4,8 @@
 This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
 Package, Release 3e, by John R. Hauser.
 
-Copyright 2011, 2012, 2013, 2014 The Regents of the University of California.
+Copyright 2011, 2012, 2013, 2014, 2015 The Regents of the University of
-All rights reserved.
+California.  All rights reserved.
 
 Redistribution and use in source and binary forms, with or without
 modification, are permitted provided that the following conditions are met:
@@ -34,10 +34,8 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 =============================================================================*/
 
-#include <stdbool.h>
 #include <stdint.h>
 #include "platform.h"
-#include "internals.h"
 #include "specialize.h"
 #include "softfloat.h"
 
@@ -50,14 +48,11 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 uint_fast64_t
  softfloat_propagateNaNF64UI( uint_fast64_t uiA, uint_fast64_t uiB )
 {
-    bool isSigNaNA;
 
-    isSigNaNA = softfloat_isSigNaNF64UI( uiA );
+    if ( softfloat_isSigNaNF64UI( uiA ) || softfloat_isSigNaNF64UI( uiB ) ) {
-    if ( isSigNaNA || softfloat_isSigNaNF64UI( uiB ) ) {
         softfloat_raiseFlags( softfloat_flag_invalid );
-        if ( isSigNaNA ) return uiA | UINT64_C( 0x0008000000000000 );
     }
-    return (isNaNF64UI( uiA ) ? uiA : uiB) | UINT64_C( 0x0008000000000000 );
+    return defaultNaNF64UI;
 
 }
 

softfloat/source/RISCV/specialize.h

@@ -37,10 +37,10 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #ifndef specialize_h
 #define specialize_h 1
 
-#include <stdbool.h>
-#include <stdint.h>
 #include "primitiveTypes.h"
 #include "softfloat.h"
+#include <stdbool.h>
+#include <stdint.h>
 
 /*----------------------------------------------------------------------------
 | Default value for 'softfloat_detectTininess'.
@@ -51,48 +51,50 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 | The values to return on conversions to 32-bit integer formats that raise an
 | invalid exception.
 *----------------------------------------------------------------------------*/
-#define ui32_fromPosOverflow UINT32_C(0xFFFFFFFF)
+#define ui32_fromPosOverflow 0xFFFFFFFF
-#define ui32_fromNegOverflow UINT32_C(0x0)
+#define ui32_fromNegOverflow 0
-#define ui32_fromNaN         UINT32_C(0xFFFFFFFF)
+#define ui32_fromNaN 0xFFFFFFFF
-#define i32_fromPosOverflow   INT64_C(0x7FFFFFFF)
+#define i32_fromPosOverflow 0x7FFFFFFF
-#define i32_fromNegOverflow  (-INT64_C(0x7FFFFFFF)-1)
+#define i32_fromNegOverflow (-0x7FFFFFFF - 1)
-#define i32_fromNaN           INT64_C(0x7FFFFFFF)
+#define i32_fromNaN 0x7FFFFFFF
 
 /*----------------------------------------------------------------------------
 | The values to return on conversions to 64-bit integer formats that raise an
 | invalid exception.
 *----------------------------------------------------------------------------*/
-#define ui64_fromPosOverflow UINT64_C( 0xFFFFFFFFFFFFFFFF )
+#define ui64_fromPosOverflow UINT64_C(0xFFFFFFFFFFFFFFFF)
-#define ui64_fromNegOverflow UINT64_C( 0x0 )
+#define ui64_fromNegOverflow 0
-#define ui64_fromNaN         UINT64_C( 0xFFFFFFFFFFFFFFFF)
+#define ui64_fromNaN UINT64_C(0xFFFFFFFFFFFFFFFF)
-#define i64_fromPosOverflow   INT64_C( 0x7FFFFFFFFFFFFFFF)
+#define i64_fromPosOverflow INT64_C(0x7FFFFFFFFFFFFFFF)
-#define i64_fromNegOverflow  (-INT64_C( 0x7FFFFFFFFFFFFFFF)-1)
+#define i64_fromNegOverflow (-INT64_C(0x7FFFFFFFFFFFFFFF) - 1)
-#define i64_fromNaN           INT64_C( 0x7FFFFFFFFFFFFFFF)
+#define i64_fromNaN INT64_C(0x7FFFFFFFFFFFFFFF)
 
 /*----------------------------------------------------------------------------
 | "Common NaN" structure, used to transfer NaN representations from one format
 | to another.
 *----------------------------------------------------------------------------*/
 struct commonNaN {
-    bool sign;
+    char _unused;
-#ifdef LITTLEENDIAN
-    uint64_t v0, v64;
-#else
-    uint64_t v64, v0;
-#endif
 };
 
 /*----------------------------------------------------------------------------
 | The bit pattern for a default generated 16-bit floating-point NaN.
 *----------------------------------------------------------------------------*/
-#define defaultNaNF16UI 0xFE00
+#define defaultNaNF16UI 0x7E00
 
 /*----------------------------------------------------------------------------
 | Returns true when 16-bit unsigned integer 'uiA' has the bit pattern of a
 | 16-bit floating-point signaling NaN.
 | Note:  This macro evaluates its argument more than once.
 *----------------------------------------------------------------------------*/
-#define softfloat_isSigNaNF16UI( uiA ) ((((uiA) & 0x7E00) == 0x7C00) && ((uiA) & 0x01FF))
+#define softfloat_isSigNaNF16UI(uiA) ((((uiA)&0x7E00) == 0x7C00) && ((uiA)&0x01FF))
+
+/*----------------------------------------------------------------------------
+| Returns true when 16-bit unsigned integer 'uiA' has the bit pattern of a
+| 16-bit brain floating-point (BF16) signaling NaN.
+| Note:  This macro evaluates its argument more than once.
+*----------------------------------------------------------------------------*/
+#define softfloat_isSigNaNBF16UI(uiA) ((((uiA)&0x7FC0) == 0x7F80) && ((uiA)&0x003F))
 
 /*----------------------------------------------------------------------------
 | Assuming 'uiA' has the bit pattern of a 16-bit floating-point NaN, converts
@@ -100,13 +102,25 @@ struct commonNaN {
 | location pointed to by 'zPtr'.  If the NaN is a signaling NaN, the invalid
 | exception is raised.
 *----------------------------------------------------------------------------*/
-void softfloat_f16UIToCommonNaN( uint_fast16_t uiA, struct commonNaN *zPtr );
+#define softfloat_f16UIToCommonNaN(uiA, zPtr)                                                                                              \
+    if(!((uiA)&0x0200))                                                                                                                    \
+    softfloat_raiseFlags(softfloat_flag_invalid)
+
+/*----------------------------------------------------------------------------
+| Assuming 'uiA' has the bit pattern of a 16-bit BF16 floating-point NaN, converts
+| this NaN to the common NaN form, and stores the resulting common NaN at the
+| location pointed to by 'zPtr'.  If the NaN is a signaling NaN, the invalid
+| exception is raised.
+*----------------------------------------------------------------------------*/
+#define softfloat_bf16UIToCommonNaN(uiA, zPtr)                                                                                             \
+    if(!((uiA)&0x0040))                                                                                                                    \
+    softfloat_raiseFlags(softfloat_flag_invalid)
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into a 16-bit floating-point
 | NaN, and returns the bit pattern of this value as an unsigned integer.
 *----------------------------------------------------------------------------*/
-uint_fast16_t softfloat_commonNaNToF16UI( const struct commonNaN *aPtr );
+#define softfloat_commonNaNToF16UI(aPtr) ((uint_fast16_t)defaultNaNF16UI)
 
 /*----------------------------------------------------------------------------
 | Interpreting 'uiA' and 'uiB' as the bit patterns of two 16-bit floating-
@@ -114,8 +128,18 @@ uint_fast16_t softfloat_commonNaNToF16UI( const struct commonNaN *aPtr );
 | the combined NaN result.  If either 'uiA' or 'uiB' has the pattern of a
 | signaling NaN, the invalid exception is raised.
 *----------------------------------------------------------------------------*/
-uint_fast16_t
+uint_fast16_t softfloat_propagateNaNF16UI(uint_fast16_t uiA, uint_fast16_t uiB);
- softfloat_propagateNaNF16UI( uint_fast16_t uiA, uint_fast16_t uiB );
+
+/*----------------------------------------------------------------------------
+| The bit pattern for a default generated 16-bit BF16 floating-point NaN.
+*----------------------------------------------------------------------------*/
+#define defaultNaNBF16UI 0x7FC0
+
+/*----------------------------------------------------------------------------
+| Converts the common NaN pointed to by 'aPtr' into a 16-bit floating-point
+| NaN, and returns the bit pattern of this value as an unsigned integer.
+*----------------------------------------------------------------------------*/
+#define softfloat_commonNaNToBF16UI(aPtr) ((uint_fast16_t)defaultNaNBF16UI)
 
 /*----------------------------------------------------------------------------
 | The bit pattern for a default generated 32-bit floating-point NaN.
@@ -127,7 +151,7 @@ uint_fast16_t
 | 32-bit floating-point signaling NaN.
 | Note:  This macro evaluates its argument more than once.
 *----------------------------------------------------------------------------*/
-#define softfloat_isSigNaNF32UI( uiA ) ((((uiA) & 0x7FC00000) == 0x7F800000) && ((uiA) & 0x003FFFFF))
+#define softfloat_isSigNaNF32UI(uiA) ((((uiA)&0x7FC00000) == 0x7F800000) && ((uiA)&0x003FFFFF))
 
 /*----------------------------------------------------------------------------
 | Assuming 'uiA' has the bit pattern of a 32-bit floating-point NaN, converts
@@ -135,13 +159,15 @@ uint_fast16_t
 | location pointed to by 'zPtr'.  If the NaN is a signaling NaN, the invalid
 | exception is raised.
 *----------------------------------------------------------------------------*/
-void softfloat_f32UIToCommonNaN( uint_fast32_t uiA, struct commonNaN *zPtr );
+#define softfloat_f32UIToCommonNaN(uiA, zPtr)                                                                                              \
+    if(!((uiA)&0x00400000))                                                                                                                \
+    softfloat_raiseFlags(softfloat_flag_invalid)
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into a 32-bit floating-point
 | NaN, and returns the bit pattern of this value as an unsigned integer.
 *----------------------------------------------------------------------------*/
-uint_fast32_t softfloat_commonNaNToF32UI( const struct commonNaN *aPtr );
+#define softfloat_commonNaNToF32UI(aPtr) ((uint_fast32_t)defaultNaNF32UI)
 
 /*----------------------------------------------------------------------------
 | Interpreting 'uiA' and 'uiB' as the bit patterns of two 32-bit floating-
@@ -149,20 +175,20 @@ uint_fast32_t softfloat_commonNaNToF32UI( const struct commonNaN *aPtr );
 | the combined NaN result.  If either 'uiA' or 'uiB' has the pattern of a
 | signaling NaN, the invalid exception is raised.
 *----------------------------------------------------------------------------*/
-uint_fast32_t
+uint_fast32_t softfloat_propagateNaNF32UI(uint_fast32_t uiA, uint_fast32_t uiB);
- softfloat_propagateNaNF32UI( uint_fast32_t uiA, uint_fast32_t uiB );
 
 /*----------------------------------------------------------------------------
 | The bit pattern for a default generated 64-bit floating-point NaN.
 *----------------------------------------------------------------------------*/
-#define defaultNaNF64UI UINT64_C( 0x7FF8000000000000 )
+#define defaultNaNF64UI UINT64_C(0x7FF8000000000000)
 
 /*----------------------------------------------------------------------------
 | Returns true when 64-bit unsigned integer 'uiA' has the bit pattern of a
 | 64-bit floating-point signaling NaN.
 | Note:  This macro evaluates its argument more than once.
 *----------------------------------------------------------------------------*/
-#define softfloat_isSigNaNF64UI( uiA ) ((((uiA) & UINT64_C( 0x7FF8000000000000 )) == UINT64_C( 0x7FF0000000000000 )) && ((uiA) & UINT64_C( 0x0007FFFFFFFFFFFF )))
+#define softfloat_isSigNaNF64UI(uiA)                                                                                                       \
+    ((((uiA)&UINT64_C(0x7FF8000000000000)) == UINT64_C(0x7FF0000000000000)) && ((uiA)&UINT64_C(0x0007FFFFFFFFFFFF)))
 
 /*----------------------------------------------------------------------------
 | Assuming 'uiA' has the bit pattern of a 64-bit floating-point NaN, converts
@@ -170,13 +196,15 @@ uint_fast32_t
 | location pointed to by 'zPtr'.  If the NaN is a signaling NaN, the invalid
 | exception is raised.
 *----------------------------------------------------------------------------*/
-void softfloat_f64UIToCommonNaN( uint_fast64_t uiA, struct commonNaN *zPtr );
+#define softfloat_f64UIToCommonNaN(uiA, zPtr)                                                                                              \
+    if(!((uiA)&UINT64_C(0x0008000000000000)))                                                                                              \
+    softfloat_raiseFlags(softfloat_flag_invalid)
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into a 64-bit floating-point
 | NaN, and returns the bit pattern of this value as an unsigned integer.
 *----------------------------------------------------------------------------*/
-uint_fast64_t softfloat_commonNaNToF64UI( const struct commonNaN *aPtr );
+#define softfloat_commonNaNToF64UI(aPtr) ((uint_fast64_t)defaultNaNF64UI)
 
 /*----------------------------------------------------------------------------
 | Interpreting 'uiA' and 'uiB' as the bit patterns of two 64-bit floating-
@@ -184,14 +212,13 @@ uint_fast64_t softfloat_commonNaNToF64UI( const struct commonNaN *aPtr );
 | the combined NaN result.  If either 'uiA' or 'uiB' has the pattern of a
 | signaling NaN, the invalid exception is raised.
 *----------------------------------------------------------------------------*/
-uint_fast64_t
+uint_fast64_t softfloat_propagateNaNF64UI(uint_fast64_t uiA, uint_fast64_t uiB);
- softfloat_propagateNaNF64UI( uint_fast64_t uiA, uint_fast64_t uiB );
 
 /*----------------------------------------------------------------------------
 | The bit pattern for a default generated 80-bit extended floating-point NaN.
 *----------------------------------------------------------------------------*/
-#define defaultNaNExtF80UI64 0xFFFF
+#define defaultNaNExtF80UI64 0x7FFF
-#define defaultNaNExtF80UI0  UINT64_C( 0xC000000000000000 )
+#define defaultNaNExtF80UI0 UINT64_C(0xC000000000000000)
 
 /*----------------------------------------------------------------------------
 | Returns true when the 80-bit unsigned integer formed from concatenating
@@ -199,7 +226,8 @@ uint_fast64_t
 | floating-point signaling NaN.
 | Note:  This macro evaluates its arguments more than once.
 *----------------------------------------------------------------------------*/
-#define softfloat_isSigNaNExtF80UI( uiA64, uiA0 ) ((((uiA64) & 0x7FFF) == 0x7FFF) && ! ((uiA0) & UINT64_C( 0x4000000000000000 )) && ((uiA0) & UINT64_C( 0x3FFFFFFFFFFFFFFF )))
+#define softfloat_isSigNaNExtF80UI(uiA64, uiA0)                                                                                            \
+    ((((uiA64)&0x7FFF) == 0x7FFF) && !((uiA0)&UINT64_C(0x4000000000000000)) && ((uiA0)&UINT64_C(0x3FFFFFFFFFFFFFFF)))
 
 #ifdef SOFTFLOAT_FAST_INT64
 
@@ -215,16 +243,26 @@ uint_fast64_t
 | location pointed to by 'zPtr'.  If the NaN is a signaling NaN, the invalid
 | exception is raised.
 *----------------------------------------------------------------------------*/
-void
+#define softfloat_extF80UIToCommonNaN(uiA64, uiA0, zPtr)                                                                                   \
- softfloat_extF80UIToCommonNaN(
+    if(!((uiA0)&UINT64_C(0x4000000000000000)))                                                                                             \
-     uint_fast16_t uiA64, uint_fast64_t uiA0, struct commonNaN *zPtr );
+    softfloat_raiseFlags(softfloat_flag_invalid)
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into an 80-bit extended
 | floating-point NaN, and returns the bit pattern of this value as an unsigned
 | integer.
 *----------------------------------------------------------------------------*/
-struct uint128 softfloat_commonNaNToExtF80UI( const struct commonNaN *aPtr );
+#if defined INLINE && !defined softfloat_commonNaNToExtF80UI
+INLINE
+struct uint128 softfloat_commonNaNToExtF80UI(const struct commonNaN* aPtr) {
+    struct uint128 uiZ;
+    uiZ.v64 = defaultNaNExtF80UI64;
+    uiZ.v0 = defaultNaNExtF80UI0;
+    return uiZ;
+}
+#else
+struct uint128 softfloat_commonNaNToExtF80UI(const struct commonNaN* aPtr);
+#endif
 
 /*----------------------------------------------------------------------------
 | Interpreting the unsigned integer formed from concatenating 'uiA64' and
@@ -235,19 +273,13 @@ struct uint128 softfloat_commonNaNToExtF80UI( const struct commonNaN *aPtr );
 | result.  If either original floating-point value is a signaling NaN, the
 | invalid exception is raised.
 *----------------------------------------------------------------------------*/
-struct uint128
+struct uint128 softfloat_propagateNaNExtF80UI(uint_fast16_t uiA64, uint_fast64_t uiA0, uint_fast16_t uiB64, uint_fast64_t uiB0);
- softfloat_propagateNaNExtF80UI(
-     uint_fast16_t uiA64,
-     uint_fast64_t uiA0,
-     uint_fast16_t uiB64,
-     uint_fast64_t uiB0
- );
 
 /*----------------------------------------------------------------------------
 | The bit pattern for a default generated 128-bit floating-point NaN.
 *----------------------------------------------------------------------------*/
-#define defaultNaNF128UI64 UINT64_C( 0xFFFF800000000000 )
+#define defaultNaNF128UI64 UINT64_C(0x7FFF800000000000)
-#define defaultNaNF128UI0  UINT64_C( 0 )
+#define defaultNaNF128UI0 UINT64_C(0)
 
 /*----------------------------------------------------------------------------
 | Returns true when the 128-bit unsigned integer formed from concatenating
@@ -255,7 +287,8 @@ struct uint128
 | point signaling NaN.
 | Note:  This macro evaluates its arguments more than once.
 *----------------------------------------------------------------------------*/
-#define softfloat_isSigNaNF128UI( uiA64, uiA0 ) ((((uiA64) & UINT64_C( 0x7FFF800000000000 )) == UINT64_C( 0x7FFF000000000000 )) && ((uiA0) || ((uiA64) & UINT64_C( 0x00007FFFFFFFFFFF ))))
+#define softfloat_isSigNaNF128UI(uiA64, uiA0)                                                                                              \
+    ((((uiA64)&UINT64_C(0x7FFF800000000000)) == UINT64_C(0x7FFF000000000000)) && ((uiA0) || ((uiA64)&UINT64_C(0x00007FFFFFFFFFFF))))
 
 /*----------------------------------------------------------------------------
 | Assuming the unsigned integer formed from concatenating 'uiA64' and 'uiA0'
@@ -264,15 +297,25 @@ struct uint128
 | pointed to by 'zPtr'.  If the NaN is a signaling NaN, the invalid exception
 | is raised.
 *----------------------------------------------------------------------------*/
-void
+#define softfloat_f128UIToCommonNaN(uiA64, uiA0, zPtr)                                                                                     \
- softfloat_f128UIToCommonNaN(
+    if(!((uiA64)&UINT64_C(0x0000800000000000)))                                                                                            \
-     uint_fast64_t uiA64, uint_fast64_t uiA0, struct commonNaN *zPtr );
+    softfloat_raiseFlags(softfloat_flag_invalid)
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into a 128-bit floating-point
 | NaN, and returns the bit pattern of this value as an unsigned integer.
 *----------------------------------------------------------------------------*/
-struct uint128 softfloat_commonNaNToF128UI( const struct commonNaN * );
+#if defined INLINE && !defined softfloat_commonNaNToF128UI
+INLINE
+struct uint128 softfloat_commonNaNToF128UI(const struct commonNaN* aPtr) {
+    struct uint128 uiZ;
+    uiZ.v64 = defaultNaNF128UI64;
+    uiZ.v0 = defaultNaNF128UI0;
+    return uiZ;
+}
+#else
+struct uint128 softfloat_commonNaNToF128UI(const struct commonNaN*);
+#endif
 
 /*----------------------------------------------------------------------------
 | Interpreting the unsigned integer formed from concatenating 'uiA64' and
@@ -283,13 +326,7 @@ struct uint128 softfloat_commonNaNToF128UI( const struct commonNaN * );
 | If either original floating-point value is a signaling NaN, the invalid
 | exception is raised.
 *----------------------------------------------------------------------------*/
-struct uint128
+struct uint128 softfloat_propagateNaNF128UI(uint_fast64_t uiA64, uint_fast64_t uiA0, uint_fast64_t uiB64, uint_fast64_t uiB0);
- softfloat_propagateNaNF128UI(
-     uint_fast64_t uiA64,
-     uint_fast64_t uiA0,
-     uint_fast64_t uiB64,
-     uint_fast64_t uiB0
- );
 
 #else
 
@@ -304,18 +341,24 @@ struct uint128
 | common NaN at the location pointed to by 'zPtr'.  If the NaN is a signaling
 | NaN, the invalid exception is raised.
 *----------------------------------------------------------------------------*/
-void
+#define softfloat_extF80MToCommonNaN(aSPtr, zPtr)                                                                                          \
- softfloat_extF80MToCommonNaN(
+    if(!((aSPtr)->signif & UINT64_C(0x4000000000000000)))                                                                                  \
-     const struct extFloat80M *aSPtr, struct commonNaN *zPtr );
+    softfloat_raiseFlags(softfloat_flag_invalid)
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into an 80-bit extended
 | floating-point NaN, and stores this NaN at the location pointed to by
 | 'zSPtr'.
 *----------------------------------------------------------------------------*/
-void
+#if defined INLINE && !defined softfloat_commonNaNToExtF80M
- softfloat_commonNaNToExtF80M(
+INLINE
-     const struct commonNaN *aPtr, struct extFloat80M *zSPtr );
+void softfloat_commonNaNToExtF80M(const struct commonNaN* aPtr, struct extFloat80M* zSPtr) {
+    zSPtr->signExp = defaultNaNExtF80UI64;
+    zSPtr->signif = defaultNaNExtF80UI0;
+}
+#else
+void softfloat_commonNaNToExtF80M(const struct commonNaN* aPtr, struct extFloat80M* zSPtr);
+#endif
 
 /*----------------------------------------------------------------------------
 | Assuming at least one of the two 80-bit extended floating-point values
@@ -323,17 +366,12 @@ void
 | at the location pointed to by 'zSPtr'.  If either original floating-point
 | value is a signaling NaN, the invalid exception is raised.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_propagateNaNExtF80M(const struct extFloat80M* aSPtr, const struct extFloat80M* bSPtr, struct extFloat80M* zSPtr);
- softfloat_propagateNaNExtF80M(
-     const struct extFloat80M *aSPtr,
-     const struct extFloat80M *bSPtr,
-     struct extFloat80M *zSPtr
- );
 
 /*----------------------------------------------------------------------------
 | The bit pattern for a default generated 128-bit floating-point NaN.
 *----------------------------------------------------------------------------*/
-#define defaultNaNF128UI96 0xFFFF8000
+#define defaultNaNF128UI96 0x7FFF8000
 #define defaultNaNF128UI64 0
 #define defaultNaNF128UI32 0
 #define defaultNaNF128UI0 0
@@ -346,8 +384,9 @@ void
 | four 32-bit elements that concatenate in the platform's normal endian order
 | to form a 128-bit floating-point value.
 *----------------------------------------------------------------------------*/
-void
+#define softfloat_f128MToCommonNaN(aWPtr, zPtr)                                                                                            \
- softfloat_f128MToCommonNaN( const uint32_t *aWPtr, struct commonNaN *zPtr );
+    if(!((aWPtr)[indexWordHi(4)] & UINT64_C(0x0000800000000000)))                                                                          \
+    softfloat_raiseFlags(softfloat_flag_invalid)
 
 /*----------------------------------------------------------------------------
 | Converts the common NaN pointed to by 'aPtr' into a 128-bit floating-point
@@ -355,8 +394,17 @@ void
 | 'zWPtr' points to an array of four 32-bit elements that concatenate in the
 | platform's normal endian order to form a 128-bit floating-point value.
 *----------------------------------------------------------------------------*/
-void
+#if defined INLINE && !defined softfloat_commonNaNToF128M
- softfloat_commonNaNToF128M( const struct commonNaN *aPtr, uint32_t *zWPtr );
+INLINE
+void softfloat_commonNaNToF128M(const struct commonNaN* aPtr, uint32_t* zWPtr) {
+    zWPtr[indexWord(4, 3)] = defaultNaNF128UI96;
+    zWPtr[indexWord(4, 2)] = defaultNaNF128UI64;
+    zWPtr[indexWord(4, 1)] = defaultNaNF128UI32;
+    zWPtr[indexWord(4, 0)] = defaultNaNF128UI0;
+}
+#else
+void softfloat_commonNaNToF128M(const struct commonNaN* aPtr, uint32_t* zWPtr);
+#endif
 
 /*----------------------------------------------------------------------------
 | Assuming at least one of the two 128-bit floating-point values pointed to by
@@ -366,11 +414,8 @@ void
 | and 'zWPtr' points to an array of four 32-bit elements that concatenate in
 | the platform's normal endian order to form a 128-bit floating-point value.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_propagateNaNF128M(const uint32_t* aWPtr, const uint32_t* bWPtr, uint32_t* zWPtr);
- softfloat_propagateNaNF128M(
-     const uint32_t *aWPtr, const uint32_t *bWPtr, uint32_t *zWPtr );
 
 #endif
 
 #endif
-

softfloat/source/bf16_isSignalingNaN.c

@@ -0,0 +1,51 @@
+
+/*============================================================================
+
+This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
+Package, Release 3e, by John R. Hauser.
+
+Copyright 2011, 2012, 2013, 2014, 2015 The Regents of the University of
+California.  All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+
+ 1. Redistributions of source code must retain the above copyright notice,
+    this list of conditions, and the following disclaimer.
+
+ 2. Redistributions in binary form must reproduce the above copyright notice,
+    this list of conditions, and the following disclaimer in the documentation
+    and/or other materials provided with the distribution.
+
+ 3. Neither the name of the University nor the names of its contributors may
+    be used to endorse or promote products derived from this software without
+    specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS "AS IS", AND ANY
+EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, ARE
+DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY
+DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
+ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+=============================================================================*/
+
+#include <stdbool.h>
+#include "platform.h"
+#include "internals.h"
+#include "specialize.h"
+#include "softfloat.h"
+
+bool bf16_isSignalingNaN( bfloat16_t a )
+{
+    union ui16_bf16 uA;
+
+    uA.f = a;
+    return softfloat_isSigNaNBF16UI( uA.ui );
+
+}
+

softfloat/source/bf16_to_f32.c

@@ -0,0 +1,90 @@
+
+/*============================================================================
+
+This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
+Package, Release 3e, by John R. Hauser.
+
+Copyright 2011, 2012, 2013, 2014, 2015 The Regents of the University of
+California.  All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+
+ 1. Redistributions of source code must retain the above copyright notice,
+    this list of conditions, and the following disclaimer.
+
+ 2. Redistributions in binary form must reproduce the above copyright notice,
+    this list of conditions, and the following disclaimer in the documentation
+    and/or other materials provided with the distribution.
+
+ 3. Neither the name of the University nor the names of its contributors may
+    be used to endorse or promote products derived from this software without
+    specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS "AS IS", AND ANY
+EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, ARE
+DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY
+DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
+ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+=============================================================================*/
+
+#include <stdbool.h>
+#include <stdint.h>
+#include "platform.h"
+#include "internals.h"
+#include "specialize.h"
+#include "softfloat.h"
+
+float32_t bf16_to_f32( bfloat16_t a )
+{
+    union ui16_bf16 uA;
+    uint_fast16_t uiA;
+    bool sign;
+    int_fast16_t exp;
+    uint_fast16_t frac;
+    struct commonNaN commonNaN;
+    uint_fast32_t uiZ;
+    struct exp8_sig16 normExpSig;
+    union ui32_f32 uZ;
+
+    /*------------------------------------------------------------------------
+    *------------------------------------------------------------------------*/
+    uA.f = a;
+    uiA = uA.ui;
+    sign = signBF16UI( uiA );
+    exp  = expBF16UI( uiA );
+    frac = fracBF16UI( uiA );
+    /*------------------------------------------------------------------------
+    *------------------------------------------------------------------------*/
+    // NaN or Inf
+    if ( exp == 0xFF ) {
+        if ( frac ) {
+            softfloat_bf16UIToCommonNaN( uiA, &commonNaN );
+            uiZ = softfloat_commonNaNToF32UI( &commonNaN );
+        } else {
+            uiZ = packToF32UI( sign, 0xFF, 0 );
+        }
+        goto uiZ;
+    }
+    /*------------------------------------------------------------------------
+    *------------------------------------------------------------------------*/
+    // packToF32UI simply packs bitfields without any numerical change
+    // which means it can be used directly for any BF16 to f32 conversions which
+    // does not require bits manipulation
+    // (that is everything where the 16-bit are just padded right with 16 zeros, including
+    //  subnormal numbers)
+    uiZ = packToF32UI( sign, exp, ((uint_fast32_t) frac) <<16 );
+ uiZ:
+    uZ.ui = uiZ;
+    return uZ.f;
+
+}
+
+
+

softfloat/source/f32_to_bf16.c

@@ -0,0 +1,105 @@
+
+/*============================================================================
+
+This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
+Package, Release 3e, by John R. Hauser.
+
+Copyright 2011, 2012, 2013, 2014, 2015 The Regents of the University of
+California.  All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+
+ 1. Redistributions of source code must retain the above copyright notice,
+    this list of conditions, and the following disclaimer.
+
+ 2. Redistributions in binary form must reproduce the above copyright notice,
+    this list of conditions, and the following disclaimer in the documentation
+    and/or other materials provided with the distribution.
+
+ 3. Neither the name of the University nor the names of its contributors may
+    be used to endorse or promote products derived from this software without
+    specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS "AS IS", AND ANY
+EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, ARE
+DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY
+DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
+ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+=============================================================================*/
+
+#include <stdbool.h>
+#include <stdint.h>
+#include "platform.h"
+#include "internals.h"
+#include "specialize.h"
+#include "softfloat.h"
+
+#include <inttypes.h>
+#include <stdio.h>
+
+bfloat16_t f32_to_bf16( float32_t a )
+{
+    union ui32_f32 uA;
+    uint_fast32_t uiA;
+    bool sign;
+    int_fast16_t exp;
+    uint_fast32_t frac;
+    struct commonNaN commonNaN;
+    uint_fast16_t uiZ, frac16;
+    union ui16_bf16 uZ;
+
+    /*------------------------------------------------------------------------
+    *------------------------------------------------------------------------*/
+    uA.f = a;
+    uiA = uA.ui;
+    sign = signF32UI( uiA );
+    exp  = expF32UI( uiA );
+    frac = fracF32UI( uiA );
+    /*------------------------------------------------------------------------
+    *------------------------------------------------------------------------*/
+    // infinity or NaN cases
+    if ( exp == 0xFF ) {
+        if ( frac ) {
+            // NaN case
+            softfloat_f32UIToCommonNaN( uiA, &commonNaN );
+            uiZ = softfloat_commonNaNToBF16UI( &commonNaN );
+        } else {
+            // infinity case
+            uiZ = packToBF16UI( sign, 0xFF, 0 );
+        }
+        goto uiZ;
+    }
+    /*------------------------------------------------------------------------
+    *------------------------------------------------------------------------*/
+    // frac is a 24-bit mantissa, right shifted by 9
+    // In the normal case, (24-9) = 15 are set 
+    frac16 = frac>>9 | ((frac & 0x1FF) != 0);
+    if ( ! (exp | frac16) ) {
+        uiZ = packToBF16UI( sign, 0, 0 );
+        goto uiZ;
+    }
+    /*------------------------------------------------------------------------
+    *------------------------------------------------------------------------*/
+    // softfloat_roundPackToBF16 exponent argument (2nd argument)
+    // must correspond to the exponent of fracIn[13] bits
+    // (fracIn is the 3rd and last argument) 
+    uint_fast32_t mask = exp ? 0x4000 : 0x0; // implicit one mask added if input is a normal number
+    // exponent for the lowest normal and largest subnormal should be equal
+    // but is not in IEEE encoding so mantissa must be partially normalized
+    // (by one bit) for subnormal numbers. Such that (exp - 1) corresponds
+    // to the exponent of frac16[13]
+    frac16 = frac16 << (exp ? 0 : 1);
+    return softfloat_roundPackToBF16( sign, exp - 1, frac16 | mask );
+ uiZ:
+    uZ.ui = uiZ;
+    return uZ.f;
+
+}
+

softfloat/source/f32_to_f16.c

@@ -72,6 +72,9 @@ float16_t f32_to_f16( float32_t a )
     }
     /*------------------------------------------------------------------------
     *------------------------------------------------------------------------*/
+    // frac is a 24-bit significand, the bottom 9 bits LSB are extracted and OR-red
+    // into a sticky flag, the top 15 MSBs are extracted, the LSB of this top slice
+    // is OR-red with the sticky 
     frac16 = frac>>9 | ((frac & 0x1FF) != 0);
     if ( ! (exp | frac16) ) {
         uiZ = packToF16UI( sign, 0, 0 );

softfloat/source/include/internals.h

@@ -37,242 +37,221 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #ifndef internals_h
 #define internals_h 1
 
