changes the io_buf

CMakeLists.txt

@@ -109,16 +109,6 @@ if(TARGET yaml-cpp::yaml-cpp)
     target_link_libraries(${PROJECT_NAME} PUBLIC yaml-cpp::yaml-cpp)
 endif()
 
-if(WITH_LLVM)
-    find_package(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
@@ -262,3 +252,9 @@ if(TARGET scc-sysc)
         INCLUDES DESTINATION ${CMAKE_INSTALL_INCLUDEDIR} # headers
     )
 endif()
+
+project(elfio-test)
+find_package(Boost COMPONENTS program_options thread REQUIRED)
+
+add_executable(${PROJECT_NAME} src/elfio.cpp)
+target_link_libraries(${PROJECT_NAME} PUBLIC elfio::elfio)

gen_input/templates/CORENAME.h.gtl

@@ -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 {<%

gen_input/templates/CORENAME_sysc.cpp.gtl

@@ -45,17 +45,17 @@ namespace interp {
 using namespace sysc;
 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*>(data);
+            auto* cc = reinterpret_cast<sysc::tgfs::core_complex_if*>(data);
             auto* cpu = new sc_core_adapter<arch::riscv_hart_mu_p<arch::${coreDef.name.toLowerCase()}, (iss::arch::features_e)(iss::arch::FEAT_PMP | iss::arch::FEAT_EXT_N | iss::arch::FEAT_CLIC)>>(cc);
             return {sysc::sc_cpu_ptr{cpu}, vm_ptr{create(static_cast<arch::${coreDef.name.toLowerCase()}*>(cpu), gdb_port)}};
         })<%}%>
@@ -66,17 +66,17 @@ 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*>(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("${coreDef.name.toLowerCase()}|mu_p|llvm", [](unsigned gdb_port, void* data) -> iss_factory::base_t {
-            auto* cc = reinterpret_cast<sysc::tgfs::core_complex*>(data);
+            auto* cc = reinterpret_cast<sysc::tgfs::core_complex_if*>(data);
             auto* cpu = new sc_core_adapter<arch::riscv_hart_mu_p<arch::${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|llvm", [](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::${coreDef.name.toLowerCase()}, (iss::arch::features_e)(iss::arch::FEAT_PMP | iss::arch::FEAT_EXT_N | iss::arch::FEAT_CLIC)>>(cc);
             return {sysc::sc_cpu_ptr{cpu}, vm_ptr{create(static_cast<arch::${coreDef.name.toLowerCase()}*>(cpu), gdb_port)}};
         })<%}%>
@@ -88,17 +88,17 @@ namespace tcc {
 using namespace sysc;
 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*>(data);
+            auto* cc = reinterpret_cast<sysc::tgfs::core_complex_if*>(data);
             auto* cpu = new sc_core_adapter<arch::riscv_hart_mu_p<arch::${coreDef.name.toLowerCase()}, (iss::arch::features_e)(iss::arch::FEAT_PMP | iss::arch::FEAT_EXT_N | iss::arch::FEAT_CLIC)>>(cc);
             return {sysc::sc_cpu_ptr{cpu}, vm_ptr{create(static_cast<arch::${coreDef.name.toLowerCase()}*>(cpu), gdb_port)}};
         })<%}%>
@@ -110,17 +110,17 @@ 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*>(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("${coreDef.name.toLowerCase()}|mu_p|asmjit", [](unsigned gdb_port, void* data) -> iss_factory::base_t {
-            auto* cc = reinterpret_cast<sysc::tgfs::core_complex*>(data);
+            auto* cc = reinterpret_cast<sysc::tgfs::core_complex_if*>(data);
             auto* cpu = new sc_core_adapter<arch::riscv_hart_mu_p<arch::${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|asmjit", [](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::${coreDef.name.toLowerCase()}, (iss::arch::features_e)(iss::arch::FEAT_PMP | iss::arch::FEAT_EXT_N | iss::arch::FEAT_CLIC)>>(cc);
             return {sysc::sc_cpu_ptr{cpu}, vm_ptr{create(static_cast<arch::${coreDef.name.toLowerCase()}*>(cpu), gdb_port)}};
         })<%}%>

gen_input/templates/asmjit/CORENAME.cpp.gtl

@@ -96,7 +96,7 @@ protected:
     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;
+    continuation_e gen_single_inst_behavior(virt_addr_t&, 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;
@@ -221,7 +221,7 @@ vm_impl<ARCH>::vm_impl(ARCH &core, unsigned core_id, unsigned cluster_id)
     }()) {}
 
 template <typename ARCH>
-continuation_e vm_impl<ARCH>::gen_single_inst_behavior(virt_addr_t &pc, unsigned int &inst_cnt, jit_holder& jh) {
+continuation_e vm_impl<ARCH>::gen_single_inst_behavior(virt_addr_t &pc, jit_holder& jh) {
     enum {TRAP_ID=1<<16};
     code_word_t instr = 0;
     phys_addr_t paddr(pc);
@@ -233,7 +233,6 @@ continuation_e vm_impl<ARCH>::gen_single_inst_behavior(virt_addr_t &pc, unsigned
         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())
@@ -263,6 +262,7 @@ void vm_impl<ARCH>::gen_instr_epilogue(jit_holder& jh) {
     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){
@@ -308,6 +308,7 @@ inline void vm_impl<ARCH>::gen_raise(jit_holder& jh, uint16_t trap_id, uint16_t
     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>

gen_input/templates/interp/CORENAME.cpp.gtl

@@ -199,9 +199,6 @@ template <typename CODE_WORD> void debug_fn(CODE_WORD insn) {
     volatile CODE_WORD x = insn;
     insn = 2 * x;
 }
-
-template <typename ARCH> vm_impl<ARCH>::vm_impl() { this(new ARCH()); }
-
 // according to
 // https://stackoverflow.com/questions/8871204/count-number-of-1s-in-binary-representation
 #ifdef __GCC__

gen_input/templates/llvm/CORENAME.cpp.gtl

@@ -101,7 +101,7 @@ protected:
         return super::gen_cond_assign(cond, this->gen_ext(trueVal, size), this->gen_ext(falseVal, size));
     }
 
-    std::tuple<continuation_e, BasicBlock *> gen_single_inst_behavior(virt_addr_t &, unsigned int &, BasicBlock *) override;
+    std::tuple<continuation_e, BasicBlock *> gen_single_inst_behavior(virt_addr_t &, BasicBlock *) override;
 
     void gen_leave_behavior(BasicBlock *leave_blk) override;
     void gen_raise_trap(uint16_t trap_id, uint16_t cause);
@@ -244,7 +244,7 @@ vm_impl<ARCH>::vm_impl(ARCH &core, unsigned core_id, unsigned cluster_id)
 
 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) {
+vm_impl<ARCH>::gen_single_inst_behavior(virt_addr_t &pc, BasicBlock *this_block) {
     // we fetch at max 4 byte, alignment is 2
     enum {TRAP_ID=1<<16};
     code_word_t instr = 0;
@@ -256,9 +256,10 @@ vm_impl<ARCH>::gen_single_inst_behavior(virt_addr_t &pc, unsigned int &inst_cnt,
     auto res = this->core.read(paddr, 4, data);
     if (res != iss::Ok) 
         return std::make_tuple(ILLEGAL_FETCH, nullptr);
-    if (instr == 0x0000006f || (instr&0xffff)==0xa001)
+    if (instr == 0x0000006f || (instr&0xffff)==0xa001){
+        this->builder.CreateBr(this->leave_blk);
         return std::make_tuple(JUMP_TO_SELF, nullptr);
-    ++inst_cnt;
+        }
     uint32_t inst_index = instr_decoder.decode_instr(instr);
     compile_func f = nullptr;
     if(inst_index < instr_descr.size())
@@ -279,6 +280,7 @@ 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::TRAP_STATE), true);
+    this->builder.CreateBr(this->trap_blk);
 }
 
 template <typename ARCH>
@@ -339,6 +341,10 @@ void vm_impl<ARCH>::gen_instr_epilogue(BasicBlock *bb) {
     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);
+    //increment cyclecount
+    auto* cycle_val = this->builder.CreateAdd(
+        this->builder.CreateLoad(this->get_typeptr(arch::traits<ARCH>::CYCLE), get_reg_ptr(arch::traits<ARCH>::CYCLE)), this->gen_const(64U, 1));
+    this->builder.CreateStore(cycle_val, get_reg_ptr(arch::traits<ARCH>::CYCLE), false);
 }
 
 } // namespace ${coreDef.name.toLowerCase()}

gen_input/templates/tcc/CORENAME.cpp.gtl

@@ -83,21 +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 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;
 
@@ -176,6 +176,7 @@ private:
         auto cur_pc_val = tu.constant(pc.val, traits::reg_bit_widths[traits::PC]);
         pc=pc+ ${instr.length/8};
         gen_set_pc(tu, pc, traits::NEXT_PC);
+        tu("(*cycle)++;");
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         <%instr.behavior.eachLine{%>${it}
@@ -225,8 +226,8 @@ vm_impl<ARCH>::vm_impl(ARCH &core, unsigned core_id, unsigned cluster_id)
     }()) {}
 
