fixes missing namespaces

CMakeLists.txt

@@ -1,75 +1,80 @@
 cmake_minimum_required(VERSION 3.12)
-list(APPEND CMAKE_MODULE_PATH ${CMAKE_CURRENT_SOURCE_DIR}/cmake)
 ###############################################################################
 #
 ###############################################################################
 project(dbt-rise-tgc VERSION 1.0.0)
 
 include(GNUInstallDirs)
-include(flink)
 
 find_package(elfio QUIET)
 find_package(Boost COMPONENTS coroutine)
 find_package(jsoncpp)
 
-if(TARGET tcc::tcc)
-  set(WITH_TCC ON)
-endif()
-
 if(WITH_LLVM)
     if(DEFINED ENV{LLVM_HOME})
         find_path (LLVM_DIR LLVM-Config.cmake $ENV{LLVM_HOME}/lib/cmake/llvm)
-    endif()
+    endif(DEFINED ENV{LLVM_HOME})
-    find_package(LLVM QUIET CONFIG)
+    find_package(LLVM REQUIRED CONFIG)
-    if(LLVM_FOUND)
+    message(STATUS "Found LLVM ${LLVM_PACKAGE_VERSION}")
-    	message(STATUS "Found LLVM ${LLVM_PACKAGE_VERSION}")
+    message(STATUS "Using LLVMConfig.cmake in: ${LLVM_DIR}")
-    	message(STATUS "Using LLVMConfig.cmake in: ${LLVM_DIR}")
+    llvm_map_components_to_libnames(llvm_libs support core mcjit x86codegen x86asmparser)
-    	llvm_map_components_to_libnames(llvm_libs support core mcjit x86codegen x86asmparser)
-	else()
-		find_package(LLVM REQUIRED LLVMSupport LLVMCore LLVMMCJIT LLVMX86CodeGen LLVMX86AsmParser)
-	endif()
 endif()
 
+#Mac needed variables (adapt for your needs - http://www.cmake.org/Wiki/CMake_RPATH_handling#Mac_OS_X_and_the_RPATH)
+#set(CMAKE_MACOSX_RPATH ON)
+#set(CMAKE_SKIP_BUILD_RPATH FALSE)
+#set(CMAKE_BUILD_WITH_INSTALL_RPATH FALSE)
+#set(CMAKE_INSTALL_RPATH "${CMAKE_INSTALL_PREFIX}/lib")
+#set(CMAKE_INSTALL_RPATH_USE_LINK_PATH TRUE)
+
 add_subdirectory(softfloat)
 
 set(LIB_SOURCES 
     src/iss/plugin/instruction_count.cpp
 	src/iss/arch/tgc_c.cpp
-	src/vm/tcc/vm_tgc_c.cpp
 	src/vm/interp/vm_tgc_c.cpp
 	src/vm/fp_functions.cpp
 )
-if(WITH_TCC)
-	list(APPEND LIB_SOURCES src/vm/tcc/vm_tgc_c.cpp)
-endif()
 
 # library files
-FILE(GLOB GEN_ISS_SOURCES ${CMAKE_CURRENT_SOURCE_DIR}/src-gen/iss/arch/*.cpp)
+if(TARGET ${CORE_NAME}_cpp)
-FILE(GLOB GEN_VM_SOURCES ${CMAKE_CURRENT_SOURCE_DIR}/src-gen/vm/interp/vm_*.cpp)
+    list(APPEND LIB_SOURCES ${${CORE_NAME}_OUTPUT_FILES})
-list(APPEND LIB_SOURCES ${GEN_ISS_SOURCES} ${GEN_VM_SOURCES})
+else()
-foreach(FILEPATH ${GEN_ISS_SOURCES})
+    FILE(GLOB GEN_ISS_SOURCES ${CMAKE_CURRENT_SOURCE_DIR}/src-gen/iss/arch/*.cpp)
-    get_filename_component(CORE ${FILEPATH} NAME_WE)
+    FILE(GLOB GEN_VM_SOURCES ${CMAKE_CURRENT_SOURCE_DIR}/src-gen/vm/interp/vm_*.cpp)
-    string(TOUPPER ${CORE} CORE)
+    list(APPEND LIB_SOURCES ${GEN_ISS_SOURCES} ${GEN_VM_SOURCES})
-    list(APPEND LIB_DEFINES CORE_${CORE})
+    foreach(FILEPATH ${GEN_ISS_SOURCES})
-endforeach()
+        get_filename_component(CORE ${FILEPATH} NAME_WE)
-message("Core defines are ${LIB_DEFINES}")
+        string(TOUPPER ${CORE} CORE)
-
+        list(APPEND LIB_DEFINES CORE_${CORE})
-if(WITH_LLVM)
+    endforeach()
-	FILE(GLOB LLVM_GEN_SOURCES ${CMAKE_CURRENT_SOURCE_DIR}/src-gen/vm/llvm/vm_*.cpp)
+    message("Defines are ${LIB_DEFINES}")
-	list(APPEND LIB_SOURCES ${LLVM_GEN_SOURCES})
-endif()
-
-if(WITH_TCC)
-	FILE(GLOB TCC_GEN_SOURCES ${CMAKE_CURRENT_SOURCE_DIR}/src-gen/vm/tcc/vm_*.cpp)
-	list(APPEND LIB_SOURCES ${TCC_GEN_SOURCES})
 endif()
 
 if(TARGET RapidJSON OR TARGET RapidJSON::RapidJSON)
     list(APPEND LIB_SOURCES src/iss/plugin/cycle_estimate.cpp src/iss/plugin/pctrace.cpp)
 endif()
 
+if(WITH_LLVM)
+	FILE(GLOB LLVM_GEN_SOURCES
+	    ${CMAKE_CURRENT_SOURCE_DIR}/src-gen/vm/llvm/vm_*.cpp
+	)
+	list(APPEND LIB_SOURCES ${LLVM_GEN_SOURCES})
+endif()
+
+if(WITH_TCC)
+	FILE(GLOB TCC_GEN_SOURCES
+	    ${CMAKE_CURRENT_SOURCE_DIR}/src/vm/tcc/vm_*.cpp
+	)
+	list(APPEND LIB_SOURCES ${TCC_GEN_SOURCES})
+endif()
+
 # Define the library
 add_library(${PROJECT_NAME} ${LIB_SOURCES})
+# list code gen dependencies
+if(TARGET ${CORE_NAME}_cpp)
+    add_dependencies(${PROJECT_NAME} ${CORE_NAME}_cpp)
+endif()
 
 if("${CMAKE_CXX_COMPILER_ID}" STREQUAL "GNU")
      target_compile_options(${PROJECT_NAME} PRIVATE -Wno-shift-count-overflow)
@@ -84,8 +89,8 @@ if(TARGET jsoncpp::jsoncpp)
 else()
 	target_link_libraries(${PROJECT_NAME} PUBLIC jsoncpp)
 endif()
-if(BUILD_SHARED_LIBS)
+if("${CMAKE_CXX_COMPILER_ID}" STREQUAL "GNU" AND BUILD_SHARED_LIBS)
-    target_force_link_libraries(${PROJECT_NAME} PUBLIC dbt-rise-core)
+    target_link_libraries(${PROJECT_NAME} PUBLIC -Wl,--whole-archive dbt-rise-core -Wl,--no-whole-archive)
 else()
     target_link_libraries(${PROJECT_NAME} PUBLIC dbt-rise-core)
 endif()
@@ -156,7 +161,7 @@ if(WITH_TCC)
     target_compile_definitions(${PROJECT_NAME} PRIVATE WITH_TCC)
 endif()
 # Links the target exe against the libraries
-target_force_link_libraries(${PROJECT_NAME} PUBLIC dbt-rise-tgc)
+target_link_libraries(${PROJECT_NAME} PUBLIC dbt-rise-tgc)
 if(TARGET Boost::program_options)
     target_link_libraries(${PROJECT_NAME} PUBLIC Boost::program_options)
 else()
@@ -181,13 +186,10 @@ install(TARGETS tgc-sim
 ###############################################################################
 if(TARGET scc-sysc)
 	project(dbt-rise-tgc_sc VERSION 1.0.0)
-	set(LIB_SOURCES 
+    add_library(${PROJECT_NAME} 
     	src/sysc/core_complex.cpp
     	src/sysc/register_tgc_c.cpp
-	)
+    )
-	FILE(GLOB GEN_SC_SOURCES ${CMAKE_CURRENT_SOURCE_DIR}/src-gen/sysc/register_*.cpp)
-	list(APPEND LIB_SOURCES ${GEN_SC_SOURCES})
-    add_library(${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)