-#include <stdbool.h>
-#include <stdint.h>
 #include "primitives.h"
 #include "softfloat_types.h"
+#include <stdbool.h>
+#include <stdint.h>
 
-union ui16_f16 { uint16_t ui; float16_t f; };
+union ui16_f16 {
-union ui32_f32 { uint32_t ui; float32_t f; };
+    uint16_t ui;
-union ui64_f64 { uint64_t ui; float64_t f; };
+    float16_t f;
-
+};
-#ifdef SOFTFLOAT_FAST_INT64
+union ui16_bf16 {
-union extF80M_extF80 { struct extFloat80M fM; extFloat80_t f; };
+    uint16_t ui;
-union ui128_f128 { struct uint128 ui; float128_t f; };
+    bfloat16_t f;
-#endif
+};
-
+union ui32_f32 {
-enum {
+    uint32_t ui;
-    softfloat_mulAdd_subC    = 1,
+    float32_t f;
-    softfloat_mulAdd_subProd = 2
+};
+union ui64_f64 {
+    uint64_t ui;
+    float64_t f;
 };
 
-/*----------------------------------------------------------------------------
-*----------------------------------------------------------------------------*/
-uint_fast32_t softfloat_roundToUI32( bool, uint_fast64_t, uint_fast8_t, bool );
-
 #ifdef SOFTFLOAT_FAST_INT64
-uint_fast64_t
+union extF80M_extF80 {
- softfloat_roundToUI64(
+    struct extFloat80M fM;
-     bool, uint_fast64_t, uint_fast64_t, uint_fast8_t, bool );
+    extFloat80_t f;
-#else
+};
-uint_fast64_t softfloat_roundMToUI64( bool, uint32_t *, uint_fast8_t, bool );
+union ui128_f128 {
+    struct uint128 ui;
+    float128_t f;
+};
 #endif
 
-int_fast32_t softfloat_roundToI32( bool, uint_fast64_t, uint_fast8_t, bool );
+enum { softfloat_mulAdd_subC = 1, softfloat_mulAdd_subProd = 2 };
+
+/*----------------------------------------------------------------------------
+ *----------------------------------------------------------------------------*/
+uint_fast32_t softfloat_roundToUI32(bool, uint_fast64_t, uint_fast8_t, bool);
 
 #ifdef SOFTFLOAT_FAST_INT64
-int_fast64_t
+uint_fast64_t softfloat_roundToUI64(bool, uint_fast64_t, uint_fast64_t, uint_fast8_t, bool);
- softfloat_roundToI64(
-     bool, uint_fast64_t, uint_fast64_t, uint_fast8_t, bool );
 #else
-int_fast64_t softfloat_roundMToI64( bool, uint32_t *, uint_fast8_t, bool );
+uint_fast64_t softfloat_roundMToUI64(bool, uint32_t*, uint_fast8_t, bool);
+#endif
+
+int_fast32_t softfloat_roundToI32(bool, uint_fast64_t, uint_fast8_t, bool);
+
+#ifdef SOFTFLOAT_FAST_INT64
+int_fast64_t softfloat_roundToI64(bool, uint_fast64_t, uint_fast64_t, uint_fast8_t, bool);
+#else
+int_fast64_t softfloat_roundMToI64(bool, uint32_t*, uint_fast8_t, bool);
 #endif
 
 /*----------------------------------------------------------------------------
-*----------------------------------------------------------------------------*/
+ *----------------------------------------------------------------------------*/
-#define signF16UI( a ) ((bool) ((uint16_t) (a)>>15))
+#define signF16UI(a) ((bool)((uint16_t)(a) >> 15))
-#define expF16UI( a ) ((int_fast8_t) ((a)>>10) & 0x1F)
+#define expF16UI(a) ((int_fast8_t)((a) >> 10) & 0x1F)
-#define fracF16UI( a ) ((a) & 0x03FF)
+#define fracF16UI(a) ((a)&0x03FF)
-#define packToF16UI( sign, exp, sig ) (((uint16_t) (sign)<<15) + ((uint16_t) (exp)<<10) + (sig))
+#define packToF16UI(sign, exp, sig) (((uint16_t)(sign) << 15) + ((uint16_t)(exp) << 10) + (sig))
 
-#define isNaNF16UI( a ) (((~(a) & 0x7C00) == 0) && ((a) & 0x03FF))
+#define isNaNF16UI(a) (((~(a)&0x7C00) == 0) && ((a)&0x03FF))
 
-struct exp8_sig16 { int_fast8_t exp; uint_fast16_t sig; };
+struct exp8_sig16 {
-struct exp8_sig16 softfloat_normSubnormalF16Sig( uint_fast16_t );
+    int_fast8_t exp;
+    uint_fast16_t sig;
+};
+struct exp8_sig16 softfloat_normSubnormalF16Sig(uint_fast16_t);
 
-float16_t softfloat_roundPackToF16( bool, int_fast16_t, uint_fast16_t );
+float16_t softfloat_roundPackToF16(bool, int_fast16_t, uint_fast16_t);
-float16_t softfloat_normRoundPackToF16( bool, int_fast16_t, uint_fast16_t );
+float16_t softfloat_normRoundPackToF16(bool, int_fast16_t, uint_fast16_t);
 
-float16_t softfloat_addMagsF16( uint_fast16_t, uint_fast16_t );
+float16_t softfloat_addMagsF16(uint_fast16_t, uint_fast16_t);
-float16_t softfloat_subMagsF16( uint_fast16_t, uint_fast16_t );
+float16_t softfloat_subMagsF16(uint_fast16_t, uint_fast16_t);
-float16_t
+float16_t softfloat_mulAddF16(uint_fast16_t, uint_fast16_t, uint_fast16_t, uint_fast8_t);
- softfloat_mulAddF16(
-     uint_fast16_t, uint_fast16_t, uint_fast16_t, uint_fast8_t );
 
 /*----------------------------------------------------------------------------
-*----------------------------------------------------------------------------*/
+ *----------------------------------------------------------------------------*/
-#define signF32UI( a ) ((bool) ((uint32_t) (a)>>31))
+#define signBF16UI(a) ((bool)((uint16_t)(a) >> 15))
-#define expF32UI( a ) ((int_fast16_t) ((a)>>23) & 0xFF)
+#define expBF16UI(a) ((int_fast16_t)((a) >> 7) & 0xFF)
-#define fracF32UI( a ) ((a) & 0x007FFFFF)
+#define fracBF16UI(a) ((a)&0x07F)
-#define packToF32UI( sign, exp, sig ) (((uint32_t) (sign)<<31) + ((uint32_t) (exp)<<23) + (sig))
+#define packToBF16UI(sign, exp, sig) (((uint16_t)(sign) << 15) + ((uint16_t)(exp) << 7) + (sig))
 
-#define isNaNF32UI( a ) (((~(a) & 0x7F800000) == 0) && ((a) & 0x007FFFFF))
+#define isNaNBF16UI(a) (((~(a)&0x7FC0) == 0) && ((a)&0x07F))
 
-struct exp16_sig32 { int_fast16_t exp; uint_fast32_t sig; };
+bfloat16_t softfloat_roundPackToBF16(bool, int_fast16_t, uint_fast16_t);
-struct exp16_sig32 softfloat_normSubnormalF32Sig( uint_fast32_t );
+struct exp8_sig16 softfloat_normSubnormalBF16Sig(uint_fast16_t);
-
-float32_t softfloat_roundPackToF32( bool, int_fast16_t, uint_fast32_t );
-float32_t softfloat_normRoundPackToF32( bool, int_fast16_t, uint_fast32_t );
-
-float32_t softfloat_addMagsF32( uint_fast32_t, uint_fast32_t );
-float32_t softfloat_subMagsF32( uint_fast32_t, uint_fast32_t );
-float32_t
- softfloat_mulAddF32(
-     uint_fast32_t, uint_fast32_t, uint_fast32_t, uint_fast8_t );
 
 /*----------------------------------------------------------------------------
-*----------------------------------------------------------------------------*/
+ *----------------------------------------------------------------------------*/
-#define signF64UI( a ) ((bool) ((uint64_t) (a)>>63))
+#define signF32UI(a) ((bool)((uint32_t)(a) >> 31))
-#define expF64UI( a ) ((int_fast16_t) ((a)>>52) & 0x7FF)
+#define expF32UI(a) ((int_fast16_t)((a) >> 23) & 0xFF)
-#define fracF64UI( a ) ((a) & UINT64_C( 0x000FFFFFFFFFFFFF ))
+#define fracF32UI(a) ((a)&0x007FFFFF)
-#define packToF64UI( sign, exp, sig ) ((uint64_t) (((uint_fast64_t) (sign)<<63) + ((uint_fast64_t) (exp)<<52) + (sig)))
+#define packToF32UI(sign, exp, sig) (((uint32_t)(sign) << 31) + ((uint32_t)(exp) << 23) + (sig))
 
-#define isNaNF64UI( a ) (((~(a) & UINT64_C( 0x7FF0000000000000 )) == 0) && ((a) & UINT64_C( 0x000FFFFFFFFFFFFF )))
+#define isNaNF32UI(a) (((~(a)&0x7F800000) == 0) && ((a)&0x007FFFFF))
 
-struct exp16_sig64 { int_fast16_t exp; uint_fast64_t sig; };
+struct exp16_sig32 {
-struct exp16_sig64 softfloat_normSubnormalF64Sig( uint_fast64_t );
+    int_fast16_t exp;
+    uint_fast32_t sig;
+};
+struct exp16_sig32 softfloat_normSubnormalF32Sig(uint_fast32_t);
 
-float64_t softfloat_roundPackToF64( bool, int_fast16_t, uint_fast64_t );
+float32_t softfloat_roundPackToF32(bool, int_fast16_t, uint_fast32_t);
-float64_t softfloat_normRoundPackToF64( bool, int_fast16_t, uint_fast64_t );
+float32_t softfloat_normRoundPackToF32(bool, int_fast16_t, uint_fast32_t);
 
-float64_t softfloat_addMagsF64( uint_fast64_t, uint_fast64_t, bool );
+float32_t softfloat_addMagsF32(uint_fast32_t, uint_fast32_t);
-float64_t softfloat_subMagsF64( uint_fast64_t, uint_fast64_t, bool );
+float32_t softfloat_subMagsF32(uint_fast32_t, uint_fast32_t);
-float64_t
+float32_t softfloat_mulAddF32(uint_fast32_t, uint_fast32_t, uint_fast32_t, uint_fast8_t);
- softfloat_mulAddF64(
-     uint_fast64_t, uint_fast64_t, uint_fast64_t, uint_fast8_t );
 
 /*----------------------------------------------------------------------------
-*----------------------------------------------------------------------------*/
+ *----------------------------------------------------------------------------*/
-#define signExtF80UI64( a64 ) ((bool) ((uint16_t) (a64)>>15))
+#define signF64UI(a) ((bool)((uint64_t)(a) >> 63))
-#define expExtF80UI64( a64 ) ((a64) & 0x7FFF)
+#define expF64UI(a) ((int_fast16_t)((a) >> 52) & 0x7FF)
-#define packToExtF80UI64( sign, exp ) ((uint_fast16_t) (sign)<<15 | (exp))
+#define fracF64UI(a) ((a)&UINT64_C(0x000FFFFFFFFFFFFF))
+#define packToF64UI(sign, exp, sig) ((uint64_t)(((uint_fast64_t)(sign) << 63) + ((uint_fast64_t)(exp) << 52) + (sig)))
 
-#define isNaNExtF80UI( a64, a0 ) ((((a64) & 0x7FFF) == 0x7FFF) && ((a0) & UINT64_C( 0x7FFFFFFFFFFFFFFF )))
+#define isNaNF64UI(a) (((~(a)&UINT64_C(0x7FF0000000000000)) == 0) && ((a)&UINT64_C(0x000FFFFFFFFFFFFF)))
+
+struct exp16_sig64 {
+    int_fast16_t exp;
+    uint_fast64_t sig;
+};
+struct exp16_sig64 softfloat_normSubnormalF64Sig(uint_fast64_t);
+
+float64_t softfloat_roundPackToF64(bool, int_fast16_t, uint_fast64_t);
+float64_t softfloat_normRoundPackToF64(bool, int_fast16_t, uint_fast64_t);
+
+float64_t softfloat_addMagsF64(uint_fast64_t, uint_fast64_t, bool);
+float64_t softfloat_subMagsF64(uint_fast64_t, uint_fast64_t, bool);
+float64_t softfloat_mulAddF64(uint_fast64_t, uint_fast64_t, uint_fast64_t, uint_fast8_t);
+
+/*----------------------------------------------------------------------------
+ *----------------------------------------------------------------------------*/
+#define signExtF80UI64(a64) ((bool)((uint16_t)(a64) >> 15))
+#define expExtF80UI64(a64) ((a64)&0x7FFF)
+#define packToExtF80UI64(sign, exp) ((uint_fast16_t)(sign) << 15 | (exp))
+
+#define isNaNExtF80UI(a64, a0) ((((a64)&0x7FFF) == 0x7FFF) && ((a0)&UINT64_C(0x7FFFFFFFFFFFFFFF)))
 
 #ifdef SOFTFLOAT_FAST_INT64
 
 /*----------------------------------------------------------------------------
-*----------------------------------------------------------------------------*/
+ *----------------------------------------------------------------------------*/
 
-struct exp32_sig64 { int_fast32_t exp; uint64_t sig; };
+struct exp32_sig64 {
-struct exp32_sig64 softfloat_normSubnormalExtF80Sig( uint_fast64_t );
+    int_fast32_t exp;
+    uint64_t sig;
+};
+struct exp32_sig64 softfloat_normSubnormalExtF80Sig(uint_fast64_t);
 
-extFloat80_t
+extFloat80_t softfloat_roundPackToExtF80(bool, int_fast32_t, uint_fast64_t, uint_fast64_t, uint_fast8_t);
- softfloat_roundPackToExtF80(
+extFloat80_t softfloat_normRoundPackToExtF80(bool, int_fast32_t, uint_fast64_t, uint_fast64_t, uint_fast8_t);
-     bool, int_fast32_t, uint_fast64_t, uint_fast64_t, uint_fast8_t );
-extFloat80_t
- softfloat_normRoundPackToExtF80(
-     bool, int_fast32_t, uint_fast64_t, uint_fast64_t, uint_fast8_t );
 
-extFloat80_t
+extFloat80_t softfloat_addMagsExtF80(uint_fast16_t, uint_fast64_t, uint_fast16_t, uint_fast64_t, bool);
- softfloat_addMagsExtF80(
+extFloat80_t softfloat_subMagsExtF80(uint_fast16_t, uint_fast64_t, uint_fast16_t, uint_fast64_t, bool);
-     uint_fast16_t, uint_fast64_t, uint_fast16_t, uint_fast64_t, bool );
-extFloat80_t
- softfloat_subMagsExtF80(
-     uint_fast16_t, uint_fast64_t, uint_fast16_t, uint_fast64_t, bool );
 
 /*----------------------------------------------------------------------------
-*----------------------------------------------------------------------------*/
+ *----------------------------------------------------------------------------*/
-#define signF128UI64( a64 ) ((bool) ((uint64_t) (a64)>>63))
+#define signF128UI64(a64) ((bool)((uint64_t)(a64) >> 63))
-#define expF128UI64( a64 ) ((int_fast32_t) ((a64)>>48) & 0x7FFF)
+#define expF128UI64(a64) ((int_fast32_t)((a64) >> 48) & 0x7FFF)
-#define fracF128UI64( a64 ) ((a64) & UINT64_C( 0x0000FFFFFFFFFFFF ))
+#define fracF128UI64(a64) ((a64)&UINT64_C(0x0000FFFFFFFFFFFF))
-#define packToF128UI64( sign, exp, sig64 ) (((uint_fast64_t) (sign)<<63) + ((uint_fast64_t) (exp)<<48) + (sig64))
+#define packToF128UI64(sign, exp, sig64) (((uint_fast64_t)(sign) << 63) + ((uint_fast64_t)(exp) << 48) + (sig64))
 
-#define isNaNF128UI( a64, a0 ) (((~(a64) & UINT64_C( 0x7FFF000000000000 )) == 0) && (a0 || ((a64) & UINT64_C( 0x0000FFFFFFFFFFFF ))))
+#define isNaNF128UI(a64, a0) (((~(a64)&UINT64_C(0x7FFF000000000000)) == 0) && (a0 || ((a64)&UINT64_C(0x0000FFFFFFFFFFFF))))
 
-struct exp32_sig128 { int_fast32_t exp; struct uint128 sig; };
+struct exp32_sig128 {
-struct exp32_sig128
+    int_fast32_t exp;
- softfloat_normSubnormalF128Sig( uint_fast64_t, uint_fast64_t );
+    struct uint128 sig;
+};
+struct exp32_sig128 softfloat_normSubnormalF128Sig(uint_fast64_t, uint_fast64_t);
 
-float128_t
+float128_t softfloat_roundPackToF128(bool, int_fast32_t, uint_fast64_t, uint_fast64_t, uint_fast64_t);
- softfloat_roundPackToF128(
+float128_t softfloat_normRoundPackToF128(bool, int_fast32_t, uint_fast64_t, uint_fast64_t);
-     bool, int_fast32_t, uint_fast64_t, uint_fast64_t, uint_fast64_t );
-float128_t
- softfloat_normRoundPackToF128(
-     bool, int_fast32_t, uint_fast64_t, uint_fast64_t );
 
-float128_t
+float128_t softfloat_addMagsF128(uint_fast64_t, uint_fast64_t, uint_fast64_t, uint_fast64_t, bool);
- softfloat_addMagsF128(
+float128_t softfloat_subMagsF128(uint_fast64_t, uint_fast64_t, uint_fast64_t, uint_fast64_t, bool);
-     uint_fast64_t, uint_fast64_t, uint_fast64_t, uint_fast64_t, bool );
+float128_t softfloat_mulAddF128(uint_fast64_t, uint_fast64_t, uint_fast64_t, uint_fast64_t, uint_fast64_t, uint_fast64_t, uint_fast8_t);
-float128_t
- softfloat_subMagsF128(
-     uint_fast64_t, uint_fast64_t, uint_fast64_t, uint_fast64_t, bool );
-float128_t
- softfloat_mulAddF128(
-     uint_fast64_t,
-     uint_fast64_t,
-     uint_fast64_t,
-     uint_fast64_t,
-     uint_fast64_t,
-     uint_fast64_t,
-     uint_fast8_t
- );
 
 #else
 
 /*----------------------------------------------------------------------------
-*----------------------------------------------------------------------------*/
+ *----------------------------------------------------------------------------*/
 
-bool
+bool softfloat_tryPropagateNaNExtF80M(const struct extFloat80M*, const struct extFloat80M*, struct extFloat80M*);
- softfloat_tryPropagateNaNExtF80M(
+void softfloat_invalidExtF80M(struct extFloat80M*);
-     const struct extFloat80M *,
-     const struct extFloat80M *,
-     struct extFloat80M *
- );
-void softfloat_invalidExtF80M( struct extFloat80M * );
 
-int softfloat_normExtF80SigM( uint64_t * );
+int softfloat_normExtF80SigM(uint64_t*);
 
-void
+void softfloat_roundPackMToExtF80M(bool, int32_t, uint32_t*, uint_fast8_t, struct extFloat80M*);
- softfloat_roundPackMToExtF80M(
+void softfloat_normRoundPackMToExtF80M(bool, int32_t, uint32_t*, uint_fast8_t, struct extFloat80M*);
-     bool, int32_t, uint32_t *, uint_fast8_t, struct extFloat80M * );
-void
- softfloat_normRoundPackMToExtF80M(
-     bool, int32_t, uint32_t *, uint_fast8_t, struct extFloat80M * );
 
-void
+void softfloat_addExtF80M(const struct extFloat80M*, const struct extFloat80M*, struct extFloat80M*, bool);
- softfloat_addExtF80M(
-     const struct extFloat80M *,
-     const struct extFloat80M *,
-     struct extFloat80M *,
-     bool
- );
 
-int
+int softfloat_compareNonnormExtF80M(const struct extFloat80M*, const struct extFloat80M*);
- softfloat_compareNonnormExtF80M(
-     const struct extFloat80M *, const struct extFloat80M * );
 
 /*----------------------------------------------------------------------------
-*----------------------------------------------------------------------------*/
+ *----------------------------------------------------------------------------*/
-#define signF128UI96( a96 ) ((bool) ((uint32_t) (a96)>>31))
+#define signF128UI96(a96) ((bool)((uint32_t)(a96) >> 31))
-#define expF128UI96( a96 ) ((int32_t) ((a96)>>16) & 0x7FFF)
+#define expF128UI96(a96) ((int32_t)((a96) >> 16) & 0x7FFF)
-#define fracF128UI96( a96 ) ((a96) & 0x0000FFFF)
+#define fracF128UI96(a96) ((a96)&0x0000FFFF)
-#define packToF128UI96( sign, exp, sig96 ) (((uint32_t) (sign)<<31) + ((uint32_t) (exp)<<16) + (sig96))
+#define packToF128UI96(sign, exp, sig96) (((uint32_t)(sign) << 31) + ((uint32_t)(exp) << 16) + (sig96))
 
-bool softfloat_isNaNF128M( const uint32_t * );
+bool softfloat_isNaNF128M(const uint32_t*);
 
-bool
+bool softfloat_tryPropagateNaNF128M(const uint32_t*, const uint32_t*, uint32_t*);
- softfloat_tryPropagateNaNF128M(
+void softfloat_invalidF128M(uint32_t*);
-     const uint32_t *, const uint32_t *, uint32_t * );
-void softfloat_invalidF128M( uint32_t * );
 
-int softfloat_shiftNormSigF128M( const uint32_t *, uint_fast8_t, uint32_t * );
+int softfloat_shiftNormSigF128M(const uint32_t*, uint_fast8_t, uint32_t*);
 
-void softfloat_roundPackMToF128M( bool, int32_t, uint32_t *, uint32_t * );
+void softfloat_roundPackMToF128M(bool, int32_t, uint32_t*, uint32_t*);
-void softfloat_normRoundPackMToF128M( bool, int32_t, uint32_t *, uint32_t * );
+void softfloat_normRoundPackMToF128M(bool, int32_t, uint32_t*, uint32_t*);
 
-void
+void softfloat_addF128M(const uint32_t*, const uint32_t*, uint32_t*, bool);
- softfloat_addF128M( const uint32_t *, const uint32_t *, uint32_t *, bool );
+void softfloat_mulAddF128M(const uint32_t*, const uint32_t*, const uint32_t*, uint32_t*, uint_fast8_t);
-void
- softfloat_mulAddF128M(
-     const uint32_t *,
-     const uint32_t *,
-     const uint32_t *,
-     uint32_t *,
-     uint_fast8_t
- );
 
 #endif
 
 #endif
-

softfloat/source/include/opts-GCC.h

@@ -39,70 +39,70 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 #ifdef INLINE
 
-#include <stdint.h>
 #include "primitiveTypes.h"
+#include <stdint.h>
 
 #ifdef SOFTFLOAT_BUILTIN_CLZ
 
-INLINE uint_fast8_t softfloat_countLeadingZeros16( uint16_t a )
+INLINE uint_fast8_t softfloat_countLeadingZeros16(uint16_t a) { return a ? __builtin_clz(a) - 16 : 16; }
-    { return a ? __builtin_clz( a ) - 16 : 16; }
 #define softfloat_countLeadingZeros16 softfloat_countLeadingZeros16
 
-INLINE uint_fast8_t softfloat_countLeadingZeros32( uint32_t a )
+INLINE uint_fast8_t softfloat_countLeadingZeros32(uint32_t a) { return a ? __builtin_clz(a) : 32; }
-    { return a ? __builtin_clz( a ) : 32; }
 #define softfloat_countLeadingZeros32 softfloat_countLeadingZeros32
 
-INLINE uint_fast8_t softfloat_countLeadingZeros64( uint64_t a )
+INLINE uint_fast8_t softfloat_countLeadingZeros64(uint64_t a) { return a ? __builtin_clzll(a) : 64; }
-    { return a ? __builtin_clzll( a ) : 64; }
 #define softfloat_countLeadingZeros64 softfloat_countLeadingZeros64
 
 #endif
 
 #ifdef SOFTFLOAT_INTRINSIC_INT128
 
-INLINE struct uint128 softfloat_mul64ByShifted32To128( uint64_t a, uint32_t b )
+INLINE struct uint128 softfloat_mul64ByShifted32To128(uint64_t a, uint32_t b) {
-{
+    union {
-    union { unsigned __int128 ui; struct uint128 s; } uZ;
+        unsigned __int128 ui;
-    uZ.ui = (unsigned __int128) a * ((uint_fast64_t) b<<32);
+        struct uint128 s;
+    } uZ;
+    uZ.ui = (unsigned __int128)a * ((uint_fast64_t)b << 32);
     return uZ.s;
 }
 #define softfloat_mul64ByShifted32To128 softfloat_mul64ByShifted32To128
 
-INLINE struct uint128 softfloat_mul64To128( uint64_t a, uint64_t b )
+INLINE struct uint128 softfloat_mul64To128(uint64_t a, uint64_t b) {
-{
+    union {
-    union { unsigned __int128 ui; struct uint128 s; } uZ;
+        unsigned __int128 ui;
-    uZ.ui = (unsigned __int128) a * b;
+        struct uint128 s;
+    } uZ;
+    uZ.ui = (unsigned __int128)a * b;
     return uZ.s;
 }
 #define softfloat_mul64To128 softfloat_mul64To128
 
 INLINE
-struct uint128 softfloat_mul128By32( uint64_t a64, uint64_t a0, uint32_t b )
+struct uint128 softfloat_mul128By32(uint64_t a64, uint64_t a0, uint32_t b) {
-{
+    union {
-    union { unsigned __int128 ui; struct uint128 s; } uZ;
+        unsigned __int128 ui;
-    uZ.ui = ((unsigned __int128) a64<<64 | a0) * b;
+        struct uint128 s;
+    } uZ;
+    uZ.ui = ((unsigned __int128)a64 << 64 | a0) * b;
     return uZ.s;
 }
 #define softfloat_mul128By32 softfloat_mul128By32
 
 INLINE
-void
+void softfloat_mul128To256M(uint64_t a64, uint64_t a0, uint64_t b64, uint64_t b0, uint64_t* zPtr) {
- softfloat_mul128To256M(
-     uint64_t a64, uint64_t a0, uint64_t b64, uint64_t b0, uint64_t *zPtr )
-{
     unsigned __int128 z0, mid1, mid, z128;
-    z0 = (unsigned __int128) a0 * b0;
+    z0 = (unsigned __int128)a0 * b0;
-    mid1 = (unsigned __int128) a64 * b0;
+    mid1 = (unsigned __int128)a64 * b0;
-    mid = mid1 + (unsigned __int128) a0 * b64;
+    mid = mid1 + (unsigned __int128)a0 * b64;
-    z128 = (unsigned __int128) a64 * b64;
+    z128 = (unsigned __int128)a64 * b64;
-    z128 += (unsigned __int128) (mid < mid1)<<64 | mid>>64;
+    z128 += (unsigned __int128)(mid < mid1) << 64 | mid >> 64;
     mid <<= 64;
     z0 += mid;
     z128 += (z0 < mid);
-    zPtr[indexWord( 4, 0 )] = z0;
+    zPtr[indexWord(4, 0)] = z0;
-    zPtr[indexWord( 4, 1 )] = z0>>64;
+    zPtr[indexWord(4, 1)] = z0 >> 64;
-    zPtr[indexWord( 4, 2 )] = z128;
+    zPtr[indexWord(4, 2)] = z128;
-    zPtr[indexWord( 4, 3 )] = z128>>64;
+    zPtr[indexWord(4, 3)] = z128 >> 64;
 }
 #define softfloat_mul128To256M softfloat_mul128To256M
 
@@ -111,4 +111,3 @@ void
 #endif
 
 #endif
-

softfloat/source/include/primitiveTypes.h

@@ -42,13 +42,27 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #ifdef SOFTFLOAT_FAST_INT64
 
 #ifdef LITTLEENDIAN
-struct uint128 { uint64_t v0, v64; };
+struct uint128 {
-struct uint64_extra { uint64_t extra, v; };
+    uint64_t v0, v64;
-struct uint128_extra { uint64_t extra; struct uint128 v; };
+};
+struct uint64_extra {
+    uint64_t extra, v;
+};
+struct uint128_extra {
+    uint64_t extra;
+    struct uint128 v;
+};
 #else
-struct uint128 { uint64_t v64, v0; };
+struct uint128 {
-struct uint64_extra { uint64_t v, extra; };
+    uint64_t v64, v0;
-struct uint128_extra { struct uint128 v; uint64_t extra; };
+};
+struct uint64_extra {
+    uint64_t v, extra;
+};
+struct uint128_extra {
+    struct uint128 v;
+    uint64_t extra;
+};
 #endif
 
 #endif
@@ -59,27 +73,28 @@ struct uint128_extra { struct uint128 v; uint64_t extra; };
 *----------------------------------------------------------------------------*/
 #ifdef LITTLEENDIAN
 #define wordIncr 1
-#define indexWord( total, n ) (n)
+#define indexWord(total, n) (n)
-#define indexWordHi( total ) ((total) - 1)
+#define indexWordHi(total) ((total)-1)
-#define indexWordLo( total ) 0
+#define indexWordLo(total) 0
-#define indexMultiword( total, m, n ) (n)
+#define indexMultiword(total, m, n) (n)
-#define indexMultiwordHi( total, n ) ((total) - (n))
+#define indexMultiwordHi(total, n) ((total) - (n))
-#define indexMultiwordLo( total, n ) 0
+#define indexMultiwordLo(total, n) 0
-#define indexMultiwordHiBut( total, n ) (n)
+#define indexMultiwordHiBut(total, n) (n)
-#define indexMultiwordLoBut( total, n ) 0
+#define indexMultiwordLoBut(total, n) 0
-#define INIT_UINTM4( v3, v2, v1, v0 ) { v0, v1, v2, v3 }
+#define INIT_UINTM4(v3, v2, v1, v0)                                                                                                        \
+    { v0, v1, v2, v3 }
 #else
 #define wordIncr -1
-#define indexWord( total, n ) ((total) - 1 - (n))
+#define indexWord(total, n) ((total)-1 - (n))
-#define indexWordHi( total ) 0
+#define indexWordHi(total) 0
-#define indexWordLo( total ) ((total) - 1)
+#define indexWordLo(total) ((total)-1)
-#define indexMultiword( total, m, n ) ((total) - 1 - (m))
+#define indexMultiword(total, m, n) ((total)-1 - (m))
-#define indexMultiwordHi( total, n ) 0
+#define indexMultiwordHi(total, n) 0
-#define indexMultiwordLo( total, n ) ((total) - (n))
+#define indexMultiwordLo(total, n) ((total) - (n))
-#define indexMultiwordHiBut( total, n ) 0
+#define indexMultiwordHiBut(total, n) 0
-#define indexMultiwordLoBut( total, n ) (n)
+#define indexMultiwordLoBut(total, n) (n)
-#define INIT_UINTM4( v3, v2, v1, v0 ) { v3, v2, v1, v0 }
+#define INIT_UINTM4(v3, v2, v1, v0)                                                                                                        \
+    { v3, v2, v1, v0 }
 #endif
 
 #endif
-

softfloat/source/include/primitives.h

@@ -37,9 +37,9 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #ifndef primitives_h
 #define primitives_h 1
 
+#include "primitiveTypes.h"
 #include <stdbool.h>
 #include <stdint.h>
-#include "primitiveTypes.h"
 
 #ifndef softfloat_shortShiftRightJam64
 /*----------------------------------------------------------------------------
@@ -50,10 +50,9 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 *----------------------------------------------------------------------------*/
 #if defined INLINE_LEVEL && (2 <= INLINE_LEVEL)
 INLINE
-uint64_t softfloat_shortShiftRightJam64( uint64_t a, uint_fast8_t dist )
+uint64_t softfloat_shortShiftRightJam64(uint64_t a, uint_fast8_t dist) { return a >> dist | ((a & (((uint_fast64_t)1 << dist) - 1)) != 0); }
-    { return a>>dist | ((a & (((uint_fast64_t) 1<<dist) - 1)) != 0); }
 #else
-uint64_t softfloat_shortShiftRightJam64( uint64_t a, uint_fast8_t dist );
+uint64_t softfloat_shortShiftRightJam64(uint64_t a, uint_fast8_t dist);
 #endif
 #endif
 