 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;
@@ -238,7 +239,6 @@ vm_impl<ARCH>::gen_single_inst_behavior(virt_addr_t &pc, unsigned int &inst_cnt,
         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())
@@ -273,10 +273,12 @@ template <typename ARCH> void vm_impl<ARCH>::gen_trap_behavior(tu_builder& tu) {
     tu.store(traits::LAST_BRANCH, tu.constant(static_cast<int>(UNKNOWN_JUMP),32));
     tu("return *next_pc;");
 }
-<%
-if(fcsr != null) {%>
 template <typename ARCH> void vm_impl<ARCH>::add_prologue(tu_builder& tu){
     std::ostringstream os;
+    os << tu.add_reg_ptr("trap_state", arch::traits<ARCH>::TRAP_STATE, this->regs_base_ptr);
+    os << tu.add_reg_ptr("pending_trap", arch::traits<ARCH>::PENDING_TRAP, this->regs_base_ptr);
+    os << tu.add_reg_ptr("cycle", arch::traits<ARCH>::CYCLE, this->regs_base_ptr);
+<%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";
@@ -303,9 +305,9 @@ template <typename ARCH> void vm_impl<ARCH>::add_prologue(tu_builder& tu){
     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()}
 

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/riscv_hart_common.h

@@ -35,11 +35,15 @@
 #ifndef _RISCV_HART_COMMON
 #define _RISCV_HART_COMMON
 
+#include "iss/vm_types.h"
+#include <array>
 #include <cstdint>
 #include <elfio/elfio.hpp>
 #include <fmt/format.h>
 #include <iss/arch_if.h>
 #include <iss/log_categories.h>
+#include <limits>
+#include <sstream>
 #include <string>
 #include <unordered_map>
 #include <util/logging.h>
@@ -55,8 +59,6 @@
 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 riscv_csr {
@@ -312,33 +314,43 @@ inline void write_reg_uint32(uint64_t offs, uint32_t& reg, const uint8_t* const
 }
 struct riscv_hart_common {
     riscv_hart_common(){};
-    ~riscv_hart_common(){};
+    ~riscv_hart_common() {
+        if(io_buf.str().length()) {
+            CPPLOG(INFO) << "tohost send '" << io_buf.str() << "'";
+        }
+    };
     std::unordered_map<std::string, uint64_t> symbol_table;
+    uint64_t entry_address{0};
+    uint64_t tohost = std::numeric_limits<uint64_t>::max();
+    uint64_t fromhost = std::numeric_limits<uint64_t>::max();
+    std::stringstream io_buf;
 
-    std::unordered_map<std::string, uint64_t> get_sym_table(std::string name) {
+    bool read_elf_file(std::string name, uint8_t expected_elf_class,
-        if(!symbol_table.empty())
+                       std::function<iss::status(uint64_t, uint64_t, const uint8_t* const)> cb) {
-            return symbol_table;
-        FILE* fp = fopen(name.c_str(), "r");
-        if(fp) {
-            std::array<char, 5> buf;
-            auto n = fread(buf.data(), 1, 4, fp);
-            fclose(fp);
-            if(n != 4)
-                throw std::runtime_error("input file has insufficient size");
-            buf[4] = 0;
-            if(strcmp(buf.data() + 1, "ELF") == 0) {
         // Create elfio reader
         ELFIO::elfio reader;
         // Load ELF data
-                if(!reader.load(name))
+        if(reader.load(name)) {
-                    throw std::runtime_error("could not process elf file");
             // check elf properties
-                if(reader.get_type() != ET_EXEC)
+            if(reader.get_class() != expected_elf_class)
-                    throw std::runtime_error("wrong elf type in file");
+                return false;
-                if(reader.get_machine() != EM_RISCV)
+            if(reader.get_type() != ELFIO::ET_EXEC)
-                    throw std::runtime_error("wrong elf machine in file");
+                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(SHT_SYMTAB == sym_sec->get_type() || SHT_DYNSYM == sym_sec->get_type()) {
+            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;
@@ -357,13 +369,46 @@ struct riscv_hart_common {
 #endif
                     }
                 }
+                try {
+                    tohost = symbol_table.at("tohost");
+                } catch(std::out_of_range& e) {
                 }
-                return symbol_table;
+                try {
+                    fromhost = symbol_table.at("fromhost");
+                } catch(std::out_of_range& e) {
                 }
-            throw std::runtime_error(fmt::format("memory load file {} is not a valid elf file", name));
+            }
-        } else
+            return true;
-            throw std::runtime_error(fmt::format("memory load file not found, check if {} is a valid file", name));
+        }
+        return false;
     };
+    iss::status execute_sys_write(arch_if* aif, const std::array<uint64_t, 8>& loaded_payload, unsigned mem_type) {
+        uint64_t fd = loaded_payload[1];
+        uint64_t buf_ptr = loaded_payload[2];
+        uint64_t len = loaded_payload[3];
+        std::vector<char> buf(len);
+        if(aif->read(address_type::PHYSICAL, access_type::DEBUG_READ, mem_type, buf_ptr, len, reinterpret_cast<uint8_t*>(buf.data()))) {
+            CPPLOG(ERR) << "SYS_WRITE buffer read went wrong";
+            return iss::Err;
+        }
+        // we disregard the fd and just log to stdout
+        for(size_t i = 0; i < len; i++) {
+            if(buf[i] == '\n' || buf[i] == '\0') {
+                CPPLOG(INFO) << "tohost send '" << io_buf.str() << "'";
+                io_buf.str("");
+            } else
+                io_buf << buf[i];
+        }
+
+        // Not sure what the correct return value should be
+        uint8_t ret_val = 1;
+        if(fromhost != std::numeric_limits<uint64_t>::max())
+            if(aif->write(address_type::PHYSICAL, access_type::DEBUG_WRITE, mem_type, fromhost, 1, &ret_val)) {
+                CPPLOG(ERR) << "Fromhost write went wrong";
+                return iss::Err;
+            }
+        return iss::Ok;
+    }
 };
 
 } // namespace arch

src/iss/arch/riscv_hart_m_p.h

@@ -41,6 +41,11 @@
 #include "iss/vm_if.h"
 #include "iss/vm_types.h"
 #include "riscv_hart_common.h"
+#include "util/logging.h"
+#include <algorithm>
+#include <cstdint>
+#include <elfio/elf_types.hpp>
+#include <limits>
 #include <stdexcept>
 #ifndef FMT_HEADER_ONLY
 #define FMT_HEADER_ONLY
@@ -321,7 +326,7 @@ protected:
 
         unsigned get_reg_size(unsigned num) override { return traits<BASE>::reg_bit_widths[num]; }
 
-        std::unordered_map<std::string, uint64_t> get_symbol_table(std::string name) override { return arch.get_sym_table(name); }
+        std::unordered_map<std::string, uint64_t> const& get_symbol_table(std::string name) override { return arch.symbol_table; }
 
         riscv_hart_m_p<BASE, FEAT, LOGCAT>& arch;
     };
@@ -343,9 +348,6 @@ protected:
     int64_t instret_offset{0};
     uint64_t minstret_csr{0};
     reg_t fault_data;
-    uint64_t tohost = tohost_dflt;
-    uint64_t fromhost = fromhost_dflt;
-    bool tohost_lower_written = false;
     riscv_instrumentation_if instr_if;
 
     semihosting_cb_t<reg_t> semihosting_cb;
@@ -355,7 +357,6 @@ protected:
     using csr_page_type = typename csr_type::page_type;
     mem_type mem;
     csr_type csr;
-    std::stringstream uart_buf;
     std::unordered_map<reg_t, uint64_t> ptw;
     std::unordered_map<uint64_t, uint8_t> atomic_reservation;
     std::unordered_map<unsigned, rd_csr_f> csr_rd_cb;
@@ -447,7 +448,6 @@ riscv_hart_m_p<BASE, FEAT, LOGCAT>::riscv_hart_m_p(feature_config cfg)
     csr[marchid] = traits<BASE>::MARCHID_VAL;
     csr[mimpid] = 1;
 