cmake/flink.cmake

@@ -1,35 +0,0 @@
-# according to https://github.com/horance-liu/flink.cmake/tree/master
-# SPDX-License-Identifier: Apache-2.0
-
-include(CMakeParseArguments)
-
-function(target_do_force_link_libraries target visibility lib)
-  if(MSVC)
-    target_link_libraries(${target} ${visibility} "/WHOLEARCHIVE:${lib}")
-  elseif(APPLE)
-    target_link_libraries(${target} ${visibility} -Wl,-force_load ${lib})
-  else()
-    target_link_libraries(${target} ${visibility} -Wl,--whole-archive ${lib} -Wl,--no-whole-archive)
-  endif()
-endfunction()
-
-function(target_force_link_libraries target)
-  cmake_parse_arguments(FLINK
-    ""
-    ""
-    "PUBLIC;INTERFACE;PRIVATE"
-    ${ARGN}
-  )
-  
-  foreach(lib IN LISTS FLINK_PUBLIC)
-    target_do_force_link_libraries(${target} PUBLIC ${lib})
-  endforeach()
-
-  foreach(lib IN LISTS FLINK_INTERFACE)
-    target_do_force_link_libraries(${target} INTERFACE ${lib})
-  endforeach()
-  
-  foreach(lib IN LISTS FLINK_PRIVATE)
-    target_do_force_link_libraries(${target} PRIVATE ${lib})
-  endforeach()
-endfunction()

gen_input/templates/CORENAME.h.gtl

@@ -55,7 +55,7 @@ def byteSize(int size){
     return 128;
 }
 def getCString(def val){
-    return val.toString()+'ULL'
+    return val.toString()
 }
 %>
 #ifndef _${coreDef.name.toUpperCase()}_H_

gen_input/templates/CORENAME_sysc.cpp.gtl

@@ -1,72 +0,0 @@
-/*******************************************************************************
- * Copyright (C) 2023 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.
- *
- *******************************************************************************/
-
-#include "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 "core_complex.h"
-
-namespace iss {
-namespace interp {
-using namespace sysc;
-volatile std::array<bool, 2> ${coreDef.name.toLowerCase()}_init = {
-        iss_factory::instance().register_creator("${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* 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("${coreDef.name.toLowerCase()}|mu_p|interp", [](unsigned gdb_port, void* data) -> iss_factory::base_t {
-            auto* cc = reinterpret_cast<sysc::tgfs::core_complex*>(data);
-            auto* cpu = new sc_core_adapter<arch::riscv_hart_mu_p<arch::${coreDef.name.toLowerCase()}>>(cc);
-            return {sysc::sc_cpu_ptr{cpu}, vm_ptr{create(static_cast<arch::${coreDef.name.toLowerCase()}*>(cpu), gdb_port)}};
-        })
-};
-}
-#if defined(WITH_TCC)
-namespace tcc {
-volatile std::array<bool, 2> ${coreDef.name.toLowerCase()}_init = {
-        core_factory::instance().register_creator("${coreDef.name.toLowerCase()}|m_p|tcc", [](unsigned gdb_port, void* data) -> std::tuple<cpu_ptr, vm_ptr>{
-            auto cc = reinterpret_cast<sysc::tgfs::core_complex*>(data);
-            auto* cpu = new sc_core_adapter<arch::riscv_hart_m_p<arch::${coreDef.name.toLowerCase()}>>(cc);
-            return {cpu_ptr{cpu}, vm_ptr{create(static_cast<arch::${coreDef.name.toLowerCase()}*>(cpu), gdb_port)}};
-        }),
-        core_factory::instance().register_creator("${coreDef.name.toLowerCase()}|mu_p|tcc", [](unsigned gdb_port, void* data) -> std::tuple<cpu_ptr, vm_ptr>{
-            auto cc = reinterpret_cast<sysc::tgfs::core_complex*>(data);
-            auto* cpu = new sc_core_adapter<arch::riscv_hart_mu_p<arch::${coreDef.name.toLowerCase()}>>(cc);
-            return {cpu_ptr{cpu}, vm_ptr{create(static_cast<arch::${coreDef.name.toLowerCase()}*>(cpu), gdb_port)}};
-        })
-};
-}
-#endif
-}

gen_input/templates/interp/CORENAME.cpp.gtl

@@ -158,81 +158,30 @@ private:
     /****************************************************************************
      * start opcode definitions
      ****************************************************************************/
-    struct instruction_descriptor {
+    struct InstructionDesriptor {
         size_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<InstructionDesriptor, ${instructions.size}> instr_descr = {{
-    const std::array<instruction_descriptor, ${instructions.size}> instr_descr = {{
          /* entries are: size, valid value, valid mask, function ptr */<%instructions.each{instr -> %>
         {${instr.length}, ${instr.encoding}, ${instr.mask}, arch::traits<ARCH>::opcode_e::${instr.instruction.name}},<%}%>
     }};
 
+    //static constexpr typename traits::addr_t upper_bits = ~traits::PGMASK;
     iss::status fetch_ins(virt_addr_t pc, uint8_t * data){
         auto phys_pc = this->core.v2p(pc);
+        //if ((pc.val & upper_bits) != ((pc.val + 2) & upper_bits)) { // we may cross a page boundary
+        //    if (this->core.read(phys_pc, 2, data) != iss::Ok) return iss::Err;
+        //    if ((data[0] & 0x3) == 0x3) // this is a 32bit instruction
+        //        if (this->core.read(this->core.v2p(pc + 2), 2, data + 2) != iss::Ok) return iss::Err;
+        //} else {
             if (this->core.read(phys_pc, 4, data) != iss::Ok)  return iss::Err;
+        //}
         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) {
@@ -259,11 +208,16 @@ 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) {
-    root = new decoding_tree_node(std::numeric_limits<uint32_t>::max());
+    unsigned id=0;
-    for(auto instr:instr_descr){
+    for (auto instr : instr_descr) {
-        root->instrs.push_back(instr);
+        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);
+        });
     }
-    populate_decoding_tree(root);
 }
 
 inline bool is_count_limit_enabled(finish_cond_e cond){
@@ -274,6 +228,14 @@ inline bool is_jump_to_self_enabled(finish_cond_e cond){
     return (cond & finish_cond_e::JUMP_TO_SELF) == finish_cond_e::JUMP_TO_SELF;
 }
 
+template <typename ARCH>
+typename arch::traits<ARCH>::opcode_e vm_impl<ARCH>::decode_inst_id(code_word_t instr){
+    for(auto& e: qlut[instr&0x3]){
+        if(!((instr&e.mask) ^ e.value )) return e.id;
+    }
+    return arch::traits<ARCH>::opcode_e::MAX_OPCODE;
+}
+
 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){
     auto pc=start;
@@ -295,7 +257,7 @@ typename vm_base<ARCH>::virt_addr_t vm_impl<ARCH>::execute_inst(finish_cond_e co
         } 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);
+            auto inst_id = decode_inst_id(instr);
             // 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));
@@ -342,7 +304,7 @@ typename vm_base<ARCH>::virt_addr_t vm_impl<ARCH>::execute_inst(finish_cond_e co
     return pc;
 }
 