@@ -68,13 +67,11 @@ uint64_t softfloat_shortShiftRightJam64( uint64_t a, uint_fast8_t dist );
 | is zero or nonzero.
 *----------------------------------------------------------------------------*/
 #if defined INLINE_LEVEL && (2 <= INLINE_LEVEL)
-INLINE uint32_t softfloat_shiftRightJam32( uint32_t a, uint_fast16_t dist )
+INLINE uint32_t softfloat_shiftRightJam32(uint32_t a, uint_fast16_t dist) {
-{
+    return (dist < 31) ? a >> dist | ((uint32_t)(a << (-dist & 31)) != 0) : (a != 0);
-    return
-        (dist < 31) ? a>>dist | ((uint32_t) (a<<(-dist & 31)) != 0) : (a != 0);
 }
 #else
-uint32_t softfloat_shiftRightJam32( uint32_t a, uint_fast16_t dist );
+uint32_t softfloat_shiftRightJam32(uint32_t a, uint_fast16_t dist);
 #endif
 #endif
 
@@ -89,13 +86,11 @@ uint32_t softfloat_shiftRightJam32( uint32_t a, uint_fast16_t dist );
 | is zero or nonzero.
 *----------------------------------------------------------------------------*/
 #if defined INLINE_LEVEL && (3 <= INLINE_LEVEL)
-INLINE uint64_t softfloat_shiftRightJam64( uint64_t a, uint_fast32_t dist )
+INLINE uint64_t softfloat_shiftRightJam64(uint64_t a, uint_fast32_t dist) {
-{
+    return (dist < 63) ? a >> dist | ((uint64_t)(a << (-dist & 63)) != 0) : (a != 0);
-    return
-        (dist < 63) ? a>>dist | ((uint64_t) (a<<(-dist & 63)) != 0) : (a != 0);
 }
 #else
-uint64_t softfloat_shiftRightJam64( uint64_t a, uint_fast32_t dist );
+uint64_t softfloat_shiftRightJam64(uint64_t a, uint_fast32_t dist);
 #endif
 #endif
 
@@ -112,10 +107,9 @@ extern const uint_least8_t softfloat_countLeadingZeros8[256];
 | 'a'.  If 'a' is zero, 16 is returned.
 *----------------------------------------------------------------------------*/
 #if defined INLINE_LEVEL && (2 <= INLINE_LEVEL)
-INLINE uint_fast8_t softfloat_countLeadingZeros16( uint16_t a )
+INLINE uint_fast8_t softfloat_countLeadingZeros16(uint16_t a) {
-{
     uint_fast8_t count = 8;
-    if ( 0x100 <= a ) {
+    if(0x100 <= a) {
         count = 0;
         a >>= 8;
     }
@@ -123,7 +117,7 @@ INLINE uint_fast8_t softfloat_countLeadingZeros16( uint16_t a )
     return count;
 }
 #else
-uint_fast8_t softfloat_countLeadingZeros16( uint16_t a );
+uint_fast8_t softfloat_countLeadingZeros16(uint16_t a);
 #endif
 #endif
 
@@ -133,22 +127,21 @@ uint_fast8_t softfloat_countLeadingZeros16( uint16_t a );
 | 'a'.  If 'a' is zero, 32 is returned.
 *----------------------------------------------------------------------------*/
 #if defined INLINE_LEVEL && (3 <= INLINE_LEVEL)
-INLINE uint_fast8_t softfloat_countLeadingZeros32( uint32_t a )
+INLINE uint_fast8_t softfloat_countLeadingZeros32(uint32_t a) {
-{
     uint_fast8_t count = 0;
-    if ( a < 0x10000 ) {
+    if(a < 0x10000) {
         count = 16;
         a <<= 16;
     }
-    if ( a < 0x1000000 ) {
+    if(a < 0x1000000) {
         count += 8;
         a <<= 8;
     }
-    count += softfloat_countLeadingZeros8[a>>24];
+    count += softfloat_countLeadingZeros8[a >> 24];
     return count;
 }
 #else
-uint_fast8_t softfloat_countLeadingZeros32( uint32_t a );
+uint_fast8_t softfloat_countLeadingZeros32(uint32_t a);
 #endif
 #endif
 
@@ -157,7 +150,7 @@ uint_fast8_t softfloat_countLeadingZeros32( uint32_t a );
 | Returns the number of leading 0 bits before the most-significant 1 bit of
 | 'a'.  If 'a' is zero, 64 is returned.
 *----------------------------------------------------------------------------*/
-uint_fast8_t softfloat_countLeadingZeros64( uint64_t a );
+uint_fast8_t softfloat_countLeadingZeros64(uint64_t a);
 #endif
 
 extern const uint16_t softfloat_approxRecip_1k0s[16];
@@ -176,9 +169,9 @@ extern const uint16_t softfloat_approxRecip_1k1s[16];
 | (units in the last place).
 *----------------------------------------------------------------------------*/
 #ifdef SOFTFLOAT_FAST_DIV64TO32
-#define softfloat_approxRecip32_1( a ) ((uint32_t) (UINT64_C( 0x7FFFFFFFFFFFFFFF ) / (uint32_t) (a)))
+#define softfloat_approxRecip32_1(a) ((uint32_t)(UINT64_C(0x7FFFFFFFFFFFFFFF) / (uint32_t)(a)))
 #else
-uint32_t softfloat_approxRecip32_1( uint32_t a );
+uint32_t softfloat_approxRecip32_1(uint32_t a);
 #endif
 #endif
 
@@ -204,7 +197,7 @@ extern const uint16_t softfloat_approxRecipSqrt_1k1s[16];
 | returned is also always within the range 0.5 to 1; thus, the most-
 | significant bit of the result is always set.
 *----------------------------------------------------------------------------*/
-uint32_t softfloat_approxRecipSqrt32_1( unsigned int oddExpA, uint32_t a );
+uint32_t softfloat_approxRecipSqrt32_1(unsigned int oddExpA, uint32_t a);
 #endif
 
 #ifdef SOFTFLOAT_FAST_INT64
@@ -222,10 +215,9 @@ uint32_t softfloat_approxRecipSqrt32_1( unsigned int oddExpA, uint32_t a );
 *----------------------------------------------------------------------------*/
 #if defined INLINE_LEVEL && (1 <= INLINE_LEVEL)
 INLINE
-bool softfloat_eq128( uint64_t a64, uint64_t a0, uint64_t b64, uint64_t b0 )
+bool softfloat_eq128(uint64_t a64, uint64_t a0, uint64_t b64, uint64_t b0) { return (a64 == b64) && (a0 == b0); }
-    { return (a64 == b64) && (a0 == b0); }
 #else
-bool softfloat_eq128( uint64_t a64, uint64_t a0, uint64_t b64, uint64_t b0 );
+bool softfloat_eq128(uint64_t a64, uint64_t a0, uint64_t b64, uint64_t b0);
 #endif
 #endif
 
@@ -237,10 +229,9 @@ bool softfloat_eq128( uint64_t a64, uint64_t a0, uint64_t b64, uint64_t b0 );
 *----------------------------------------------------------------------------*/
 #if defined INLINE_LEVEL && (2 <= INLINE_LEVEL)
 INLINE
-bool softfloat_le128( uint64_t a64, uint64_t a0, uint64_t b64, uint64_t b0 )
+bool softfloat_le128(uint64_t a64, uint64_t a0, uint64_t b64, uint64_t b0) { return (a64 < b64) || ((a64 == b64) && (a0 <= b0)); }
-    { return (a64 < b64) || ((a64 == b64) && (a0 <= b0)); }
 #else
-bool softfloat_le128( uint64_t a64, uint64_t a0, uint64_t b64, uint64_t b0 );
+bool softfloat_le128(uint64_t a64, uint64_t a0, uint64_t b64, uint64_t b0);
 #endif
 #endif
 
@@ -252,10 +243,9 @@ bool softfloat_le128( uint64_t a64, uint64_t a0, uint64_t b64, uint64_t b0 );
 *----------------------------------------------------------------------------*/
 #if defined INLINE_LEVEL && (2 <= INLINE_LEVEL)
 INLINE
-bool softfloat_lt128( uint64_t a64, uint64_t a0, uint64_t b64, uint64_t b0 )
+bool softfloat_lt128(uint64_t a64, uint64_t a0, uint64_t b64, uint64_t b0) { return (a64 < b64) || ((a64 == b64) && (a0 < b0)); }
-    { return (a64 < b64) || ((a64 == b64) && (a0 < b0)); }
 #else
-bool softfloat_lt128( uint64_t a64, uint64_t a0, uint64_t b64, uint64_t b0 );
+bool softfloat_lt128(uint64_t a64, uint64_t a0, uint64_t b64, uint64_t b0);
 #endif
 #endif
 
@@ -266,17 +256,14 @@ bool softfloat_lt128( uint64_t a64, uint64_t a0, uint64_t b64, uint64_t b0 );
 *----------------------------------------------------------------------------*/
 #if defined INLINE_LEVEL && (2 <= INLINE_LEVEL)
 INLINE
-struct uint128
+struct uint128 softfloat_shortShiftLeft128(uint64_t a64, uint64_t a0, uint_fast8_t dist) {
- softfloat_shortShiftLeft128( uint64_t a64, uint64_t a0, uint_fast8_t dist )
-{
     struct uint128 z;
-    z.v64 = a64<<dist | a0>>(-dist & 63);
+    z.v64 = a64 << dist | a0 >> (-dist & 63);
-    z.v0 = a0<<dist;
+    z.v0 = a0 << dist;
     return z;
 }
 #else
-struct uint128
+struct uint128 softfloat_shortShiftLeft128(uint64_t a64, uint64_t a0, uint_fast8_t dist);
- softfloat_shortShiftLeft128( uint64_t a64, uint64_t a0, uint_fast8_t dist );
 #endif
 #endif
 
@@ -287,17 +274,14 @@ struct uint128
 *----------------------------------------------------------------------------*/
 #if defined INLINE_LEVEL && (2 <= INLINE_LEVEL)
 INLINE
-struct uint128
+struct uint128 softfloat_shortShiftRight128(uint64_t a64, uint64_t a0, uint_fast8_t dist) {
- softfloat_shortShiftRight128( uint64_t a64, uint64_t a0, uint_fast8_t dist )
-{
     struct uint128 z;
-    z.v64 = a64>>dist;
+    z.v64 = a64 >> dist;
-    z.v0 = a64<<(-dist & 63) | a0>>dist;
+    z.v0 = a64 << (-dist & 63) | a0 >> dist;
     return z;
 }
 #else
-struct uint128
+struct uint128 softfloat_shortShiftRight128(uint64_t a64, uint64_t a0, uint_fast8_t dist);
- softfloat_shortShiftRight128( uint64_t a64, uint64_t a0, uint_fast8_t dist );
 #endif
 #endif
 
@@ -308,19 +292,14 @@ struct uint128
 *----------------------------------------------------------------------------*/
 #if defined INLINE_LEVEL && (2 <= INLINE_LEVEL)
 INLINE
-struct uint64_extra
+struct uint64_extra softfloat_shortShiftRightJam64Extra(uint64_t a, uint64_t extra, uint_fast8_t dist) {
- softfloat_shortShiftRightJam64Extra(
-     uint64_t a, uint64_t extra, uint_fast8_t dist )
-{
     struct uint64_extra z;
-    z.v = a>>dist;
+    z.v = a >> dist;
-    z.extra = a<<(-dist & 63) | (extra != 0);
+    z.extra = a << (-dist & 63) | (extra != 0);
     return z;
 }
 #else
-struct uint64_extra
+struct uint64_extra softfloat_shortShiftRightJam64Extra(uint64_t a, uint64_t extra, uint_fast8_t dist);
- softfloat_shortShiftRightJam64Extra(
-     uint64_t a, uint64_t extra, uint_fast8_t dist );
 #endif
 #endif
 
@@ -334,22 +313,15 @@ struct uint64_extra
 *----------------------------------------------------------------------------*/
 #if defined INLINE_LEVEL && (3 <= INLINE_LEVEL)
 INLINE
-struct uint128
+struct uint128 softfloat_shortShiftRightJam128(uint64_t a64, uint64_t a0, uint_fast8_t dist) {
- softfloat_shortShiftRightJam128(
-     uint64_t a64, uint64_t a0, uint_fast8_t dist )
-{
     uint_fast8_t negDist = -dist;
     struct uint128 z;
-    z.v64 = a64>>dist;
+    z.v64 = a64 >> dist;
-    z.v0 =
+    z.v0 = a64 << (negDist & 63) | a0 >> dist | ((uint64_t)(a0 << (negDist & 63)) != 0);
-        a64<<(negDist & 63) | a0>>dist
-            | ((uint64_t) (a0<<(negDist & 63)) != 0);
     return z;
 }
 #else
-struct uint128
+struct uint128 softfloat_shortShiftRightJam128(uint64_t a64, uint64_t a0, uint_fast8_t dist);
- softfloat_shortShiftRightJam128(
-     uint64_t a64, uint64_t a0, uint_fast8_t dist );
 #endif
 #endif
 
@@ -360,21 +332,16 @@ struct uint128
 *----------------------------------------------------------------------------*/
 #if defined INLINE_LEVEL && (3 <= INLINE_LEVEL)
 INLINE
-struct uint128_extra
+struct uint128_extra softfloat_shortShiftRightJam128Extra(uint64_t a64, uint64_t a0, uint64_t extra, uint_fast8_t dist) {
- softfloat_shortShiftRightJam128Extra(
-     uint64_t a64, uint64_t a0, uint64_t extra, uint_fast8_t dist )
-{
     uint_fast8_t negDist = -dist;
     struct uint128_extra z;
-    z.v.v64 = a64>>dist;
+    z.v.v64 = a64 >> dist;
-    z.v.v0 = a64<<(negDist & 63) | a0>>dist;
+    z.v.v0 = a64 << (negDist & 63) | a0 >> dist;
-    z.extra = a0<<(negDist & 63) | (extra != 0);
+    z.extra = a0 << (negDist & 63) | (extra != 0);
     return z;
 }
 #else
-struct uint128_extra
+struct uint128_extra softfloat_shortShiftRightJam128Extra(uint64_t a64, uint64_t a0, uint64_t extra, uint_fast8_t dist);
- softfloat_shortShiftRightJam128Extra(
-     uint64_t a64, uint64_t a0, uint64_t extra, uint_fast8_t dist );
 #endif
 #endif
 
@@ -397,14 +364,11 @@ struct uint128_extra
 *----------------------------------------------------------------------------*/
 #if defined INLINE_LEVEL && (4 <= INLINE_LEVEL)
 INLINE
-struct uint64_extra
+struct uint64_extra softfloat_shiftRightJam64Extra(uint64_t a, uint64_t extra, uint_fast32_t dist) {
- softfloat_shiftRightJam64Extra(
-     uint64_t a, uint64_t extra, uint_fast32_t dist )
-{
     struct uint64_extra z;
-    if ( dist < 64 ) {
+    if(dist < 64) {
-        z.v = a>>dist;
+        z.v = a >> dist;
-        z.extra = a<<(-dist & 63);
+        z.extra = a << (-dist & 63);
     } else {
         z.v = 0;
         z.extra = (dist == 64) ? a : (a != 0);
@@ -413,9 +377,7 @@ struct uint64_extra
     return z;
 }
 #else
-struct uint64_extra
+struct uint64_extra softfloat_shiftRightJam64Extra(uint64_t a, uint64_t extra, uint_fast32_t dist);
- softfloat_shiftRightJam64Extra(
-     uint64_t a, uint64_t extra, uint_fast32_t dist );
 #endif
 #endif
 
@@ -430,8 +392,7 @@ struct uint64_extra
 | greater than 128, the result will be either 0 or 1, depending on whether the
 | original 128 bits are all zeros.
 *----------------------------------------------------------------------------*/
-struct uint128
+struct uint128 softfloat_shiftRightJam128(uint64_t a64, uint64_t a0, uint_fast32_t dist);
- softfloat_shiftRightJam128( uint64_t a64, uint64_t a0, uint_fast32_t dist );
 #endif
 
 #ifndef softfloat_shiftRightJam128Extra
@@ -452,9 +413,7 @@ struct uint128
 | is modified as described above and returned in the 'extra' field of the
 | result.)
 *----------------------------------------------------------------------------*/
-struct uint128_extra
+struct uint128_extra softfloat_shiftRightJam128Extra(uint64_t a64, uint64_t a0, uint64_t extra, uint_fast32_t dist);
- softfloat_shiftRightJam128Extra(
-     uint64_t a64, uint64_t a0, uint64_t extra, uint_fast32_t dist );
 #endif
 
 #ifndef softfloat_shiftRightJam256M
@@ -470,9 +429,7 @@ struct uint128_extra
 | is greater than 256, the stored result will be either 0 or 1, depending on
 | whether the original 256 bits are all zeros.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_shiftRightJam256M(const uint64_t* aPtr, uint_fast32_t dist, uint64_t* zPtr);
- softfloat_shiftRightJam256M(
-     const uint64_t *aPtr, uint_fast32_t dist, uint64_t *zPtr );
 #endif
 
 #ifndef softfloat_add128
@@ -483,17 +440,14 @@ void
 *----------------------------------------------------------------------------*/
 #if defined INLINE_LEVEL && (2 <= INLINE_LEVEL)
 INLINE
-struct uint128
+struct uint128 softfloat_add128(uint64_t a64, uint64_t a0, uint64_t b64, uint64_t b0) {
- softfloat_add128( uint64_t a64, uint64_t a0, uint64_t b64, uint64_t b0 )
-{
     struct uint128 z;
     z.v0 = a0 + b0;
     z.v64 = a64 + b64 + (z.v0 < a0);
     return z;
 }
 #else
-struct uint128
+struct uint128 softfloat_add128(uint64_t a64, uint64_t a0, uint64_t b64, uint64_t b0);
- softfloat_add128( uint64_t a64, uint64_t a0, uint64_t b64, uint64_t b0 );
 #endif
 #endif
 
@@ -505,9 +459,7 @@ struct uint128
 | an array of four 64-bit elements that concatenate in the platform's normal
 | endian order to form a 256-bit integer.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_add256M(const uint64_t* aPtr, const uint64_t* bPtr, uint64_t* zPtr);
- softfloat_add256M(
-     const uint64_t *aPtr, const uint64_t *bPtr, uint64_t *zPtr );
 #endif
 
 #ifndef softfloat_sub128
@@ -518,9 +470,7 @@ void
 *----------------------------------------------------------------------------*/
 #if defined INLINE_LEVEL && (2 <= INLINE_LEVEL)
 INLINE
-struct uint128
+struct uint128 softfloat_sub128(uint64_t a64, uint64_t a0, uint64_t b64, uint64_t b0) {
- softfloat_sub128( uint64_t a64, uint64_t a0, uint64_t b64, uint64_t b0 )
-{
     struct uint128 z;
     z.v0 = a0 - b0;
     z.v64 = a64 - b64;
@@ -528,8 +478,7 @@ struct uint128
     return z;
 }
 #else
-struct uint128
+struct uint128 softfloat_sub128(uint64_t a64, uint64_t a0, uint64_t b64, uint64_t b0);
- softfloat_sub128( uint64_t a64, uint64_t a0, uint64_t b64, uint64_t b0 );
 #endif
 #endif
 
@@ -542,9 +491,7 @@ struct uint128
 | 64-bit elements that concatenate in the platform's normal endian order to
 | form a 256-bit integer.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_sub256M(const uint64_t* aPtr, const uint64_t* bPtr, uint64_t* zPtr);
- softfloat_sub256M(
-     const uint64_t *aPtr, const uint64_t *bPtr, uint64_t *zPtr );
 #endif
 
 #ifndef softfloat_mul64ByShifted32To128
@@ -552,17 +499,16 @@ void
 | Returns the 128-bit product of 'a', 'b', and 2^32.
 *----------------------------------------------------------------------------*/
 #if defined INLINE_LEVEL && (3 <= INLINE_LEVEL)
-INLINE struct uint128 softfloat_mul64ByShifted32To128( uint64_t a, uint32_t b )
+INLINE struct uint128 softfloat_mul64ByShifted32To128(uint64_t a, uint32_t b) {
-{
     uint_fast64_t mid;
     struct uint128 z;
-    mid = (uint_fast64_t) (uint32_t) a * b;
+    mid = (uint_fast64_t)(uint32_t)a * b;
-    z.v0 = mid<<32;
+    z.v0 = mid << 32;
-    z.v64 = (uint_fast64_t) (uint32_t) (a>>32) * b + (mid>>32);
+    z.v64 = (uint_fast64_t)(uint32_t)(a >> 32) * b + (mid >> 32);
     return z;
 }
 #else
-struct uint128 softfloat_mul64ByShifted32To128( uint64_t a, uint32_t b );
+struct uint128 softfloat_mul64ByShifted32To128(uint64_t a, uint32_t b);
 #endif
 #endif
 
@@ -570,7 +516,7 @@ struct uint128 softfloat_mul64ByShifted32To128( uint64_t a, uint32_t b );
 /*----------------------------------------------------------------------------
 | Returns the 128-bit product of 'a' and 'b'.
 *----------------------------------------------------------------------------*/
-struct uint128 softfloat_mul64To128( uint64_t a, uint64_t b );
+struct uint128 softfloat_mul64To128(uint64_t a, uint64_t b);
 #endif
 
 #ifndef softfloat_mul128By32
@@ -581,19 +527,18 @@ struct uint128 softfloat_mul64To128( uint64_t a, uint64_t b );
 *----------------------------------------------------------------------------*/
 #if defined INLINE_LEVEL && (4 <= INLINE_LEVEL)
 INLINE
-struct uint128 softfloat_mul128By32( uint64_t a64, uint64_t a0, uint32_t b )
+struct uint128 softfloat_mul128By32(uint64_t a64, uint64_t a0, uint32_t b) {
-{
     struct uint128 z;
     uint_fast64_t mid;
     uint_fast32_t carry;
     z.v0 = a0 * b;
-    mid = (uint_fast64_t) (uint32_t) (a0>>32) * b;
+    mid = (uint_fast64_t)(uint32_t)(a0 >> 32) * b;
-    carry = (uint32_t) ((uint_fast32_t) (z.v0>>32) - (uint_fast32_t) mid);
+    carry = (uint32_t)((uint_fast32_t)(z.v0 >> 32) - (uint_fast32_t)mid);
-    z.v64 = a64 * b + (uint_fast32_t) ((mid + carry)>>32);
+    z.v64 = a64 * b + (uint_fast32_t)((mid + carry) >> 32);
     return z;
 }
 #else
-struct uint128 softfloat_mul128By32( uint64_t a64, uint64_t a0, uint32_t b );
+struct uint128 softfloat_mul128By32(uint64_t a64, uint64_t a0, uint32_t b);
 #endif
 #endif
 
@@ -605,9 +550,7 @@ struct uint128 softfloat_mul128By32( uint64_t a64, uint64_t a0, uint32_t b );
 | Argument 'zPtr' points to an array of four 64-bit elements that concatenate
 | in the platform's normal endian order to form a 256-bit integer.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_mul128To256M(uint64_t a64, uint64_t a0, uint64_t b64, uint64_t b0, uint64_t* zPtr);
- softfloat_mul128To256M(
-     uint64_t a64, uint64_t a0, uint64_t b64, uint64_t b0, uint64_t *zPtr );
 #endif
 
 #else
@@ -626,7 +569,7 @@ void
 | Each of 'aPtr' and 'bPtr' points to an array of three 32-bit elements that
 | concatenate in the platform's normal endian order to form a 96-bit integer.
 *----------------------------------------------------------------------------*/
-int_fast8_t softfloat_compare96M( const uint32_t *aPtr, const uint32_t *bPtr );
+int_fast8_t softfloat_compare96M(const uint32_t* aPtr, const uint32_t* bPtr);
 #endif
 
 #ifndef softfloat_compare128M
@@ -638,8 +581,7 @@ int_fast8_t softfloat_compare96M( const uint32_t *aPtr, const uint32_t *bPtr );
 | Each of 'aPtr' and 'bPtr' points to an array of four 32-bit elements that
 | concatenate in the platform's normal endian order to form a 128-bit integer.
 *----------------------------------------------------------------------------*/
-int_fast8_t
+int_fast8_t softfloat_compare128M(const uint32_t* aPtr, const uint32_t* bPtr);
- softfloat_compare128M( const uint32_t *aPtr, const uint32_t *bPtr );
 #endif
 
 #ifndef softfloat_shortShiftLeft64To96M
@@ -652,19 +594,14 @@ int_fast8_t
 *----------------------------------------------------------------------------*/
 #if defined INLINE_LEVEL && (2 <= INLINE_LEVEL)
 INLINE
-void
+void softfloat_shortShiftLeft64To96M(uint64_t a, uint_fast8_t dist, uint32_t* zPtr) {
- softfloat_shortShiftLeft64To96M(
+    zPtr[indexWord(3, 0)] = (uint32_t)a << dist;
-     uint64_t a, uint_fast8_t dist, uint32_t *zPtr )
-{
-    zPtr[indexWord( 3, 0 )] = (uint32_t) a<<dist;
     a >>= 32 - dist;
-    zPtr[indexWord( 3, 2 )] = a>>32;
+    zPtr[indexWord(3, 2)] = a >> 32;
-    zPtr[indexWord( 3, 1 )] = a;
+    zPtr[indexWord(3, 1)] = a;
 }
 #else
-void
+void softfloat_shortShiftLeft64To96M(uint64_t a, uint_fast8_t dist, uint32_t* zPtr);
- softfloat_shortShiftLeft64To96M(
-     uint64_t a, uint_fast8_t dist, uint32_t *zPtr );
 #endif
 #endif
 
@@ -678,13 +615,7 @@ void
 | that concatenate in the platform's normal endian order to form an N-bit
 | integer.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_shortShiftLeftM(uint_fast8_t size_words, const uint32_t* aPtr, uint_fast8_t dist, uint32_t* zPtr);
- softfloat_shortShiftLeftM(
-     uint_fast8_t size_words,
-     const uint32_t *aPtr,
-     uint_fast8_t dist,
-     uint32_t *zPtr
- );
 #endif
 
 #ifndef softfloat_shortShiftLeft96M
@@ -692,7 +623,7 @@ void
 | This function or macro is the same as 'softfloat_shortShiftLeftM' with
 | 'size_words' = 3 (N = 96).
 *----------------------------------------------------------------------------*/
-#define softfloat_shortShiftLeft96M( aPtr, dist, zPtr ) softfloat_shortShiftLeftM( 3, aPtr, dist, zPtr )
+#define softfloat_shortShiftLeft96M(aPtr, dist, zPtr) softfloat_shortShiftLeftM(3, aPtr, dist, zPtr)
 #endif
 
 #ifndef softfloat_shortShiftLeft128M
@@ -700,7 +631,7 @@ void
 | This function or macro is the same as 'softfloat_shortShiftLeftM' with
 | 'size_words' = 4 (N = 128).
 *----------------------------------------------------------------------------*/
-#define softfloat_shortShiftLeft128M( aPtr, dist, zPtr ) softfloat_shortShiftLeftM( 4, aPtr, dist, zPtr )
+#define softfloat_shortShiftLeft128M(aPtr, dist, zPtr) softfloat_shortShiftLeftM(4, aPtr, dist, zPtr)
 #endif
 
 #ifndef softfloat_shortShiftLeft160M
@@ -708,7 +639,7 @@ void
 | This function or macro is the same as 'softfloat_shortShiftLeftM' with
 | 'size_words' = 5 (N = 160).
 *----------------------------------------------------------------------------*/
-#define softfloat_shortShiftLeft160M( aPtr, dist, zPtr ) softfloat_shortShiftLeftM( 5, aPtr, dist, zPtr )
+#define softfloat_shortShiftLeft160M(aPtr, dist, zPtr) softfloat_shortShiftLeftM(5, aPtr, dist, zPtr)
 #endif
 
 #ifndef softfloat_shiftLeftM
@@ -722,13 +653,7 @@ void
 |   The value of 'dist' can be arbitrarily large.  In particular, if 'dist' is
 | greater than N, the stored result will be 0.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_shiftLeftM(uint_fast8_t size_words, const uint32_t* aPtr, uint32_t dist, uint32_t* zPtr);
- softfloat_shiftLeftM(
-     uint_fast8_t size_words,
-     const uint32_t *aPtr,
-     uint32_t dist,
-     uint32_t *zPtr
- );
 #endif
 
 #ifndef softfloat_shiftLeft96M
@@ -736,7 +661,7 @@ void
 | This function or macro is the same as 'softfloat_shiftLeftM' with
 | 'size_words' = 3 (N = 96).
 *----------------------------------------------------------------------------*/
-#define softfloat_shiftLeft96M( aPtr, dist, zPtr ) softfloat_shiftLeftM( 3, aPtr, dist, zPtr )
+#define softfloat_shiftLeft96M(aPtr, dist, zPtr) softfloat_shiftLeftM(3, aPtr, dist, zPtr)
 #endif
 
 #ifndef softfloat_shiftLeft128M
@@ -744,7 +669,7 @@ void
 | This function or macro is the same as 'softfloat_shiftLeftM' with
 | 'size_words' = 4 (N = 128).
 *----------------------------------------------------------------------------*/
-#define softfloat_shiftLeft128M( aPtr, dist, zPtr ) softfloat_shiftLeftM( 4, aPtr, dist, zPtr )
+#define softfloat_shiftLeft128M(aPtr, dist, zPtr) softfloat_shiftLeftM(4, aPtr, dist, zPtr)
 #endif
 
 #ifndef softfloat_shiftLeft160M
@@ -752,7 +677,7 @@ void
 | This function or macro is the same as 'softfloat_shiftLeftM' with
 | 'size_words' = 5 (N = 160).
 *----------------------------------------------------------------------------*/
-#define softfloat_shiftLeft160M( aPtr, dist, zPtr ) softfloat_shiftLeftM( 5, aPtr, dist, zPtr )
+#define softfloat_shiftLeft160M(aPtr, dist, zPtr) softfloat_shiftLeftM(5, aPtr, dist, zPtr)
 #endif
 
 #ifndef softfloat_shortShiftRightM
@@ -765,13 +690,7 @@ void
 | that concatenate in the platform's normal endian order to form an N-bit
 | integer.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_shortShiftRightM(uint_fast8_t size_words, const uint32_t* aPtr, uint_fast8_t dist, uint32_t* zPtr);
- softfloat_shortShiftRightM(
-     uint_fast8_t size_words,
-     const uint32_t *aPtr,
-     uint_fast8_t dist,
-     uint32_t *zPtr
- );
 #endif
 
 #ifndef softfloat_shortShiftRight128M
@@ -779,7 +698,7 @@ void
 | This function or macro is the same as 'softfloat_shortShiftRightM' with
 | 'size_words' = 4 (N = 128).
 *----------------------------------------------------------------------------*/
-#define softfloat_shortShiftRight128M( aPtr, dist, zPtr ) softfloat_shortShiftRightM( 4, aPtr, dist, zPtr )
+#define softfloat_shortShiftRight128M(aPtr, dist, zPtr) softfloat_shortShiftRightM(4, aPtr, dist, zPtr)
 #endif
 
 #ifndef softfloat_shortShiftRight160M
@@ -787,7 +706,7 @@ void
 | This function or macro is the same as 'softfloat_shortShiftRightM' with
 | 'size_words' = 5 (N = 160).
 *----------------------------------------------------------------------------*/
-#define softfloat_shortShiftRight160M( aPtr, dist, zPtr ) softfloat_shortShiftRightM( 5, aPtr, dist, zPtr )
+#define softfloat_shortShiftRight160M(aPtr, dist, zPtr) softfloat_shortShiftRightM(5, aPtr, dist, zPtr)
 #endif
 
 #ifndef softfloat_shortShiftRightJamM
@@ -801,9 +720,7 @@ void
 | to a 'size_words'-long array of 32-bit elements that concatenate in the
 | platform's normal endian order to form an N-bit integer.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_shortShiftRightJamM(uint_fast8_t, const uint32_t*, uint_fast8_t, uint32_t*);
- softfloat_shortShiftRightJamM(
-     uint_fast8_t, const uint32_t *, uint_fast8_t, uint32_t * );
 #endif
 
 #ifndef softfloat_shortShiftRightJam160M
@@ -811,7 +728,7 @@ void
 | This function or macro is the same as 'softfloat_shortShiftRightJamM' with
 | 'size_words' = 5 (N = 160).
 *----------------------------------------------------------------------------*/
-#define softfloat_shortShiftRightJam160M( aPtr, dist, zPtr ) softfloat_shortShiftRightJamM( 5, aPtr, dist, zPtr )
+#define softfloat_shortShiftRightJam160M(aPtr, dist, zPtr) softfloat_shortShiftRightJamM(5, aPtr, dist, zPtr)
 #endif
 