-    uart_buf.str("");
     if(traits<BASE>::FLEN > 0) {
         csr_rd_cb[fcsr] = &this_class::read_fcsr;
         csr_wr_cb[fcsr] = &this_class::write_fcsr;
@@ -573,57 +573,14 @@ riscv_hart_m_p<BASE, FEAT, LOGCAT>::riscv_hart_m_p(feature_config cfg)
 
 template <typename BASE, features_e FEAT, typename LOGCAT>
 std::pair<uint64_t, bool> riscv_hart_m_p<BASE, FEAT, LOGCAT>::load_file(std::string name, int type) {
-    get_sym_table(name);
+    if(read_elf_file(name, sizeof(reg_t) == 4 ? ELFIO::ELFCLASS32 : ELFIO::ELFCLASS64,
-    try {
+                     [this](uint64_t addr, uint64_t size, const uint8_t* const data) -> iss::status {
-        tohost = symbol_table.at("tohost");
+                         return this->write(iss::address_type::PHYSICAL, iss::access_type::DEBUG_WRITE, traits<BASE>::MEM, addr, size,
-        fromhost = symbol_table.at("fromhost");
+                                            data);
-    } catch(std::out_of_range& e) {
+                     })) {
+        return std::make_pair(entry_address, true);
     }
-    FILE* fp = fopen(name.c_str(), "r");
+    return std::make_pair(entry_address, false);
-    if(fp) {
-        std::array<char, 5> buf;
-        auto n = fread(buf.data(), 1, 4, fp);
-        fclose(fp);
-        if(n != 4)
-            throw std::runtime_error("input file has insufficient size");
-        buf[4] = 0;
-        if(strcmp(buf.data() + 1, "ELF") == 0) {
-            // Create elfio reader
-            ELFIO::elfio reader;
-            // Load ELF data
-            if(!reader.load(name))
-                throw std::runtime_error("could not process elf file");
-            // check elf properties
-            if(reader.get_class() != ELFCLASS32)
-                if(sizeof(reg_t) == 4)
-                    throw std::runtime_error("wrong elf class in file");
-            if(reader.get_type() != ET_EXEC)
-                throw std::runtime_error("wrong elf type in file");
-            if(reader.get_machine() != EM_RISCV)
-                throw std::runtime_error("wrong elf machine in file");
-            auto entry = 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 = this->write(iss::address_type::PHYSICAL, iss::access_type::DEBUG_WRITE, traits<BASE>::MEM,
-                                           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();
-                }
-            }
-            for(const auto& sec : reader.sections) {
-                if(sec->get_name() == ".tohost") {
-                    tohost = sec->get_address();
-                    fromhost = tohost + 0x40;
-                }
-            }
-            return std::make_pair(entry, true);
-        }
-        throw std::runtime_error(fmt::format("memory load file {} is not a valid elf file", name));
-    }
-    throw std::runtime_error(fmt::format("memory load file not found, check if {} is a valid file", name));
 }
 
 template <typename BASE, features_e FEAT, typename LOGCAT>
@@ -654,7 +611,7 @@ iss::status riscv_hart_m_p<BASE, FEAT, LOGCAT>::read(const address_type type, co
     try {
         switch(space) {
         case traits<BASE>::MEM: {
-            auto alignment = is_fetch(access) ? (has_compressed() ? 2 : 4) : length;
+            auto alignment = is_fetch(access) ? (has_compressed() ? 2 : 4) : std::min<unsigned>(length, sizeof(reg_t));
             if(unlikely(is_fetch(access) && (addr & (alignment - 1)))) {
                 fault_data = addr;
                 if(is_debug(access))
@@ -750,7 +707,7 @@ iss::status riscv_hart_m_p<BASE, FEAT, LOGCAT>::write(const address_type type, c
                       << std::hex << addr;
         break;
     default:
-        CPPLOG(TRACE) << prefix << "write of " << length << " bytes @addr " << addr;
+        CPPLOG(TRACE) << prefix << "write of " << length << " bytes @addr 0x" << std::hex << addr;
     }
 #endif
     try {
@@ -764,7 +721,8 @@ iss::status riscv_hart_m_p<BASE, FEAT, LOGCAT>::write(const address_type type, c
                 return iss::Err;
             }
             try {
-                if(length > 1 && (addr & (length - 1)) && (access & access_type::DEBUG) != access_type::DEBUG) {
+                auto alignment = std::min<unsigned>(length, sizeof(reg_t));
+                if(length > 1 && (addr & (alignment - 1)) && !is_debug(access)) {
                     this->reg.trap_state = (1UL << 31) | 6 << 16;
                     fault_data = addr;
                     return iss::Err;
@@ -784,7 +742,7 @@ iss::status riscv_hart_m_p<BASE, FEAT, LOGCAT>::write(const address_type type, c
                 } else {
                     res = write_mem(phys_addr, length, data);
                 }
-                if(unlikely(res != iss::Ok && (access & access_type::DEBUG) == 0)) {
+                if(unlikely(res != iss::Ok && !is_debug(access))) {
                     this->reg.trap_state = (1UL << 31) | (7UL << 16); // issue trap 7 (Store/AMO access fault)
                     fault_data = addr;
                 }
@@ -794,38 +752,6 @@ iss::status riscv_hart_m_p<BASE, FEAT, LOGCAT>::write(const address_type type, c
                 fault_data = ta.addr;
                 return iss::Err;
             }
-
-            if((addr + length) > mem.size())
-                return iss::Err;
-            switch(addr) {
-            case 0x10013000: // UART0 base, TXFIFO reg
-            case 0x10023000: // UART1 base, TXFIFO reg
-                uart_buf << (char)data[0];
-                if(((char)data[0]) == '\n' || data[0] == 0) {
-                    std::cout << uart_buf.str();
-                    uart_buf.str("");
-                }
-                return iss::Ok;
-            case 0x10008000: { // HFROSC base, hfrosccfg reg
-                auto& p = mem(addr / mem.page_size);
-                auto offs = addr & mem.page_addr_mask;
-                std::copy(data, data + length, p.data() + offs);
-                auto& x = *(p.data() + offs + 3);
-                if(x & 0x40)
-                    x |= 0x80; // hfroscrdy = 1 if hfroscen==1
-                return iss::Ok;
-            }
-            case 0x10008008: { // HFROSC base, pllcfg reg
-                auto& p = mem(addr / mem.page_size);
-                auto offs = addr & mem.page_addr_mask;
-                std::copy(data, data + length, p.data() + offs);
-                auto& x = *(p.data() + offs + 3);
-                x |= 0x80; // set pll lock upon writing
-                return iss::Ok;
-            } break;
-            default: {
-            }
-            }
         } break;
         case traits<BASE>::CSR: {
             if(length != sizeof(reg_t))
@@ -1138,60 +1064,52 @@ iss::status riscv_hart_m_p<BASE, FEAT, LOGCAT>::read_mem(phys_addr_t paddr, unsi
 
 template <typename BASE, features_e FEAT, typename LOGCAT>
 iss::status riscv_hart_m_p<BASE, FEAT, LOGCAT>::write_mem(phys_addr_t paddr, unsigned length, const uint8_t* const data) {
-    switch(paddr.val) {
-    // TODO remove UART, Peripherals should not be part of the ISS
-    case 0xFFFF0000: // UART0 base, TXFIFO reg
-        if(((char)data[0]) == '\n' || data[0] == 0) {
-            CPPLOG(INFO) << "UART" << ((paddr.val >> 12) & 0x3) << " send '" << uart_buf.str() << "'";
-            uart_buf.str("");
-        } else if(((char)data[0]) != '\r')
-            uart_buf << (char)data[0];
-        break;
-    default: {
     mem_type::page_type& p = mem(paddr.val / mem.page_size);
     std::copy(data, data + length, p.data() + (paddr.val & mem.page_addr_mask));
     // tohost handling in case of riscv-test
+    // according to https://github.com/riscv-software-src/riscv-isa-sim/issues/364#issuecomment-607657754:
     if(paddr.access && iss::access_type::FUNC) {
-            auto tohost_upper =
+        if(paddr.val == tohost) {
-                (traits<BASE>::XLEN == 32 && paddr.val == (tohost + 4)) || (traits<BASE>::XLEN == 64 && paddr.val == tohost);
+            reg_t cur_data = *reinterpret_cast<const reg_t*>(data);
-            auto tohost_lower = (traits<BASE>::XLEN == 32 && paddr.val == tohost) || (traits<BASE>::XLEN == 64 && paddr.val == tohost);
+            // Extract Device (bits 63:56)
-            if(tohost_lower || tohost_upper) {
+            uint8_t device = traits<BASE>::XLEN == 32 ? 0 : (cur_data >> 56) & 0xFF;
-                uint64_t hostvar = *reinterpret_cast<uint64_t*>(p.data() + (tohost & mem.page_addr_mask));
+            // Extract Command (bits 55:48)
-                // in case of 32 bit system, two writes to tohost are needed, only evaluate on the second (high) write
+            uint8_t command = traits<BASE>::XLEN == 32 ? 0 : (cur_data >> 48) & 0xFF;
-                if(tohost_upper && (tohost_lower || tohost_lower_written)) {
+            // Extract payload (bits 47:0)
-                    switch(hostvar >> 48) {
+            uint64_t payload_addr = cur_data & 0xFFFFFFFFFFFFULL;
-                    case 0:
+            if(payload_addr & 1) {
-                        if(hostvar != 0x1) {
+                CPPLOG(FATAL) << "tohost value is 0x" << std::hex << payload_addr << std::dec << " (" << payload_addr
-                            CPPLOG(FATAL) << "tohost value is 0x" << std::hex << hostvar << std::dec << " (" << hostvar
                               << "), stopping simulation";
-                        } else {
-                            CPPLOG(INFO) << "tohost value is 0x" << std::hex << hostvar << std::dec << " (" << hostvar
-                                         << "), stopping simulation";
-                        }
                 this->reg.trap_state = std::numeric_limits<uint32_t>::max();
-                        this->interrupt_sim = hostvar;
+                this->interrupt_sim = payload_addr;
-                        break;
+                return iss::Ok;
-                    case 0x0101: {
+            } else if(device == 0 && command == 0) {
-                        char c = static_cast<char>(hostvar & 0xff);
+                std::array<uint64_t, 8> loaded_payload;
-                        if(c == '\n' || c == 0) {
+                if(read(address_type::PHYSICAL, access_type::DEBUG_READ, traits<BASE>::MEM, payload_addr, 8 * sizeof(uint64_t),
-                            CPPLOG(INFO) << "tohost send '" << uart_buf.str() << "'";
+                        reinterpret_cast<uint8_t*>(loaded_payload.data())) == iss::Err)
-                            uart_buf.str("");
+                    CPPLOG(ERR) << "Syscall read went wrong";
-                        } else
+                uint64_t syscall_num = loaded_payload.at(0);
-                            uart_buf << c;
+                if(syscall_num == 64) { // SYS_WRITE
-                    } break;
+                    return execute_sys_write(this, loaded_payload, traits<BASE>::MEM);
-                    default:
+                } else {
-                        break;
+                    CPPLOG(ERR) << "tohost syscall with number 0x" << std::hex << syscall_num << std::dec << " (" << syscall_num
+                                << ") not implemented";
+                    this->reg.trap_state = std::numeric_limits<uint32_t>::max();
+                    this->interrupt_sim = payload_addr;
+                    return iss::Ok;
                 }
-                    tohost_lower_written = false;
+            } else {
-                } else if(tohost_lower)
+                CPPLOG(ERR) << "tohost functionality not implemented for device " << device << " and command " << command;
-                    tohost_lower_written = true;
+                this->reg.trap_state = std::numeric_limits<uint32_t>::max();
-            } else if((traits<BASE>::XLEN == 32 && paddr.val == fromhost + 4) || (traits<BASE>::XLEN == 64 && paddr.val == fromhost)) {
+                this->interrupt_sim = payload_addr;
+                return iss::Ok;
+            }
+        }
+        if((traits<BASE>::XLEN == 32 && paddr.val == fromhost + 4) || (traits<BASE>::XLEN == 64 && paddr.val == fromhost)) {
             uint64_t fhostvar = *reinterpret_cast<uint64_t*>(p.data() + (fromhost & mem.page_addr_mask));
             *reinterpret_cast<uint64_t*>(p.data() + (tohost & mem.page_addr_mask)) = fhostvar;
         }
     }
-    }
-    }
     return iss::Ok;
 }
 