-} // namespace ${coreDef.name.toLowerCase()}
+}
 
 template <>
 std::unique_ptr<vm_if> create<arch::${coreDef.name.toLowerCase()}>(arch::${coreDef.name.toLowerCase()} *core, unsigned short port, bool dump) {
@@ -358,14 +320,14 @@ std::unique_ptr<vm_if> create<arch::${coreDef.name.toLowerCase()}>(arch::${coreD
 #include <iss/arch/riscv_hart_mu_p.h>
 namespace iss {
 namespace {
-volatile std::array<bool, 2> dummy = {
+std::array<bool, 2> dummy = {
-        core_factory::instance().register_creator<core_factory::CPP>("${coreDef.name.toLowerCase()}|m_p|interp", [](unsigned port, void*) -> std::tuple<cpu_ptr, vm_ptr>{
+        core_factory::instance().register_creator("${coreDef.name.toLowerCase()}|m_p|interp", [](unsigned port, void*) -> 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::${coreDef.name.toLowerCase()}>(*cpu, false);
 		    if (port != 0) debugger::server<debugger::gdb_session>::run_server(vm, port);
             return {cpu_ptr{cpu}, vm_ptr{vm}};
         }),
-        core_factory::instance().register_creator<core_factory::CPP>("${coreDef.name.toLowerCase()}|mu_p|interp", [](unsigned port, void*) -> std::tuple<cpu_ptr, vm_ptr>{
+        core_factory::instance().register_creator("${coreDef.name.toLowerCase()}|mu_p|interp", [](unsigned port, void*) -> std::tuple<cpu_ptr, vm_ptr>{
             auto* cpu = new iss::arch::riscv_hart_mu_p<iss::arch::${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);
@@ -374,3 +336,8 @@ volatile std::array<bool, 2> dummy = {
 };
 }
 }
+extern "C" {
+	bool* get_${coreDef.name.toLowerCase()}_interp_creators() {
+		return iss::dummy.data();
+	}
+}

gen_input/templates/tcc/CORENAME.cpp.gtl

@@ -120,7 +120,57 @@ protected:
         }
     }
 
-    
+    // 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;
+
+    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);
+                }
+            }
+        }
+    }
+
+    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;
+    }
     template<unsigned W, typename U, typename S = typename std::make_signed<U>::type>
     inline S sext(U from) {
         auto mask = (1ULL<<W) - 1;
@@ -132,23 +182,14 @@ private:
     /****************************************************************************
      * start opcode definitions
      ****************************************************************************/
-    struct instruction_descriptor {
+    struct InstructionDesriptor {
         size_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<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 ${instr.instruction.name}, encoding '${instr.encoding}' */
         {${instr.length}, ${instr.encoding}, ${instr.mask}, &this_class::__${generator.functionName(instr.name)}},<%}%>
@@ -186,64 +227,11 @@ private:
         vm_impl::gen_trap_check(tu);
         return BRANCH;
     }
-    
-    //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;
-    }
 };
 
-template <typename CODE_WORD> void debug_fn(CODE_WORD instr) {
+template <typename CODE_WORD> void debug_fn(CODE_WORD insn) {
-    volatile CODE_WORD x = instr;
+    volatile CODE_WORD x = insn;
-    instr = 2 * x;
+    insn = 2 * x;
 }
 
 template <typename ARCH> vm_impl<ARCH>::vm_impl() { this(new ARCH()); }
@@ -251,11 +239,14 @@ 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) {
-    root = new decoding_tree_node(std::numeric_limits<uint32_t>::max());
+    qlut[0] = lut_00.data();
-    for(auto instr:instr_descr){
+    qlut[1] = lut_01.data();
-        root->instrs.push_back(instr);
+    qlut[2] = lut_10.data();
+    qlut[3] = lut_11.data();
+    for (auto instr : instr_descr) {
+        auto quantrant = instr.value & 0x3;
+        expand_bit_mask(29, lutmasks[quantrant], instr.value >> 2, instr.mask >> 2, 0, qlut[quantrant], instr.op);
     }
-    populate_decoding_tree(root);
 }
 
 template <typename ARCH>
@@ -263,19 +254,30 @@ std::tuple<continuation_e>
 vm_impl<ARCH>::gen_single_inst_behavior(virt_addr_t &pc, unsigned int &inst_cnt, tu_builder& tu) {
     // we fetch at max 4 byte, alignment is 2
     enum {TRAP_ID=1<<16};
-    code_word_t instr = 0;
+    code_word_t insn = 0;
+    // const typename traits::addr_t upper_bits = ~traits::PGMASK;
     phys_addr_t paddr(pc);
+    auto *const data = (uint8_t *)&insn;
     paddr = this->core.v2p(pc);
-    auto res = this->core.read(paddr, 4, reinterpret_cast<uint8_t*>(&instr));
+//    if ((pc.val & upper_bits) != ((pc.val + 2) & upper_bits)) { // we may cross a page boundary
-    if (res != iss::Ok) throw trap_access(TRAP_ID, pc.val);
+//        auto res = this->core.read(paddr, 2, data);
-    if (instr == 0x0000006f || (instr&0xffff)==0xa001) throw simulation_stopped(0); // 'J 0' or 'C.J 0'
+//        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 (insn == 0x0000006f || (insn&0xffff)==0xa001) throw simulation_stopped(0); // 'J 0' or 'C.J 0'
     // curr pc on stack
     ++inst_cnt;
-    auto f = decode_instr(root, instr);
+    auto lut_val = extract_fields(insn);
+    auto f = qlut[insn & 0x3][lut_val];
     if (f == nullptr) {
         f = &this_class::illegal_intruction;
     }
-    return (this->*f)(pc, instr, tu);
+    return (this->*f)(pc, insn, tu);
 }
 
 template <typename ARCH> void vm_impl<ARCH>::gen_raise_trap(tu_builder& tu, uint16_t trap_id, uint16_t cause) {
@@ -299,7 +301,7 @@ template <typename ARCH> void vm_impl<ARCH>::gen_trap_behavior(tu_builder& tu) {
     tu("return *next_pc;");
 }
 
-} // namespace ${coreDef.name.toLowerCase()}
+} // namespace mnrv32
 
 template <>
 std::unique_ptr<vm_if> create<arch::${coreDef.name.toLowerCase()}>(arch::${coreDef.name.toLowerCase()} *core, unsigned short port, bool dump) {
@@ -315,14 +317,14 @@ std::unique_ptr<vm_if> create<arch::${coreDef.name.toLowerCase()}>(arch::${coreD
 #include <iss/arch/riscv_hart_mu_p.h>
 namespace iss {
 namespace {
-volatile std::array<bool, 2> dummy = {
+std::array<bool, 2> dummy = {
-        core_factory::instance().register_creator<core_factory::CPP>("${coreDef.name.toLowerCase()}|m_p|tcc", [](unsigned port, void*) -> std::tuple<cpu_ptr, vm_ptr>{
+        core_factory::instance().register_creator("${coreDef.name.toLowerCase()}|m_p|tcc", [](unsigned port, void*) -> 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::${coreDef.name.toLowerCase()}>(*cpu, false);
 		    if (port != 0) debugger::server<debugger::gdb_session>::run_server(vm, port);
             return {cpu_ptr{cpu}, vm_ptr{vm}};
         }),
-        core_factory::instance().register_creator<core_factory::CPP>("${coreDef.name.toLowerCase()}|mu_p|tcc", [](unsigned port, void*) -> std::tuple<cpu_ptr, vm_ptr>{
+        core_factory::instance().register_creator("${coreDef.name.toLowerCase()}|mu_p|tcc", [](unsigned port, void*) -> std::tuple<cpu_ptr, vm_ptr>{
             auto* cpu = new iss::arch::riscv_hart_mu_p<iss::arch::${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);
@@ -331,3 +333,8 @@ volatile std::array<bool, 2> dummy = {
 };
 }
 }
+extern "C" {
+	bool* get_${coreDef.name.toLowerCase()}_tcc_creators() {
+		return iss::dummy.data();
+	}
+}

src-gen/.gitignore

@@ -1,3 +1,2 @@
 /iss
-/vm
+/vm
-/sysc

src/iss/arch/tgc_c.h

@@ -53,7 +53,7 @@ template <> struct traits<tgc_c> {
     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};
+    enum constants {MISA_VAL=1073746180, MARCHID_VAL=2147483651, XLEN=32, INSTR_ALIGNMENT=2, RFS=32, fence=0, fencei=1, fencevmal=2, fencevmau=3, CSR_SIZE=4096, MUL_LEN=64};
 
     constexpr static unsigned FP_REGS_SIZE = 0;
 

src/iss/factory.h

@@ -80,17 +80,9 @@ class core_factory {
 public:
     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 &, create_fn const&);
-        registry[className] = fn;
-        return true;
-    }
 