 #ifndef softfloat_shiftRightM
@@ -825,13 +742,7 @@ void
 |   The value of 'dist' can be arbitrarily large.  In particular, if 'dist' is
 | greater than N, the stored result will be 0.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_shiftRightM(uint_fast8_t size_words, const uint32_t* aPtr, uint32_t dist, uint32_t* zPtr);
- softfloat_shiftRightM(
-     uint_fast8_t size_words,
-     const uint32_t *aPtr,
-     uint32_t dist,
-     uint32_t *zPtr
- );
 #endif
 
 #ifndef softfloat_shiftRight96M
@@ -839,7 +750,7 @@ void
 | This function or macro is the same as 'softfloat_shiftRightM' with
 | 'size_words' = 3 (N = 96).
 *----------------------------------------------------------------------------*/
-#define softfloat_shiftRight96M( aPtr, dist, zPtr ) softfloat_shiftRightM( 3, aPtr, dist, zPtr )
+#define softfloat_shiftRight96M(aPtr, dist, zPtr) softfloat_shiftRightM(3, aPtr, dist, zPtr)
 #endif
 
 #ifndef softfloat_shiftRightJamM
@@ -856,13 +767,7 @@ void
 | is greater than N, the stored result will be either 0 or 1, depending on
 | whether the original N bits are all zeros.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_shiftRightJamM(uint_fast8_t size_words, const uint32_t* aPtr, uint32_t dist, uint32_t* zPtr);
- softfloat_shiftRightJamM(
-     uint_fast8_t size_words,
-     const uint32_t *aPtr,
-     uint32_t dist,
-     uint32_t *zPtr
- );
 #endif
 
 #ifndef softfloat_shiftRightJam96M
@@ -870,7 +775,7 @@ void
 | This function or macro is the same as 'softfloat_shiftRightJamM' with
 | 'size_words' = 3 (N = 96).
 *----------------------------------------------------------------------------*/
-#define softfloat_shiftRightJam96M( aPtr, dist, zPtr ) softfloat_shiftRightJamM( 3, aPtr, dist, zPtr )
+#define softfloat_shiftRightJam96M(aPtr, dist, zPtr) softfloat_shiftRightJamM(3, aPtr, dist, zPtr)
 #endif
 
 #ifndef softfloat_shiftRightJam128M
@@ -878,7 +783,7 @@ void
 | This function or macro is the same as 'softfloat_shiftRightJamM' with
 | 'size_words' = 4 (N = 128).
 *----------------------------------------------------------------------------*/
-#define softfloat_shiftRightJam128M( aPtr, dist, zPtr ) softfloat_shiftRightJamM( 4, aPtr, dist, zPtr )
+#define softfloat_shiftRightJam128M(aPtr, dist, zPtr) softfloat_shiftRightJamM(4, aPtr, dist, zPtr)
 #endif
 
 #ifndef softfloat_shiftRightJam160M
@@ -886,7 +791,7 @@ void
 | This function or macro is the same as 'softfloat_shiftRightJamM' with
 | 'size_words' = 5 (N = 160).
 *----------------------------------------------------------------------------*/
-#define softfloat_shiftRightJam160M( aPtr, dist, zPtr ) softfloat_shiftRightJamM( 5, aPtr, dist, zPtr )
+#define softfloat_shiftRightJam160M(aPtr, dist, zPtr) softfloat_shiftRightJamM(5, aPtr, dist, zPtr)
 #endif
 
 #ifndef softfloat_addM
@@ -898,13 +803,7 @@ void
 | elements that concatenate in the platform's normal endian order to form an
 | N-bit integer.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_addM(uint_fast8_t size_words, const uint32_t* aPtr, const uint32_t* bPtr, uint32_t* zPtr);
- softfloat_addM(
-     uint_fast8_t size_words,
-     const uint32_t *aPtr,
-     const uint32_t *bPtr,
-     uint32_t *zPtr
- );
 #endif
 
 #ifndef softfloat_add96M
@@ -912,7 +811,7 @@ void
 | This function or macro is the same as 'softfloat_addM' with 'size_words'
 | = 3 (N = 96).
 *----------------------------------------------------------------------------*/
-#define softfloat_add96M( aPtr, bPtr, zPtr ) softfloat_addM( 3, aPtr, bPtr, zPtr )
+#define softfloat_add96M(aPtr, bPtr, zPtr) softfloat_addM(3, aPtr, bPtr, zPtr)
 #endif
 
 #ifndef softfloat_add128M
@@ -920,7 +819,7 @@ void
 | This function or macro is the same as 'softfloat_addM' with 'size_words'
 | = 4 (N = 128).
 *----------------------------------------------------------------------------*/
-#define softfloat_add128M( aPtr, bPtr, zPtr ) softfloat_addM( 4, aPtr, bPtr, zPtr )
+#define softfloat_add128M(aPtr, bPtr, zPtr) softfloat_addM(4, aPtr, bPtr, zPtr)
 #endif
 
 #ifndef softfloat_add160M
@@ -928,7 +827,7 @@ void
 | This function or macro is the same as 'softfloat_addM' with 'size_words'
 | = 5 (N = 160).
 *----------------------------------------------------------------------------*/
-#define softfloat_add160M( aPtr, bPtr, zPtr ) softfloat_addM( 5, aPtr, bPtr, zPtr )
+#define softfloat_add160M(aPtr, bPtr, zPtr) softfloat_addM(5, aPtr, bPtr, zPtr)
 #endif
 
 #ifndef softfloat_addCarryM
@@ -940,14 +839,7 @@ void
 | points to a 'size_words'-long array of 32-bit elements that concatenate in
 | the platform's normal endian order to form an N-bit integer.
 *----------------------------------------------------------------------------*/
-uint_fast8_t
+uint_fast8_t softfloat_addCarryM(uint_fast8_t size_words, const uint32_t* aPtr, const uint32_t* bPtr, uint_fast8_t carry, uint32_t* zPtr);
- softfloat_addCarryM(
-     uint_fast8_t size_words,
-     const uint32_t *aPtr,
-     const uint32_t *bPtr,
-     uint_fast8_t carry,
-     uint32_t *zPtr
- );
 #endif
 
 #ifndef softfloat_addComplCarryM
@@ -956,14 +848,8 @@ uint_fast8_t
 | the value of the unsigned integer pointed to by 'bPtr' is bit-wise completed
 | before the addition.
 *----------------------------------------------------------------------------*/
-uint_fast8_t
+uint_fast8_t softfloat_addComplCarryM(uint_fast8_t size_words, const uint32_t* aPtr, const uint32_t* bPtr, uint_fast8_t carry,
- softfloat_addComplCarryM(
+                                      uint32_t* zPtr);
-     uint_fast8_t size_words,
-     const uint32_t *aPtr,
-     const uint32_t *bPtr,
-     uint_fast8_t carry,
-     uint32_t *zPtr
- );
 #endif
 
 #ifndef softfloat_addComplCarry96M
@@ -971,7 +857,7 @@ uint_fast8_t
 | This function or macro is the same as 'softfloat_addComplCarryM' with
 | 'size_words' = 3 (N = 96).
 *----------------------------------------------------------------------------*/
-#define softfloat_addComplCarry96M( aPtr, bPtr, carry, zPtr ) softfloat_addComplCarryM( 3, aPtr, bPtr, carry, zPtr )
+#define softfloat_addComplCarry96M(aPtr, bPtr, carry, zPtr) softfloat_addComplCarryM(3, aPtr, bPtr, carry, zPtr)
 #endif
 
 #ifndef softfloat_negXM
@@ -981,7 +867,7 @@ uint_fast8_t
 | points to a 'size_words'-long array of 32-bit elements that concatenate in
 | the platform's normal endian order to form an N-bit integer.
 *----------------------------------------------------------------------------*/
-void softfloat_negXM( uint_fast8_t size_words, uint32_t *zPtr );
+void softfloat_negXM(uint_fast8_t size_words, uint32_t* zPtr);
 #endif
 
 #ifndef softfloat_negX96M
@@ -989,7 +875,7 @@ void softfloat_negXM( uint_fast8_t size_words, uint32_t *zPtr );
 | This function or macro is the same as 'softfloat_negXM' with 'size_words'
 | = 3 (N = 96).
 *----------------------------------------------------------------------------*/
-#define softfloat_negX96M( zPtr ) softfloat_negXM( 3, zPtr )
+#define softfloat_negX96M(zPtr) softfloat_negXM(3, zPtr)
 #endif
 
 #ifndef softfloat_negX128M
@@ -997,7 +883,7 @@ void softfloat_negXM( uint_fast8_t size_words, uint32_t *zPtr );
 | This function or macro is the same as 'softfloat_negXM' with 'size_words'
 | = 4 (N = 128).
 *----------------------------------------------------------------------------*/
-#define softfloat_negX128M( zPtr ) softfloat_negXM( 4, zPtr )
+#define softfloat_negX128M(zPtr) softfloat_negXM(4, zPtr)
 #endif
 
 #ifndef softfloat_negX160M
@@ -1005,7 +891,7 @@ void softfloat_negXM( uint_fast8_t size_words, uint32_t *zPtr );
 | This function or macro is the same as 'softfloat_negXM' with 'size_words'
 | = 5 (N = 160).
 *----------------------------------------------------------------------------*/
-#define softfloat_negX160M( zPtr ) softfloat_negXM( 5, zPtr )
+#define softfloat_negX160M(zPtr) softfloat_negXM(5, zPtr)
 #endif
 
 #ifndef softfloat_negX256M
@@ -1013,7 +899,7 @@ void softfloat_negXM( uint_fast8_t size_words, uint32_t *zPtr );
 | This function or macro is the same as 'softfloat_negXM' with 'size_words'
 | = 8 (N = 256).
 *----------------------------------------------------------------------------*/
-#define softfloat_negX256M( zPtr ) softfloat_negXM( 8, zPtr )
+#define softfloat_negX256M(zPtr) softfloat_negXM(8, zPtr)
 #endif
 
 #ifndef softfloat_sub1XM
@@ -1024,7 +910,7 @@ void softfloat_negXM( uint_fast8_t size_words, uint32_t *zPtr );
 | elements that concatenate in the platform's normal endian order to form an
 | N-bit integer.
 *----------------------------------------------------------------------------*/
-void softfloat_sub1XM( uint_fast8_t size_words, uint32_t *zPtr );
+void softfloat_sub1XM(uint_fast8_t size_words, uint32_t* zPtr);
 #endif
 
 #ifndef softfloat_sub1X96M
@@ -1032,7 +918,7 @@ void softfloat_sub1XM( uint_fast8_t size_words, uint32_t *zPtr );
 | This function or macro is the same as 'softfloat_sub1XM' with 'size_words'
 | = 3 (N = 96).
 *----------------------------------------------------------------------------*/
-#define softfloat_sub1X96M( zPtr ) softfloat_sub1XM( 3, zPtr )
+#define softfloat_sub1X96M(zPtr) softfloat_sub1XM(3, zPtr)
 #endif
 
 #ifndef softfloat_sub1X160M
@@ -1040,7 +926,7 @@ void softfloat_sub1XM( uint_fast8_t size_words, uint32_t *zPtr );
 | This function or macro is the same as 'softfloat_sub1XM' with 'size_words'
 | = 5 (N = 160).
 *----------------------------------------------------------------------------*/
-#define softfloat_sub1X160M( zPtr ) softfloat_sub1XM( 5, zPtr )
+#define softfloat_sub1X160M(zPtr) softfloat_sub1XM(5, zPtr)
 #endif
 
 #ifndef softfloat_subM
@@ -1052,13 +938,7 @@ void softfloat_sub1XM( uint_fast8_t size_words, uint32_t *zPtr );
 | array of 32-bit elements that concatenate in the platform's normal endian
 | order to form an N-bit integer.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_subM(uint_fast8_t size_words, const uint32_t* aPtr, const uint32_t* bPtr, uint32_t* zPtr);
- softfloat_subM(
-     uint_fast8_t size_words,
-     const uint32_t *aPtr,
-     const uint32_t *bPtr,
-     uint32_t *zPtr
- );
 #endif
 
 #ifndef softfloat_sub96M
@@ -1066,7 +946,7 @@ void
 | This function or macro is the same as 'softfloat_subM' with 'size_words'
 | = 3 (N = 96).
 *----------------------------------------------------------------------------*/
-#define softfloat_sub96M( aPtr, bPtr, zPtr ) softfloat_subM( 3, aPtr, bPtr, zPtr )
+#define softfloat_sub96M(aPtr, bPtr, zPtr) softfloat_subM(3, aPtr, bPtr, zPtr)
 #endif
 
 #ifndef softfloat_sub128M
@@ -1074,7 +954,7 @@ void
 | This function or macro is the same as 'softfloat_subM' with 'size_words'
 | = 4 (N = 128).
 *----------------------------------------------------------------------------*/
-#define softfloat_sub128M( aPtr, bPtr, zPtr ) softfloat_subM( 4, aPtr, bPtr, zPtr )
+#define softfloat_sub128M(aPtr, bPtr, zPtr) softfloat_subM(4, aPtr, bPtr, zPtr)
 #endif
 
 #ifndef softfloat_sub160M
@@ -1082,7 +962,7 @@ void
 | This function or macro is the same as 'softfloat_subM' with 'size_words'
 | = 5 (N = 160).
 *----------------------------------------------------------------------------*/
-#define softfloat_sub160M( aPtr, bPtr, zPtr ) softfloat_subM( 5, aPtr, bPtr, zPtr )
+#define softfloat_sub160M(aPtr, bPtr, zPtr) softfloat_subM(5, aPtr, bPtr, zPtr)
 #endif
 
 #ifndef softfloat_mul64To128M
@@ -1092,7 +972,7 @@ void
 | elements that concatenate in the platform's normal endian order to form a
 | 128-bit integer.
 *----------------------------------------------------------------------------*/
-void softfloat_mul64To128M( uint64_t a, uint64_t b, uint32_t *zPtr );
+void softfloat_mul64To128M(uint64_t a, uint64_t b, uint32_t* zPtr);
 #endif
 
 #ifndef softfloat_mul128MTo256M
@@ -1104,9 +984,7 @@ void softfloat_mul64To128M( uint64_t a, uint64_t b, uint32_t *zPtr );
 | Argument 'zPtr' points to an array of eight 32-bit elements that concatenate
 | to form a 256-bit integer.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_mul128MTo256M(const uint32_t* aPtr, const uint32_t* bPtr, uint32_t* zPtr);
- softfloat_mul128MTo256M(
-     const uint32_t *aPtr, const uint32_t *bPtr, uint32_t *zPtr );
 #endif
 
 #ifndef softfloat_remStepMBy32
@@ -1119,15 +997,8 @@ void
 | to a 'size_words'-long array of 32-bit elements that concatenate in the
 | platform's normal endian order to form an N-bit integer.
 *----------------------------------------------------------------------------*/
-void
+void softfloat_remStepMBy32(uint_fast8_t size_words, const uint32_t* remPtr, uint_fast8_t dist, const uint32_t* bPtr, uint32_t q,
- softfloat_remStepMBy32(
+                            uint32_t* zPtr);
-     uint_fast8_t size_words,
-     const uint32_t *remPtr,
-     uint_fast8_t dist,
-     const uint32_t *bPtr,
-     uint32_t q,
-     uint32_t *zPtr
- );
 #endif
 
 #ifndef softfloat_remStep96MBy32
@@ -1135,7 +1006,7 @@ void
 | This function or macro is the same as 'softfloat_remStepMBy32' with
 | 'size_words' = 3 (N = 96).
 *----------------------------------------------------------------------------*/
-#define softfloat_remStep96MBy32( remPtr, dist, bPtr, q, zPtr ) softfloat_remStepMBy32( 3, remPtr, dist, bPtr, q, zPtr )
+#define softfloat_remStep96MBy32(remPtr, dist, bPtr, q, zPtr) softfloat_remStepMBy32(3, remPtr, dist, bPtr, q, zPtr)
 #endif
 
 #ifndef softfloat_remStep128MBy32
@@ -1143,7 +1014,7 @@ void
 | This function or macro is the same as 'softfloat_remStepMBy32' with
 | 'size_words' = 4 (N = 128).
 *----------------------------------------------------------------------------*/
-#define softfloat_remStep128MBy32( remPtr, dist, bPtr, q, zPtr ) softfloat_remStepMBy32( 4, remPtr, dist, bPtr, q, zPtr )
+#define softfloat_remStep128MBy32(remPtr, dist, bPtr, q, zPtr) softfloat_remStepMBy32(4, remPtr, dist, bPtr, q, zPtr)
 #endif
 
 #ifndef softfloat_remStep160MBy32
@@ -1151,10 +1022,9 @@ void
 | This function or macro is the same as 'softfloat_remStepMBy32' with
 | 'size_words' = 5 (N = 160).
 *----------------------------------------------------------------------------*/
-#define softfloat_remStep160MBy32( remPtr, dist, bPtr, q, zPtr ) softfloat_remStepMBy32( 5, remPtr, dist, bPtr, q, zPtr )
+#define softfloat_remStep160MBy32(remPtr, dist, bPtr, q, zPtr) softfloat_remStepMBy32(5, remPtr, dist, bPtr, q, zPtr)
 #endif
 
 #endif
 
 #endif
-

softfloat/source/include/softfloat.h

@@ -34,7 +34,6 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 =============================================================================*/
 
-
 /*============================================================================
 | Note:  If SoftFloat is made available as a general library for programs to
 | use, it is strongly recommended that a platform-specific version of this
@@ -42,13 +41,12 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 | eliminates all dependencies on compile-time macros.
 *============================================================================*/
 
-
 #ifndef softfloat_h
 #define softfloat_h 1
 
+#include "softfloat_types.h"
 #include <stdbool.h>
 #include <stdint.h>
-#include "softfloat_types.h"
 
 #ifndef THREAD_LOCAL
 #define THREAD_LOCAL
@@ -58,10 +56,7 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 | Software floating-point underflow tininess-detection mode.
 *----------------------------------------------------------------------------*/
 extern THREAD_LOCAL uint_fast8_t softfloat_detectTininess;
-enum {
+enum { softfloat_tininess_beforeRounding = 0, softfloat_tininess_afterRounding = 1 };
-    softfloat_tininess_beforeRounding = 0,
-    softfloat_tininess_afterRounding  = 1
-};
 
 /*----------------------------------------------------------------------------
 | Software floating-point rounding mode.  (Mode "odd" is supported only if
@@ -81,163 +76,170 @@ enum {
 | Software floating-point exception flags.
 *----------------------------------------------------------------------------*/
 extern THREAD_LOCAL uint_fast8_t softfloat_exceptionFlags;
-enum {
+typedef enum {
     softfloat_flag_inexact = 1,
     softfloat_flag_underflow = 2,
     softfloat_flag_overflow = 4,
     softfloat_flag_infinite = 8,
     softfloat_flag_invalid = 16
-};
+} exceptionFlag_t;
 
 /*----------------------------------------------------------------------------
 | Routine to raise any or all of the software floating-point exception flags.
 *----------------------------------------------------------------------------*/
-void softfloat_raiseFlags( uint_fast8_t );
+void softfloat_raiseFlags(uint_fast8_t);
 
 /*----------------------------------------------------------------------------
 | Integer-to-floating-point conversion routines.
 *----------------------------------------------------------------------------*/
-float16_t ui32_to_f16( uint32_t );
+float16_t ui32_to_f16(uint32_t);
-float32_t ui32_to_f32( uint32_t );
+float32_t ui32_to_f32(uint32_t);
-float64_t ui32_to_f64( uint32_t );
+float64_t ui32_to_f64(uint32_t);
 #ifdef SOFTFLOAT_FAST_INT64
-extFloat80_t ui32_to_extF80( uint32_t );
+extFloat80_t ui32_to_extF80(uint32_t);
-float128_t ui32_to_f128( uint32_t );
+float128_t ui32_to_f128(uint32_t);
 #endif
-void ui32_to_extF80M( uint32_t, extFloat80_t * );
+void ui32_to_extF80M(uint32_t, extFloat80_t*);
-void ui32_to_f128M( uint32_t, float128_t * );
+void ui32_to_f128M(uint32_t, float128_t*);
-float16_t ui64_to_f16( uint64_t );
+float16_t ui64_to_f16(uint64_t);
-float32_t ui64_to_f32( uint64_t );
+float32_t ui64_to_f32(uint64_t);
-float64_t ui64_to_f64( uint64_t );
+float64_t ui64_to_f64(uint64_t);
 #ifdef SOFTFLOAT_FAST_INT64
-extFloat80_t ui64_to_extF80( uint64_t );
+extFloat80_t ui64_to_extF80(uint64_t);
-float128_t ui64_to_f128( uint64_t );
+float128_t ui64_to_f128(uint64_t);
 #endif
-void ui64_to_extF80M( uint64_t, extFloat80_t * );
+void ui64_to_extF80M(uint64_t, extFloat80_t*);
-void ui64_to_f128M( uint64_t, float128_t * );
+void ui64_to_f128M(uint64_t, float128_t*);
-float16_t i32_to_f16( int32_t );
+float16_t i32_to_f16(int32_t);
-float32_t i32_to_f32( int32_t );
+float32_t i32_to_f32(int32_t);
-float64_t i32_to_f64( int32_t );
+float64_t i32_to_f64(int32_t);
 #ifdef SOFTFLOAT_FAST_INT64
-extFloat80_t i32_to_extF80( int32_t );
+extFloat80_t i32_to_extF80(int32_t);
-float128_t i32_to_f128( int32_t );
+float128_t i32_to_f128(int32_t);
 #endif
-void i32_to_extF80M( int32_t, extFloat80_t * );
+void i32_to_extF80M(int32_t, extFloat80_t*);
-void i32_to_f128M( int32_t, float128_t * );
+void i32_to_f128M(int32_t, float128_t*);
-float16_t i64_to_f16( int64_t );
+float16_t i64_to_f16(int64_t);
-float32_t i64_to_f32( int64_t );
+float32_t i64_to_f32(int64_t);
-float64_t i64_to_f64( int64_t );
+float64_t i64_to_f64(int64_t);
 #ifdef SOFTFLOAT_FAST_INT64
-extFloat80_t i64_to_extF80( int64_t );
+extFloat80_t i64_to_extF80(int64_t);
-float128_t i64_to_f128( int64_t );
+float128_t i64_to_f128(int64_t);
 #endif
-void i64_to_extF80M( int64_t, extFloat80_t * );
+void i64_to_extF80M(int64_t, extFloat80_t*);
-void i64_to_f128M( int64_t, float128_t * );
+void i64_to_f128M(int64_t, float128_t*);
 
 /*----------------------------------------------------------------------------
 | 16-bit (half-precision) floating-point operations.
 *----------------------------------------------------------------------------*/
-uint_fast32_t f16_to_ui32( float16_t, uint_fast8_t, bool );
+uint_fast32_t f16_to_ui32(float16_t, uint_fast8_t, bool);
-uint_fast64_t f16_to_ui64( float16_t, uint_fast8_t, bool );
+uint_fast64_t f16_to_ui64(float16_t, uint_fast8_t, bool);
-int_fast32_t f16_to_i32( float16_t, uint_fast8_t, bool );
+int_fast32_t f16_to_i32(float16_t, uint_fast8_t, bool);
-int_fast64_t f16_to_i64( float16_t, uint_fast8_t, bool );
+int_fast64_t f16_to_i64(float16_t, uint_fast8_t, bool);
-uint_fast32_t f16_to_ui32_r_minMag( float16_t, bool );
+uint_fast32_t f16_to_ui32_r_minMag(float16_t, bool);
-uint_fast64_t f16_to_ui64_r_minMag( float16_t, bool );
+uint_fast64_t f16_to_ui64_r_minMag(float16_t, bool);
-int_fast32_t f16_to_i32_r_minMag( float16_t, bool );
+int_fast32_t f16_to_i32_r_minMag(float16_t, bool);
-int_fast64_t f16_to_i64_r_minMag( float16_t, bool );
+int_fast64_t f16_to_i64_r_minMag(float16_t, bool);
-float32_t f16_to_f32( float16_t );
+float32_t f16_to_f32(float16_t);
-float64_t f16_to_f64( float16_t );
+float64_t f16_to_f64(float16_t);
 #ifdef SOFTFLOAT_FAST_INT64
-extFloat80_t f16_to_extF80( float16_t );
+extFloat80_t f16_to_extF80(float16_t);
-float128_t f16_to_f128( float16_t );
+float128_t f16_to_f128(float16_t);
 #endif
-void f16_to_extF80M( float16_t, extFloat80_t * );
+void f16_to_extF80M(float16_t, extFloat80_t*);
-void f16_to_f128M( float16_t, float128_t * );
+void f16_to_f128M(float16_t, float128_t*);
-float16_t f16_roundToInt( float16_t, uint_fast8_t, bool );
+float16_t f16_roundToInt(float16_t, uint_fast8_t, bool);
-float16_t f16_add( float16_t, float16_t );
+float16_t f16_add(float16_t, float16_t);
-float16_t f16_sub( float16_t, float16_t );
+float16_t f16_sub(float16_t, float16_t);
-float16_t f16_mul( float16_t, float16_t );
+float16_t f16_mul(float16_t, float16_t);
-float16_t f16_mulAdd( float16_t, float16_t, float16_t );
+float16_t f16_mulAdd(float16_t, float16_t, float16_t);
-float16_t f16_div( float16_t, float16_t );
+float16_t f16_div(float16_t, float16_t);
-float16_t f16_rem( float16_t, float16_t );
+float16_t f16_rem(float16_t, float16_t);
-float16_t f16_sqrt( float16_t );
+float16_t f16_sqrt(float16_t);
-bool f16_eq( float16_t, float16_t );
+bool f16_eq(float16_t, float16_t);
-bool f16_le( float16_t, float16_t );
+bool f16_le(float16_t, float16_t);
-bool f16_lt( float16_t, float16_t );
+bool f16_lt(float16_t, float16_t);
-bool f16_eq_signaling( float16_t, float16_t );
+bool f16_eq_signaling(float16_t, float16_t);
-bool f16_le_quiet( float16_t, float16_t );
+bool f16_le_quiet(float16_t, float16_t);
-bool f16_lt_quiet( float16_t, float16_t );
+bool f16_lt_quiet(float16_t, float16_t);
-bool f16_isSignalingNaN( float16_t );
+bool f16_isSignalingNaN(float16_t);
+
+/*----------------------------------------------------------------------------
+| 16-bit (brain float 16) floating-point operations.
+*----------------------------------------------------------------------------*/
+float32_t bf16_to_f32(bfloat16_t);
+bfloat16_t f32_to_bf16(float32_t);
+bool bf16_isSignalingNaN(bfloat16_t);
 
 /*----------------------------------------------------------------------------
 | 32-bit (single-precision) floating-point operations.
 *----------------------------------------------------------------------------*/
-uint_fast32_t f32_to_ui32( float32_t, uint_fast8_t, bool );
+uint_fast32_t f32_to_ui32(float32_t, uint_fast8_t, bool);
-uint_fast64_t f32_to_ui64( float32_t, uint_fast8_t, bool );
+uint_fast64_t f32_to_ui64(float32_t, uint_fast8_t, bool);
-int_fast32_t f32_to_i32( float32_t, uint_fast8_t, bool );
+int_fast32_t f32_to_i32(float32_t, uint_fast8_t, bool);
-int_fast64_t f32_to_i64( float32_t, uint_fast8_t, bool );
+int_fast64_t f32_to_i64(float32_t, uint_fast8_t, bool);
-uint_fast32_t f32_to_ui32_r_minMag( float32_t, bool );
+uint_fast32_t f32_to_ui32_r_minMag(float32_t, bool);
-uint_fast64_t f32_to_ui64_r_minMag( float32_t, bool );
+uint_fast64_t f32_to_ui64_r_minMag(float32_t, bool);
-int_fast32_t f32_to_i32_r_minMag( float32_t, bool );
+int_fast32_t f32_to_i32_r_minMag(float32_t, bool);
-int_fast64_t f32_to_i64_r_minMag( float32_t, bool );
+int_fast64_t f32_to_i64_r_minMag(float32_t, bool);
-float16_t f32_to_f16( float32_t );
+float16_t f32_to_f16(float32_t);
-float64_t f32_to_f64( float32_t );
+float64_t f32_to_f64(float32_t);
 #ifdef SOFTFLOAT_FAST_INT64
-extFloat80_t f32_to_extF80( float32_t );
+extFloat80_t f32_to_extF80(float32_t);
-float128_t f32_to_f128( float32_t );
+float128_t f32_to_f128(float32_t);
 #endif
-void f32_to_extF80M( float32_t, extFloat80_t * );
+void f32_to_extF80M(float32_t, extFloat80_t*);
-void f32_to_f128M( float32_t, float128_t * );
+void f32_to_f128M(float32_t, float128_t*);
-float32_t f32_roundToInt( float32_t, uint_fast8_t, bool );
+float32_t f32_roundToInt(float32_t, uint_fast8_t, bool);
-float32_t f32_add( float32_t, float32_t );
+float32_t f32_add(float32_t, float32_t);
-float32_t f32_sub( float32_t, float32_t );
+float32_t f32_sub(float32_t, float32_t);
-float32_t f32_mul( float32_t, float32_t );
+float32_t f32_mul(float32_t, float32_t);
-float32_t f32_mulAdd( float32_t, float32_t, float32_t );
+float32_t f32_mulAdd(float32_t, float32_t, float32_t);
-float32_t f32_div( float32_t, float32_t );
+float32_t f32_div(float32_t, float32_t);
-float32_t f32_rem( float32_t, float32_t );
+float32_t f32_rem(float32_t, float32_t);
-float32_t f32_sqrt( float32_t );
+float32_t f32_sqrt(float32_t);
-bool f32_eq( float32_t, float32_t );
+bool f32_eq(float32_t, float32_t);
-bool f32_le( float32_t, float32_t );
+bool f32_le(float32_t, float32_t);
-bool f32_lt( float32_t, float32_t );
+bool f32_lt(float32_t, float32_t);
-bool f32_eq_signaling( float32_t, float32_t );
+bool f32_eq_signaling(float32_t, float32_t);
-bool f32_le_quiet( float32_t, float32_t );
+bool f32_le_quiet(float32_t, float32_t);
-bool f32_lt_quiet( float32_t, float32_t );
+bool f32_lt_quiet(float32_t, float32_t);
-bool f32_isSignalingNaN( float32_t );
+bool f32_isSignalingNaN(float32_t);
 
 /*----------------------------------------------------------------------------
 | 64-bit (double-precision) floating-point operations.
 *----------------------------------------------------------------------------*/
-uint_fast32_t f64_to_ui32( float64_t, uint_fast8_t, bool );
+uint_fast32_t f64_to_ui32(float64_t, uint_fast8_t, bool);
-uint_fast64_t f64_to_ui64( float64_t, uint_fast8_t, bool );
+uint_fast64_t f64_to_ui64(float64_t, uint_fast8_t, bool);
-int_fast32_t f64_to_i32( float64_t, uint_fast8_t, bool );
+int_fast32_t f64_to_i32(float64_t, uint_fast8_t, bool);
-int_fast64_t f64_to_i64( float64_t, uint_fast8_t, bool );
+int_fast64_t f64_to_i64(float64_t, uint_fast8_t, bool);
-uint_fast32_t f64_to_ui32_r_minMag( float64_t, bool );
+uint_fast32_t f64_to_ui32_r_minMag(float64_t, bool);
-uint_fast64_t f64_to_ui64_r_minMag( float64_t, bool );
+uint_fast64_t f64_to_ui64_r_minMag(float64_t, bool);
-int_fast32_t f64_to_i32_r_minMag( float64_t, bool );
+int_fast32_t f64_to_i32_r_minMag(float64_t, bool);
-int_fast64_t f64_to_i64_r_minMag( float64_t, bool );
+int_fast64_t f64_to_i64_r_minMag(float64_t, bool);
-float16_t f64_to_f16( float64_t );
+float16_t f64_to_f16(float64_t);
-float32_t f64_to_f32( float64_t );
+float32_t f64_to_f32(float64_t);
 #ifdef SOFTFLOAT_FAST_INT64
-extFloat80_t f64_to_extF80( float64_t );
+extFloat80_t f64_to_extF80(float64_t);
-float128_t f64_to_f128( float64_t );
+float128_t f64_to_f128(float64_t);
 #endif
-void f64_to_extF80M( float64_t, extFloat80_t * );
+void f64_to_extF80M(float64_t, extFloat80_t*);
-void f64_to_f128M( float64_t, float128_t * );
+void f64_to_f128M(float64_t, float128_t*);
-float64_t f64_roundToInt( float64_t, uint_fast8_t, bool );
+float64_t f64_roundToInt(float64_t, uint_fast8_t, bool);
-float64_t f64_add( float64_t, float64_t );
+float64_t f64_add(float64_t, float64_t);
-float64_t f64_sub( float64_t, float64_t );
+float64_t f64_sub(float64_t, float64_t);
-float64_t f64_mul( float64_t, float64_t );
+float64_t f64_mul(float64_t, float64_t);
-float64_t f64_mulAdd( float64_t, float64_t, float64_t );
+float64_t f64_mulAdd(float64_t, float64_t, float64_t);
-float64_t f64_div( float64_t, float64_t );
+float64_t f64_div(float64_t, float64_t);
-float64_t f64_rem( float64_t, float64_t );
+float64_t f64_rem(float64_t, float64_t);
-float64_t f64_sqrt( float64_t );
+float64_t f64_sqrt(float64_t);
-bool f64_eq( float64_t, float64_t );
+bool f64_eq(float64_t, float64_t);
-bool f64_le( float64_t, float64_t );
+bool f64_le(float64_t, float64_t);
-bool f64_lt( float64_t, float64_t );
+bool f64_lt(float64_t, float64_t);
-bool f64_eq_signaling( float64_t, float64_t );
+bool f64_eq_signaling(float64_t, float64_t);
-bool f64_le_quiet( float64_t, float64_t );
+bool f64_le_quiet(float64_t, float64_t);
-bool f64_lt_quiet( float64_t, float64_t );
+bool f64_lt_quiet(float64_t, float64_t);
-bool f64_isSignalingNaN( float64_t );
+bool f64_isSignalingNaN(float64_t);
 