src/iss/arch/riscv_hart_msu_vp.h

@@ -41,6 +41,11 @@
 #include "iss/vm_if.h"
 #include "iss/vm_types.h"
 #include "riscv_hart_common.h"
+#include "util/logging.h"
+#include <algorithm>
+#include <cstdint>
+#include <elfio/elf_types.hpp>
+#include <limits>
 #include <stdexcept>
 #ifndef FMT_HEADER_ONLY
 #define FMT_HEADER_ONLY
@@ -371,7 +376,7 @@ protected:
 
         unsigned get_reg_size(unsigned num) override { return traits<BASE>::reg_bit_widths[num]; }
 
-        std::unordered_map<std::string, uint64_t> get_symbol_table(std::string name) override { return arch.get_sym_table(name); }
+        std::unordered_map<std::string, uint64_t> const& get_symbol_table(std::string name) override { return arch.symbol_table; }
 
         riscv_hart_msu_vp<BASE>& arch;
     };
@@ -393,9 +398,6 @@ protected:
     uint64_t minstret_csr{0};
     reg_t fault_data;
     std::array<vm_info, 2> vm;
-    uint64_t tohost = tohost_dflt;
-    uint64_t fromhost = fromhost_dflt;
-    bool tohost_lower_written = false;
     riscv_instrumentation_if instr_if;
 
     std::function<void(arch_if*, reg_t, reg_t)> semihosting_cb;
@@ -406,7 +408,6 @@ protected:
     mem_type mem;
     csr_type csr;
     void update_vm_info();
-    std::stringstream uart_buf;
     std::unordered_map<reg_t, uint64_t> ptw;
     std::unordered_map<uint64_t, uint8_t> atomic_reservation;
     std::unordered_map<unsigned, rd_csr_f> csr_rd_cb;
@@ -461,7 +462,6 @@ riscv_hart_msu_vp<BASE>::riscv_hart_msu_vp()
     csr[marchid] = traits<BASE>::MARCHID_VAL;
     csr[mimpid] = 1;
 
-    uart_buf.str("");
     for(unsigned addr = mhpmcounter3; addr <= mhpmcounter31; ++addr) {
         csr_rd_cb[addr] = &this_class::read_null;
         csr_wr_cb[addr] = &this_class::write_csr_reg;
@@ -557,71 +557,14 @@ riscv_hart_msu_vp<BASE>::riscv_hart_msu_vp()
 }
 
 template <typename BASE> std::pair<uint64_t, bool> riscv_hart_msu_vp<BASE>::load_file(std::string name, int type) {
-    FILE* fp = fopen(name.c_str(), "r");
+    if(read_elf_file(name, sizeof(reg_t) == 4 ? ELFIO::ELFCLASS32 : ELFIO::ELFCLASS64,
-    if(fp) {
+                     [this](uint64_t addr, uint64_t size, const uint8_t* const data) -> iss::status {
-        std::array<char, 5> buf;
+                         return this->write(iss::address_type::PHYSICAL, iss::access_type::DEBUG_WRITE, traits<BASE>::MEM, addr, size,
-        auto n = fread(buf.data(), 1, 4, fp);
+                                            data);
-        fclose(fp);
+                     })) {
-        if(n != 4)
+        return std::make_pair(entry_address, true);
-            throw std::runtime_error("input file has insufficient size");
-        buf[4] = 0;
-        if(strcmp(buf.data() + 1, "ELF") == 0) {
-            // Create elfio reader
-            ELFIO::elfio reader;
-            // Load ELF data
-            if(!reader.load(name))
-                throw std::runtime_error("could not process elf file");
-            // check elf properties
-            if(reader.get_class() != ELFCLASS32)
-                if(sizeof(reg_t) == 4)
-                    throw std::runtime_error("wrong elf class in file");
-            if(reader.get_type() != ET_EXEC)
-                throw std::runtime_error("wrong elf type in file");
-            if(reader.get_machine() != EM_RISCV)
-                throw std::runtime_error("wrong elf machine in file");
-            auto entry = 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 = this->write(iss::address_type::PHYSICAL, iss::access_type::DEBUG_WRITE, traits<BASE>::MEM,
-                                           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();
     }
-            }
+    return std::make_pair(entry_address, false);
-            for(const auto sec : reader.sections) {
-                if(sec->get_name() == ".symtab") {
-                    if(SHT_SYMTAB == sec->get_type() || SHT_DYNSYM == sec->get_type()) {
-                        ELFIO::symbol_section_accessor symbols(reader, 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 == "tohost") {
-                                tohost = value;
-                            } else if(name == "fromhost") {
-                                fromhost = value;
-                            }
-                        }
-                    }
-                } else if(sec->get_name() == ".tohost") {
-                    tohost = sec->get_address();
-                    fromhost = tohost + 0x40;
-                }
-            }
-            return std::make_pair(entry, true);
-        }
-        throw std::runtime_error(fmt::format("memory load file {} is not a valid elf file", name));
-    }
-    throw std::runtime_error(fmt::format("memory load file not found, check if {} is a valid file", name));
 }
 
 template <typename BASE>
@@ -639,7 +582,7 @@ iss::status riscv_hart_msu_vp<BASE>::read(const address_type type, const access_
     try {
         switch(space) {
         case traits<BASE>::MEM: {
-            auto alignment = is_fetch(access) ? (traits<BASE>::MISA_VAL & 0x100 ? 2 : 4) : length;
+            auto alignment = is_fetch(access) ? (has_compressed() ? 2 : 4) : std::min<unsigned>(length, sizeof(reg_t));
             if(unlikely(is_fetch(access) && (addr & (alignment - 1)))) {
                 fault_data = addr;
                 if(access && iss::access_type::DEBUG)
@@ -758,6 +701,7 @@ iss::status riscv_hart_msu_vp<BASE>::write(const address_type type, const access
             }
             phys_addr_t paddr = BASE::v2p(iss::addr_t{access, type, space, addr});
             try {
+                // TODO: There is no check for alignment
                 if(unlikely((addr & ~PGMASK) != ((addr + length - 1) & ~PGMASK))) { // we may cross a page boundary
                     vm_info vm = hart_state_type::decode_vm_info(this->reg.PRIV, state.satp);
                     if(vm.levels != 0) { // VM is active
@@ -780,40 +724,6 @@ iss::status riscv_hart_msu_vp<BASE>::write(const address_type type, const access
                 fault_data = ta.addr;
                 return iss::Err;
             }
-
-            if((paddr.val + length) > mem.size())
-                return iss::Err;
-            switch(paddr.val) {
-            case 0x10013000: // UART0 base, TXFIFO reg
-            case 0x10023000: // UART1 base, TXFIFO reg
-                uart_buf << (char)data[0];
-                if(((char)data[0]) == '\n' || data[0] == 0) {
-                    // CPPLOG(INFO)<<"UART"<<((paddr.val>>16)&0x3)<<" send
-                    // '"<<uart_buf.str()<<"'";
-                    std::cout << uart_buf.str();
-                    uart_buf.str("");
-                }
-                return iss::Ok;
-            case 0x10008000: { // HFROSC base, hfrosccfg reg
-                auto& p = mem(paddr.val / mem.page_size);
-                auto offs = paddr.val & mem.page_addr_mask;
-                std::copy(data, data + length, p.data() + offs);
-                auto& x = *(p.data() + offs + 3);
-                if(x & 0x40)
-                    x |= 0x80; // hfroscrdy = 1 if hfroscen==1
-                return iss::Ok;
-            }
-            case 0x10008008: { // HFROSC base, pllcfg reg
-                auto& p = mem(paddr.val / mem.page_size);
-                auto offs = paddr.val & mem.page_addr_mask;
-                std::copy(data, data + length, p.data() + offs);
-                auto& x = *(p.data() + offs + 3);
-                x |= 0x80; // set pll lock upon writing
-                return iss::Ok;
-            } break;
-            default: {
-            }
-            }
         } break;
         case traits<BASE>::CSR: {
             if(length != sizeof(reg_t))
@@ -1083,62 +993,52 @@ template <typename BASE> iss::status riscv_hart_msu_vp<BASE>::read_mem(phys_addr
 }
 