-    base_t create(std::string const& className, unsigned gdb_port=0, void* init_data=nullptr) const {
+    base_t create(const std::string &, unsigned gdb_port=0, void* init_data=nullptr) const;
-        registry_t::const_iterator regEntry = registry.find(className);
-        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()};
@@ -101,6 +93,18 @@ public:
     }
 };
 
+inline bool core_factory::register_creator(const std::string & className, create_fn const& fn) {
+    registry[className] = fn;
+    return true;
+}
+
+inline core_factory::base_t core_factory::create(const std::string &className, unsigned gdb_port, void* data) const {
+    registry_t::const_iterator regEntry = registry.find(className);
+    if (regEntry != registry.end())
+        return regEntry->second(gdb_port, data);
+    return {nullptr, nullptr};
+}
+
 }
 
 #endif /* _ISS_FACTORY_H_ */

src/main.cpp

@@ -39,7 +39,7 @@
 #include <boost/program_options.hpp>
 #include "iss/arch/tgc_mapper.h"
 #ifdef WITH_LLVM
-#include <iss/llvm/jit_init.h>
+#include <iss/llvm/jit_helper.h>
 #endif
 #include <iss/log_categories.h>
 #include "iss/plugin/cycle_estimate.h"

src/sysc/core_complex.cpp

@@ -37,11 +37,10 @@
 #include <iss/debugger/target_adapter_if.h>
 #include <iss/iss.h>
 #include <iss/vm_types.h>
-#include "iss_factory.h"
 #ifndef WIN32
 #include <iss/plugin/loader.h>
 #endif
-#include "sc_core_adapter_if.h"
+#include "core_complex.h"
 #include <iss/arch/tgc_mapper.h>
 #include <scc/report.h>
 #include <util/ities.h>
@@ -86,9 +85,136 @@ using namespace sc_core;
 
 namespace {
 iss::debugger::encoder_decoder encdec;
+
 std::array<const char, 4> lvl = {{'U', 'S', 'H', 'M'}};
 }
 
+template<typename PLAT>
+class core_wrapper_t : public PLAT {
+public:
+    using reg_t       = typename arch::traits<typename PLAT::core>::reg_t;
+    using phys_addr_t = typename arch::traits<typename PLAT::core>::phys_addr_t;
+    using heart_state_t = typename PLAT::hart_state_type;
+    core_wrapper_t(core_complex *owner)
+    : owner(owner) { }
+
+    uint32_t get_mode() { return this->reg.PRIV; }
+
+    inline void set_interrupt_execution(bool v) { this->interrupt_sim = v?1:0; }
+
+    inline bool get_interrupt_execution() { return this->interrupt_sim; }
+
+    heart_state_t &get_state() { return this->state; }
+
+    void notify_phase(iss::arch_if::exec_phase p) override {
+        if (p == iss::arch_if::ISTART)
+            owner->sync(this->instr_if.get_total_cycles());
+    }
+
+    sync_type needed_sync() const override { return PRE_SYNC; }
+
+    void disass_output(uint64_t pc, const std::string instr) override {
+        if (!owner->disass_output(pc, instr)) {
+            std::stringstream s;
+            s << "[p:" << lvl[this->reg.PRIV] << ";s:0x" << std::hex << std::setfill('0')
+              << std::setw(sizeof(reg_t) * 2) << (reg_t)this->state.mstatus << std::dec << ";c:"
+              << this->reg.icount + this->cycle_offset << "]";
+            SCCDEBUG(owner->name())<<"disass: "
+                << "0x" << std::setw(16) << std::right << std::setfill('0') << std::hex << pc << "\t\t" << std::setw(40)
+                << std::setfill(' ') << std::left << instr << s.str();
+        }
+    };
+
+    status read_mem(phys_addr_t addr, unsigned length, uint8_t *const data) override {
+        if (addr.access && access_type::DEBUG)
+            return owner->read_mem_dbg(addr.val, length, data) ? Ok : Err;
+        else {
+            return owner->read_mem(addr.val, length, data, is_fetch(addr.access)) ? Ok : Err;
+        }
+    }
+
+    status write_mem(phys_addr_t addr, unsigned length, const uint8_t *const data) override {
+        if (addr.access && access_type::DEBUG)
+            return owner->write_mem_dbg(addr.val, length, data) ? Ok : Err;
+        else {
+            auto res = owner->write_mem(addr.val, length, data) ? Ok : Err;
+            // clear MTIP on mtimecmp write
+            if (addr.val == 0x2004000) {
+                reg_t val;
+                this->read_csr(arch::mip, val);
+                if (val & (1ULL << 7)) this->write_csr(arch::mip, val & ~(1ULL << 7));
+            }
+            return res;
+        }
+    }
+
+    status read_csr(unsigned addr, reg_t &val) override {
+#ifndef CWR_SYSTEMC
+        if((addr==arch::time || addr==arch::timeh) && owner->mtime_o.get_interface(0)){
+            uint64_t time_val;
+            bool ret = owner->mtime_o->nb_peek(time_val);
+            if (addr == iss::arch::time) {
+                val = static_cast<reg_t>(time_val);
+            } else if (addr == iss::arch::timeh) {
+                if (sizeof(reg_t) != 4) return iss::Err;
+                val = static_cast<reg_t>(time_val >> 32);
+            }
+            return ret?Ok:Err;
+#else
+		if((addr==arch::time || addr==arch::timeh)){
+			uint64_t time_val = owner->mtime_i.read();
+			if (addr == iss::arch::time) {
+				val = static_cast<reg_t>(time_val);
+			} else if (addr == iss::arch::timeh) {
+				if (sizeof(reg_t) != 4) return iss::Err;
+				val = static_cast<reg_t>(time_val >> 32);
+			}
+			return Ok;
+#endif
+        } else {
+            return PLAT::read_csr(addr, val);
+        }
+    }
+
+    void wait_until(uint64_t flags) override {
+        SCCDEBUG(owner->name()) << "Sleeping until interrupt";
+        while(this->reg.pending_trap == 0 && (this->csr[arch::mip] & this->csr[arch::mie]) == 0) {
+            sc_core::wait(wfi_evt);
+        }
+        PLAT::wait_until(flags);
+    }
+
+    void local_irq(short id, bool value) {
+        reg_t mask = 0;
+        switch (id) {
+        case 3: // SW
+            mask = 1 << 3;
+            break;
+        case 7: // timer
+            mask = 1 << 7;
+            break;
+        case 11: // external
+            mask = 1 << 11;
+            break;
+        default:
+            if(id>15) mask = 1 << id;
+            break;
+        }
+        if (value) {
+            this->csr[arch::mip] |= mask;
+            wfi_evt.notify();
+        } else
+            this->csr[arch::mip] &= ~mask;
+        this->check_interrupt();
+        if(value)
+            SCCTRACE(owner->name()) << "Triggering interrupt " << id << " Pending trap: " << this->reg.pending_trap;
+    }
+
+private:
+    core_complex *const owner;
+    sc_event wfi_evt;
+};
+
 int cmd_sysc(int argc, char *argv[], debugger::out_func of, debugger::data_func df,
              debugger::target_adapter_if *tgt_adapter) {
     if (argc > 1) {
@@ -128,9 +254,7 @@ public:
 
     void reset(uint64_t addr){vm->reset(addr);}
     inline void start(){vm->start();}
-    inline std::pair<uint64_t, bool> load_file(std::string const& name){
+    inline std::pair<uint64_t, bool> load_file(std::string const& name){ return cpu->load_file(name);};
-        iss::arch_if* cc = cpu->get_arch_if();
-        return cc->load_file(name);};
 
     std::function<unsigned(void)> get_mode;
     std::function<uint64_t(void)> get_state;
@@ -138,35 +262,45 @@ public:
     std::function<void(bool)> set_interrupt_execution;
     std::function<void(short, bool)> local_irq;
 