 /*----------------------------------------------------------------------------
 | Rounding precision for 80-bit extended double-precision floating-point.
@@ -249,124 +251,118 @@ extern THREAD_LOCAL uint_fast8_t extF80_roundingPrecision;
 | 80-bit extended double-precision floating-point operations.
 *----------------------------------------------------------------------------*/
 #ifdef SOFTFLOAT_FAST_INT64
-uint_fast32_t extF80_to_ui32( extFloat80_t, uint_fast8_t, bool );
+uint_fast32_t extF80_to_ui32(extFloat80_t, uint_fast8_t, bool);
-uint_fast64_t extF80_to_ui64( extFloat80_t, uint_fast8_t, bool );
+uint_fast64_t extF80_to_ui64(extFloat80_t, uint_fast8_t, bool);
-int_fast32_t extF80_to_i32( extFloat80_t, uint_fast8_t, bool );
+int_fast32_t extF80_to_i32(extFloat80_t, uint_fast8_t, bool);
-int_fast64_t extF80_to_i64( extFloat80_t, uint_fast8_t, bool );
+int_fast64_t extF80_to_i64(extFloat80_t, uint_fast8_t, bool);
-uint_fast32_t extF80_to_ui32_r_minMag( extFloat80_t, bool );
+uint_fast32_t extF80_to_ui32_r_minMag(extFloat80_t, bool);
-uint_fast64_t extF80_to_ui64_r_minMag( extFloat80_t, bool );
+uint_fast64_t extF80_to_ui64_r_minMag(extFloat80_t, bool);
-int_fast32_t extF80_to_i32_r_minMag( extFloat80_t, bool );
+int_fast32_t extF80_to_i32_r_minMag(extFloat80_t, bool);
-int_fast64_t extF80_to_i64_r_minMag( extFloat80_t, bool );
+int_fast64_t extF80_to_i64_r_minMag(extFloat80_t, bool);
-float16_t extF80_to_f16( extFloat80_t );
+float16_t extF80_to_f16(extFloat80_t);
-float32_t extF80_to_f32( extFloat80_t );
+float32_t extF80_to_f32(extFloat80_t);
-float64_t extF80_to_f64( extFloat80_t );
+float64_t extF80_to_f64(extFloat80_t);
-float128_t extF80_to_f128( extFloat80_t );
+float128_t extF80_to_f128(extFloat80_t);
-extFloat80_t extF80_roundToInt( extFloat80_t, uint_fast8_t, bool );
+extFloat80_t extF80_roundToInt(extFloat80_t, uint_fast8_t, bool);
-extFloat80_t extF80_add( extFloat80_t, extFloat80_t );
+extFloat80_t extF80_add(extFloat80_t, extFloat80_t);
-extFloat80_t extF80_sub( extFloat80_t, extFloat80_t );
+extFloat80_t extF80_sub(extFloat80_t, extFloat80_t);
-extFloat80_t extF80_mul( extFloat80_t, extFloat80_t );
+extFloat80_t extF80_mul(extFloat80_t, extFloat80_t);
-extFloat80_t extF80_div( extFloat80_t, extFloat80_t );
+extFloat80_t extF80_div(extFloat80_t, extFloat80_t);
-extFloat80_t extF80_rem( extFloat80_t, extFloat80_t );
+extFloat80_t extF80_rem(extFloat80_t, extFloat80_t);
-extFloat80_t extF80_sqrt( extFloat80_t );
+extFloat80_t extF80_sqrt(extFloat80_t);
-bool extF80_eq( extFloat80_t, extFloat80_t );
+bool extF80_eq(extFloat80_t, extFloat80_t);
-bool extF80_le( extFloat80_t, extFloat80_t );
+bool extF80_le(extFloat80_t, extFloat80_t);
-bool extF80_lt( extFloat80_t, extFloat80_t );
+bool extF80_lt(extFloat80_t, extFloat80_t);
-bool extF80_eq_signaling( extFloat80_t, extFloat80_t );
+bool extF80_eq_signaling(extFloat80_t, extFloat80_t);
-bool extF80_le_quiet( extFloat80_t, extFloat80_t );
+bool extF80_le_quiet(extFloat80_t, extFloat80_t);
-bool extF80_lt_quiet( extFloat80_t, extFloat80_t );
+bool extF80_lt_quiet(extFloat80_t, extFloat80_t);
-bool extF80_isSignalingNaN( extFloat80_t );
+bool extF80_isSignalingNaN(extFloat80_t);
 #endif
-uint_fast32_t extF80M_to_ui32( const extFloat80_t *, uint_fast8_t, bool );
+uint_fast32_t extF80M_to_ui32(const extFloat80_t*, uint_fast8_t, bool);
-uint_fast64_t extF80M_to_ui64( const extFloat80_t *, uint_fast8_t, bool );
+uint_fast64_t extF80M_to_ui64(const extFloat80_t*, uint_fast8_t, bool);
-int_fast32_t extF80M_to_i32( const extFloat80_t *, uint_fast8_t, bool );
+int_fast32_t extF80M_to_i32(const extFloat80_t*, uint_fast8_t, bool);
-int_fast64_t extF80M_to_i64( const extFloat80_t *, uint_fast8_t, bool );
+int_fast64_t extF80M_to_i64(const extFloat80_t*, uint_fast8_t, bool);
-uint_fast32_t extF80M_to_ui32_r_minMag( const extFloat80_t *, bool );
+uint_fast32_t extF80M_to_ui32_r_minMag(const extFloat80_t*, bool);
-uint_fast64_t extF80M_to_ui64_r_minMag( const extFloat80_t *, bool );
+uint_fast64_t extF80M_to_ui64_r_minMag(const extFloat80_t*, bool);
-int_fast32_t extF80M_to_i32_r_minMag( const extFloat80_t *, bool );
+int_fast32_t extF80M_to_i32_r_minMag(const extFloat80_t*, bool);
-int_fast64_t extF80M_to_i64_r_minMag( const extFloat80_t *, bool );
+int_fast64_t extF80M_to_i64_r_minMag(const extFloat80_t*, bool);
-float16_t extF80M_to_f16( const extFloat80_t * );
+float16_t extF80M_to_f16(const extFloat80_t*);
-float32_t extF80M_to_f32( const extFloat80_t * );
+float32_t extF80M_to_f32(const extFloat80_t*);
-float64_t extF80M_to_f64( const extFloat80_t * );
+float64_t extF80M_to_f64(const extFloat80_t*);
-void extF80M_to_f128M( const extFloat80_t *, float128_t * );
+void extF80M_to_f128M(const extFloat80_t*, float128_t*);
-void
+void extF80M_roundToInt(const extFloat80_t*, uint_fast8_t, bool, extFloat80_t*);
- extF80M_roundToInt(
+void extF80M_add(const extFloat80_t*, const extFloat80_t*, extFloat80_t*);
-     const extFloat80_t *, uint_fast8_t, bool, extFloat80_t * );
+void extF80M_sub(const extFloat80_t*, const extFloat80_t*, extFloat80_t*);
-void extF80M_add( const extFloat80_t *, const extFloat80_t *, extFloat80_t * );
+void extF80M_mul(const extFloat80_t*, const extFloat80_t*, extFloat80_t*);
-void extF80M_sub( const extFloat80_t *, const extFloat80_t *, extFloat80_t * );
+void extF80M_div(const extFloat80_t*, const extFloat80_t*, extFloat80_t*);
-void extF80M_mul( const extFloat80_t *, const extFloat80_t *, extFloat80_t * );
+void extF80M_rem(const extFloat80_t*, const extFloat80_t*, extFloat80_t*);
-void extF80M_div( const extFloat80_t *, const extFloat80_t *, extFloat80_t * );
+void extF80M_sqrt(const extFloat80_t*, extFloat80_t*);
-void extF80M_rem( const extFloat80_t *, const extFloat80_t *, extFloat80_t * );
+bool extF80M_eq(const extFloat80_t*, const extFloat80_t*);
-void extF80M_sqrt( const extFloat80_t *, extFloat80_t * );
+bool extF80M_le(const extFloat80_t*, const extFloat80_t*);
-bool extF80M_eq( const extFloat80_t *, const extFloat80_t * );
+bool extF80M_lt(const extFloat80_t*, const extFloat80_t*);
-bool extF80M_le( const extFloat80_t *, const extFloat80_t * );
+bool extF80M_eq_signaling(const extFloat80_t*, const extFloat80_t*);
-bool extF80M_lt( const extFloat80_t *, const extFloat80_t * );
+bool extF80M_le_quiet(const extFloat80_t*, const extFloat80_t*);
-bool extF80M_eq_signaling( const extFloat80_t *, const extFloat80_t * );
+bool extF80M_lt_quiet(const extFloat80_t*, const extFloat80_t*);
-bool extF80M_le_quiet( const extFloat80_t *, const extFloat80_t * );
+bool extF80M_isSignalingNaN(const extFloat80_t*);
-bool extF80M_lt_quiet( const extFloat80_t *, const extFloat80_t * );
-bool extF80M_isSignalingNaN( const extFloat80_t * );
 
 /*----------------------------------------------------------------------------
 | 128-bit (quadruple-precision) floating-point operations.
 *----------------------------------------------------------------------------*/
 #ifdef SOFTFLOAT_FAST_INT64
-uint_fast32_t f128_to_ui32( float128_t, uint_fast8_t, bool );
+uint_fast32_t f128_to_ui32(float128_t, uint_fast8_t, bool);
-uint_fast64_t f128_to_ui64( float128_t, uint_fast8_t, bool );
+uint_fast64_t f128_to_ui64(float128_t, uint_fast8_t, bool);
-int_fast32_t f128_to_i32( float128_t, uint_fast8_t, bool );
+int_fast32_t f128_to_i32(float128_t, uint_fast8_t, bool);
-int_fast64_t f128_to_i64( float128_t, uint_fast8_t, bool );
+int_fast64_t f128_to_i64(float128_t, uint_fast8_t, bool);
-uint_fast32_t f128_to_ui32_r_minMag( float128_t, bool );
+uint_fast32_t f128_to_ui32_r_minMag(float128_t, bool);
-uint_fast64_t f128_to_ui64_r_minMag( float128_t, bool );
+uint_fast64_t f128_to_ui64_r_minMag(float128_t, bool);
-int_fast32_t f128_to_i32_r_minMag( float128_t, bool );
+int_fast32_t f128_to_i32_r_minMag(float128_t, bool);
-int_fast64_t f128_to_i64_r_minMag( float128_t, bool );
+int_fast64_t f128_to_i64_r_minMag(float128_t, bool);
-float16_t f128_to_f16( float128_t );
+float16_t f128_to_f16(float128_t);
-float32_t f128_to_f32( float128_t );
+float32_t f128_to_f32(float128_t);
-float64_t f128_to_f64( float128_t );
+float64_t f128_to_f64(float128_t);
-extFloat80_t f128_to_extF80( float128_t );
+extFloat80_t f128_to_extF80(float128_t);
-float128_t f128_roundToInt( float128_t, uint_fast8_t, bool );
+float128_t f128_roundToInt(float128_t, uint_fast8_t, bool);
-float128_t f128_add( float128_t, float128_t );
+float128_t f128_add(float128_t, float128_t);
-float128_t f128_sub( float128_t, float128_t );
+float128_t f128_sub(float128_t, float128_t);
-float128_t f128_mul( float128_t, float128_t );
+float128_t f128_mul(float128_t, float128_t);
-float128_t f128_mulAdd( float128_t, float128_t, float128_t );
+float128_t f128_mulAdd(float128_t, float128_t, float128_t);
-float128_t f128_div( float128_t, float128_t );
+float128_t f128_div(float128_t, float128_t);
-float128_t f128_rem( float128_t, float128_t );
+float128_t f128_rem(float128_t, float128_t);
-float128_t f128_sqrt( float128_t );
+float128_t f128_sqrt(float128_t);
-bool f128_eq( float128_t, float128_t );
+bool f128_eq(float128_t, float128_t);
-bool f128_le( float128_t, float128_t );
+bool f128_le(float128_t, float128_t);
-bool f128_lt( float128_t, float128_t );
+bool f128_lt(float128_t, float128_t);
-bool f128_eq_signaling( float128_t, float128_t );
+bool f128_eq_signaling(float128_t, float128_t);
-bool f128_le_quiet( float128_t, float128_t );
+bool f128_le_quiet(float128_t, float128_t);
-bool f128_lt_quiet( float128_t, float128_t );
+bool f128_lt_quiet(float128_t, float128_t);
-bool f128_isSignalingNaN( float128_t );
+bool f128_isSignalingNaN(float128_t);
 #endif
-uint_fast32_t f128M_to_ui32( const float128_t *, uint_fast8_t, bool );
+uint_fast32_t f128M_to_ui32(const float128_t*, uint_fast8_t, bool);
-uint_fast64_t f128M_to_ui64( const float128_t *, uint_fast8_t, bool );
+uint_fast64_t f128M_to_ui64(const float128_t*, uint_fast8_t, bool);
-int_fast32_t f128M_to_i32( const float128_t *, uint_fast8_t, bool );
+int_fast32_t f128M_to_i32(const float128_t*, uint_fast8_t, bool);
-int_fast64_t f128M_to_i64( const float128_t *, uint_fast8_t, bool );
+int_fast64_t f128M_to_i64(const float128_t*, uint_fast8_t, bool);
-uint_fast32_t f128M_to_ui32_r_minMag( const float128_t *, bool );
+uint_fast32_t f128M_to_ui32_r_minMag(const float128_t*, bool);
-uint_fast64_t f128M_to_ui64_r_minMag( const float128_t *, bool );
+uint_fast64_t f128M_to_ui64_r_minMag(const float128_t*, bool);
-int_fast32_t f128M_to_i32_r_minMag( const float128_t *, bool );
+int_fast32_t f128M_to_i32_r_minMag(const float128_t*, bool);
-int_fast64_t f128M_to_i64_r_minMag( const float128_t *, bool );
+int_fast64_t f128M_to_i64_r_minMag(const float128_t*, bool);
-float16_t f128M_to_f16( const float128_t * );
+float16_t f128M_to_f16(const float128_t*);
-float32_t f128M_to_f32( const float128_t * );
+float32_t f128M_to_f32(const float128_t*);
-float64_t f128M_to_f64( const float128_t * );
+float64_t f128M_to_f64(const float128_t*);
-void f128M_to_extF80M( const float128_t *, extFloat80_t * );
+void f128M_to_extF80M(const float128_t*, extFloat80_t*);
-void f128M_roundToInt( const float128_t *, uint_fast8_t, bool, float128_t * );
+void f128M_roundToInt(const float128_t*, uint_fast8_t, bool, float128_t*);
-void f128M_add( const float128_t *, const float128_t *, float128_t * );
+void f128M_add(const float128_t*, const float128_t*, float128_t*);
-void f128M_sub( const float128_t *, const float128_t *, float128_t * );
+void f128M_sub(const float128_t*, const float128_t*, float128_t*);
-void f128M_mul( const float128_t *, const float128_t *, float128_t * );
+void f128M_mul(const float128_t*, const float128_t*, float128_t*);
-void
+void f128M_mulAdd(const float128_t*, const float128_t*, const float128_t*, float128_t*);
- f128M_mulAdd(
+void f128M_div(const float128_t*, const float128_t*, float128_t*);
-     const float128_t *, const float128_t *, const float128_t *, float128_t *
+void f128M_rem(const float128_t*, const float128_t*, float128_t*);
- );
+void f128M_sqrt(const float128_t*, float128_t*);
-void f128M_div( const float128_t *, const float128_t *, float128_t * );
+bool f128M_eq(const float128_t*, const float128_t*);
-void f128M_rem( const float128_t *, const float128_t *, float128_t * );
+bool f128M_le(const float128_t*, const float128_t*);
-void f128M_sqrt( const float128_t *, float128_t * );
+bool f128M_lt(const float128_t*, const float128_t*);
-bool f128M_eq( const float128_t *, const float128_t * );
+bool f128M_eq_signaling(const float128_t*, const float128_t*);
-bool f128M_le( const float128_t *, const float128_t * );
+bool f128M_le_quiet(const float128_t*, const float128_t*);
-bool f128M_lt( const float128_t *, const float128_t * );
+bool f128M_lt_quiet(const float128_t*, const float128_t*);
-bool f128M_eq_signaling( const float128_t *, const float128_t * );
+bool f128M_isSignalingNaN(const float128_t*);
-bool f128M_le_quiet( const float128_t *, const float128_t * );
-bool f128M_lt_quiet( const float128_t *, const float128_t * );
-bool f128M_isSignalingNaN( const float128_t * );
 
 #endif
-

softfloat/source/include/softfloat_types.h

@@ -47,10 +47,21 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 | the types below may, if desired, be defined as aliases for the native types
 | (typically 'float' and 'double', and possibly 'long double').
 *----------------------------------------------------------------------------*/
-typedef struct { uint16_t v; } float16_t;
+typedef struct {
-typedef struct { uint32_t v; } float32_t;
+    uint16_t v;
-typedef struct { uint64_t v; } float64_t;
+} float16_t;
-typedef struct { uint64_t v[2]; } float128_t;
+typedef struct {
+    uint16_t v;
+} bfloat16_t;
+typedef struct {
+    uint32_t v;
+} float32_t;
+typedef struct {
+    uint64_t v;
+} float64_t;
+typedef struct {
+    uint64_t v[2];
+} float128_t;
 
 /*----------------------------------------------------------------------------
 | The format of an 80-bit extended floating-point number in memory.  This
@@ -58,9 +69,15 @@ typedef struct { uint64_t v[2]; } float128_t;
 | named 'signif'.
 *----------------------------------------------------------------------------*/
 #ifdef LITTLEENDIAN
-struct extFloat80M { uint64_t signif; uint16_t signExp; };
+struct extFloat80M {
+    uint64_t signif;
+    uint16_t signExp;
+};
 #else
-struct extFloat80M { uint16_t signExp; uint64_t signif; };
+struct extFloat80M {
+    uint16_t signExp;
+    uint64_t signif;
+};
 #endif
 
 /*----------------------------------------------------------------------------
@@ -78,4 +95,3 @@ struct extFloat80M { uint16_t signExp; uint64_t signif; };
 typedef struct extFloat80M extFloat80_t;
 
 #endif
-

softfloat/source/s_mulAddF32.c

@@ -221,4 +221,3 @@ float32_t
     return uZ.f;
 
 }
-

softfloat/source/s_normSubnormalBF16Sig.c

@@ -0,0 +1,52 @@
+
+/*============================================================================
+
+This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
+Package, Release 3e, by John R. Hauser.
+
+Copyright 2011, 2012, 2013, 2014, 2015, 2016 The Regents of the University of
+California.  All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+
+ 1. Redistributions of source code must retain the above copyright notice,
+    this list of conditions, and the following disclaimer.
+
+ 2. Redistributions in binary form must reproduce the above copyright notice,
+    this list of conditions, and the following disclaimer in the documentation
+    and/or other materials provided with the distribution.
+
+ 3. Neither the name of the University nor the names of its contributors may
+    be used to endorse or promote products derived from this software without
+    specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS "AS IS", AND ANY
+EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, ARE
+DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY
+DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
+ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+=============================================================================*/
+
+#include <stdint.h>
+#include "platform.h"
+#include "internals.h"
+
+struct exp8_sig16 softfloat_normSubnormalBF16Sig( uint_fast16_t sig )
+{
+    int_fast8_t shiftDist;
+    struct exp8_sig16 z;
+
+    shiftDist = softfloat_countLeadingZeros16( sig ) - 8;
+    z.exp = 1 - shiftDist;
+    z.sig = sig<<shiftDist;
+    return z;
+
+}
+

softfloat/source/s_roundPackToBF16.c

@@ -0,0 +1,114 @@
+
+/*============================================================================
+
+This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
+Package, Release 3e, by John R. Hauser.
+
+Copyright 2011, 2012, 2013, 2014, 2015, 2017 The Regents of the University of
+California.  All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+
+ 1. Redistributions of source code must retain the above copyright notice,
+    this list of conditions, and the following disclaimer.
+
+ 2. Redistributions in binary form must reproduce the above copyright notice,
+    this list of conditions, and the following disclaimer in the documentation
+    and/or other materials provided with the distribution.
+
+ 3. Neither the name of the University nor the names of its contributors may
+    be used to endorse or promote products derived from this software without
+    specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS "AS IS", AND ANY
+EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, ARE
+DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY
+DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
+ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+=============================================================================*/
+
+#include <stdbool.h>
+#include <stdint.h>
+#include "platform.h"
+#include "internals.h"
+#include "softfloat.h"
+
+/** sig last significant bit is sig[7], the 7 LSBs will be used for rounding */
+bfloat16_t
+ softfloat_roundPackToBF16( bool sign, int_fast16_t exp, uint_fast16_t sig )
+{
+    uint_fast8_t roundingMode;
+    bool roundNearEven;
+    uint_fast8_t roundIncrement, roundBits;
+    bool isTiny;
+    uint_fast16_t uiZ;
+    union ui16_bf16 uZ;
+
+    /*------------------------------------------------------------------------
+    *------------------------------------------------------------------------*/
+    roundingMode = softfloat_roundingMode;
+    roundNearEven = (roundingMode == softfloat_round_near_even);
+    roundIncrement = 0x40;
+    if ( ! roundNearEven && (roundingMode != softfloat_round_near_maxMag) ) {
+        roundIncrement =
+            (roundingMode
+                 == (sign ? softfloat_round_min : softfloat_round_max))
+                ? 0x7F
+                : 0;
+    }
+    roundBits = sig & 0x7F;
+    /*------------------------------------------------------------------------
+    *------------------------------------------------------------------------*/
+    if ( 0xFD <= (unsigned int) exp ) {
+        if ( exp < 0 ) {
+            /*----------------------------------------------------------------
+            *----------------------------------------------------------------*/
+            isTiny =
+                (softfloat_detectTininess == softfloat_tininess_beforeRounding)
+                    || (exp < -1) || (sig + roundIncrement < 0x8000);
+            sig = softfloat_shiftRightJam32( sig, -exp );
+            exp = 0;
+            roundBits = sig & 0x7F;
+            if ( isTiny && roundBits ) {
+                softfloat_raiseFlags( softfloat_flag_underflow );
+            }
+        } else if ( (0xFD < exp) || (0x8000 <= sig + roundIncrement) ) {
+            /*----------------------------------------------------------------
+            *----------------------------------------------------------------*/
+            softfloat_raiseFlags(
+                softfloat_flag_overflow | softfloat_flag_inexact );
+            uiZ = packToBF16UI( sign, 0xFF, 0 ) - ! roundIncrement;
+            goto uiZ;
+        }
+    }
+    /*------------------------------------------------------------------------
+    *------------------------------------------------------------------------*/
+    sig = (sig + roundIncrement)>>7;
+    if ( roundBits ) {
+        softfloat_exceptionFlags |= softfloat_flag_inexact;
+#ifdef SOFTFLOAT_ROUND_ODD
+        if ( roundingMode == softfloat_round_odd ) {
+            sig |= 1;
+            goto packReturn;
+        }
+#endif
+    }
+    sig &= ~(uint_fast16_t) (! (roundBits ^ 0x40) & roundNearEven);
+    if ( ! sig ) exp = 0;
+    /*------------------------------------------------------------------------
+    *------------------------------------------------------------------------*/
+ packReturn:
+    uiZ = packToBF16UI( sign, exp, sig );
+ uiZ:
+    uZ.ui = uiZ;
+    return uZ.f;
+
+}
+

src/elfio.cpp

@@ -0,0 +1,35 @@
+#ifdef _MSC_VER
+#define _SCL_SECURE_NO_WARNINGS
+#define ELFIO_NO_INTTYPES
+#endif
+
+#include <elfio/elfio_dump.hpp>
+#include <iostream>
+
+using namespace ELFIO;
+
+int main(int argc, char** argv) {
+    if(argc != 2) {
+        printf("Usage: elfdump <file_name>\n");
+        return 1;
+    }
+
+    elfio reader;
+
+    if(!reader.load(argv[1])) {
+        printf("File %s is not found or it is not an ELF file\n", argv[1]);
+        return 1;
+    }
+
+    dump::header(std::cout, reader);
+    dump::section_headers(std::cout, reader);
+    dump::segment_headers(std::cout, reader);
+    dump::symbol_tables(std::cout, reader);
+    dump::notes(std::cout, reader);
+    dump::modinfo(std::cout, reader);
+    dump::dynamic_tags(std::cout, reader);
+    dump::section_datas(std::cout, reader);
+    dump::segment_datas(std::cout, reader);
+
+    return 0;
+}

src/iss/arch/hwl.h

@@ -35,6 +35,7 @@
 #ifndef _RISCV_HART_M_P_HWL_H
 #define _RISCV_HART_M_P_HWL_H
 
+#include "riscv_hart_common.h"
 #include <iss/vm_types.h>
 
 namespace iss {
@@ -46,49 +47,71 @@ public:
     using this_class = hwl<BASE>;
     using reg_t = typename BASE::reg_t;
 
-    hwl();
+    hwl(feature_config cfg = feature_config{});
     virtual ~hwl() = default;
 
 protected:
-    iss::status read_custom_csr_reg(unsigned addr, reg_t &val) override;
+    iss::status read_custom_csr(unsigned addr, reg_t& val) override;
-    iss::status write_custom_csr_reg(unsigned addr, reg_t val) override;
+    iss::status write_custom_csr(unsigned addr, reg_t val) override;
 };
 
-
+template <typename BASE>
-template<typename BASE>
+inline hwl<BASE>::hwl(feature_config cfg)
-inline hwl<BASE>::hwl() {
+: BASE(cfg) {
-    for (unsigned addr = 0x800; addr < 0x803; ++addr){
+    for(unsigned addr = 0x800; addr < 0x803; ++addr) {
         this->register_custom_csr_rd(addr);
         this->register_custom_csr_wr(addr);
     }
-    for (unsigned addr = 0x804; addr < 0x807; ++addr){
+    for(unsigned addr = 0x804; addr < 0x807; ++addr) {
         this->register_custom_csr_rd(addr);
         this->register_custom_csr_wr(addr);
     }
 }
 
-template<typename BASE>
+template <typename BASE> inline iss::status iss::arch::hwl<BASE>::read_custom_csr(unsigned addr, reg_t& val) {
-inline iss::status iss::arch::hwl<BASE>::read_custom_csr_reg(unsigned addr, reg_t &val) {
+    switch(addr) {
-    switch(addr){
+    case 0x800:
-    case 0x800: val = this->reg.lpstart0; break;
+        val = this->reg.lpstart0;
-    case 0x801: val = this->reg.lpend0;   break;
+        break;
-    case 0x802: val = this->reg.lpcount0; break;
+    case 0x801:
-    case 0x804: val = this->reg.lpstart1; break;
+        val = this->reg.lpend0;
-    case 0x805: val = this->reg.lpend1;   break;
+        break;
-    case 0x806: val = this->reg.lpcount1; break;
+    case 0x802:
+        val = this->reg.lpcount0;
+        break;
+    case 0x804:
+        val = this->reg.lpstart1;
+        break;
+    case 0x805:
+        val = this->reg.lpend1;
+        break;
+    case 0x806:
+        val = this->reg.lpcount1;
+        break;
     }
     return iss::Ok;
 }
 
-template<typename BASE>
+template <typename BASE> inline iss::status iss::arch::hwl<BASE>::write_custom_csr(unsigned addr, reg_t val) {
-inline iss::status iss::arch::hwl<BASE>::write_custom_csr_reg(unsigned addr, reg_t val) {
+    switch(addr) {
-    switch(addr){
+    case 0x800:
-    case 0x800: this->reg.lpstart0 = val; break;
+        this->reg.lpstart0 = val;
-    case 0x801: this->reg.lpend0   = val; break;
+        break;
-    case 0x802: this->reg.lpcount0 = val; break;
+    case 0x801:
-    case 0x804: this->reg.lpstart1 = val; break;
+        this->reg.lpend0 = val;
-    case 0x805: this->reg.lpend1   = val; break;
+        break;
-    case 0x806: this->reg.lpcount1 = val; break;
+    case 0x802:
+        this->reg.lpcount0 = val;
+        break;
+    case 0x804:
+        this->reg.lpstart1 = val;
+        break;
+    case 0x805:
+        this->reg.lpend1 = val;
+        break;
+    case 0x806:
+        this->reg.lpcount1 = val;
+        break;
     }
     return iss::Ok;
 }
@@ -96,5 +119,4 @@ inline iss::status iss::arch::hwl<BASE>::write_custom_csr_reg(unsigned addr, reg
 } // namespace arch
 } // namespace iss
 
-
 #endif /* _RISCV_HART_M_P_H */

src/iss/arch/riscv_hart_common.h

@@ -35,15 +35,30 @@
 #ifndef _RISCV_HART_COMMON
 #define _RISCV_HART_COMMON
 
-#include "iss/arch_if.h"
+#include "iss/vm_types.h"
 #include <cstdint>
+#include <elfio/elfio.hpp>
+#include <fmt/format.h>
+#include <iss/arch_if.h>
+#include <iss/log_categories.h>
+#include <string>
+#include <unordered_map>
+#include <util/logging.h>
+
+#if defined(__GNUC__)
+#define likely(x) ::__builtin_expect(!!(x), 1)
+#define unlikely(x) ::__builtin_expect(!!(x), 0)
+#else
+#define likely(x) x
+#define unlikely(x) x
+#endif
 
 namespace iss {
 namespace arch {
 
 enum { tohost_dflt = 0xF0001000, fromhost_dflt = 0xF0001040 };
 
-enum features_e{FEAT_NONE, FEAT_PMP=1, FEAT_EXT_N=2, FEAT_CLIC=4, FEAT_DEBUG=8, FEAT_TCM=16};
+enum features_e { FEAT_NONE, FEAT_PMP = 1, FEAT_EXT_N = 2, FEAT_CLIC = 4, FEAT_DEBUG = 8, FEAT_TCM = 16 };
 
 enum riscv_csr {
     /* user-level CSR */
@@ -51,14 +66,14 @@ enum riscv_csr {
     ustatus = 0x000,
     uie = 0x004,
     utvec = 0x005,
-    utvt = 0x007, //CLIC
+    utvt = 0x007, // CLIC
     // User Trap Handling
     uscratch = 0x040,
     uepc = 0x041,
     ucause = 0x042,
     utval = 0x043,
     uip = 0x044,
-    uxnti = 0x045, //CLIC
+    uxnti = 0x045,        // CLIC
     uintstatus = 0xCB1,   // MRW Current interrupt levels (CLIC) - addr subject to change
     uintthresh = 0x047,   // MRW Interrupt-level threshold (CLIC) - addr subject to change
     uscratchcsw = 0x048,  // MRW Conditional scratch swap on priv mode change (CLIC)
@@ -112,14 +127,14 @@ enum riscv_csr {
     mie = 0x304,
     mtvec = 0x305,
     mcounteren = 0x306,
-    mtvt = 0x307, //CLIC
+    mtvt = 0x307, // CLIC
     // Machine Trap Handling
     mscratch = 0x340,
     mepc = 0x341,
     mcause = 0x342,
     mtval = 0x343,
     mip = 0x344,
-    mxnti = 0x345, //CLIC
+    mxnti = 0x345,        // CLIC
     mintstatus = 0xFB1,   // MRW Current interrupt levels (CLIC) - addr subject to change
     mintthresh = 0x347,   // MRW Interrupt-level threshold (CLIC) - addr subject to change
     mscratchcsw = 0x348,  // MRW Conditional scratch swap on priv mode change (CLIC)
@@ -175,7 +190,6 @@ enum riscv_csr {
     dscratch1 = 0x7B3
 };
 