 template <typename BASE> iss::status riscv_hart_msu_vp<BASE>::write_mem(phys_addr_t paddr, unsigned length, const uint8_t* const data) {
-    switch(paddr.val) {
-    case 0xFFFF0000: // UART0 base, TXFIFO reg
-        if(((char)data[0]) == '\n' || data[0] == 0) {
-            CPPLOG(INFO) << "UART" << ((paddr.val >> 12) & 0x3) << " send '" << uart_buf.str() << "'";
-            uart_buf.str("");
-        } else if(((char)data[0]) != '\r')
-            uart_buf << (char)data[0];
-        break;
-    default: {
     mem_type::page_type& p = mem(paddr.val / mem.page_size);
     std::copy(data, data + length, p.data() + (paddr.val & mem.page_addr_mask));
     // tohost handling in case of riscv-test
+    // according to https://github.com/riscv-software-src/riscv-isa-sim/issues/364#issuecomment-607657754:
     if(paddr.access && iss::access_type::FUNC) {
-            auto tohost_upper =
+        if(paddr.val == tohost) {
-                (traits<BASE>::XLEN == 32 && paddr.val == (tohost + 4)) || (traits<BASE>::XLEN == 64 && paddr.val == tohost);
+            reg_t cur_data = *reinterpret_cast<const reg_t*>(data);
-            auto tohost_lower = (traits<BASE>::XLEN == 32 && paddr.val == tohost) || (traits<BASE>::XLEN == 64 && paddr.val == tohost);
+            // Extract Device (bits 63:56)
-            if(tohost_lower || tohost_upper) {
+            uint8_t device = traits<BASE>::XLEN == 32 ? 0 : (cur_data >> 56) & 0xFF;
-                uint64_t hostvar = *reinterpret_cast<uint64_t*>(p.data() + (tohost & mem.page_addr_mask));
+            // Extract Command (bits 55:48)
-                // in case of 32 bit system, two writes to tohost are needed, only evaluate on the second (high) write
+            uint8_t command = traits<BASE>::XLEN == 32 ? 0 : (cur_data >> 48) & 0xFF;
-                if(tohost_upper && (tohost_lower || tohost_lower_written)) {
+            // Extract payload (bits 47:0)
-                    switch(hostvar >> 48) {
+            uint64_t payload_addr = cur_data & 0xFFFFFFFFFFFFULL;
-                    case 0:
+            if(payload_addr & 1) {
-                        if(hostvar != 0x1) {
+                CPPLOG(FATAL) << "tohost value is 0x" << std::hex << payload_addr << std::dec << " (" << payload_addr
-                            CPPLOG(FATAL) << "tohost value is 0x" << std::hex << hostvar << std::dec << " (" << hostvar
                               << "), stopping simulation";
-                        } else {
-                            CPPLOG(INFO) << "tohost value is 0x" << std::hex << hostvar << std::dec << " (" << hostvar
-                                         << "), stopping simulation";
-                        }
                 this->reg.trap_state = std::numeric_limits<uint32_t>::max();
-                        this->interrupt_sim = hostvar;
+                this->interrupt_sim = payload_addr;
-#ifndef WITH_TCC
+                return iss::Ok;
-                        throw(iss::simulation_stopped(hostvar));
+            } else if(device == 0 && command == 0) {
-#endif
+                std::array<uint64_t, 8> loaded_payload;
-                        break;
+                if(read(address_type::PHYSICAL, access_type::DEBUG_READ, traits<BASE>::MEM, payload_addr, 8 * sizeof(uint64_t),
-                    case 0x0101: {
+                        reinterpret_cast<uint8_t*>(loaded_payload.data())) == iss::Err)
-                        char c = static_cast<char>(hostvar & 0xff);
+                    CPPLOG(ERR) << "Syscall read went wrong";
-                        if(c == '\n' || c == 0) {
+                uint64_t syscall_num = loaded_payload.at(0);
-                            CPPLOG(INFO) << "tohost send '" << uart_buf.str() << "'";
+                if(syscall_num == 64) { // SYS_WRITE
-                            uart_buf.str("");
+                    return execute_sys_write(this, loaded_payload, traits<BASE>::MEM);
-                        } else
+                } else {
-                            uart_buf << c;
+                    CPPLOG(ERR) << "tohost syscall with number 0x" << std::hex << syscall_num << std::dec << " (" << syscall_num
-                    } break;
+                                << ") not implemented";
-                    default:
+                    this->reg.trap_state = std::numeric_limits<uint32_t>::max();
-                        break;
+                    this->interrupt_sim = payload_addr;
+                    return iss::Ok;
                 }
-                    tohost_lower_written = false;
+            } else {
-                } else if(tohost_lower)
+                CPPLOG(ERR) << "tohost functionality not implemented for device " << device << " and command " << command;
-                    tohost_lower_written = true;
+                this->reg.trap_state = std::numeric_limits<uint32_t>::max();
-            } else if((traits<BASE>::XLEN == 32 && paddr.val == fromhost + 4) || (traits<BASE>::XLEN == 64 && paddr.val == fromhost)) {
+                this->interrupt_sim = payload_addr;
+                return iss::Ok;
+            }
+        }
+        if((traits<BASE>::XLEN == 32 && paddr.val == fromhost + 4) || (traits<BASE>::XLEN == 64 && paddr.val == fromhost)) {
             uint64_t fhostvar = *reinterpret_cast<uint64_t*>(p.data() + (fromhost & mem.page_addr_mask));
             *reinterpret_cast<uint64_t*>(p.data() + (tohost & mem.page_addr_mask)) = fhostvar;
         }
     }
-    }
-    }
     return iss::Ok;
 }
 

src/iss/arch/riscv_hart_mu_p.h

@@ -41,6 +41,11 @@
 #include "iss/vm_if.h"
 #include "iss/vm_types.h"
 #include "riscv_hart_common.h"
+#include "util/logging.h"
+#include <algorithm>
+#include <cstdint>
+#include <elfio/elf_types.hpp>
+#include <limits>
 #include <stdexcept>
 #ifndef FMT_HEADER_ONLY
 #define FMT_HEADER_ONLY
@@ -348,7 +353,7 @@ protected:
 
         unsigned get_reg_size(unsigned num) override { return traits<BASE>::reg_bit_widths[num]; }
 
-        std::unordered_map<std::string, uint64_t> get_symbol_table(std::string name) override { return arch.get_sym_table(name); }
+        std::unordered_map<std::string, uint64_t> const& get_symbol_table(std::string name) override { return arch.symbol_table; }
 
         riscv_hart_mu_p<BASE, FEAT, LOGCAT>& arch;
     };
@@ -370,9 +375,6 @@ protected:
     int64_t instret_offset{0};
     uint64_t minstret_csr{0};
     reg_t fault_data;
-    uint64_t tohost = tohost_dflt;
-    uint64_t fromhost = fromhost_dflt;
-    bool tohost_lower_written = false;
     riscv_instrumentation_if instr_if;
 
     semihosting_cb_t<reg_t> semihosting_cb;
@@ -382,7 +384,6 @@ protected:
     using csr_page_type = typename csr_type::page_type;
     mem_type mem;
     csr_type csr;
-    std::stringstream uart_buf;
     std::unordered_map<reg_t, uint64_t> ptw;
     std::unordered_map<uint64_t, uint8_t> atomic_reservation;
     std::unordered_map<unsigned, rd_csr_f> csr_rd_cb;
@@ -477,7 +478,6 @@ riscv_hart_mu_p<BASE, FEAT, LOGCAT>::riscv_hart_mu_p(feature_config cfg)
     csr[marchid] = traits<BASE>::MARCHID_VAL;
     csr[mimpid] = 1;
 