-    void create_cpu(std::string const& type, std::string const& backend, unsigned gdb_port, uint32_t hart_id){
+    template<typename PLAT>
-        auto & f = sysc::iss_factory::instance();
+    std::tuple<cpu_ptr, vm_ptr> create_core(std::string const& backend, unsigned gdb_port, uint32_t hart_id){
-        if(type.size()==0 || type == "?") {
+        auto* lcpu = new core_wrapper_t<PLAT>(owner);
-            std::cout<<"Available cores: "<<util::join(f.get_names(), ", ")<<std::endl;
+        lcpu->set_mhartid(hart_id);
-            sc_core::sc_stop();
+        get_mode = [lcpu]() { return lcpu->get_mode(); };
-        } else if (type.find('|') != std::string::npos) {
+        get_state = [lcpu]() { return lcpu->get_state().mstatus.backing.val; };
-            std::tie(cpu, vm) = f.create(type+"|"+backend);
+        get_interrupt_execution = [lcpu]() { return lcpu->get_interrupt_execution(); };
-        } else {
+        set_interrupt_execution = [lcpu](bool b) { return lcpu->set_interrupt_execution(b); };
-            auto base_isa = type.substr(0, 5);
+        local_irq = [lcpu](short s, bool b) { return lcpu->local_irq(s, b); };
-            if(base_isa=="tgc_d" || base_isa=="tgc_e") {
+        if(backend == "interp")
-                std::tie(cpu, vm) = f.create(type + "|mu_p_clic_pmp|" + backend, gdb_port, owner);
+            return {cpu_ptr{lcpu}, vm_ptr{iss::interp::create(static_cast<typename PLAT::core*>(lcpu), gdb_port)}};
-            } else {
+#ifdef WITH_LLVM
-                std::tie(cpu, vm) = f.create(type + "|m_p|" + backend, gdb_port, owner);
+        if(backend == "llvm")
-           }
+            return {cpu_ptr{lcpu}, vm_ptr{iss::llvm::create(lcpu, gdb_port)}};
-        }
+#endif
-        if(!cpu ){
+#ifdef WITH_TCC
-            SCCFATAL() << "Could not create cpu for isa " << type << " and backend " <<backend;
+        if(backend == "tcc")
-        }
+    s        return {cpu_ptr{lcpu}, vm_ptr{iss::tcc::create(lcpu, gdb_port)}};
-        if(!vm ){
+#endif
-            SCCFATAL() << "Could not create vm for isa " << type << " and backend " <<backend;
+        return {nullptr, nullptr};
-        }
+    }
-        auto* sc_cpu_if = reinterpret_cast<sc_core_adapter_if*>(cpu.get());
-        sc_cpu_if->set_mhartid(hart_id);
-        get_mode = [sc_cpu_if]() { return sc_cpu_if->get_mode(); };
-        get_state = [sc_cpu_if]() { return sc_cpu_if->get_state(); };
-        get_interrupt_execution = [sc_cpu_if]() { return sc_cpu_if->get_interrupt_execution(); };
-        set_interrupt_execution = [sc_cpu_if](bool b) { return sc_cpu_if->set_interrupt_execution(b); };
-        local_irq = [sc_cpu_if](short s, bool b) { return sc_cpu_if->local_irq(s, b); };
 
+    void create_cpu(std::string const& type, std::string const& backend, unsigned gdb_port, uint32_t hart_id){
+        CREATE_CORE(tgc_c)
+#ifdef CORE_TGC_B
+        CREATE_CORE(tgc_b)
+#endif
+#ifdef CORE_TGC_D
+        CREATE_CORE(tgc_d)
+#endif
+#ifdef CORE_TGC_D_XRB_MAC
+        CREATE_CORE(tgc_d_xrb_mac)
+#endif
+#ifdef CORE_TGC_D_XRB_NN
+        CREATE_CORE(tgc_d_xrb_nn)
+#endif
+        {
+            LOG(ERR) << "Illegal argument value for core type: " << type << std::endl;
+        }
         auto *srv = debugger::server<debugger::gdb_session>::get();
         if (srv) tgt_adapter = srv->get_target();
         if (tgt_adapter)
@@ -179,7 +313,7 @@ public:
 
     core_complex * const owner;
     vm_ptr vm{nullptr};
-    sc_cpu_ptr cpu{nullptr};
+    cpu_ptr cpu{nullptr};
     iss::debugger::target_adapter_if *tgt_adapter{nullptr};
 };
 
@@ -500,5 +634,5 @@ bool core_complex::write_mem_dbg(uint64_t addr, unsigned length, const uint8_t *
     gp.set_streaming_width(length);
     return dbus->transport_dbg(gp) == length;
 }
-} /* namespace tgfs */
+} /* namespace SiFive */
 } /* namespace sysc */

src/sysc/iss_factory.h

@@ -1,88 +0,0 @@
-/*******************************************************************************
- * Copyright (C) 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 _ISS_FACTORY_H_
-#define _ISS_FACTORY_H_
-
-#include <iss/iss.h>
-#include "sc_core_adapter_if.h"
-#include <memory>
-#include <unordered_map>
-#include <functional>
-#include <string>
-#include <algorithm>
-#include <vector>
-
-namespace sysc {
-
-using sc_cpu_ptr = std::unique_ptr<sc_core_adapter_if>;
-using vm_ptr= std::unique_ptr<iss::vm_if>;
-
-class iss_factory {
-public:
-    using base_t = std::tuple<sc_cpu_ptr, vm_ptr>;
-    using create_fn = std::function<base_t(unsigned, void*) >;
-    using registry_t = std::unordered_map<std::string, create_fn> ;
-
-    iss_factory() = default;
-    iss_factory(const iss_factory &) = delete;
-    iss_factory & operator=(const iss_factory &) = delete;
-
-    static iss_factory & instance() { static iss_factory bf; return bf; }
-
-    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 {
-        registry_t::const_iterator regEntry = registry.find(className);
-        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){
-            return p.first;
-        });
-        return keys;
-    }
-private:
-    registry_t registry;
-
-};
-
-}
-
-#endif /* _ISS_FACTORY_H_ */

src/sysc/register_tgc_c.cpp

@@ -1,36 +1,14 @@
-/*******************************************************************************
+/*
- * Copyright (C) 2023 MINRES Technologies GmbH
+ * register_tgc_c.cpp
- * All rights reserved.
  *
- * Redistribution and use in source and binary forms, with or without
+ *  Created on: Jul 5, 2023
- * modification, are permitted provided that the following conditions are met:
+ *      Author: eyck
- *
+ */
- * 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.
- *
- *******************************************************************************/
 