-
 enum {
     PGSHIFT = 12,
     PTE_PPN_SHIFT = 10,
@@ -193,7 +207,7 @@ enum {
 
 template <typename T> inline bool PTE_TABLE(T PTE) { return (((PTE) & (PTE_V | PTE_R | PTE_W | PTE_X)) == PTE_V); }
 
-enum { PRIV_U = 0, PRIV_S = 1, PRIV_M = 3, PRIV_D = 4};
+enum { PRIV_U = 0, PRIV_S = 1, PRIV_M = 3, PRIV_D = 4 };
 
 enum {
     ISA_A = 1,
@@ -226,6 +240,8 @@ struct feature_config {
     unsigned clic_num_trigger{0};
     uint64_t tcm_base{0x10000000};
     uint64_t tcm_size{0x8000};
+    uint64_t io_address{0xf0000000};
+    uint64_t io_addr_mask{0xf0000000};
 };
 
 class trap_load_access_fault : public trap_access {
@@ -254,19 +270,19 @@ public:
     : trap_access(15 << 16, badaddr) {}
 };
 
-inline void read_reg_uint32(uint64_t offs, uint32_t& reg, uint8_t *const data, unsigned length) {
+inline void read_reg_uint32(uint64_t offs, uint32_t& reg, uint8_t* const data, unsigned length) {
     auto reg_ptr = reinterpret_cast<uint8_t*>(&reg);
-    switch (offs & 0x3) {
+    switch(offs & 0x3) {
     case 0:
-        for (auto i = 0U; i < length; ++i)
+        for(auto i = 0U; i < length; ++i)
             *(data + i) = *(reg_ptr + i);
         break;
     case 1:
-        for (auto i = 0U; i < length; ++i)
+        for(auto i = 0U; i < length; ++i)
             *(data + i) = *(reg_ptr + 1 + i);
         break;
     case 2:
-        for (auto i = 0U; i < length; ++i)
+        for(auto i = 0U; i < length; ++i)
             *(data + i) = *(reg_ptr + 2 + i);
         break;
     case 3:
@@ -275,28 +291,95 @@ inline void read_reg_uint32(uint64_t offs, uint32_t& reg, uint8_t *const data, u
     }
 }
 
-inline void write_reg_uint32(uint64_t offs, uint32_t& reg, const uint8_t *const data, unsigned length) {
+inline void write_reg_uint32(uint64_t offs, uint32_t& reg, const uint8_t* const data, unsigned length) {
     auto reg_ptr = reinterpret_cast<uint8_t*>(&reg);
-    switch (offs & 0x3) {
+    switch(offs & 0x3) {
     case 0:
-        for (auto i = 0U; i < length; ++i)
+        for(auto i = 0U; i < length; ++i)
             *(reg_ptr + i) = *(data + i);
         break;
     case 1:
-        for (auto i = 0U; i < length; ++i)
+        for(auto i = 0U; i < length; ++i)
             *(reg_ptr + 1 + i) = *(data + i);
         break;
     case 2:
-        for (auto i = 0U; i < length; ++i)
+        for(auto i = 0U; i < length; ++i)
             *(reg_ptr + 2 + i) = *(data + i);
         break;
     case 3:
-        *(reg_ptr + 3) = *data ;
+        *(reg_ptr + 3) = *data;
         break;
     }
 }
+struct riscv_hart_common {
+    riscv_hart_common(){};
+    ~riscv_hart_common(){};
+    std::unordered_map<std::string, uint64_t> symbol_table;
+    uint64_t entry_address{0};
+    uint64_t tohost = tohost_dflt;
+    uint64_t fromhost = fromhost_dflt;
 
-}
+    bool read_elf_file(std::string name, uint8_t expected_elf_class,
-}
+                       std::function<iss::status(uint64_t, uint64_t, const uint8_t* const)> cb) {
+        // Create elfio reader
+        ELFIO::elfio reader;
+        // Load ELF data
+        if(reader.load(name)) {
+            // check elf properties
+            if(reader.get_class() != expected_elf_class)
+                return false;
+            if(reader.get_type() != ELFIO::ET_EXEC)
+                return false;
+            if(reader.get_machine() != ELFIO::EM_RISCV)
+                return false;
+            entry_address = reader.get_entry();
+            for(const auto& pseg : reader.segments) {
+                const auto fsize = pseg->get_file_size(); // 0x42c/0x0
+                const auto seg_data = pseg->get_data();
+                const auto type = pseg->get_type();
+                if(type == 1 && fsize > 0) {
+                    auto res = cb(pseg->get_physical_address(), fsize, reinterpret_cast<const uint8_t* const>(seg_data));
+                    if(res != iss::Ok)
+                        CPPLOG(ERR) << "problem writing " << fsize << "bytes to 0x" << std::hex << pseg->get_physical_address();
+                }
+            }
+            const auto sym_sec = reader.sections[".symtab"];
+            if(ELFIO::SHT_SYMTAB == sym_sec->get_type() || ELFIO::SHT_DYNSYM == sym_sec->get_type()) {
+                ELFIO::symbol_section_accessor symbols(reader, sym_sec);
+                auto sym_no = symbols.get_symbols_num();
+                std::string name;
+                ELFIO::Elf64_Addr value = 0;
+                ELFIO::Elf_Xword size = 0;
+                unsigned char bind = 0;
+                unsigned char type = 0;
+                ELFIO::Elf_Half section = 0;
+                unsigned char other = 0;
+                for(auto i = 0U; i < sym_no; ++i) {
+                    symbols.get_symbol(i, name, value, size, bind, type, section, other);
+                    if(name != "") {
+                        this->symbol_table[name] = value;
+#ifndef NDEBUG
+                        CPPLOG(DEBUG) << "Found Symbol " << name;
+#endif
+                    }
+                }
+                try {
+                    tohost = symbol_table.at("tohost");
+                    try {
+                        fromhost = symbol_table.at("fromhost");
+                    } catch(std::out_of_range& e) {
+                        fromhost = tohost + 0x40;
+                    }
+                } catch(std::out_of_range& e) {
+                }
+            }
+            return true;
+        }
+        return false;
+    };
+};
+
+} // namespace arch
+} // namespace iss
 
 #endif

src/iss/arch/tgc5c.cpp

@@ -1,5 +1,5 @@
 /*******************************************************************************
- * Copyright (C) 2017 - 2020 MINRES Technologies GmbH
+ * Copyright (C) 2024 MINRES Technologies GmbH
  * All rights reserved.
  *
  * Redistribution and use in source and binary forms, with or without
@@ -30,7 +30,8 @@
  *
  *******************************************************************************/
 
-#include "tgc_c.h"
+// clang-format off
+#include "tgc5c.h"
 #include "util/ities.h"
 #include <util/logging.h>
 #include <cstdio>
@@ -39,18 +40,18 @@
 
 using namespace iss::arch;
 
-constexpr std::array<const char*, 36>    iss::arch::traits<iss::arch::tgc_c>::reg_names;
+constexpr std::array<const char*, 36>    iss::arch::traits<iss::arch::tgc5c>::reg_names;
-constexpr std::array<const char*, 36>    iss::arch::traits<iss::arch::tgc_c>::reg_aliases;
+constexpr std::array<const char*, 36>    iss::arch::traits<iss::arch::tgc5c>::reg_aliases;
-constexpr std::array<const uint32_t, 43> iss::arch::traits<iss::arch::tgc_c>::reg_bit_widths;
+constexpr std::array<const uint32_t, 43> iss::arch::traits<iss::arch::tgc5c>::reg_bit_widths;
-constexpr std::array<const uint32_t, 43> iss::arch::traits<iss::arch::tgc_c>::reg_byte_offsets;
+constexpr std::array<const uint32_t, 43> iss::arch::traits<iss::arch::tgc5c>::reg_byte_offsets;
 
-tgc_c::tgc_c()  = default;
+tgc5c::tgc5c()  = default;
 
-tgc_c::~tgc_c() = default;
+tgc5c::~tgc5c() = default;
 
-void tgc_c::reset(uint64_t address) {
+void tgc5c::reset(uint64_t address) {
-    auto base_ptr = reinterpret_cast<traits<tgc_c>::reg_t*>(get_regs_base_ptr());
+    auto base_ptr = reinterpret_cast<traits<tgc5c>::reg_t*>(get_regs_base_ptr());
-    for(size_t i=0; i<traits<tgc_c>::NUM_REGS; ++i)
+    for(size_t i=0; i<traits<tgc5c>::NUM_REGS; ++i)
         *(base_ptr+i)=0;
     reg.PC=address;
     reg.NEXT_PC=reg.PC;
@@ -59,11 +60,11 @@ void tgc_c::reset(uint64_t address) {
     reg.icount=0;
 }
 
-uint8_t *tgc_c::get_regs_base_ptr() {
+uint8_t *tgc5c::get_regs_base_ptr() {
 	return reinterpret_cast<uint8_t*>(&reg);
 }
 
-tgc_c::phys_addr_t tgc_c::virt2phys(const iss::addr_t &addr) {
+tgc5c::phys_addr_t tgc5c::virt2phys(const iss::addr_t &addr) {
-    return phys_addr_t(addr.access, addr.space, addr.val&traits<tgc_c>::addr_mask);
+    return phys_addr_t(addr.access, addr.space, addr.val&traits<tgc5c>::addr_mask);
 }
-
+// clang-format on

src/iss/arch/tgc_c.h

@@ -1,261 +0,0 @@
-/*******************************************************************************
- * Copyright (C) 2017 - 2021 MINRES Technologies GmbH
- * All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions are met:
- *
- * 1. Redistributions of source code must retain the above copyright notice,
- *    this list of conditions and the following disclaimer.
- *
- * 2. Redistributions in binary form must reproduce the above copyright notice,
- *    this list of conditions and the following disclaimer in the documentation
- *    and/or other materials provided with the distribution.
- *
- * 3. Neither the name of the copyright holder nor the names of its contributors
- *    may be used to endorse or promote products derived from this software
- *    without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
- * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
- * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
- * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
- * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
- * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
- * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
- * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
- * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
- * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
- * POSSIBILITY OF SUCH DAMAGE.
- *
- *******************************************************************************/
-
-#ifndef _TGC_C_H_
-#define _TGC_C_H_
-
-#include <array>
-#include <iss/arch/traits.h>
-#include <iss/arch_if.h>
-#include <iss/vm_if.h>
-
-namespace iss {
-namespace arch {
-
-struct tgc_c;
-
-template <> struct traits<tgc_c> {
-
-    constexpr static char const* const core_type = "TGC_C";
-    
-    static constexpr std::array<const char*, 36> reg_names{
-        {"X0", "X1", "X2", "X3", "X4", "X5", "X6", "X7", "X8", "X9", "X10", "X11", "X12", "X13", "X14", "X15", "X16", "X17", "X18", "X19", "X20", "X21", "X22", "X23", "X24", "X25", "X26", "X27", "X28", "X29", "X30", "X31", "PC", "NEXT_PC", "PRIV", "DPC"}};
- 
-    static constexpr std::array<const char*, 36> reg_aliases{
-        {"ZERO", "RA", "SP", "GP", "TP", "T0", "T1", "T2", "S0", "S1", "A0", "A1", "A2", "A3", "A4", "A5", "A6", "A7", "S2", "S3", "S4", "S5", "S6", "S7", "S8", "S9", "S10", "S11", "T3", "T4", "T5", "T6", "PC", "NEXT_PC", "PRIV", "DPC"}};
-
-    enum constants {MISA_VAL=1073746180ULL, MARCHID_VAL=2147483651ULL, XLEN=32ULL, INSTR_ALIGNMENT=2ULL, RFS=32ULL, fence=0ULL, fencei=1ULL, fencevmal=2ULL, fencevmau=3ULL, CSR_SIZE=4096ULL, MUL_LEN=64ULL};
-
-    constexpr static unsigned FP_REGS_SIZE = 0;
-
-    enum reg_e {
-        X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X12, X13, X14, X15, X16, X17, X18, X19, X20, X21, X22, X23, X24, X25, X26, X27, X28, X29, X30, X31, PC, NEXT_PC, PRIV, DPC, NUM_REGS, TRAP_STATE=NUM_REGS, PENDING_TRAP, ICOUNT, CYCLE, INSTRET, INSTRUCTION, LAST_BRANCH
-    };
-
-    using reg_t = uint32_t;
-
-    using addr_t = uint32_t;
-
-    using code_word_t = uint32_t; //TODO: check removal
-
-    using virt_addr_t = iss::typed_addr_t<iss::address_type::VIRTUAL>;
-
-    using phys_addr_t = iss::typed_addr_t<iss::address_type::PHYSICAL>;
-
-    static constexpr std::array<const uint32_t, 43> reg_bit_widths{
-        {32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,8,32,32,32,64,64,64,32,32}};
-
-    static constexpr std::array<const uint32_t, 43> reg_byte_offsets{
-        {0,4,8,12,16,20,24,28,32,36,40,44,48,52,56,60,64,68,72,76,80,84,88,92,96,100,104,108,112,116,120,124,128,132,136,137,141,145,149,157,165,173,177}};
-
-    static const uint64_t addr_mask = (reg_t(1) << (XLEN - 1)) | ((reg_t(1) << (XLEN - 1)) - 1);
-
-    enum sreg_flag_e { FLAGS };
-
-    enum mem_type_e { MEM, FENCE, RES, CSR };
-    
-    enum class opcode_e {
-        LUI = 0,
-        AUIPC = 1,
-        JAL = 2,
-        JALR = 3,
-        BEQ = 4,
-        BNE = 5,
-        BLT = 6,
-        BGE = 7,
-        BLTU = 8,
-        BGEU = 9,
-        LB = 10,
-        LH = 11,
-        LW = 12,
-        LBU = 13,
-        LHU = 14,
-        SB = 15,
-        SH = 16,
-        SW = 17,
-        ADDI = 18,
-        SLTI = 19,
-        SLTIU = 20,
-        XORI = 21,
-        ORI = 22,
-        ANDI = 23,
-        SLLI = 24,
-        SRLI = 25,
-        SRAI = 26,
-        ADD = 27,
-        SUB = 28,
-        SLL = 29,
-        SLT = 30,
-        SLTU = 31,
-        XOR = 32,
-        SRL = 33,
-        SRA = 34,
-        OR = 35,
-        AND = 36,
-        FENCE = 37,
-        ECALL = 38,
-        EBREAK = 39,
-        MRET = 40,
-        WFI = 41,
-        CSRRW = 42,
-        CSRRS = 43,
-        CSRRC = 44,
-        CSRRWI = 45,
-        CSRRSI = 46,
-        CSRRCI = 47,
-        FENCE_I = 48,
-        MUL = 49,
-        MULH = 50,
-        MULHSU = 51,
-        MULHU = 52,
-        DIV = 53,
-        DIVU = 54,
-        REM = 55,
-        REMU = 56,
-        CADDI4SPN = 57,
-        CLW = 58,
-        CSW = 59,
-        CADDI = 60,
-        CNOP = 61,
-        CJAL = 62,
-        CLI = 63,
-        CLUI = 64,
-        CADDI16SP = 65,
-        __reserved_clui = 66,
-        CSRLI = 67,
-        CSRAI = 68,
-        CANDI = 69,
-        CSUB = 70,
-        CXOR = 71,
-        COR = 72,
-        CAND = 73,
-        CJ = 74,
-        CBEQZ = 75,
-        CBNEZ = 76,
-        CSLLI = 77,
-        CLWSP = 78,
-        CMV = 79,
-        CJR = 80,
-        __reserved_cmv = 81,
-        CADD = 82,
-        CJALR = 83,
-        CEBREAK = 84,
-        CSWSP = 85,
-        DII = 86,
-        MAX_OPCODE
-    };
-};
-
-struct tgc_c: public arch_if {
-
-    using virt_addr_t = typename traits<tgc_c>::virt_addr_t;
-    using phys_addr_t = typename traits<tgc_c>::phys_addr_t;
-    using reg_t =  typename traits<tgc_c>::reg_t;
-    using addr_t = typename traits<tgc_c>::addr_t;
-
-    tgc_c();
-    ~tgc_c();
-
-    void reset(uint64_t address=0) override;
-
-    uint8_t* get_regs_base_ptr() override;
-
-    inline uint64_t get_icount() { return reg.icount; }
-
-    inline bool should_stop() { return interrupt_sim; }
-
-    inline uint64_t stop_code() { return interrupt_sim; }
-
-    virtual phys_addr_t virt2phys(const iss::addr_t& addr);
-
-    virtual iss::sync_type needed_sync() const { return iss::NO_SYNC; }
-
-    inline uint32_t get_last_branch() { return reg.last_branch; }
-
-
-#pragma pack(push, 1)
-    struct TGC_C_regs { 
-        uint32_t X0 = 0; 
-        uint32_t X1 = 0; 
-        uint32_t X2 = 0; 
-        uint32_t X3 = 0; 
-        uint32_t X4 = 0; 
-        uint32_t X5 = 0; 
-        uint32_t X6 = 0; 
-        uint32_t X7 = 0; 
-        uint32_t X8 = 0; 
-        uint32_t X9 = 0; 
-        uint32_t X10 = 0; 
-        uint32_t X11 = 0; 
-        uint32_t X12 = 0; 
-        uint32_t X13 = 0; 
-        uint32_t X14 = 0; 
-        uint32_t X15 = 0; 
-        uint32_t X16 = 0; 
-        uint32_t X17 = 0; 
-        uint32_t X18 = 0; 
-        uint32_t X19 = 0; 
-        uint32_t X20 = 0; 
-        uint32_t X21 = 0; 
-        uint32_t X22 = 0; 
-        uint32_t X23 = 0; 
-        uint32_t X24 = 0; 
-        uint32_t X25 = 0; 
-        uint32_t X26 = 0; 
-        uint32_t X27 = 0; 
-        uint32_t X28 = 0; 
-        uint32_t X29 = 0; 
-        uint32_t X30 = 0; 
-        uint32_t X31 = 0; 
-        uint32_t PC = 0; 
-        uint32_t NEXT_PC = 0; 
-        uint8_t PRIV = 0; 
-        uint32_t DPC = 0;
-        uint32_t trap_state = 0, pending_trap = 0;
-        uint64_t icount = 0;
-        uint64_t cycle = 0;
-        uint64_t instret = 0;
-        uint32_t instruction = 0;
-        uint32_t last_branch = 0;
-    } reg;
-#pragma pack(pop)
-    std::array<address_type, 4> addr_mode;
-    
-    uint64_t interrupt_sim=0;
-
-    uint32_t get_fcsr(){return 0;}
-    void set_fcsr(uint32_t val){}
-};
-
-}
-}            
-#endif /* _TGC_C_H_ */

src/iss/arch/tgc_c_decoder.cpp

@@ -1,175 +0,0 @@
-#include "tgc_c.h"
-#include <vector>
-#include <array>
-#include <cstdlib>
-#include <algorithm>
-
-namespace iss {
-namespace arch {
-namespace {
-// according to
-// https://stackoverflow.com/questions/8871204/count-number-of-1s-in-binary-representation
-#ifdef __GCC__
-constexpr size_t bit_count(uint32_t u) { return __builtin_popcount(u); }
-#elif __cplusplus < 201402L
-constexpr size_t uCount(uint32_t u) { return u - ((u >> 1) & 033333333333) - ((u >> 2) & 011111111111); }
-constexpr size_t bit_count(uint32_t u) { return ((uCount(u) + (uCount(u) >> 3)) & 030707070707) % 63; }
-#else
-constexpr size_t bit_count(uint32_t u) {
-    size_t uCount = u - ((u >> 1) & 033333333333) - ((u >> 2) & 011111111111);
-    return ((uCount + (uCount >> 3)) & 030707070707) % 63;
-}
-#endif
-
-using opcode_e = traits<tgc_c>::opcode_e;
-
-/****************************************************************************
- * start opcode definitions
- ****************************************************************************/
-struct instruction_desriptor {
-    size_t length;
-    uint32_t value;
-    uint32_t mask;
-    opcode_e op;
-};
-
-const std::array<instruction_desriptor, 90> instr_descr = {{
-     /* entries are: size, valid value, valid mask, function ptr */
-    {32, 0b00000000000000000000000000110111, 0b00000000000000000000000001111111, opcode_e::LUI},
-    {32, 0b00000000000000000000000000010111, 0b00000000000000000000000001111111, opcode_e::AUIPC},
-    {32, 0b00000000000000000000000001101111, 0b00000000000000000000000001111111, opcode_e::JAL},
-    {32, 0b00000000000000000000000001100111, 0b00000000000000000111000001111111, opcode_e::JALR},
-    {32, 0b00000000000000000000000001100011, 0b00000000000000000111000001111111, opcode_e::BEQ},
-    {32, 0b00000000000000000001000001100011, 0b00000000000000000111000001111111, opcode_e::BNE},
-    {32, 0b00000000000000000100000001100011, 0b00000000000000000111000001111111, opcode_e::BLT},
-    {32, 0b00000000000000000101000001100011, 0b00000000000000000111000001111111, opcode_e::BGE},
-    {32, 0b00000000000000000110000001100011, 0b00000000000000000111000001111111, opcode_e::BLTU},
-    {32, 0b00000000000000000111000001100011, 0b00000000000000000111000001111111, opcode_e::BGEU},
-    {32, 0b00000000000000000000000000000011, 0b00000000000000000111000001111111, opcode_e::LB},
-    {32, 0b00000000000000000001000000000011, 0b00000000000000000111000001111111, opcode_e::LH},
-    {32, 0b00000000000000000010000000000011, 0b00000000000000000111000001111111, opcode_e::LW},
-    {32, 0b00000000000000000100000000000011, 0b00000000000000000111000001111111, opcode_e::LBU},
-    {32, 0b00000000000000000101000000000011, 0b00000000000000000111000001111111, opcode_e::LHU},
-    {32, 0b00000000000000000000000000100011, 0b00000000000000000111000001111111, opcode_e::SB},
-    {32, 0b00000000000000000001000000100011, 0b00000000000000000111000001111111, opcode_e::SH},
-    {32, 0b00000000000000000010000000100011, 0b00000000000000000111000001111111, opcode_e::SW},
-    {32, 0b00000000000000000000000000010011, 0b00000000000000000111000001111111, opcode_e::ADDI},
-    {32, 0b00000000000000000010000000010011, 0b00000000000000000111000001111111, opcode_e::SLTI},
-    {32, 0b00000000000000000011000000010011, 0b00000000000000000111000001111111, opcode_e::SLTIU},
-    {32, 0b00000000000000000100000000010011, 0b00000000000000000111000001111111, opcode_e::XORI},
-    {32, 0b00000000000000000110000000010011, 0b00000000000000000111000001111111, opcode_e::ORI},
-    {32, 0b00000000000000000111000000010011, 0b00000000000000000111000001111111, opcode_e::ANDI},
-    {32, 0b00000000000000000001000000010011, 0b11111110000000000111000001111111, opcode_e::SLLI},
-    {32, 0b00000000000000000101000000010011, 0b11111110000000000111000001111111, opcode_e::SRLI},
-    {32, 0b01000000000000000101000000010011, 0b11111110000000000111000001111111, opcode_e::SRAI},
-    {32, 0b00000000000000000000000000110011, 0b11111110000000000111000001111111, opcode_e::ADD},
-    {32, 0b01000000000000000000000000110011, 0b11111110000000000111000001111111, opcode_e::SUB},
-    {32, 0b00000000000000000001000000110011, 0b11111110000000000111000001111111, opcode_e::SLL},
-    {32, 0b00000000000000000010000000110011, 0b11111110000000000111000001111111, opcode_e::SLT},
-    {32, 0b00000000000000000011000000110011, 0b11111110000000000111000001111111, opcode_e::SLTU},
-    {32, 0b00000000000000000100000000110011, 0b11111110000000000111000001111111, opcode_e::XOR},
-    {32, 0b00000000000000000101000000110011, 0b11111110000000000111000001111111, opcode_e::SRL},
-    {32, 0b01000000000000000101000000110011, 0b11111110000000000111000001111111, opcode_e::SRA},
-    {32, 0b00000000000000000110000000110011, 0b11111110000000000111000001111111, opcode_e::OR},
-    {32, 0b00000000000000000111000000110011, 0b11111110000000000111000001111111, opcode_e::AND},
-    {32, 0b00000000000000000000000000001111, 0b00000000000000000111000001111111, opcode_e::FENCE},
-    {32, 0b00000000000000000000000001110011, 0b11111111111111111111111111111111, opcode_e::ECALL},
-    {32, 0b00000000000100000000000001110011, 0b11111111111111111111111111111111, opcode_e::EBREAK},
-    {32, 0b00000000001000000000000001110011, 0b11111111111111111111111111111111, opcode_e::URET},
-    {32, 0b00010000001000000000000001110011, 0b11111111111111111111111111111111, opcode_e::SRET},
-    {32, 0b00110000001000000000000001110011, 0b11111111111111111111111111111111, opcode_e::MRET},
-    {32, 0b00010000010100000000000001110011, 0b11111111111111111111111111111111, opcode_e::WFI},
-    {32, 0b01111011001000000000000001110011, 0b11111111111111111111111111111111, opcode_e::DRET},
-    {32, 0b00000000000000000001000001110011, 0b00000000000000000111000001111111, opcode_e::CSRRW},
-    {32, 0b00000000000000000010000001110011, 0b00000000000000000111000001111111, opcode_e::CSRRS},
-    {32, 0b00000000000000000011000001110011, 0b00000000000000000111000001111111, opcode_e::CSRRC},
-    {32, 0b00000000000000000101000001110011, 0b00000000000000000111000001111111, opcode_e::CSRRWI},
-    {32, 0b00000000000000000110000001110011, 0b00000000000000000111000001111111, opcode_e::CSRRSI},
-    {32, 0b00000000000000000111000001110011, 0b00000000000000000111000001111111, opcode_e::CSRRCI},
-    {32, 0b00000000000000000001000000001111, 0b00000000000000000111000001111111, opcode_e::FENCE_I},
-    {32, 0b00000010000000000000000000110011, 0b11111110000000000111000001111111, opcode_e::MUL},
-    {32, 0b00000010000000000001000000110011, 0b11111110000000000111000001111111, opcode_e::MULH},
-    {32, 0b00000010000000000010000000110011, 0b11111110000000000111000001111111, opcode_e::MULHSU},
-    {32, 0b00000010000000000011000000110011, 0b11111110000000000111000001111111, opcode_e::MULHU},
-    {32, 0b00000010000000000100000000110011, 0b11111110000000000111000001111111, opcode_e::DIV},
-    {32, 0b00000010000000000101000000110011, 0b11111110000000000111000001111111, opcode_e::DIVU},
-    {32, 0b00000010000000000110000000110011, 0b11111110000000000111000001111111, opcode_e::REM},
-    {32, 0b00000010000000000111000000110011, 0b11111110000000000111000001111111, opcode_e::REMU},
-    {16, 0b0000000000000000, 0b1110000000000011, opcode_e::CADDI4SPN},
-    {16, 0b0100000000000000, 0b1110000000000011, opcode_e::CLW},
-    {16, 0b1100000000000000, 0b1110000000000011, opcode_e::CSW},
-    {16, 0b0000000000000001, 0b1110000000000011, opcode_e::CADDI},
-    {16, 0b0000000000000001, 0b1110111110000011, opcode_e::CNOP},
-    {16, 0b0010000000000001, 0b1110000000000011, opcode_e::CJAL},
-    {16, 0b0100000000000001, 0b1110000000000011, opcode_e::CLI},
-    {16, 0b0110000000000001, 0b1110000000000011, opcode_e::CLUI},
-    {16, 0b0110000100000001, 0b1110111110000011, opcode_e::CADDI16SP},
-    {16, 0b0110000000000001, 0b1111000001111111, opcode_e::__reserved_clui},
-    {16, 0b1000000000000001, 0b1111110000000011, opcode_e::CSRLI},
-    {16, 0b1000010000000001, 0b1111110000000011, opcode_e::CSRAI},
-    {16, 0b1000100000000001, 0b1110110000000011, opcode_e::CANDI},
-    {16, 0b1000110000000001, 0b1111110001100011, opcode_e::CSUB},
-    {16, 0b1000110000100001, 0b1111110001100011, opcode_e::CXOR},
-    {16, 0b1000110001000001, 0b1111110001100011, opcode_e::COR},
-    {16, 0b1000110001100001, 0b1111110001100011, opcode_e::CAND},
-    {16, 0b1010000000000001, 0b1110000000000011, opcode_e::CJ},
-    {16, 0b1100000000000001, 0b1110000000000011, opcode_e::CBEQZ},
-    {16, 0b1110000000000001, 0b1110000000000011, opcode_e::CBNEZ},
-    {16, 0b0000000000000010, 0b1111000000000011, opcode_e::CSLLI},
-    {16, 0b0100000000000010, 0b1110000000000011, opcode_e::CLWSP},
-    {16, 0b1000000000000010, 0b1111000000000011, opcode_e::CMV},
-    {16, 0b1000000000000010, 0b1111000001111111, opcode_e::CJR},
-    {16, 0b1000000000000010, 0b1111111111111111, opcode_e::__reserved_cmv},
-    {16, 0b1001000000000010, 0b1111000000000011, opcode_e::CADD},
-    {16, 0b1001000000000010, 0b1111000001111111, opcode_e::CJALR},
-    {16, 0b1001000000000010, 0b1111111111111111, opcode_e::CEBREAK},
-    {16, 0b1100000000000010, 0b1110000000000011, opcode_e::CSWSP},
-    {16, 0b0000000000000000, 0b1111111111111111, opcode_e::DII},
-}};
-
-}
-
-template<>
-struct instruction_decoder<tgc_c> {
-    using opcode_e = traits<tgc_c>::opcode_e;
-    using code_word_t=traits<tgc_c>::code_word_t;
-
-    struct instruction_pattern {
-        uint32_t value;
-        uint32_t mask;
-        opcode_e id;
-    };
-
-    std::array<std::vector<instruction_pattern>, 4> qlut;
-
-    template<typename T>
-    unsigned decode_instruction(T);
-
-    instruction_decoder() {
-        for (auto instr : instr_descr) {
-            auto quadrant = instr.value & 0x3;
-            qlut[quadrant].push_back(instruction_pattern{instr.value, instr.mask, instr.op});
-        }
-        for(auto& lut: qlut){
-            std::sort(std::begin(lut), std::end(lut), [](instruction_pattern const& a, instruction_pattern const& b){
-                return bit_count(a.mask) > bit_count(b.mask);
-            });
-        }
-    }
-};
-
-template<>
-unsigned instruction_decoder<tgc_c>::decode_instruction<traits<tgc_c>::code_word_t>(traits<tgc_c>::code_word_t instr){
-    auto res = std::find_if(std::begin(qlut[instr&0x3]), std::end(qlut[instr&0x3]), [instr](instruction_pattern const& e){
-        return !((instr&e.mask) ^ e.value );
-    });
-    return static_cast<unsigned>(res!=std::end(qlut[instr&0x3])? res->id : opcode_e::MAX_OPCODE);
-}
-
-
-std::unique_ptr<instruction_decoder<tgc_c>> traits<tgc_c>::get_decoder(){
-    return std::make_unique<instruction_decoder<tgc_c>>();
-}
-
-}
-}