-    uart_buf.str("");
     if(traits<BASE>::FLEN > 0) {
         csr_rd_cb[fcsr] = &this_class::read_fcsr;
         csr_wr_cb[fcsr] = &this_class::write_fcsr;
@@ -651,71 +651,14 @@ riscv_hart_mu_p<BASE, FEAT, LOGCAT>::riscv_hart_mu_p(feature_config cfg)
 
 template <typename BASE, features_e FEAT, typename LOGCAT>
 std::pair<uint64_t, bool> riscv_hart_mu_p<BASE, FEAT, LOGCAT>::load_file(std::string name, int type) {
-    FILE* fp = fopen(name.c_str(), "r");
+    if(read_elf_file(name, sizeof(reg_t) == 4 ? ELFIO::ELFCLASS32 : ELFIO::ELFCLASS64,
-    if(fp) {
+                     [this](uint64_t addr, uint64_t size, const uint8_t* const data) -> iss::status {
-        std::array<char, 5> buf;
+                         return this->write(iss::address_type::PHYSICAL, iss::access_type::DEBUG_WRITE, traits<BASE>::MEM, addr, size,
-        auto n = fread(buf.data(), 1, 4, fp);
+                                            data);
-        fclose(fp);
+                     })) {
-        if(n != 4)
+        return std::make_pair(entry_address, true);
-            throw std::runtime_error("input file has insufficient size");
-        buf[4] = 0;
-        if(strcmp(buf.data() + 1, "ELF") == 0) {
-            // Create elfio reader
-            ELFIO::elfio reader;
-            // Load ELF data
-            if(!reader.load(name))
-                throw std::runtime_error("could not process elf file");
-            // check elf properties
-            if(reader.get_class() != ELFCLASS32)
-                if(sizeof(reg_t) == 4)
-                    throw std::runtime_error("wrong elf class in file");
-            if(reader.get_type() != ET_EXEC)
-                throw std::runtime_error("wrong elf type in file");
-            if(reader.get_machine() != EM_RISCV)
-                throw std::runtime_error("wrong elf machine in file");
-            auto entry = 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 = this->write(iss::address_type::PHYSICAL, iss::access_type::DEBUG_WRITE, traits<BASE>::MEM,
-                                           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();
     }
-            }
+    return std::make_pair(entry_address, false);
-            for(const auto& sec : reader.sections) {
-                if(sec->get_name() == ".symtab") {
-                    if(SHT_SYMTAB == sec->get_type() || SHT_DYNSYM == sec->get_type()) {
-                        ELFIO::symbol_section_accessor symbols(reader, 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 == "tohost") {
-                                tohost = value;
-                            } else if(name == "fromhost") {
-                                fromhost = value;
-                            }
-                        }
-                    }
-                } else if(sec->get_name() == ".tohost") {
-                    tohost = sec->get_address();
-                    fromhost = tohost + 0x40;
-                }
-            }
-            return std::make_pair(entry, true);
-        }
-        throw std::runtime_error(fmt::format("memory load file {} is not a valid elf file", name));
-    }
-    throw std::runtime_error(fmt::format("memory load file not found, check if {} is a valid file", name));
 }
 
 template <typename BASE, features_e FEAT, typename LOGCAT>
@@ -842,7 +785,7 @@ iss::status riscv_hart_mu_p<BASE, FEAT, LOGCAT>::read(const address_type type, c
                     return iss::Err;
                 }
             }
-            auto alignment = is_fetch(access) ? (has_compressed() ? 2 : 4) : length;
+            auto alignment = is_fetch(access) ? (has_compressed() ? 2 : 4) : std::min<unsigned>(length, sizeof(reg_t));
             if(unlikely(is_fetch(access) && (addr & (alignment - 1)))) {
                 fault_data = addr;
                 if(is_debug(access))
@@ -961,7 +904,8 @@ iss::status riscv_hart_mu_p<BASE, FEAT, LOGCAT>::write(const address_type type,
                 return iss::Err;
             }
             try {
-                if(length > 1 && (addr & (length - 1)) && (access & access_type::DEBUG) != access_type::DEBUG) {
+                auto alignment = std::min<unsigned>(length, sizeof(reg_t));
+                if(length > 1 && (addr & (alignment - 1)) && !is_debug(access)) {
                     this->reg.trap_state = (1UL << 31) | 6 << 16;
                     fault_data = addr;
                     return iss::Err;
@@ -991,38 +935,6 @@ iss::status riscv_hart_mu_p<BASE, FEAT, LOGCAT>::write(const address_type type,
                 fault_data = ta.addr;
                 return iss::Err;
             }
-
-            if((addr + length) > mem.size())
-                return iss::Err;
-            switch(addr) {
-            case 0x10013000: // UART0 base, TXFIFO reg
-            case 0x10023000: // UART1 base, TXFIFO reg
-                uart_buf << (char)data[0];
-                if(((char)data[0]) == '\n' || data[0] == 0) {
-                    std::cout << uart_buf.str();
-                    uart_buf.str("");
-                }
-                return iss::Ok;
-            case 0x10008000: { // HFROSC base, hfrosccfg reg
-                auto& p = mem(addr / mem.page_size);
-                auto offs = addr & mem.page_addr_mask;
-                std::copy(data, data + length, p.data() + offs);
-                auto& x = *(p.data() + offs + 3);
-                if(x & 0x40)
-                    x |= 0x80; // hfroscrdy = 1 if hfroscen==1
-                return iss::Ok;
-            }
-            case 0x10008008: { // HFROSC base, pllcfg reg
-                auto& p = mem(addr / mem.page_size);
-                auto offs = addr & mem.page_addr_mask;
-                std::copy(data, data + length, p.data() + offs);
-                auto& x = *(p.data() + offs + 3);
-                x |= 0x80; // set pll lock upon writing
-                return iss::Ok;
-            } break;
-            default: {
-            }
-            }
         } break;
         case traits<BASE>::CSR: {
             if(length != sizeof(reg_t))
@@ -1371,66 +1283,54 @@ iss::status riscv_hart_mu_p<BASE, FEAT, LOGCAT>::read_mem(phys_addr_t paddr, uns
     }
     return iss::Ok;
 }
-
 template <typename BASE, features_e FEAT, typename LOGCAT>
 iss::status riscv_hart_mu_p<BASE, FEAT, LOGCAT>::write_mem(phys_addr_t paddr, unsigned length, const uint8_t* const data) {
-    switch(paddr.val) {
-    // TODO remove UART, Peripherals should not be part of the ISS
-    case 0xFFFF0000: // UART0 base, TXFIFO reg
-        if(((char)data[0]) == '\n' || data[0] == 0) {
-            CPPLOG(INFO) << "UART" << ((paddr.val >> 12) & 0x3) << " send '" << uart_buf.str() << "'";
-            uart_buf.str("");
-        } else if(((char)data[0]) != '\r')
-            uart_buf << (char)data[0];
-        break;
-    default: {
     mem_type::page_type& p = mem(paddr.val / mem.page_size);
     std::copy(data, data + length, p.data() + (paddr.val & mem.page_addr_mask));
     // tohost handling in case of riscv-test
+    // according to https://github.com/riscv-software-src/riscv-isa-sim/issues/364#issuecomment-607657754:
     if(paddr.access && iss::access_type::FUNC) {
-            auto tohost_upper =
+        if(paddr.val == tohost) {
-                (traits<BASE>::XLEN == 32 && paddr.val == (tohost + 4)) || (traits<BASE>::XLEN == 64 && paddr.val == tohost);
+            reg_t cur_data = *reinterpret_cast<const reg_t*>(data);
-            auto tohost_lower = (traits<BASE>::XLEN == 32 && paddr.val == tohost) || (traits<BASE>::XLEN == 64 && paddr.val == tohost);
+            // Extract Device (bits 63:56)
-            if(tohost_lower || tohost_upper) {
+            uint8_t device = traits<BASE>::XLEN == 32 ? 0 : (cur_data >> 56) & 0xFF;
-                uint64_t hostvar = *reinterpret_cast<uint64_t*>(p.data() + (tohost & mem.page_addr_mask));
+            // Extract Command (bits 55:48)
-                // in case of 32 bit system, two writes to tohost are needed, only evaluate on the second (high) write
+            uint8_t command = traits<BASE>::XLEN == 32 ? 0 : (cur_data >> 48) & 0xFF;
-                if(tohost_upper && (tohost_lower || tohost_lower_written)) {
+            // Extract payload (bits 47:0)
-                    switch(hostvar >> 48) {
+            uint64_t payload_addr = cur_data & 0xFFFFFFFFFFFFULL;
-                    case 0:
+            if(payload_addr & 1) {
-                        if(hostvar != 0x1) {
+                CPPLOG(FATAL) << "tohost value is 0x" << std::hex << payload_addr << std::dec << " (" << payload_addr
-                            CPPLOG(FATAL) << "tohost value is 0x" << std::hex << hostvar << std::dec << " (" << hostvar
                               << "), stopping simulation";
-                        } else {
-                            CPPLOG(INFO) << "tohost value is 0x" << std::hex << hostvar << std::dec << " (" << hostvar
-                                         << "), stopping simulation";
-                        }
                 this->reg.trap_state = std::numeric_limits<uint32_t>::max();
-                        this->interrupt_sim = hostvar;
+                this->interrupt_sim = payload_addr;
-#ifndef WITH_TCC
+                return iss::Ok;
-                        throw(iss::simulation_stopped(hostvar));
+            } else if(device == 0 && command == 0) {
-#endif
+                std::array<uint64_t, 8> loaded_payload;
-                        break;
+                if(read(address_type::PHYSICAL, access_type::DEBUG_READ, traits<BASE>::MEM, payload_addr, 8 * sizeof(uint64_t),
-                    case 0x0101: {
+                        reinterpret_cast<uint8_t*>(loaded_payload.data())) == iss::Err)
-                        char c = static_cast<char>(hostvar & 0xff);
+                    CPPLOG(ERR) << "Syscall read went wrong";
-                        if(c == '\n' || c == 0) {
+                uint64_t syscall_num = loaded_payload.at(0);
-                            CPPLOG(INFO) << "tohost send '" << uart_buf.str() << "'";
+                if(syscall_num == 64) { // SYS_WRITE
-                            uart_buf.str("");
+                    return execute_sys_write(this, loaded_payload, traits<BASE>::MEM);
-                        } else
+                } else {
-                            uart_buf << c;
+                    CPPLOG(ERR) << "tohost syscall with number 0x" << std::hex << syscall_num << std::dec << " (" << syscall_num
-                    } break;
+                                << ") not implemented";
-                    default:
+                    this->reg.trap_state = std::numeric_limits<uint32_t>::max();
-                        break;
+                    this->interrupt_sim = payload_addr;
+                    return iss::Ok;
                 }
-                    tohost_lower_written = false;
+            } else {
-                } else if(tohost_lower)
+                CPPLOG(ERR) << "tohost functionality not implemented for device " << device << " and command " << command;
-                    tohost_lower_written = true;
+                this->reg.trap_state = std::numeric_limits<uint32_t>::max();
-            } else if((traits<BASE>::XLEN == 32 && paddr.val == fromhost + 4) || (traits<BASE>::XLEN == 64 && paddr.val == fromhost)) {
+                this->interrupt_sim = payload_addr;
+                return iss::Ok;
+            }
+        }
+        if((traits<BASE>::XLEN == 32 && paddr.val == fromhost + 4) || (traits<BASE>::XLEN == 64 && paddr.val == fromhost)) {
             uint64_t fhostvar = *reinterpret_cast<uint64_t*>(p.data() + (fromhost & mem.page_addr_mask));
             *reinterpret_cast<uint64_t*>(p.data() + (tohost & mem.page_addr_mask)) = fhostvar;
         }
     }
-    }
-    }
     return iss::Ok;
 }
 