-#include "iss_factory.h"
+
+
+
+#include <iss/factory.h>
 #include <iss/arch/tgc_c.h>
 #include <iss/arch/riscv_hart_m_p.h>
 #include <iss/arch/riscv_hart_mu_p.h>
@@ -38,35 +16,18 @@
 #include "core_complex.h"
 
 namespace iss {
-namespace interp {
+namespace {
-using namespace sysc;
+volatile std::array<bool, 2> dummy = {
-volatile std::array<bool, 2> tgc_init = {
+        core_factory::instance().register_creator("tgc_c|m_p|interp", [](unsigned gdb_port, void* data) -> std::tuple<cpu_ptr, vm_ptr>{
-        iss_factory::instance().register_creator("tgc_c|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*>(data);
+            arch::tgc_c* lcpu = new sc_core_adapter<arch::riscv_hart_m_p<arch::tgc_c>>(cc);
-            auto* cpu = new sc_core_adapter<arch::riscv_hart_m_p<arch::tgc_c>>(cc);
+            return {cpu_ptr{lcpu}, vm_ptr{interp::create(lcpu, gdb_port)}};
-            return {sysc::sc_cpu_ptr{cpu}, vm_ptr{create(static_cast<arch::tgc_c*>(cpu), gdb_port)}};
         }),
-        iss_factory::instance().register_creator("tgc_c|mu_p|interp", [](unsigned gdb_port, void* data) -> iss_factory::base_t {
+        core_factory::instance().register_creator("tgc_c|mu_p|interp", [](unsigned gdb_port, void* data) -> std::tuple<cpu_ptr, vm_ptr>{
-            auto* cc = reinterpret_cast<sysc::tgfs::core_complex*>(data);
+            auto cc = reinterpret_cast<sysc::tgfs::core_complex*>(data);
-            auto* cpu = new sc_core_adapter<arch::riscv_hart_mu_p<arch::tgc_c>>(cc);
+            arch::tgc_c* lcpu = new sc_core_adapter<arch::riscv_hart_mu_p<arch::tgc_c>>(cc);
-            return {sysc::sc_cpu_ptr{cpu}, vm_ptr{create(static_cast<arch::tgc_c*>(cpu), gdb_port)}};
+            return {cpu_ptr{lcpu}, vm_ptr{interp::create(lcpu, gdb_port)}};
         })
 };
 }
-#if defined(WITH_TCC)
-namespace tcc {
-volatile std::array<bool, 2> tgc_init = {
-        core_factory::instance().register_creator("tgc_c|m_p|tcc", [](unsigned gdb_port, void* data) -> std::tuple<cpu_ptr, vm_ptr>{
-            auto cc = reinterpret_cast<sysc::tgfs::core_complex*>(data);
-            auto* cpu = new sc_core_adapter<arch::riscv_hart_m_p<arch::tgc_c>>(cc);
-            return {cpu_ptr{cpu}, vm_ptr{create(static_cast<arch::tgc_c*>(cpu), gdb_port)}};
-        }),
-        core_factory::instance().register_creator("tgc_c|mu_p|tcc", [](unsigned gdb_port, void* data) -> std::tuple<cpu_ptr, vm_ptr>{
-            auto cc = reinterpret_cast<sysc::tgfs::core_complex*>(data);
-            auto* cpu = new sc_core_adapter<arch::riscv_hart_mu_p<arch::tgc_c>>(cc);
-            return {cpu_ptr{cpu}, vm_ptr{create(static_cast<arch::tgc_c*>(cpu), gdb_port)}};
-        })
-};
-}
-#endif
 }

src/sysc/sc_core_adapter.h

@@ -11,14 +11,14 @@
 
 #include <scc/report.h>
 #include <util/ities.h>
-#include "sc_core_adapter_if.h"
+#include "core_complex.h"
 #include <iss/iss.h>
 #include <iss/vm_types.h>
 #include <iostream>
 
-namespace sysc {
+
 template<typename PLAT>
-class sc_core_adapter : public PLAT, public sc_core_adapter_if {
+class sc_core_adapter : public PLAT {
 public:
     using reg_t       = typename iss::arch::traits<typename PLAT::core>::reg_t;
     using phys_addr_t = typename iss::arch::traits<typename PLAT::core>::phys_addr_t;
@@ -26,17 +26,13 @@ public:
     sc_core_adapter(sysc::tgfs::core_complex *owner)
     : owner(owner) { }
 
-    iss::arch_if* get_arch_if() override { return this;}
+    uint32_t get_mode() { return this->reg.PRIV; }
 
-    void set_mhartid(unsigned id) override { PLAT::set_mhartid(id); }
+    inline void set_interrupt_execution(bool v) { this->interrupt_sim = v?1:0; }
 
-    uint32_t get_mode() override { return this->reg.PRIV; }
+    inline bool get_interrupt_execution() { return this->interrupt_sim; }
 
-    void set_interrupt_execution(bool v) override { this->interrupt_sim = v?1:0; }
+    heart_state_t &get_state() { return this->state; }
-
-    bool get_interrupt_execution() override { return this->interrupt_sim; }
-
-    uint64_t get_state() override { return this->state.mstatus.backing.val; }
 
     void notify_phase(iss::arch_if::exec_phase p) override {
         if (p == iss::arch_if::ISTART)
@@ -117,7 +113,7 @@ public:
         PLAT::wait_until(flags);
     }
 
-    void local_irq(short id, bool value) override {
+    void local_irq(short id, bool value) {
         reg_t mask = 0;
         switch (id) {
         case 3: // SW
@@ -147,5 +143,6 @@ private:
     sysc::tgfs::core_complex *const owner;
     sc_core::sc_event wfi_evt;
 };
-}
+
+
 #endif /* _SYSC_SC_CORE_ADAPTER_H_ */

src/sysc/sc_core_adapter_if.h

@@ -1,31 +0,0 @@
-/*
- * sc_core_adapter.h
- *
- *  Created on: Jul 5, 2023
- *      Author: eyck
- */
-
-#ifndef _SYSC_SC_CORE_ADAPTER_IF_H_
-#define _SYSC_SC_CORE_ADAPTER_IF_H_
-
-
-#include <scc/report.h>
-#include <util/ities.h>
-#include "core_complex.h"
-#include <iss/iss.h>
-#include <iss/vm_types.h>
-#include <iostream>
-
-namespace sysc {
-struct sc_core_adapter_if {
-    virtual iss::arch_if* get_arch_if() = 0;
-    virtual void set_mhartid(unsigned) = 0;
-    virtual uint32_t get_mode() = 0;
-    virtual uint64_t get_state() = 0;
-    virtual bool get_interrupt_execution() = 0;
-    virtual void set_interrupt_execution(bool v) = 0;
-    virtual void local_irq(short id, bool value) = 0;
-    virtual ~sc_core_adapter_if() = default;
-};
-}
-#endif /* _SYSC_SC_CORE_ADAPTER_IF_H_ */

src/vm/interp/vm_tgc_c.cpp

@@ -152,22 +152,14 @@ private:
     /****************************************************************************
      * start opcode definitions
      ****************************************************************************/
-    struct instruction_descriptor {
+    struct InstructionDesriptor {
         size_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<InstructionDesriptor, 87> instr_descr = {{
-    const std::array<instruction_descriptor, 87> instr_descr = {{
          /* entries are: size, valid value, valid mask, function ptr */
         {32, 0b00000000000000000000000000110111, 0b00000000000000000000000001111111, arch::traits<ARCH>::opcode_e::LUI},
         {32, 0b00000000000000000000000000010111, 0b00000000000000000000000001111111, arch::traits<ARCH>::opcode_e::AUIPC},
@@ -258,61 +250,18 @@ private:
         {16, 0b0000000000000000, 0b1111111111111111, arch::traits<ARCH>::opcode_e::DII},
     }};
 
+    //static constexpr typename traits::addr_t upper_bits = ~traits::PGMASK;
     iss::status fetch_ins(virt_addr_t pc, uint8_t * data){
         auto phys_pc = this->core.v2p(pc);
+        //if ((pc.val & upper_bits) != ((pc.val + 2) & upper_bits)) { // we may cross a page boundary
+        //    if (this->core.read(phys_pc, 2, data) != iss::Ok) return iss::Err;
+        //    if ((data[0] & 0x3) == 0x3) // this is a 32bit instruction
+        //        if (this->core.read(this->core.v2p(pc + 2), 2, data + 2) != iss::Ok) return iss::Err;
+        //} else {
             if (this->core.read(phys_pc, 4, data) != iss::Ok)  return iss::Err;
+        //}
         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) {
@@ -339,11 +288,16 @@ 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) {
-    root = new decoding_tree_node(std::numeric_limits<uint32_t>::max());
+    unsigned id=0;
-    for(auto instr:instr_descr){
+    for (auto instr : instr_descr) {
-        root->instrs.push_back(instr);
+        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);
+        });
     }
-    populate_decoding_tree(root);
 }
 
 inline bool is_count_limit_enabled(finish_cond_e cond){
@@ -354,6 +308,14 @@ inline bool is_jump_to_self_enabled(finish_cond_e cond){
     return (cond & finish_cond_e::JUMP_TO_SELF) == finish_cond_e::JUMP_TO_SELF;
 }
 