src/iss/arch/tgc_mapper.h

@@ -2,49 +2,56 @@
 #define _ISS_ARCH_TGC_MAPPER_H
 
 #include "riscv_hart_m_p.h"
-#include "tgc_c.h"
+#include "tgc5c.h"
-using tgc_c_plat_type = iss::arch::riscv_hart_m_p<iss::arch::tgc_c>;
+using tgc5c_plat_type = iss::arch::riscv_hart_m_p<iss::arch::tgc5c>;
-#ifdef CORE_TGC_A
+#ifdef CORE_TGC5A
 #include "riscv_hart_m_p.h"
-#include <iss/arch/tgc_a.h>
+#include <iss/arch/tgc5a.h>
-using tgc_a_plat_type = iss::arch::riscv_hart_m_p<iss::arch::tgc_a>;
+using tgc5a_plat_type = iss::arch::riscv_hart_m_p<iss::arch::tgc5a>;
 #endif
-#ifdef CORE_TGC_B
+#ifdef CORE_TGC5B
 #include "riscv_hart_m_p.h"
-#include <iss/arch/tgc_b.h>
+#include <iss/arch/tgc5b.h>
-using tgc_b_plat_type = iss::arch::riscv_hart_m_p<iss::arch::tgc_b>;
+using tgc5b_plat_type = iss::arch::riscv_hart_m_p<iss::arch::tgc5b>;
 #endif
-#ifdef CORE_TGC_C_XRB_NN
+#ifdef CORE_TGC5C_XRB_NN
-#include "riscv_hart_m_p.h"
 #include "hwl.h"
-#include <iss/arch/tgc_c_xrb_nn.h>
+#include "riscv_hart_m_p.h"
-using tgc_c_xrb_nn_plat_type = iss::arch::hwl<iss::arch::riscv_hart_m_p<iss::arch::tgc_c_xrb_nn>>;
+#include <iss/arch/tgc5c_xrb_nn.h>
+using tgc5c_xrb_nn_plat_type = iss::arch::hwl<iss::arch::riscv_hart_m_p<iss::arch::tgc5c_xrb_nn>>;
 #endif
-#ifdef CORE_TGC_D
+#ifdef CORE_TGC5D
 #include "riscv_hart_mu_p.h"
-#include <iss/arch/tgc_d.h>
+#include <iss/arch/tgc5d.h>
-using tgc_d_plat_type = iss::arch::riscv_hart_mu_p<iss::arch::tgc_d, (iss::arch::features_e)(iss::arch::FEAT_PMP | iss::arch::FEAT_CLIC | iss::arch::FEAT_EXT_N)>;
+using tgc5d_plat_type = iss::arch::riscv_hart_mu_p<iss::arch::tgc5d, (iss::arch::features_e)(iss::arch::FEAT_PMP | iss::arch::FEAT_CLIC |
+                                                                                             iss::arch::FEAT_EXT_N)>;
 #endif
-#ifdef CORE_TGC_D_XRB_MAC
+#ifdef CORE_TGC5D_XRB_MAC
 #include "riscv_hart_mu_p.h"
-#include <iss/arch/tgc_d_xrb_mac.h>
+#include <iss/arch/tgc5d_xrb_mac.h>
-using tgc_d_xrb_mac_plat_type = iss::arch::riscv_hart_mu_p<iss::arch::tgc_d_xrb_mac, (iss::arch::features_e)(iss::arch::FEAT_PMP | iss::arch::FEAT_CLIC | iss::arch::FEAT_EXT_N)>;
+using tgc5d_xrb_mac_plat_type =
+    iss::arch::riscv_hart_mu_p<iss::arch::tgc5d_xrb_mac,
+                               (iss::arch::features_e)(iss::arch::FEAT_PMP | iss::arch::FEAT_CLIC | iss::arch::FEAT_EXT_N)>;
 #endif
-#ifdef CORE_TGC_D_XRB_NN
+#ifdef CORE_TGC5D_XRB_NN
-#include "riscv_hart_mu_p.h"
 #include "hwl.h"
-#include <iss/arch/tgc_d_xrb_nn.h>
-using tgc_d_xrb_nn_plat_type = iss::arch::hwl<iss::arch::riscv_hart_mu_p<iss::arch::tgc_d_xrb_nn, (iss::arch::features_e)(iss::arch::FEAT_PMP | iss::arch::FEAT_CLIC | iss::arch::FEAT_EXT_N)>>;
-#endif
-#ifdef CORE_TGC_E
 #include "riscv_hart_mu_p.h"
-#include <iss/arch/tgc_e.h>
+#include <iss/arch/tgc5d_xrb_nn.h>
-using tgc_e_plat_type = iss::arch::riscv_hart_mu_p<iss::arch::tgc_e, (iss::arch::features_e)(iss::arch::FEAT_PMP | iss::arch::FEAT_CLIC | iss::arch::FEAT_EXT_N)>;
+using tgc5d_xrb_nn_plat_type =
+    iss::arch::hwl<iss::arch::riscv_hart_mu_p<iss::arch::tgc5d_xrb_nn,
+                                              (iss::arch::features_e)(iss::arch::FEAT_PMP | iss::arch::FEAT_CLIC | iss::arch::FEAT_EXT_N)>>;
 #endif
-#ifdef CORE_TGC_X
+#ifdef CORE_TGC5E
 #include "riscv_hart_mu_p.h"
-#include <iss/arch/tgc_x.h>
+#include <iss/arch/tgc5e.h>
-using tgc_x_plat_type = iss::arch::riscv_hart_mu_p<iss::arch::tgc_x, (iss::arch::features_e)(iss::arch::FEAT_PMP | iss::arch::FEAT_CLIC | iss::arch::FEAT_EXT_N | iss::arch::FEAT_TCM)>;
+using tgc5e_plat_type = iss::arch::riscv_hart_mu_p<iss::arch::tgc5e, (iss::arch::features_e)(iss::arch::FEAT_PMP | iss::arch::FEAT_CLIC |
+                                                                                             iss::arch::FEAT_EXT_N)>;
+#endif
+#ifdef CORE_TGC5X
+#include "riscv_hart_mu_p.h"
+#include <iss/arch/tgc5x.h>
+using tgc5x_plat_type = iss::arch::riscv_hart_mu_p<iss::arch::tgc5x, (iss::arch::features_e)(iss::arch::FEAT_PMP | iss::arch::FEAT_CLIC |
+                                                                                             iss::arch::FEAT_EXT_N | iss::arch::FEAT_TCM)>;
 #endif
 
 #endif

src/iss/arch/wt_cache.h

@@ -36,25 +36,27 @@
 #define _RISCV_HART_M_P_WT_CACHE_H
 
 #include <iss/vm_types.h>
-#include <util/ities.h>
-#include <vector>
 #include <map>
 #include <memory>
+#include <util/ities.h>
+#include <vector>
 
 namespace iss {
 namespace arch {
 namespace cache {
 
-enum class state { INVALID, VALID};
+enum class state { INVALID, VALID };
 struct line {
     uint64_t tag_addr{0};
     state st{state::INVALID};
     std::vector<uint8_t> data;
-    line(unsigned line_sz):  data(line_sz) {}
+    line(unsigned line_sz)
+    : data(line_sz) {}
 };
 struct set {
     std::vector<line> ways;
-    set(unsigned ways_count, line const& l): ways(ways_count, l) {}
+    set(unsigned ways_count, line const& l)
+    : ways(ways_count, l) {}
 };
 struct cache {
     std::vector<set> sets;
@@ -62,14 +64,14 @@ struct cache {
     cache(unsigned size, unsigned line_sz, unsigned ways) {
         line const ref_line{line_sz};
         set const ref_set{ways, ref_line};
-        sets.resize(size/(ways*line_sz), ref_set);
+        sets.resize(size / (ways * line_sz), ref_set);
     }
 };
 
 struct wt_policy {
-    bool is_cacheline_hit(cache& c );
+    bool is_cacheline_hit(cache& c);
 };
-}
+} // namespace cache
 
 // write thru, allocate on read, direct mapped or set-associative with round-robin replacement policy
 template <typename BASE> class wt_cache : public BASE {
@@ -81,82 +83,81 @@ public:
     using mem_write_f = typename BASE::mem_write_f;
     using phys_addr_t = typename BASE::phys_addr_t;
 
-    wt_cache();
+    wt_cache(feature_config cfg = feature_config{});
     virtual ~wt_cache() = default;
 
     unsigned size{4096};
-    unsigned line_sz{32};
+    unsigned line_sz{64};
     unsigned ways{1};
     uint64_t io_address{0xf0000000};
     uint64_t io_addr_mask{0xf0000000};
+
 protected:
-    iss::status read_cache(phys_addr_t addr, unsigned, uint8_t *const);
+    iss::status read_cache(phys_addr_t addr, unsigned, uint8_t* const);
-    iss::status write_cache(phys_addr_t addr, unsigned, uint8_t const *const);
+    iss::status write_cache(phys_addr_t addr, unsigned, uint8_t const* const);
     std::function<mem_read_f> cache_mem_rd_delegate;
     std::function<mem_write_f> cache_mem_wr_delegate;
     std::unique_ptr<cache::cache> dcache_ptr;
     std::unique_ptr<cache::cache> icache_ptr;
-    size_t get_way_select() {
+    size_t get_way_select() { return 0; }
-        return 0;
-    }
 };
 
-
+template <typename BASE>
-template<typename BASE>
+inline wt_cache<BASE>::wt_cache(feature_config cfg)
-inline wt_cache<BASE>::wt_cache() {
+: BASE(cfg)
+, io_address{cfg.io_address}
+, io_addr_mask{cfg.io_addr_mask} {
     auto cb = base_class::replace_mem_access(
-            [this](phys_addr_t a, unsigned l, uint8_t* const d) -> iss::status { return read_cache(a, l,d);},
+        [this](phys_addr_t a, unsigned l, uint8_t* const d) -> iss::status { return read_cache(a, l, d); },
-            [this](phys_addr_t a, unsigned l, uint8_t const* const d) -> iss::status { return write_cache(a, l,d);});
+        [this](phys_addr_t a, unsigned l, uint8_t const* const d) -> iss::status { return write_cache(a, l, d); });
     cache_mem_rd_delegate = cb.first;
     cache_mem_wr_delegate = cb.second;
 }
 
-template<typename BASE>
+template <typename BASE> iss::status iss::arch::wt_cache<BASE>::read_cache(phys_addr_t a, unsigned l, uint8_t* const d) {
-iss::status iss::arch::wt_cache<BASE>::read_cache(phys_addr_t a, unsigned l, uint8_t* const d) {
     if(!icache_ptr) {
         icache_ptr.reset(new cache::cache(size, line_sz, ways));
         dcache_ptr.reset(new cache::cache(size, line_sz, ways));
     }
-    if((a.val&io_addr_mask) != io_address) {
+    if((a.access & iss::access_type::FETCH) == iss::access_type::FETCH || (a.val & io_addr_mask) != io_address) {
-        auto set_addr=(a.val&(size-1))>>util::ilog2(line_sz*ways);
+        auto set_addr = (a.val & (size - 1)) >> util::ilog2(line_sz * ways);
-        auto tag_addr=a.val>>util::ilog2(line_sz);
+        auto tag_addr = a.val >> util::ilog2(line_sz);
-        auto& set = (is_fetch(a.access)?icache_ptr:dcache_ptr)->sets[set_addr];
+        auto& set = (is_fetch(a.access) ? icache_ptr : dcache_ptr)->sets[set_addr];
-        for(auto& cl: set.ways) {
+        for(auto& cl : set.ways) {
-            if(cl.st==cache::state::VALID && cl.tag_addr==tag_addr) {
+            if(cl.st == cache::state::VALID && cl.tag_addr == tag_addr) {
-                auto start_addr = a.val&(line_sz-1);
+                auto start_addr = a.val & (line_sz - 1);
-                for(auto i = 0U; i<l; ++i)
+                for(auto i = 0U; i < l; ++i)
-                    d[i] = cl.data[start_addr+i];
+                    d[i] = cl.data[start_addr + i];
                 return iss::Ok;
             }
         }
         auto& cl = set.ways[get_way_select()];
         phys_addr_t cl_addr{a};
-        cl_addr.val=tag_addr<<util::ilog2(line_sz);
+        cl_addr.val = tag_addr << util::ilog2(line_sz);
         cache_mem_rd_delegate(cl_addr, line_sz, cl.data.data());
-        cl.tag_addr=tag_addr;
+        cl.tag_addr = tag_addr;
-        cl.st=cache::state::VALID;
+        cl.st = cache::state::VALID;
-        auto start_addr = a.val&(line_sz-1);
+        auto start_addr = a.val & (line_sz - 1);
-        for(auto i = 0U; i<l; ++i)
+        for(auto i = 0U; i < l; ++i)
-            d[i] = cl.data[start_addr+i];
+            d[i] = cl.data[start_addr + i];
         return iss::Ok;
     } else
         return cache_mem_rd_delegate(a, l, d);
 }
 
-template<typename BASE>
+template <typename BASE> iss::status iss::arch::wt_cache<BASE>::write_cache(phys_addr_t a, unsigned l, const uint8_t* const d) {
-iss::status iss::arch::wt_cache<BASE>::write_cache(phys_addr_t a, unsigned l, const uint8_t* const d) {
     if(!dcache_ptr)
         dcache_ptr.reset(new cache::cache(size, line_sz, ways));
     auto res = cache_mem_wr_delegate(a, l, d);
-    if(res == iss::Ok && ((a.val&io_addr_mask) != io_address)) {
+    if(res == iss::Ok && ((a.val & io_addr_mask) != io_address)) {
-        auto set_addr=(a.val&(size-1))>>util::ilog2(line_sz*ways);
+        auto set_addr = (a.val & (size - 1)) >> util::ilog2(line_sz * ways);
-        auto tag_addr=a.val>>util::ilog2(line_sz);
+        auto tag_addr = a.val >> util::ilog2(line_sz);
         auto& set = dcache_ptr->sets[set_addr];
-        for(auto& cl: set.ways) {
+        for(auto& cl : set.ways) {
-            if(cl.st==cache::state::VALID && cl.tag_addr==tag_addr) {
+            if(cl.st == cache::state::VALID && cl.tag_addr == tag_addr) {
-                auto start_addr = a.val&(line_sz-1);
+                auto start_addr = a.val & (line_sz - 1);
-                for(auto i = 0U; i<l; ++i)
+                for(auto i = 0U; i < l; ++i)
-                    cl.data[start_addr+i] = d[i];
+                    cl.data[start_addr + i] = d[i];
                 break;
             }
         }
@@ -164,8 +165,6 @@ iss::status iss::arch::wt_cache<BASE>::write_cache(phys_addr_t a, unsigned l, co
     return res;
 }
 
-
-
 } // namespace arch
 } // namespace iss
 

src/iss/debugger/riscv_target_adapter.h

@@ -30,8 +30,8 @@
  *
  *******************************************************************************/
 
-#ifndef _ISS_DEBUGGER_RISCV_TARGET_ADAPTER_H_
+#ifndef _ISS_ARCH_DEBUGGER_RISCV_TARGET_ADAPTER_H_
-#define _ISS_DEBUGGER_RISCV_TARGET_ADAPTER_H_
+#define _ISS_ARCH_DEBUGGER_RISCV_TARGET_ADAPTER_H_
 
 #include "iss/arch_if.h"
 #include <iss/arch/traits.h>
@@ -48,25 +48,29 @@
 
 namespace iss {
 namespace debugger {
+
+char const* const get_csr_name(unsigned);
+constexpr auto csr_offset = 100U;
+
 using namespace iss::arch;
 using namespace iss::debugger;
 
 template <typename ARCH> class riscv_target_adapter : public target_adapter_base {
 public:
-    riscv_target_adapter(server_if *srv, iss::arch_if *core)
+    riscv_target_adapter(server_if* srv, iss::arch_if* core)
     : target_adapter_base(srv)
     , core(core) {}
 
     /*============== Thread Control ===============================*/
 
     /* Set generic thread */
-    status set_gen_thread(rp_thread_ref &thread) override;
+    status set_gen_thread(rp_thread_ref& thread) override;
 
     /* Set control thread */
-    status set_ctrl_thread(rp_thread_ref &thread) override;
+    status set_ctrl_thread(rp_thread_ref& thread) override;
 
     /* Get thread status */
-    status is_thread_alive(rp_thread_ref &thread, bool &alive) override;
+    status is_thread_alive(rp_thread_ref& thread, bool& alive) override;
 
     /*============= Register Access ================================*/
 
@@ -74,79 +78,83 @@ public:
      target byte order. If  register is not available
      corresponding bytes in avail_buf are 0, otherwise
      avail buf is 1 */
-    status read_registers(std::vector<uint8_t> &data, std::vector<uint8_t> &avail) override;
+    status read_registers(std::vector<uint8_t>& data, std::vector<uint8_t>& avail) override;
 
     /* Write all registers. buf is 4-byte aligned and it is in target
      byte order */
-    status write_registers(const std::vector<uint8_t> &data) override;
+    status write_registers(const std::vector<uint8_t>& data) override;
 
     /* Read one register. buf is 4-byte aligned and it is in
      target byte order. If  register is not available
      corresponding bytes in avail_buf are 0, otherwise
      avail buf is 1 */
-    status read_single_register(unsigned int reg_no, std::vector<uint8_t> &buf,
+    status read_single_register(unsigned int reg_no, std::vector<uint8_t>& buf, std::vector<uint8_t>& avail_buf) override;
-            std::vector<uint8_t> &avail_buf) override;
 
     /* Write one register. buf is 4-byte aligned and it is in target byte
      order */
-    status write_single_register(unsigned int reg_no, const std::vector<uint8_t> &buf) override;
+    status write_single_register(unsigned int reg_no, const std::vector<uint8_t>& buf) override;
 
     /*=================== Memory Access =====================*/
 
     /* Read memory, buf is 4-bytes aligned and it is in target
      byte order */
-    status read_mem(uint64_t addr, std::vector<uint8_t> &buf) override;
+    status read_mem(uint64_t addr, std::vector<uint8_t>& buf) override;
 
     /* Write memory, buf is 4-bytes aligned and it is in target
      byte order */
-    status write_mem(uint64_t addr, const std::vector<uint8_t> &buf) override;
+    status write_mem(uint64_t addr, const std::vector<uint8_t>& buf) override;
 
-    status process_query(unsigned int &mask, const rp_thread_ref &arg, rp_thread_info &info) override;
+    status process_query(unsigned int& mask, const rp_thread_ref& arg, rp_thread_info& info) override;
 
-    status thread_list_query(int first, const rp_thread_ref &arg, std::vector<rp_thread_ref> &result, size_t max_num,
+    status thread_list_query(int first, const rp_thread_ref& arg, std::vector<rp_thread_ref>& result, size_t max_num, size_t& num,
-            size_t &num, bool &done) override;
+                             bool& done) override;
 
-    status current_thread_query(rp_thread_ref &thread) override;
+    status current_thread_query(rp_thread_ref& thread) override;
 
-    status offsets_query(uint64_t &text, uint64_t &data, uint64_t &bss) override;
+    status offsets_query(uint64_t& text, uint64_t& data, uint64_t& bss) override;
 
-    status crc_query(uint64_t addr, size_t len, uint32_t &val) override;
+    status crc_query(uint64_t addr, size_t len, uint32_t& val) override;
 
-    status raw_query(std::string in_buf, std::string &out_buf) override;
+    status raw_query(std::string in_buf, std::string& out_buf) override;
 
-    status threadinfo_query(int first, std::string &out_buf) override;
+    status threadinfo_query(int first, std::string& out_buf) override;
 
-    status threadextrainfo_query(const rp_thread_ref &thread, std::string &out_buf) override;
+    status threadextrainfo_query(const rp_thread_ref& thread, std::string& out_buf) override;
 
-    status packetsize_query(std::string &out_buf) override;
+    status packetsize_query(std::string& out_buf) override;
 
     status add_break(break_type type, uint64_t addr, unsigned int length) override;
 
     status remove_break(break_type type, uint64_t addr, unsigned int length) override;
 
-    status resume_from_addr(bool step, int sig, uint64_t addr, rp_thread_ref thread,
+    status resume_from_addr(bool step, int sig, uint64_t addr, rp_thread_ref thread, std::function<void(unsigned)> stop_callback) override;
-            std::function<void(unsigned)> stop_callback) override;
 
-    status target_xml_query(std::string &out_buf) override;
+    status target_xml_query(std::string& out_buf) override;
 
 protected:
-    static inline constexpr addr_t map_addr(const addr_t &i) { return i; }
+    static inline constexpr addr_t map_addr(const addr_t& i) { return i; }
-
+    std::string csr_xml;
-    iss::arch_if *core;
+    iss::arch_if* core;
     rp_thread_ref thread_idx;
 };
 
-template <typename ARCH> status riscv_target_adapter<ARCH>::set_gen_thread(rp_thread_ref &thread) {
+template <typename ARCH> typename std::enable_if<iss::arch::traits<ARCH>::FLEN != 0, unsigned>::type get_f0_offset() {
+    return iss::arch::traits<ARCH>::F0;
+}
+
+template <typename ARCH> typename std::enable_if<iss::arch::traits<ARCH>::FLEN == 0, unsigned>::type get_f0_offset() { return 0; }
+
+template <typename ARCH> status riscv_target_adapter<ARCH>::set_gen_thread(rp_thread_ref& thread) {
     thread_idx = thread;
     return Ok;
 }
 
-template <typename ARCH> status riscv_target_adapter<ARCH>::set_ctrl_thread(rp_thread_ref &thread) {
+template <typename ARCH> status riscv_target_adapter<ARCH>::set_ctrl_thread(rp_thread_ref& thread) {
     thread_idx = thread;
     return Ok;
 }
 
-template <typename ARCH> status riscv_target_adapter<ARCH>::is_thread_alive(rp_thread_ref &thread, bool &alive) {
+template <typename ARCH> status riscv_target_adapter<ARCH>::is_thread_alive(rp_thread_ref& thread, bool& alive) {
     alive = 1;
     return Ok;
 }
@@ -158,10 +166,9 @@ template <typename ARCH> status riscv_target_adapter<ARCH>::is_thread_alive(rp_t
  * set if all threads are processed.
  */
 template <typename ARCH>
-status riscv_target_adapter<ARCH>::thread_list_query(int first, const rp_thread_ref &arg,
+status riscv_target_adapter<ARCH>::thread_list_query(int first, const rp_thread_ref& arg, std::vector<rp_thread_ref>& result,
-        std::vector<rp_thread_ref> &result, size_t max_num, size_t &num,
+                                                     size_t max_num, size_t& num, bool& done) {
-        bool &done) {
+    if(first == 0) {
-    if (first == 0) {
         result.clear();
         result.push_back(thread_idx);
         num = 1;
@@ -171,80 +178,81 @@ status riscv_target_adapter<ARCH>::thread_list_query(int first, const rp_thread_
         return NotSupported;
 }
 
-template <typename ARCH> status riscv_target_adapter<ARCH>::current_thread_query(rp_thread_ref &thread) {
+template <typename ARCH> status riscv_target_adapter<ARCH>::current_thread_query(rp_thread_ref& thread) {
     thread = thread_idx;
     return Ok;
 }
 
-template <typename ARCH>
+template <typename ARCH> status riscv_target_adapter<ARCH>::read_registers(std::vector<uint8_t>& data, std::vector<uint8_t>& avail) {
-status riscv_target_adapter<ARCH>::read_registers(std::vector<uint8_t> &data, std::vector<uint8_t> &avail) {
+    CPPLOG(TRACE) << "reading target registers";
-    LOG(TRACE) << "reading target registers";
-    // return idx<0?:;
     data.clear();
     avail.clear();
-    const uint8_t *reg_base = core->get_regs_base_ptr();
+    const uint8_t* reg_base = core->get_regs_base_ptr();
-    auto start_reg=arch::traits<ARCH>::X0;
+    auto start_reg = arch::traits<ARCH>::X0;
-    for (size_t reg_no = start_reg; reg_no < start_reg+33/*arch::traits<ARCH>::NUM_REGS*/; ++reg_no) {
+    for(size_t i = 0; i < 33; ++i) {
+        if(i < arch::traits<ARCH>::RFS || i == arch::traits<ARCH>::PC) {
+            auto reg_no = i < 32 ? start_reg + i : arch::traits<ARCH>::PC;
+            unsigned offset = traits<ARCH>::reg_byte_offsets[reg_no];
+            for(size_t j = 0; j < arch::traits<ARCH>::XLEN / 8; ++j) {
+                data.push_back(*(reg_base + offset + j));
+                avail.push_back(0xff);
+            }
+        } else {
+            for(size_t j = 0; j < arch::traits<ARCH>::XLEN / 8; ++j) {
+                data.push_back(0);
+                avail.push_back(0);
+            }
+        }
+    }
+    if(iss::arch::traits<ARCH>::FLEN > 0) {
+        auto fstart_reg = get_f0_offset<ARCH>();
+        for(size_t i = 0; i < 32; ++i) {
+            auto reg_no = fstart_reg + i;
             auto reg_width = arch::traits<ARCH>::reg_bit_widths[reg_no] / 8;
             unsigned offset = traits<ARCH>::reg_byte_offsets[reg_no];
-        for (size_t j = 0; j < reg_width; ++j) {
+            for(size_t j = 0; j < reg_width; ++j) {
                 data.push_back(*(reg_base + offset + j));
                 avail.push_back(0xff);
             }
         }
-    // work around fill with F type registers
+    }
-    //    if (arch::traits<ARCH>::NUM_REGS < 65) {
-    //        auto reg_width = sizeof(typename arch::traits<ARCH>::reg_t);
-    //        for (size_t reg_no = 0; reg_no < 33; ++reg_no) {
-    //            for (size_t j = 0; j < reg_width; ++j) {
-    //                data.push_back(0x0);
-    //                avail.push_back(0x00);
-    //            }
-    //            // if(arch::traits<ARCH>::XLEN < 64)
-    //            //     for(unsigned j=0; j<4; ++j){
-    //            //         data.push_back(0x0);
-    //            //         avail.push_back(0x00);
-    //            //     }
-    //        }
-    //    }
     return Ok;
 }
 
-template <typename ARCH> status riscv_target_adapter<ARCH>::write_registers(const std::vector<uint8_t> &data) {
+template <typename ARCH> status riscv_target_adapter<ARCH>::write_registers(const std::vector<uint8_t>& data) {
-    auto start_reg=arch::traits<ARCH>::X0;
+    auto start_reg = arch::traits<ARCH>::X0;
-    auto *reg_base = core->get_regs_base_ptr();
+    auto* reg_base = core->get_regs_base_ptr();
     auto iter = data.data();
-    bool e_ext = arch::traits<ARCH>::PC<32;
+    auto iter_end = data.data() + data.size();
-    for (size_t reg_no = 0; reg_no < start_reg+33/*arch::traits<ARCH>::NUM_REGS*/; ++reg_no) {
+    for(size_t i = 0; i < 33 && iter < iter_end; ++i) {
-        if(e_ext && reg_no>15){
+        auto reg_width = arch::traits<ARCH>::XLEN / 8;
-            if(reg_no==32){
+        if(i < arch::traits<ARCH>::RFS) {
-                auto reg_width = arch::traits<ARCH>::reg_bit_widths[arch::traits<ARCH>::PC] / 8;
+            auto offset = traits<ARCH>::reg_byte_offsets[start_reg + i];
+            std::copy(iter, iter + reg_width, reg_base + offset);
+        } else if(i == 32) {
             auto offset = traits<ARCH>::reg_byte_offsets[arch::traits<ARCH>::PC];
-                std::copy(iter, iter + reg_width, reg_base);
+            std::copy(iter, iter + reg_width, reg_base + offset);
-            } else {
-                const uint64_t zero_val=0;
-                auto reg_width = arch::traits<ARCH>::reg_bit_widths[15] / 8;
-                auto iter = (uint8_t*)&zero_val;
-                std::copy(iter, iter + reg_width, reg_base);
         }
-        } else {
+        iter += reg_width;
-            auto reg_width = arch::traits<ARCH>::reg_bit_widths[reg_no] / 8;
+    }
-            auto offset = traits<ARCH>::reg_byte_offsets[reg_no];
+    if(iss::arch::traits<ARCH>::FLEN > 0) {
-            std::copy(iter, iter + reg_width, reg_base);
+        auto fstart_reg = get_f0_offset<ARCH>();
-            iter += 4;
+        auto reg_width = arch::traits<ARCH>::FLEN / 8;
-            reg_base += offset;
+        for(size_t i = 0; i < 32 && iter < iter_end; ++i) {
+            unsigned offset = traits<ARCH>::reg_byte_offsets[fstart_reg + i];
+            std::copy(iter, iter + reg_width, reg_base + offset);
+            iter += reg_width;
         }
     }
     return Ok;
 }
 
 template <typename ARCH>
-status riscv_target_adapter<ARCH>::read_single_register(unsigned int reg_no, std::vector<uint8_t> &data,
+status riscv_target_adapter<ARCH>::read_single_register(unsigned int reg_no, std::vector<uint8_t>& data, std::vector<uint8_t>& avail) {
-        std::vector<uint8_t> &avail) {
+    if(reg_no < csr_offset) {
-    if (reg_no < 65) {
         // auto reg_size = arch::traits<ARCH>::reg_bit_width(static_cast<typename
         // arch::traits<ARCH>::reg_e>(reg_no))/8;
-        auto *reg_base = core->get_regs_base_ptr();
+        auto* reg_base = core->get_regs_base_ptr();
         auto reg_width = arch::traits<ARCH>::reg_bit_widths[reg_no] / 8;
         data.resize(reg_width);
         avail.resize(reg_width);
@@ -252,64 +260,59 @@ status riscv_target_adapter<ARCH>::read_single_register(unsigned int reg_no, std
         std::copy(reg_base + offset, reg_base + offset + reg_width, data.begin());
         std::fill(avail.begin(), avail.end(), 0xff);
     } else {
-        typed_addr_t<iss::address_type::PHYSICAL> a(iss::access_type::DEBUG_READ, traits<ARCH>::CSR, reg_no - 65);
+        typed_addr_t<iss::address_type::PHYSICAL> a(iss::access_type::DEBUG_READ, traits<ARCH>::CSR, reg_no - csr_offset);
         data.resize(sizeof(typename traits<ARCH>::reg_t));
         avail.resize(sizeof(typename traits<ARCH>::reg_t));
         std::fill(avail.begin(), avail.end(), 0xff);
         core->read(a, data.size(), data.data());
+        std::fill(avail.begin(), avail.end(), 0xff);
     }
     return data.size() > 0 ? Ok : Err;
 }
 
-template <typename ARCH>
+template <typename ARCH> status riscv_target_adapter<ARCH>::write_single_register(unsigned int reg_no, const std::vector<uint8_t>& data) {
-status riscv_target_adapter<ARCH>::write_single_register(unsigned int reg_no, const std::vector<uint8_t> &data) {
+    if(reg_no < csr_offset) {
-    if (reg_no < 65) {
+        auto* reg_base = core->get_regs_base_ptr();
-        auto *reg_base = core->get_regs_base_ptr();
         auto reg_width = arch::traits<ARCH>::reg_bit_widths[static_cast<typename arch::traits<ARCH>::reg_e>(reg_no)] / 8;
         auto offset = traits<ARCH>::reg_byte_offsets[reg_no];
         std::copy(data.begin(), data.begin() + reg_width, reg_base + offset);
     } else {
-        typed_addr_t<iss::address_type::PHYSICAL> a(iss::access_type::DEBUG_WRITE, traits<ARCH>::CSR, reg_no - 65);
+        typed_addr_t<iss::address_type::PHYSICAL> a(iss::access_type::DEBUG_WRITE, traits<ARCH>::CSR, reg_no - csr_offset);
         core->write(a, data.size(), data.data());
     }
     return Ok;
 }
 
-template <typename ARCH> status riscv_target_adapter<ARCH>::read_mem(uint64_t addr, std::vector<uint8_t> &data) {
+template <typename ARCH> status riscv_target_adapter<ARCH>::read_mem(uint64_t addr, std::vector<uint8_t>& data) {
     auto a = map_addr({iss::access_type::DEBUG_READ, iss::address_type::VIRTUAL, 0, addr});
     auto f = [&]() -> status { return core->read(a, data.size(), data.data()); };
     return srv->execute_syncronized(f);
 }
 
-template <typename ARCH> status riscv_target_adapter<ARCH>::write_mem(uint64_t addr, const std::vector<uint8_t> &data) {
+template <typename ARCH> status riscv_target_adapter<ARCH>::write_mem(uint64_t addr, const std::vector<uint8_t>& data) {
-    auto a = map_addr({iss::access_type::DEBUG_READ, iss::address_type::VIRTUAL, 0, addr});
+    auto a = map_addr({iss::access_type::DEBUG_WRITE, iss::address_type::VIRTUAL, 0, addr});
     auto f = [&]() -> status { return core->write(a, data.size(), data.data()); };
     return srv->execute_syncronized(f);
 }
 
 template <typename ARCH>
-status riscv_target_adapter<ARCH>::process_query(unsigned int &mask, const rp_thread_ref &arg, rp_thread_info &info) {
+status riscv_target_adapter<ARCH>::process_query(unsigned int& mask, const rp_thread_ref& arg, rp_thread_info& info) {
     return NotSupported;
 }
 
-template <typename ARCH>
+template <typename ARCH> status riscv_target_adapter<ARCH>::offsets_query(uint64_t& text, uint64_t& data, uint64_t& bss) {
-status riscv_target_adapter<ARCH>::offsets_query(uint64_t &text, uint64_t &data, uint64_t &bss) {
     text = 0;
     data = 0;
     bss = 0;
     return Ok;
 }
 
-template <typename ARCH> status riscv_target_adapter<ARCH>::crc_query(uint64_t addr, size_t len, uint32_t &val) {
+template <typename ARCH> status riscv_target_adapter<ARCH>::crc_query(uint64_t addr, size_t len, uint32_t& val) { return NotSupported; }
-    return NotSupported;
-}
 
-template <typename ARCH> status riscv_target_adapter<ARCH>::raw_query(std::string in_buf, std::string &out_buf) {
+template <typename ARCH> status riscv_target_adapter<ARCH>::raw_query(std::string in_buf, std::string& out_buf) { return NotSupported; }
-    return NotSupported;
-}
 