src/main.cpp

@@ -206,13 +206,21 @@ int main(int argc, char* argv[]) {
         if(clim.count("elf"))
             for(std::string input : clim["elf"].as<std::vector<std::string>>()) {
                 auto start_addr = vm->get_arch()->load_file(input);
-                if(start_addr.second) // FIXME: this always evaluates to true as load file always returns <sth, true>
+                if(start_addr.second)
                     start_address = start_addr.first;
+                else {
+                    LOG(ERR) << "Error occured while loading file " << input << std::endl;
+                    return 1;
+                }
             }
         for(std::string input : args) {
             auto start_addr = vm->get_arch()->load_file(input); // treat remaining arguments as elf files
-            if(start_addr.second) // FIXME: this always evaluates to true as load file always returns <sth, true>
+            if(start_addr.second)
                 start_address = start_addr.first;
+            else {
+                LOG(ERR) << "Error occured while loading file " << input << std::endl;
+                return 1;
+            }
         }
         if(clim.count("reset")) {
             auto str = clim["reset"].as<std::string>();

src/sysc/register_tgc_c.cpp

@@ -62,12 +62,12 @@ using namespace sysc;
 volatile std::array<bool, 2> tgc_init = {
     iss_factory::instance().register_creator("tgc5c|m_p|llvm",
                                              [](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::tgc5c>>(cc);
                                                  return {sysc::sc_cpu_ptr{cpu}, vm_ptr{create(static_cast<arch::tgc5c*>(cpu), gdb_port)}};
                                              }),
     iss_factory::instance().register_creator("tgc5c|mu_p|llvm", [](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::tgc5c>>(cc);
         return {sysc::sc_cpu_ptr{cpu}, vm_ptr{create(static_cast<arch::tgc5c*>(cpu), gdb_port)}};
     })};

src/vm/asmjit/vm_tgc5c.cpp

@@ -94,7 +94,7 @@ protected:
     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;
+    continuation_e gen_single_inst_behavior(virt_addr_t&, 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;
@@ -1421,25 +1421,23 @@ private:
         }
         else{
             if(rd!=0){
-                {
+                auto label_then11 = cc.newLabel();
-                auto label_then = cc.newLabel();
+                auto label_merge11 = cc.newLabel();
-                auto label_merge = cc.newLabel();
+                auto tmp_reg11 = get_reg(cc, 8, false);
-                auto tmp_reg = get_reg_for(cc, 1);
                 cmp(cc, gen_ext(cc, 
                     load_reg_from_mem(jh, traits::X0 + rs1), 32, true), (int16_t)sext<12>(imm));
-                cc.jl(label_then);
+                cc.jl(label_then11);
-                mov(cc, tmp_reg,0);
+                mov(cc, tmp_reg11,0);
-                cc.jmp(label_merge);
+                cc.jmp(label_merge11);
-                cc.bind(label_then);
+                cc.bind(label_then11);
-                mov(cc, tmp_reg,1);
+                mov(cc, tmp_reg11, 1);
-                cc.bind(label_merge);
+                cc.bind(label_merge11);
                 mov(cc, get_ptr_for(jh, traits::X0+ rd),
-                      gen_ext(cc, tmp_reg
+                      gen_ext(cc, tmp_reg11
                       , 32, false)
                 );
             }
         }
-        }
         auto returnValue = CONT;
         
         gen_sync(jh, POST_SYNC, 19);
@@ -1484,24 +1482,22 @@ private:
         }
         else{
             if(rd!=0){
-                {
+                auto label_then12 = cc.newLabel();
-                auto label_then = cc.newLabel();
+                auto label_merge12 = cc.newLabel();
-                auto label_merge = cc.newLabel();
+                auto tmp_reg12 = get_reg(cc, 8, false);
-                auto tmp_reg = get_reg_for(cc, 1);
                 cmp(cc, load_reg_from_mem(jh, traits::X0 + rs1), (uint32_t)((int16_t)sext<12>(imm)));
-                cc.jb(label_then);
+                cc.jb(label_then12);
-                mov(cc, tmp_reg,0);
+                mov(cc, tmp_reg12,0);
-                cc.jmp(label_merge);
+                cc.jmp(label_merge12);
-                cc.bind(label_then);
+                cc.bind(label_then12);
-                mov(cc, tmp_reg,1);
+                mov(cc, tmp_reg12, 1);
-                cc.bind(label_merge);
+                cc.bind(label_merge12);
                 mov(cc, get_ptr_for(jh, traits::X0+ rd),
-                      gen_ext(cc, tmp_reg
+                      gen_ext(cc, tmp_reg12
                       , 32, false)
                 );
             }
         }
-        }
         auto returnValue = CONT;
         
         gen_sync(jh, POST_SYNC, 20);
@@ -1992,26 +1988,24 @@ private:
         }
         else{
             if(rd!=0){
-                {
+                auto label_then13 = cc.newLabel();
-                auto label_then = cc.newLabel();
+                auto label_merge13 = cc.newLabel();
-                auto label_merge = cc.newLabel();
+                auto tmp_reg13 = get_reg(cc, 8, false);
-                auto tmp_reg = get_reg_for(cc, 1);
                 cmp(cc, gen_ext(cc, 
                     load_reg_from_mem(jh, traits::X0 + rs1), 32, true), gen_ext(cc, 
                     load_reg_from_mem(jh, traits::X0 + rs2), 32, true));
-                cc.jl(label_then);
+                cc.jl(label_then13);
-                mov(cc, tmp_reg,0);
+                mov(cc, tmp_reg13,0);
-                cc.jmp(label_merge);
+                cc.jmp(label_merge13);
-                cc.bind(label_then);
+                cc.bind(label_then13);
-                mov(cc, tmp_reg,1);
+                mov(cc, tmp_reg13, 1);
-                cc.bind(label_merge);
+                cc.bind(label_merge13);
                 mov(cc, get_ptr_for(jh, traits::X0+ rd),
-                      gen_ext(cc, tmp_reg
+                      gen_ext(cc, tmp_reg13
                       , 32, false)
                 );
             }
         }
-        }
         auto returnValue = CONT;
         
         gen_sync(jh, POST_SYNC, 30);
@@ -2056,24 +2050,22 @@ private:
         }
         else{
             if(rd!=0){
-                {
+                auto label_then14 = cc.newLabel();
-                auto label_then = cc.newLabel();
+                auto label_merge14 = cc.newLabel();
-                auto label_merge = cc.newLabel();
+                auto tmp_reg14 = get_reg(cc, 8, false);
-                auto tmp_reg = get_reg_for(cc, 1);
                 cmp(cc, load_reg_from_mem(jh, traits::X0 + rs1), load_reg_from_mem(jh, traits::X0 + rs2));
-                cc.jb(label_then);
+                cc.jb(label_then14);
-                mov(cc, tmp_reg,0);
+                mov(cc, tmp_reg14,0);
-                cc.jmp(label_merge);
+                cc.jmp(label_merge14);
-                cc.bind(label_then);
+                cc.bind(label_then14);
-                mov(cc, tmp_reg,1);
+                mov(cc, tmp_reg14, 1);
-                cc.bind(label_merge);
+                cc.bind(label_merge14);
                 mov(cc, get_ptr_for(jh, traits::X0+ rd),
-                      gen_ext(cc, tmp_reg
+                      gen_ext(cc, tmp_reg14
                       , 32, false)
                 );
             }
         }
-        }
         auto returnValue = CONT;
         