+template <typename ARCH>
+typename arch::traits<ARCH>::opcode_e vm_impl<ARCH>::decode_inst_id(code_word_t instr){
+    for(auto& e: qlut[instr&0x3]){
+        if(!((instr&e.mask) ^ e.value )) return e.id;
+    }
+    return arch::traits<ARCH>::opcode_e::MAX_OPCODE;
+}
+
 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){
     auto pc=start;
@@ -375,7 +337,7 @@ typename vm_base<ARCH>::virt_addr_t vm_impl<ARCH>::execute_inst(finish_cond_e co
         } 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);
+            auto inst_id = decode_inst_id(instr);
             // 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));
@@ -2673,7 +2635,7 @@ typename vm_base<ARCH>::virt_addr_t vm_impl<ARCH>::execute_inst(finish_cond_e co
     return pc;
 }
 
-} // namespace tgc_c
+}
 
 template <>
 std::unique_ptr<vm_if> create<arch::tgc_c>(arch::tgc_c *core, unsigned short port, bool dump) {
@@ -2689,7 +2651,7 @@ std::unique_ptr<vm_if> create<arch::tgc_c>(arch::tgc_c *core, unsigned short por
 #include <iss/arch/riscv_hart_mu_p.h>
 namespace iss {
 namespace {
-volatile std::array<bool, 2> dummy = {
+std::array<bool, 2> dummy = {
         core_factory::instance().register_creator("tgc_c|m_p|interp", [](unsigned port, void*) -> std::tuple<cpu_ptr, vm_ptr>{
             auto* cpu = new iss::arch::riscv_hart_m_p<iss::arch::tgc_c>();
 		    auto vm = new interp::tgc_c::vm_impl<arch::tgc_c>(*cpu, false);
@@ -2705,3 +2667,8 @@ volatile std::array<bool, 2> dummy = {
 };
 }
 }
+extern "C" {
+	bool* get_tgc_c_interp_creators() {
+		return iss::dummy.data();
+	}
+}

src/vm/tcc/vm_tgc_c.cpp

@@ -120,7 +120,57 @@ protected:
         }
     }
 
-    
+    // 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;
+
+    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);
+                }
+            }
+        }
+    }
+
+    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;
+    }
     template<unsigned W, typename U, typename S = typename std::make_signed<U>::type>
     inline S sext(U from) {
         auto mask = (1ULL<<W) - 1;
@@ -132,23 +182,14 @@ private:
     /****************************************************************************
      * start opcode definitions
      ****************************************************************************/
-    struct instruction_descriptor {
+    struct InstructionDesriptor {
         size_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<InstructionDesriptor, 87> instr_descr = {{
-
-    const std::array<instruction_descriptor, 87> instr_descr = {{
          /* entries are: size, valid value, valid mask, function ptr */
         /* instruction LUI, encoding '0b00000000000000000000000000110111' */
         {32, 0b00000000000000000000000000110111, 0b00000000000000000000000001111111, &this_class::__lui},
@@ -452,14 +493,14 @@ private:
             this->gen_raise_trap(tu, 0,  2);
         }
         else{
-        	auto new_pc = tu.assignment(tu.ext((tu.bitwise_and((tu.add(tu.load(rs1+ traits::X0, 0),tu.constant((int16_t)sext<12>(imm),16))),tu.constant(~0x1,8))),32,true),32);
+        	auto new_pc = tu.assignment(tu.ext((tu.bitwise_and((tu.add(tu.load(rs1+ traits::X0, 0),tu.constant((int16_t)sext<12>(imm),16))),tu.constant(~ 0x1,8))),32,true),32);
         	tu.open_if(tu.srem(new_pc,tu.constant(static_cast<uint32_t>(traits:: INSTR_ALIGNMENT),32)));
         	this->gen_raise_trap(tu, 0,  0);
         	tu.open_else();
         	if(rd!= 0) {
         	    tu.store(rd + traits::X0,tu.ext((tu.add(tu.ext(cur_pc_val,32,false),tu.constant( 4,8))),32,true));
         	}
-        	auto PC_val_v = tu.assignment("PC_val", tu.bitwise_and(new_pc,tu.constant(~0x1,8)),32);
+        	auto PC_val_v = tu.assignment("PC_val", tu.bitwise_and(new_pc,tu.constant(~ 0x1,8)),32);
         	tu.store(traits::NEXT_PC, PC_val_v);
         	tu.store(traits::LAST_BRANCH, tu.constant(2U, 2));
         	tu.close_scope();
@@ -1921,7 +1962,7 @@ private:
         else{
         	auto xrd = tu.assignment(tu.read_mem(traits::CSR, csr, 32),32);
         	if(zimm!= 0) {
-        	    tu.write_mem(traits::CSR, csr, tu.bitwise_and(xrd,tu.constant(~((uint32_t)zimm),32)));
+        	    tu.write_mem(traits::CSR, csr, tu.bitwise_and(xrd,tu.constant(~ ((uint32_t)zimm),32)));
         	}
         	if(rd!= 0) {
         	    tu.store(rd + traits::X0,xrd);
@@ -2122,13 +2163,13 @@ private:
         	auto divisor = tu.assignment(tu.ext(tu.load(rs2+ traits::X0, 0),32,false),32);
         	if(rd!= 0){ tu.open_if(tu.icmp(ICmpInst::ICMP_NE,divisor,tu.constant( 0,8)));
         	auto MMIN = tu.assignment(tu.constant(((uint32_t)1)<<(static_cast<uint32_t>(traits:: XLEN)-1),32),32);
-        	tu.open_if(tu.logical_and(tu.icmp(ICmpInst::ICMP_EQ,tu.load(rs1+ traits::X0, 0),MMIN),tu.icmp(ICmpInst::ICMP_EQ,divisor,tu.constant(-1,8))));
+        	tu.open_if(tu.logical_and(tu.icmp(ICmpInst::ICMP_EQ,tu.load(rs1+ traits::X0, 0),MMIN),tu.icmp(ICmpInst::ICMP_EQ,divisor,tu.constant(- 1,8))));
         	tu.store(rd + traits::X0,MMIN);
         	tu.open_else();
         	tu.store(rd + traits::X0,tu.ext((tu.sdiv(dividend,divisor)),32,true));
         	tu.close_scope();
         	tu.open_else();
-        	tu.store(rd + traits::X0,tu.constant((uint32_t)-1,32));
+        	tu.store(rd + traits::X0,tu.constant((uint32_t)- 1,32));
         	tu.close_scope();
         	}
         }
@@ -2167,7 +2208,7 @@ private:
         	}
         	tu.open_else();
         	if(rd!=0) {
-        	    tu.store(rd + traits::X0,tu.constant((uint32_t)-1,32));
+        	    tu.store(rd + traits::X0,tu.constant((uint32_t)- 1,32));
         	}
         	tu.close_scope();
         }
@@ -2202,7 +2243,7 @@ private:
         else{
         	tu.open_if(tu.icmp(ICmpInst::ICMP_NE,tu.load(rs2+ traits::X0, 0),tu.constant( 0,8)));
         	auto MMIN = tu.assignment(tu.constant( 1<<(static_cast<uint32_t>(traits:: XLEN)-1),8),32);
-        	tu.open_if(tu.logical_and(tu.icmp(ICmpInst::ICMP_EQ,tu.load(rs1+ traits::X0, 0),MMIN),tu.icmp(ICmpInst::ICMP_EQ,tu.ext(tu.load(rs2+ traits::X0, 0),32,false),tu.constant(-1,8))));
+        	tu.open_if(tu.logical_and(tu.icmp(ICmpInst::ICMP_EQ,tu.load(rs1+ traits::X0, 0),MMIN),tu.icmp(ICmpInst::ICMP_EQ,tu.ext(tu.load(rs2+ traits::X0, 0),32,false),tu.constant(- 1,8))));
         	if(rd!=0) {
         	    tu.store(rd + traits::X0,tu.constant( 0,8));
         	}
@@ -2914,7 +2955,7 @@ private:
         gen_set_pc(tu, pc, traits::NEXT_PC);
         tu.open_scope();
         if(rs1&&rs1<static_cast<uint32_t>(traits:: RFS)) {
-            auto PC_val_v = tu.assignment("PC_val", tu.bitwise_and(tu.load(rs1%static_cast<uint32_t>(traits:: RFS)+ traits::X0, 0),tu.constant(~0x1,8)),32);
+            auto PC_val_v = tu.assignment("PC_val", tu.bitwise_and(tu.load(rs1%static_cast<uint32_t>(traits:: RFS)+ traits::X0, 0),tu.constant(~ 0x1,8)),32);
             tu.store(traits::NEXT_PC, PC_val_v);
             tu.store(traits::LAST_BRANCH, tu.constant(2U, 2));
         }
@@ -3002,7 +3043,7 @@ private:
         else{
         	auto new_pc = tu.assignment(tu.load(rs1+ traits::X0, 0),32);
         	tu.store(1 + traits::X0,tu.ext((tu.add(tu.ext(cur_pc_val,32,false),tu.constant( 2,8))),32,true));
-        	auto PC_val_v = tu.assignment("PC_val", tu.bitwise_and(new_pc,tu.constant(~0x1,8)),32);
+        	auto PC_val_v = tu.assignment("PC_val", tu.bitwise_and(new_pc,tu.constant(~ 0x1,8)),32);
         	tu.store(traits::NEXT_PC, PC_val_v);
         	tu.store(traits::LAST_BRANCH, tu.constant(2U, 2));
         }
@@ -3095,64 +3136,11 @@ private:
         vm_impl::gen_trap_check(tu);
         return BRANCH;
     }
-    
-    //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;
-    }
 };
 