-template <typename ARCH> status riscv_target_adapter<ARCH>::threadinfo_query(int first, std::string &out_buf) {
+template <typename ARCH> status riscv_target_adapter<ARCH>::threadinfo_query(int first, std::string& out_buf) {
-    if (first) {
+    if(first) {
         out_buf = fmt::format("m{:x}", thread_idx.val);
     } else {
         out_buf = "l";
@@ -317,8 +320,7 @@ template <typename ARCH> status riscv_target_adapter<ARCH>::threadinfo_query(int
     return Ok;
 }
 
-template <typename ARCH>
+template <typename ARCH> status riscv_target_adapter<ARCH>::threadextrainfo_query(const rp_thread_ref& thread, std::string& out_buf) {
-status riscv_target_adapter<ARCH>::threadextrainfo_query(const rp_thread_ref &thread, std::string &out_buf) {
     std::array<char, 20> buf;
     memset(buf.data(), 0, 20);
     sprintf(buf.data(), "%02x%02x%02x%02x%02x%02x%02x%02x%02x", 'R', 'u', 'n', 'n', 'a', 'b', 'l', 'e', 0);
@@ -326,7 +328,7 @@ status riscv_target_adapter<ARCH>::threadextrainfo_query(const rp_thread_ref &th
     return Ok;
 }
 
-template <typename ARCH> status riscv_target_adapter<ARCH>::packetsize_query(std::string &out_buf) {
+template <typename ARCH> status riscv_target_adapter<ARCH>::packetsize_query(std::string& out_buf) {
     out_buf = "PacketSize=1000";
     return Ok;
 }
@@ -340,9 +342,9 @@ template <typename ARCH> status riscv_target_adapter<ARCH>::add_break(break_type
         auto saddr = map_addr({iss::access_type::FETCH, iss::address_type::PHYSICAL, 0, addr});
         auto eaddr = map_addr({iss::access_type::FETCH, iss::address_type::PHYSICAL, 0, addr + length});
         target_adapter_base::bp_lut.addEntry(++target_adapter_base::bp_count, saddr.val, eaddr.val - saddr.val);
-        LOG(TRACE) << "Adding breakpoint with handle " << target_adapter_base::bp_count << " for addr 0x" << std::hex
+        CPPLOG(TRACE) << "Adding breakpoint with handle " << target_adapter_base::bp_count << " for addr 0x" << std::hex << saddr.val
-                << saddr.val << std::dec;
+                      << std::dec;
-        LOG(TRACE) << "Now having " << target_adapter_base::bp_lut.size() << " breakpoints";
+        CPPLOG(TRACE) << "Now having " << target_adapter_base::bp_lut.size() << " breakpoints";
         return Ok;
     }
     }
@@ -356,15 +358,14 @@ template <typename ARCH> status riscv_target_adapter<ARCH>::remove_break(break_t
     case HW_EXEC: {
         auto saddr = map_addr({iss::access_type::FETCH, iss::address_type::PHYSICAL, 0, addr});
         unsigned handle = target_adapter_base::bp_lut.getEntry(saddr.val);
-        if (handle) {
+        if(handle) {
-            LOG(TRACE) << "Removing breakpoint with handle " << handle << " for addr 0x" << std::hex << saddr.val
+            CPPLOG(TRACE) << "Removing breakpoint with handle " << handle << " for addr 0x" << std::hex << saddr.val << std::dec;
-                    << std::dec;
             // TODO: check length of addr range
             target_adapter_base::bp_lut.removeEntry(handle);
-            LOG(TRACE) << "Now having " << target_adapter_base::bp_lut.size() << " breakpoints";
+            CPPLOG(TRACE) << "Now having " << target_adapter_base::bp_lut.size() << " breakpoints";
             return Ok;
         }
-        LOG(TRACE) << "Now having " << target_adapter_base::bp_lut.size() << " breakpoints";
+        CPPLOG(TRACE) << "Now having " << target_adapter_base::bp_lut.size() << " breakpoints";
         return Err;
     }
     }
@@ -373,102 +374,66 @@ template <typename ARCH> status riscv_target_adapter<ARCH>::remove_break(break_t
 template <typename ARCH>
 status riscv_target_adapter<ARCH>::resume_from_addr(bool step, int sig, uint64_t addr, rp_thread_ref thread,
                                                     std::function<void(unsigned)> stop_callback) {
-    auto *reg_base = core->get_regs_base_ptr();
+    auto* reg_base = core->get_regs_base_ptr();
     auto reg_width = arch::traits<ARCH>::reg_bit_widths[arch::traits<ARCH>::PC] / 8;
     auto offset = traits<ARCH>::reg_byte_offsets[arch::traits<ARCH>::PC];
-    const uint8_t *iter = reinterpret_cast<const uint8_t *>(&addr);
+    const uint8_t* iter = reinterpret_cast<const uint8_t*>(&addr);
     std::copy(iter, iter + reg_width, reg_base);
     return resume_from_current(step, sig, thread, stop_callback);
 }
 
-template <typename ARCH> status riscv_target_adapter<ARCH>::target_xml_query(std::string &out_buf) {
+template <typename ARCH> status riscv_target_adapter<ARCH>::target_xml_query(std::string& out_buf) {
-    const std::string res{"<?xml version=\"1.0\"?><!DOCTYPE target SYSTEM \"gdb-target.dtd\">"
+    if(!csr_xml.size()) {
-        "<target><architecture>riscv:rv32</architecture>"
+        std::ostringstream oss;
-        //"  <feature name=\"org.gnu.gdb.riscv.rv32i\">\n"
+        oss << "<?xml version=\"1.0\"?><!DOCTYPE feature SYSTEM \"gdb-target.dtd\"><target version=\"1.0\">\n";
-        //"    <reg name=\"x0\"  bitsize=\"32\" group=\"general\"/>\n"
+        if(iss::arch::traits<ARCH>::XLEN == 32)
-        //"    <reg name=\"x1\"  bitsize=\"32\" group=\"general\"/>\n"
+            oss << "<architecture>riscv:rv32</architecture>\n";
-        //"    <reg name=\"x2\"  bitsize=\"32\" group=\"general\"/>\n"
+        else if(iss::arch::traits<ARCH>::XLEN == 64)
-        //"    <reg name=\"x3\"  bitsize=\"32\" group=\"general\"/>\n"
+            oss << "  <architectureriscv:rv64</architecture>\n";
-        //"    <reg name=\"x4\"  bitsize=\"32\" group=\"general\"/>\n"
+        oss << "  <feature name=\"org.gnu.gdb.riscv.cpu\">\n";
-        //"    <reg name=\"x5\"  bitsize=\"32\" group=\"general\"/>\n"
+        auto reg_base_num = iss::arch::traits<ARCH>::X0;
-        //"    <reg name=\"x6\"  bitsize=\"32\" group=\"general\"/>\n"
+        for(auto i = 0U; i < iss::arch::traits<ARCH>::RFS; ++i) {
-        //"    <reg name=\"x7\"  bitsize=\"32\" group=\"general\"/>\n"
+            oss << "    <reg name=\"x" << i << "\" bitsize=\"" << iss::arch::traits<ARCH>::reg_bit_widths[reg_base_num + i]
-        //"    <reg name=\"x8\"  bitsize=\"32\" group=\"general\"/>\n"
+                << "\" type=\"int\" regnum=\"" << i << "\"/>\n";
-        //"    <reg name=\"x9\"  bitsize=\"32\" group=\"general\"/>\n"
+        }
-        //"    <reg name=\"x10\" bitsize=\"32\" group=\"general\"/>\n"
+        oss << "    <reg name=\"pc\" bitsize=\"" << iss::arch::traits<ARCH>::reg_bit_widths[iss::arch::traits<ARCH>::PC]
-        //"    <reg name=\"x11\" bitsize=\"32\" group=\"general\"/>\n"
+            << "\" type=\"code_ptr\" regnum=\"" << 32U << "\"/>\n";
-        //"    <reg name=\"x12\" bitsize=\"32\" group=\"general\"/>\n"
+        oss << "  </feature>\n";
-        //"    <reg name=\"x13\" bitsize=\"32\" group=\"general\"/>\n"
+        if(iss::arch::traits<ARCH>::FLEN > 0) {
-        //"    <reg name=\"x14\" bitsize=\"32\" group=\"general\"/>\n"
+            oss << "  <feature name=\"org.gnu.gdb.riscv.fpu\">\n";
-        //"    <reg name=\"x15\" bitsize=\"32\" group=\"general\"/>\n"
+            auto reg_base_num = get_f0_offset<ARCH>();
-        //"    <reg name=\"x16\" bitsize=\"32\" group=\"general\"/>\n"
+            auto type = iss::arch::traits<ARCH>::FLEN == 32 ? "ieee_single" : "riscv_double";
-        //"    <reg name=\"x17\" bitsize=\"32\" group=\"general\"/>\n"
+            for(auto i = 0U; i < 32; ++i) {
-        //"    <reg name=\"x18\" bitsize=\"32\" group=\"general\"/>\n"
+                oss << "    <reg name=\"f" << i << "\" bitsize=\"" << iss::arch::traits<ARCH>::reg_bit_widths[reg_base_num + i]
-        //"    <reg name=\"x19\" bitsize=\"32\" group=\"general\"/>\n"
+                    << "\" type=\"" << type << "\" regnum=\"" << i + 33 << "\"/>\n";
-        //"    <reg name=\"x20\" bitsize=\"32\" group=\"general\"/>\n"
+            }
-        //"    <reg name=\"x21\" bitsize=\"32\" group=\"general\"/>\n"
+            oss << "    <reg name=\"fcsr\" bitsize=\"" << iss::arch::traits<ARCH>::XLEN << "\" regnum=\"103\" type int/>\n";
-        //"    <reg name=\"x22\" bitsize=\"32\" group=\"general\"/>\n"
+            oss << "    <reg name=\"fflags\" bitsize=\"" << iss::arch::traits<ARCH>::XLEN << "\" regnum=\"101\" type int/>\n";
-        //"    <reg name=\"x23\" bitsize=\"32\" group=\"general\"/>\n"
+            oss << "    <reg name=\"frm\" bitsize=\"" << iss::arch::traits<ARCH>::XLEN << "\" regnum=\"102\" type int/>\n";
-        //"    <reg name=\"x24\" bitsize=\"32\" group=\"general\"/>\n"
+            oss << "  </feature>\n";
-        //"    <reg name=\"x25\" bitsize=\"32\" group=\"general\"/>\n"
+        }
-        //"    <reg name=\"x26\" bitsize=\"32\" group=\"general\"/>\n"
+        oss << "  <feature name=\"org.gnu.gdb.riscv.csr\">\n";
-        //"    <reg name=\"x27\" bitsize=\"32\" group=\"general\"/>\n"
+        std::vector<uint8_t> data;
-        //"    <reg name=\"x28\" bitsize=\"32\" group=\"general\"/>\n"
+        std::vector<uint8_t> avail;
-        //"    <reg name=\"x29\" bitsize=\"32\" group=\"general\"/>\n"
+        data.resize(sizeof(typename traits<ARCH>::reg_t));
-        //"    <reg name=\"x30\" bitsize=\"32\" group=\"general\"/>\n"
+        avail.resize(sizeof(typename traits<ARCH>::reg_t));
-        //"    <reg name=\"x31\" bitsize=\"32\" group=\"general\"/>\n"
+        for(auto i = 0U; i < 4096; ++i) {
-        //"  </feature>\n"
+            typed_addr_t<iss::address_type::PHYSICAL> a(iss::access_type::DEBUG_READ, traits<ARCH>::CSR, i);
-        "</target>"};
+            std::fill(avail.begin(), avail.end(), 0xff);
-    out_buf = res;
+            auto res = core->read(a, data.size(), data.data());
+            if(res == iss::Ok) {
+                oss << "    <reg name=\"" << get_csr_name(i) << "\" bitsize=\"" << iss::arch::traits<ARCH>::XLEN
+                    << "\"  type=\"int\" regnum=\"" << (i + csr_offset) << "\"/>\n";
+            }
+        }
+        oss << "  </feature>\n";
+        oss << "</target>\n";
+        csr_xml = oss.str();
+    }
+    out_buf = csr_xml;
     return Ok;
 }
+} // namespace debugger
+} // namespace iss
 
-/*
+#endif /* _ISS_ARCH_DEBUGGER_RISCV_TARGET_ADAPTER_H_ */
- *
-<?xml version="1.0"?>
-<!DOCTYPE target SYSTEM "gdb-target.dtd">
-<target>
-  <architecture>riscv:rv32</architecture>
-
-  <feature name="org.gnu.gdb.riscv.rv32i">
-    <reg name="x0"  bitsize="32" group="general"/>
-    <reg name="x1"  bitsize="32" group="general"/>
-    <reg name="x2"  bitsize="32" group="general"/>
-    <reg name="x3"  bitsize="32" group="general"/>
-    <reg name="x4"  bitsize="32" group="general"/>
-    <reg name="x5"  bitsize="32" group="general"/>
-    <reg name="x6"  bitsize="32" group="general"/>
-    <reg name="x7"  bitsize="32" group="general"/>
-    <reg name="x8"  bitsize="32" group="general"/>
-    <reg name="x9"  bitsize="32" group="general"/>
-    <reg name="x10" bitsize="32" group="general"/>
-    <reg name="x11" bitsize="32" group="general"/>
-    <reg name="x12" bitsize="32" group="general"/>
-    <reg name="x13" bitsize="32" group="general"/>
-    <reg name="x14" bitsize="32" group="general"/>
-    <reg name="x15" bitsize="32" group="general"/>
-    <reg name="x16" bitsize="32" group="general"/>
-    <reg name="x17" bitsize="32" group="general"/>
-    <reg name="x18" bitsize="32" group="general"/>
-    <reg name="x19" bitsize="32" group="general"/>
-    <reg name="x20" bitsize="32" group="general"/>
-    <reg name="x21" bitsize="32" group="general"/>
-    <reg name="x22" bitsize="32" group="general"/>
-    <reg name="x23" bitsize="32" group="general"/>
-    <reg name="x24" bitsize="32" group="general"/>
-    <reg name="x25" bitsize="32" group="general"/>
-    <reg name="x26" bitsize="32" group="general"/>
-    <reg name="x27" bitsize="32" group="general"/>
-    <reg name="x28" bitsize="32" group="general"/>
-    <reg name="x29" bitsize="32" group="general"/>
-    <reg name="x30" bitsize="32" group="general"/>
-    <reg name="x31" bitsize="32" group="general"/>
-  </feature>
-
-</target>
-
- */
-}
-}
-
-#endif /* _ISS_DEBUGGER_RISCV_TARGET_ADAPTER_H_ */

src/iss/factory.h

@@ -33,21 +33,20 @@
 #ifndef _ISS_FACTORY_H_
 #define _ISS_FACTORY_H_
 
+#include <algorithm>
+#include <functional>
 #include <iss/iss.h>
 #include <memory>
-#include <unordered_map>
-#include <functional>
 #include <string>
-#include <algorithm>
+#include <unordered_map>
 #include <vector>
 
 namespace iss {
 
 using cpu_ptr = std::unique_ptr<iss::arch_if>;
-using vm_ptr= std::unique_ptr<iss::vm_if>;
+using vm_ptr = std::unique_ptr<iss::vm_if>;
 
-template<typename PLAT>
+template <typename PLAT> std::tuple<cpu_ptr, vm_ptr> create_cpu(std::string const& backend, unsigned gdb_port) {
-std::tuple<cpu_ptr, vm_ptr> create_cpu(std::string const& backend, unsigned gdb_port){
     using core_type = typename PLAT::core;
     core_type* lcpu = new PLAT();
     if(backend == "interp")
@@ -63,44 +62,45 @@ std::tuple<cpu_ptr, vm_ptr> create_cpu(std::string const& backend, unsigned gdb_
     return {nullptr, nullptr};
 }
 
-
 class core_factory {
     using cpu_ptr = std::unique_ptr<iss::arch_if>;
-    using vm_ptr= std::unique_ptr<iss::vm_if>;
+    using vm_ptr = std::unique_ptr<iss::vm_if>;
     using base_t = std::tuple<cpu_ptr, vm_ptr>;
-    using create_fn = std::function<base_t(unsigned, void*) >;
+    using create_fn = std::function<base_t(unsigned, void*)>;
-    using registry_t = std::unordered_map<std::string, create_fn> ;
+    using registry_t = std::unordered_map<std::string, create_fn>;
 
     registry_t registry;
 
     core_factory() = default;
-    core_factory(const core_factory &) = delete;
+    core_factory(const core_factory&) = delete;
-    core_factory & operator=(const core_factory &) = delete;
+    core_factory& operator=(const core_factory&) = delete;
 
 public:
-    static core_factory & instance() { static core_factory bf; return bf; }
+    static core_factory& instance() {
+        static core_factory bf;
+        return bf;
+    }
 
-    bool register_creator(const std::string & className, create_fn const& fn) {
+    bool register_creator(const std::string& className, create_fn const& fn) {
         registry[className] = fn;
         return true;
     }
 
-    base_t create(std::string const& className, unsigned gdb_port=0, void* init_data=nullptr) const {
+    base_t create(std::string const& className, unsigned gdb_port = 0, void* init_data = nullptr) const {
         registry_t::const_iterator regEntry = registry.find(className);
-        if (regEntry != registry.end())
+        if(regEntry != registry.end())
             return regEntry->second(gdb_port, init_data);
         return {nullptr, nullptr};
     }
 
     std::vector<std::string> get_names() {
         std::vector<std::string> keys{registry.size()};
-        std::transform(std::begin(registry), std::end(registry), std::begin(keys), [](std::pair<std::string, create_fn> const& p){
+        std::transform(std::begin(registry), std::end(registry), std::begin(keys),
-            return p.first;
+                       [](std::pair<std::string, create_fn> const& p) { return p.first; });
-        });
         return keys;
     }
 };
 
-}
+} // namespace iss
 
 #endif /* _ISS_FACTORY_H_ */

src/iss/plugin/README.md

@@ -0,0 +1,8 @@
+# pctrace
+
+Trace functionality to allow visualizing coverage in lcov and cachegrind tools. Use environment variables NOCOMPRES and REGDUMP to toggle functionality.
+- NOCOMPRES: any value turns off the LZ4 compression
+- REGDUMP: any value switches to tracing the registers instead. Also turns off compression.
+
+Known Bugs: 
+- currently does not work correctly with jit backends, the plugin cant tell if instructions are compressed. Additionaly the cost of instrs that raise a trap is not known. It takes the cost of the instrid -1 (0 at the moment).

src/iss/plugin/cycle_estimate.cpp

@@ -33,86 +33,82 @@
  ******************************************************************************/
 
 #include "cycle_estimate.h"
+#include <iss/plugin/calculator.h>
+#include <yaml-cpp/yaml.h>
 
+#include <fstream>
 #include <iss/arch_if.h>
 #include <util/logging.h>
-#include <rapidjson/document.h>
-#include <rapidjson/istreamwrapper.h>
-#include <rapidjson/writer.h>
-#include <rapidjson/stringbuffer.h>
-#include <rapidjson/ostreamwrapper.h>
-#include <rapidjson/error/en.h>
-#include <fstream>
 
-using namespace rapidjson;
 using namespace std;
 
 iss::plugin::cycle_estimate::cycle_estimate(string const& config_file_name)
 : instr_if(nullptr)
-, config_file_name(config_file_name)
+, config_file_name(config_file_name) {}
-{
-}
 
-iss::plugin::cycle_estimate::~cycle_estimate() {
+iss::plugin::cycle_estimate::~cycle_estimate() = default;
-}
 
 bool iss::plugin::cycle_estimate::registration(const char* const version, vm_if& vm) {
     instr_if = vm.get_arch()->get_instrumentation_if();
-    if(!instr_if) return false;
+    assert(instr_if && "No instrumentation interface available but callback executed");
+    reg_base_ptr = reinterpret_cast<uint32_t*>(vm.get_arch()->get_regs_base_ptr());
+    if(!instr_if)
+        return false;
     const string core_name = instr_if->core_type_name();
-    if (config_file_name.length() > 0) {
+    if(config_file_name.length() > 0) {
-        ifstream is(config_file_name);
+        std::ifstream is(config_file_name);
-        if (is.is_open()) {
+        if(is.is_open()) {
             try {
-                IStreamWrapper isw(is);
+                auto root = YAML::LoadAll(is);
-                Document d;
+                if(root.size() != 1) {
-                ParseResult ok = d.ParseStream(isw);
+                    CPPLOG(ERR) << "Too many root nodes in YAML file " << config_file_name;
-                if(ok) {
-                    Value& val = d[core_name.c_str()];
-                    if(val.IsArray()){
-                        delays.reserve(val.Size());
-                        for (auto it = val.Begin(); it != val.End(); ++it) {
-                            auto& name = (*it)["name"];
-                            auto& size = (*it)["size"];
-                            auto& delay = (*it)["delay"];
-                            auto& branch = (*it)["branch"];
-                            if(delay.IsArray()) {
-                                auto dt = delay[0].Get<unsigned>();
-                                auto dnt = delay[1].Get<unsigned>();
-                                delays.push_back(instr_desc{size.Get<unsigned>(), dt, dnt, branch.Get<bool>()});
-                            } else if(delay.Is<unsigned>()) {
-                                auto d = delay.Get<unsigned>();
-                                delays.push_back(instr_desc{size.Get<unsigned>(), d, d, branch.Get<bool>()});
-                            } else
-                                throw runtime_error("JSON parse error");
                 }
+                for(auto p : root[0]) {
+                    auto isa_subset = p.first;
+                    auto instructions = p.second;
+                    for(auto const& instr : instructions) {
+                        auto idx = instr.second["index"].as<unsigned>();
+                        if(delays.size() <= idx)
+                            delays.resize(idx + 1);
+                        auto& res = delays[idx];
+                        res.is_branch = instr.second["branch"].as<bool>();
+                        auto delay = instr.second["delay"];
+                        if(delay.IsSequence()) {
+                            res.not_taken = delay[0].as<uint64_t>();
+                            res.taken = delay[1].as<uint64_t>();
                         } else {
-                        LOG(ERR)<<"plugin cycle_estimate: could not find an entry for "<<core_name<<" in JSON file"<<endl;
+                            try {
+                                res.not_taken = delay.as<uint64_t>();
+                                res.taken = res.not_taken;
+                            } catch(const YAML::BadConversion& e) {
+                                res.f = iss::plugin::calculator(reg_base_ptr, delay.as<std::string>());
+                            }
+                        }
+                    }
+                }
+            } catch(YAML::ParserException& e) {
+                CPPLOG(ERR) << "Could not parse input file " << config_file_name << ", reason: " << e.what();
                 return false;
             }
         } else {
-                    LOG(ERR)<<"plugin cycle_estimate: could not parse in JSON file at "<< ok.Offset()<<": "<<GetParseError_En(ok.Code())<<endl;
+            CPPLOG(ERR) << "Could not open input file " << config_file_name;
-                    return false;
-               }
-            } catch (runtime_error &e) {
-                LOG(ERR) << "Could not parse input file " << config_file_name << ", reason: " << e.what();
-                return false;
-            }
-        } else {
-            LOG(ERR) << "Could not open input file " << config_file_name;
             return false;
         }
     }
     return true;
-
 }
 
 void iss::plugin::cycle_estimate::callback(instr_info_t instr_info) {
-    assert(instr_if && "No instrumentation interface available but callback executed");
+    size_t instr_id = instr_info.instr_id;
-    auto entry = delays[instr_info.instr_id];
+    auto& entry = instr_id < delays.size() ? delays[instr_id] : illegal_desc;
-    bool taken = instr_if->is_branch_taken();
+    if(instr_info.phase_id == PRE_SYNC) {
-    if (taken && (entry.taken > 1))
+        if(entry.f)
-        instr_if->update_last_instr_cycles(entry.taken);
+            current_delay = entry.f(instr_if->get_instr_word());
-    else if (entry.not_taken > 1)
+    } else {
-        instr_if->update_last_instr_cycles(entry.not_taken);
+        if(!entry.f)
+            current_delay = instr_if->is_branch_taken() ? entry.taken : entry.not_taken;
+        if(current_delay > 1)
+            instr_if->update_last_instr_cycles(current_delay);
+        current_delay = 1;
+    }
 }

src/iss/plugin/cycle_estimate.h

@@ -37,6 +37,7 @@
 
 #include "iss/instrumentation_if.h"
 #include "iss/vm_plugin.h"
+#include <functional>
 #include <string>
 #include <unordered_map>
 #include <vector>
@@ -45,46 +46,44 @@ namespace iss {
 
 namespace plugin {
 
-class cycle_estimate: public vm_plugin {
+class cycle_estimate : public vm_plugin {
-	BEGIN_BF_DECL(instr_desc, uint32_t)
+    struct instr_desc {
-		BF_FIELD(taken, 24, 8)
+        size_t size{0};
-		BF_FIELD(not_taken, 16, 8)
+        bool is_branch{false};
-        BF_FIELD(is_branch, 8, 8)
+        unsigned not_taken{1};
-        BF_FIELD(size, 0, 8)
+        unsigned taken{1};
-		instr_desc(uint32_t size, uint32_t taken, uint32_t not_taken, bool branch): instr_desc() {
+        std::function<unsigned(uint64_t)> f;
-			this->size=size;
+    };
-			this->taken=taken;
-			this->not_taken=not_taken;
-			this->is_branch=branch;
-		}
-	END_BF_DECL();
 
 public:
     cycle_estimate() = delete;
 
-    cycle_estimate(const cycle_estimate &) = delete;
+    cycle_estimate(const cycle_estimate&) = delete;
 
-    cycle_estimate(const cycle_estimate &&) = delete;
+    cycle_estimate(const cycle_estimate&&) = delete;
 
     cycle_estimate(std::string const& config_file_name);
 
     virtual ~cycle_estimate();
 
-    cycle_estimate &operator=(const cycle_estimate &) = delete;
+    cycle_estimate& operator=(const cycle_estimate&) = delete;
 
-    cycle_estimate &operator=(const cycle_estimate &&) = delete;
+    cycle_estimate& operator=(const cycle_estimate&&) = delete;
 
-    bool registration(const char *const version, vm_if &arch) override;
+    bool registration(const char* const version, vm_if& arch) override;
 
-    sync_type get_sync() override { return POST_SYNC; };
+    sync_type get_sync() override { return ALL_SYNC; };
 
     void callback(instr_info_t instr_info) override;
 
 private:
-    iss::instrumentation_if *instr_if;
+    iss::instrumentation_if* instr_if{nullptr};
+    uint32_t* reg_base_ptr{nullptr};
+    instr_desc illegal_desc{};
     std::vector<instr_desc> delays;
+    unsigned current_delay{0};
     struct pair_hash {
-        size_t operator()(const std::pair<uint64_t, uint64_t> &p) const {
+        size_t operator()(const std::pair<uint64_t, uint64_t>& p) const {
             std::hash<uint64_t> hash;
             return hash(p.first) + hash(p.second);
         }
@@ -92,7 +91,7 @@ private:
     std::unordered_map<std::pair<uint64_t, uint64_t>, uint64_t, pair_hash> blocks;
     std::string config_file_name;
 };
-}
+} // namespace plugin
-}
+} // namespace iss
 
 #endif /* _ISS_PLUGIN_CYCLE_ESTIMATE_H_ */

src/iss/plugin/instruction_count.cpp

@@ -34,62 +34,63 @@
 
 #include "instruction_count.h"
 #include <iss/instrumentation_if.h>
+#include <yaml-cpp/yaml.h>
 
+#include <fstream>
 #include <iss/arch_if.h>
 #include <util/logging.h>
-#include <fstream>
 
 iss::plugin::instruction_count::instruction_count(std::string config_file_name) {
-    if (config_file_name.length() > 0) {
+    if(config_file_name.length() > 0) {
         std::ifstream is(config_file_name);
-        if (is.is_open()) {
+        if(is.is_open()) {
             try {
-                is >> root;
+                auto root = YAML::LoadAll(is);
-            } catch (Json::RuntimeError &e) {
+                if(root.size() != 1) {
-                LOG(ERR) << "Could not parse input file " << config_file_name << ", reason: " << e.what();
+                    CPPLOG(ERR) << "Too many rro nodes in YAML file " << config_file_name;
+                }
+                for(auto p : root[0]) {
+                    auto isa_subset = p.first;
+                    auto instructions = p.second;
+                    for(auto const& instr : instructions) {
+                        instr_delay res;
+                        res.instr_name = instr.first.as<std::string>();
+                        res.size = instr.second["encoding"].as<std::string>().size() - 2; // not counting 0b
+                        auto delay = instr.second["delay"];
+                        if(delay.IsSequence()) {
+                            res.not_taken_delay = delay[0].as<uint64_t>();
+                            res.taken_delay = delay[1].as<uint64_t>();
+                        } else {
+                            res.not_taken_delay = delay.as<uint64_t>();
+                            res.taken_delay = res.not_taken_delay;
+                        }
+                        delays.push_back(std::move(res));
+                    }
+                }
+                rep_counts.resize(delays.size());
+            } catch(YAML::ParserException& e) {
+                CPPLOG(ERR) << "Could not parse input file " << config_file_name << ", reason: " << e.what();
             }
         } else {
-            LOG(ERR) << "Could not open input file " << config_file_name;
+            CPPLOG(ERR) << "Could not open input file " << config_file_name;
         }
     }
 }
 
 iss::plugin::instruction_count::~instruction_count() {
-	size_t idx=0;
+    size_t idx = 0;
-	for(auto it:delays){
+    for(auto it : delays) {
-		if(rep_counts[idx]>0)
+        if(rep_counts[idx] > 0 && it.instr_name.find("__" != 0))
-			LOG(INFO)<<it.instr_name<<";"<<rep_counts[idx];
+            CPPLOG(INFO) << it.instr_name << ";" << rep_counts[idx];
         idx++;
     }
 }
 
 bool iss::plugin::instruction_count::registration(const char* const version, vm_if& vm) {
     auto instr_if = vm.get_arch()->get_instrumentation_if();
-    if(!instr_if) return false;
+    if(!instr_if)
-	const std::string  core_name = instr_if->core_type_name();
+        return false;
-    Json::Value &val = root[core_name];
-    if(!val.isNull() && val.isArray()){
-    	delays.reserve(val.size());
-    	for(auto it:val){
-    		auto name = it["name"];
-    		auto size = it["size"];
-    		auto delay = it["delay"];
-    		if(!name.isString() || !size.isUInt() || !(delay.isUInt() || delay.isArray())) throw std::runtime_error("JSON parse error");
-    		if(delay.isUInt()){
-				const instr_delay entry{name.asCString(), size.asUInt(), delay.asUInt(), 0};
-				delays.push_back(entry);
-    		} else {
-				const instr_delay entry{name.asCString(), size.asUInt(), delay[0].asUInt(), delay[1].asUInt()};
-				delays.push_back(entry);
-    		}
-    	}
-    	rep_counts.resize(delays.size());
-    } else {
-        LOG(ERR)<<"plugin instruction_count: could not find an entry for "<<core_name<<" in JSON file"<<std::endl;
-    }
     return true;
 }
 
-void iss::plugin::instruction_count::callback(instr_info_t instr_info) {
+void iss::plugin::instruction_count::callback(instr_info_t instr_info) { rep_counts[instr_info.instr_id]++; }
-	rep_counts[instr_info.instr_id]++;
-}

src/iss/plugin/instruction_count.h

@@ -36,8 +36,8 @@
 #define _ISS_PLUGIN_INSTRUCTION_COUNTER_H_
 
 #include <iss/vm_plugin.h>
-#include <json/json.h>
 #include <string>
+#include <vector>
 
 namespace iss {
 namespace plugin {
@@ -53,30 +53,29 @@ class instruction_count : public iss::vm_plugin {
 public:
     instruction_count() = delete;
 
-    instruction_count(const instruction_count &) = delete;
+    instruction_count(const instruction_count&) = delete;
 
-    instruction_count(const instruction_count &&) = delete;
+    instruction_count(const instruction_count&&) = delete;
 
     instruction_count(std::string config_file_name);
 
     virtual ~instruction_count();
 
-    instruction_count &operator=(const instruction_count &) = delete;
+    instruction_count& operator=(const instruction_count&) = delete;
 
-    instruction_count &operator=(const instruction_count &&) = delete;
+    instruction_count& operator=(const instruction_count&&) = delete;
 
-    bool registration(const char *const version, vm_if &arch) override;
+    bool registration(const char* const version, vm_if& arch) override;
 
     sync_type get_sync() override { return POST_SYNC; };
 
     void callback(instr_info_t) override;
 
 private:
-    Json::Value root;
     std::vector<instr_delay> delays;
     std::vector<uint64_t> rep_counts;
 };
-}
+} // namespace plugin
-}
+} // namespace iss
 
 #endif /* _ISS_PLUGIN_INSTRUCTION_COUNTER_H_ */