-template <typename CODE_WORD> void debug_fn(CODE_WORD instr) {
+template <typename CODE_WORD> void debug_fn(CODE_WORD insn) {
-    volatile CODE_WORD x = instr;
+    volatile CODE_WORD x = insn;
-    instr = 2 * x;
+    insn = 2 * x;
 }
 
 template <typename ARCH> vm_impl<ARCH>::vm_impl() { this(new ARCH()); }
@@ -3160,11 +3148,14 @@ 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) {
-    root = new decoding_tree_node(std::numeric_limits<uint32_t>::max());
+    qlut[0] = lut_00.data();
-    for(auto instr:instr_descr){
+    qlut[1] = lut_01.data();
-        root->instrs.push_back(instr);
+    qlut[2] = lut_10.data();
+    qlut[3] = lut_11.data();
+    for (auto instr : instr_descr) {
+        auto quantrant = instr.value & 0x3;
+        expand_bit_mask(29, lutmasks[quantrant], instr.value >> 2, instr.mask >> 2, 0, qlut[quantrant], instr.op);
     }
-    populate_decoding_tree(root);
 }
 
 template <typename ARCH>
@@ -3172,19 +3163,30 @@ std::tuple<continuation_e>
 vm_impl<ARCH>::gen_single_inst_behavior(virt_addr_t &pc, unsigned int &inst_cnt, tu_builder& tu) {
     // we fetch at max 4 byte, alignment is 2
     enum {TRAP_ID=1<<16};
-    code_word_t instr = 0;
+    code_word_t insn = 0;
+    // const typename traits::addr_t upper_bits = ~traits::PGMASK;
     phys_addr_t paddr(pc);
+    auto *const data = (uint8_t *)&insn;
     paddr = this->core.v2p(pc);
-    auto res = this->core.read(paddr, 4, reinterpret_cast<uint8_t*>(&instr));
+//    if ((pc.val & upper_bits) != ((pc.val + 2) & upper_bits)) { // we may cross a page boundary
-    if (res != iss::Ok) throw trap_access(TRAP_ID, pc.val);
+//        auto res = this->core.read(paddr, 2, data);
-    if (instr == 0x0000006f || (instr&0xffff)==0xa001) throw simulation_stopped(0); // 'J 0' or 'C.J 0'
+//        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 (insn == 0x0000006f || (insn&0xffff)==0xa001) throw simulation_stopped(0); // 'J 0' or 'C.J 0'
     // curr pc on stack
     ++inst_cnt;
-    auto f = decode_instr(root, instr);
+    auto lut_val = extract_fields(insn);
+    auto f = qlut[insn & 0x3][lut_val];
     if (f == nullptr) {
         f = &this_class::illegal_intruction;
     }
-    return (this->*f)(pc, instr, tu);
+    return (this->*f)(pc, insn, tu);
 }
 
 template <typename ARCH> void vm_impl<ARCH>::gen_raise_trap(tu_builder& tu, uint16_t trap_id, uint16_t cause) {
@@ -3208,7 +3210,7 @@ template <typename ARCH> void vm_impl<ARCH>::gen_trap_behavior(tu_builder& tu) {
     tu("return *next_pc;");
 }
 
-} // namespace tgc_c
+} // namespace mnrv32
 
 template <>
 std::unique_ptr<vm_if> create<arch::tgc_c>(arch::tgc_c *core, unsigned short port, bool dump) {
@@ -3216,7 +3218,7 @@ std::unique_ptr<vm_if> create<arch::tgc_c>(arch::tgc_c *core, unsigned short por
     if (port != 0) debugger::server<debugger::gdb_session>::run_server(ret, port);
     return std::unique_ptr<vm_if>(ret);
 }
-} // namespace tcc
+} // namesapce tcc
 } // namespace iss
 
 #include <iss/factory.h>
@@ -3224,19 +3226,24 @@ std::unique_ptr<vm_if> create<arch::tgc_c>(arch::tgc_c *core, unsigned short por
 #include <iss/arch/riscv_hart_mu_p.h>
 namespace iss {
 namespace {
-volatile std::array<bool, 2> dummy = {
+std::array<bool, 2> dummy = {
         core_factory::instance().register_creator("tgc_c|m_p|tcc", [](unsigned port, void*) -> std::tuple<cpu_ptr, vm_ptr>{
             auto* cpu = new iss::arch::riscv_hart_m_p<iss::arch::tgc_c>();
-            auto* vm = new tcc::tgc_c::vm_impl<arch::tgc_c>(*cpu, false);
+		    auto vm = new tcc::tgc_c::vm_impl<arch::tgc_c>(*cpu, false);
 		    if (port != 0) debugger::server<debugger::gdb_session>::run_server(vm, port);
             return {cpu_ptr{cpu}, vm_ptr{vm}};
         }),
         core_factory::instance().register_creator("tgc_c|mu_p|tcc", [](unsigned port, void*) -> std::tuple<cpu_ptr, vm_ptr>{
             auto* cpu = new iss::arch::riscv_hart_mu_p<iss::arch::tgc_c>();
-            auto* vm = new tcc::tgc_c::vm_impl<arch::tgc_c>(*cpu, false);
+		    auto vm = new tcc::tgc_c::vm_impl<arch::tgc_c>(*cpu, false);
 		    if (port != 0) debugger::server<debugger::gdb_session>::run_server(vm, port);
             return {cpu_ptr{cpu}, vm_ptr{vm}};
         })
 };
 }
 }
+extern "C" {
+	bool* get_tgc_c_tcc_creators() {
+		return iss::dummy.data();
+	}
+}