         gen_sync(jh, POST_SYNC, 31);
@@ -2511,10 +2503,10 @@ private:
         gen_instr_prologue(jh);
         cc.comment("//behavior:");
         /*generate behavior*/
-        InvokeNode* call_wait;
+        InvokeNode* call_wait_15;
         jh.cc.comment("//call_wait");
-        jh.cc.invoke(&call_wait, &wait, FuncSignature::build<void, int32_t>());
+        jh.cc.invoke(&call_wait_15, &wait, FuncSignature::build<void, int32_t>());
-        setArg(call_wait, 0, 1);
+        setArg(call_wait_15, 0, 1);
         auto returnValue = CONT;
         
         gen_sync(jh, POST_SYNC, 41);
@@ -4788,7 +4780,7 @@ vm_impl<ARCH>::vm_impl(ARCH &core, unsigned core_id, unsigned cluster_id)
     }()) {}
 
 template <typename ARCH>
-continuation_e vm_impl<ARCH>::gen_single_inst_behavior(virt_addr_t &pc, unsigned int &inst_cnt, jit_holder& jh) {
+continuation_e vm_impl<ARCH>::gen_single_inst_behavior(virt_addr_t &pc, jit_holder& jh) {
     enum {TRAP_ID=1<<16};
     code_word_t instr = 0;
     phys_addr_t paddr(pc);
@@ -4800,7 +4792,6 @@ continuation_e vm_impl<ARCH>::gen_single_inst_behavior(virt_addr_t &pc, unsigned
         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())
@@ -4830,6 +4821,7 @@ void vm_impl<ARCH>::gen_instr_epilogue(jit_holder& jh) {
     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){
@@ -4875,6 +4867,7 @@ inline void vm_impl<ARCH>::gen_raise(jit_holder& jh, uint16_t trap_id, uint16_t
     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>

src/vm/fp_functions.cpp

@@ -203,8 +203,8 @@ uint32_t fclass_s(uint32_t v1) {
     uA.f = a;
     uiA = uA.ui;
 
-    uint_fast16_t infOrNaN = expF32UI(uiA) == 0xFF;
+    bool infOrNaN = expF32UI(uiA) == 0xFF;
-    uint_fast16_t subnormalOrZero = expF32UI(uiA) == 0;
+    bool subnormalOrZero = expF32UI(uiA) == 0;
     bool sign = signF32UI(uiA);
     bool fracZero = fracF32UI(uiA) == 0;
     bool isNaN = isNaNF32UI(uiA);
@@ -217,9 +217,13 @@ uint32_t fclass_s(uint32_t v1) {
 }
 
 uint32_t fconv_d2f(uint64_t v1, uint8_t mode) {
+    bool isNan = isNaNF64UI(v1);
+    bool isSNaN = softfloat_isSigNaNF64UI(v1);
     softfloat_roundingMode = rmm_map.at(mode);
-    bool nan = (v1 & defaultNaNF64UI) == defaultNaNF64UI;
+    softfloat_exceptionFlags = 0;
-    if(nan) {
+    if(isNan) {
+        if(isSNaN)
+            softfloat_raiseFlags(softfloat_flag_invalid);
         return defaultNaNF32UI;
     } else {
         float32_t res = f64_to_f32(float64_t{v1});
@@ -228,11 +232,11 @@ uint32_t fconv_d2f(uint64_t v1, uint8_t mode) {
 }
 
 uint64_t fconv_f2d(uint32_t v1, uint8_t mode) {
-    bool nan = (v1 & defaultNaNF32UI) == defaultNaNF32UI;
+    bool infOrNaN = expF32UI(v1) == 0xFF;
-    if(nan) {
+    bool subnormalOrZero = expF32UI(v1) == 0;
+    if(infOrNaN || subnormalOrZero) {
         return defaultNaNF64UI;
     } else {
-        softfloat_roundingMode = rmm_map.at(mode);
         float64_t res = f32_to_f64(float32_t{v1});
         return res.v;
     }
@@ -312,22 +316,23 @@ uint64_t fcmp_d(uint64_t v1, uint64_t v2, uint32_t op) {
 }
 
 uint64_t fcvt_d(uint64_t v1, uint32_t op, uint8_t mode) {
+
     float64_t v1f{v1};
     softfloat_exceptionFlags = 0;
     float64_t r;
     switch(op) {
-    case 0: { // l->d, fp to int32
+    case 0: { // l from d
         int64_t res = f64_to_i64(v1f, rmm_map.at(mode), true);
         return (uint64_t)res;
     }
-    case 1: { // lu->s
+    case 1: { // lu from d
         uint64_t res = f64_to_ui64(v1f, rmm_map.at(mode), true);
         return res;
     }
-    case 2: // s->l
+    case 2: // d from l
         r = i64_to_f64(v1);
         return r.v;
-    case 3: // s->lu
+    case 3: // d from lu
         r = ui64_to_f64(v1);
         return r.v;
     }
@@ -335,12 +340,24 @@ uint64_t fcvt_d(uint64_t v1, uint32_t op, uint8_t mode) {
 }
 
 uint64_t fmadd_d(uint64_t v1, uint64_t v2, uint64_t v3, uint32_t op, uint8_t mode) {
-    // op should be {softfloat_mulAdd_subProd(2), softfloat_mulAdd_subC(1)}
+    uint64_t F64_SIGN = 1ULL << 63;
+    switch(op) {
+    case 0: // FMADD_D
+        break;
+    case 1: // FMSUB_D
+        v3 ^= F64_SIGN;
+        break;
+    case 2: // FNMADD_D
+        v1 ^= F64_SIGN;
+        v3 ^= F64_SIGN;
+        break;
+    case 3: // FNMSUB_D
+        v1 ^= F64_SIGN;
+        break;
+    }
     softfloat_roundingMode = rmm_map.at(mode);
     softfloat_exceptionFlags = 0;
-    float64_t res = softfloat_mulAddF64(v1, v2, v3, op & 0x1);
+    float64_t res = softfloat_mulAddF64(v1, v2, v3, 0);
-    if(op > 1)
-        res.v ^= 1ULL << 63;
     return res.v;
 }
 
@@ -376,8 +393,8 @@ uint64_t fclass_d(uint64_t v1) {
     uA.f = a;
     uiA = uA.ui;
 
-    uint_fast16_t infOrNaN = expF64UI(uiA) == 0x7FF;
+    bool infOrNaN = expF64UI(uiA) == 0x7FF;
-    uint_fast16_t subnormalOrZero = expF64UI(uiA) == 0;
+    bool subnormalOrZero = expF64UI(uiA) == 0;
     bool sign = signF64UI(uiA);
     bool fracZero = fracF64UI(uiA) == 0;
     bool isNaN = isNaNF64UI(uiA);

src/vm/interp/vm_tgc5c.cpp

@@ -275,9 +275,6 @@ template <typename CODE_WORD> void debug_fn(CODE_WORD insn) {
     volatile CODE_WORD x = insn;
     insn = 2 * x;
 }
-
-template <typename ARCH> vm_impl<ARCH>::vm_impl() { this(new ARCH()); }
-
 // according to
 // https://stackoverflow.com/questions/8871204/count-number-of-1s-in-binary-representation
 #ifdef __GCC__
@@ -709,9 +706,9 @@ typename vm_base<ARCH>::virt_addr_t vm_impl<ARCH>::execute_inst(finish_cond_e co
                                     }
                                     else {
                                         uint32_t load_address = (uint32_t)((uint64_t)(*(X+rs1) ) + (uint64_t)((int16_t)sext<12>(imm) ));
-                                        int8_t res_27 = super::template read_mem<int8_t>(traits::MEM, load_address);
+                                        int8_t res_1 = super::template read_mem<int8_t>(traits::MEM, load_address);
                                         if(this->core.reg.trap_state>=0x80000000UL) throw memory_access_exception();
-                                        int8_t res = (int8_t)res_27;
+                                        int8_t res = (int8_t)res_1;
                                         if(rd != 0) {
                                             *(X+rd) = (uint32_t)res;
                                         }
@@ -740,9 +737,9 @@ typename vm_base<ARCH>::virt_addr_t vm_impl<ARCH>::execute_inst(finish_cond_e co
                                     }
                                     else {
                                         uint32_t load_address = (uint32_t)((uint64_t)(*(X+rs1) ) + (uint64_t)((int16_t)sext<12>(imm) ));
-                                        int16_t res_28 = super::template read_mem<int16_t>(traits::MEM, load_address);
+                                        int16_t res_2 = super::template read_mem<int16_t>(traits::MEM, load_address);
                                         if(this->core.reg.trap_state>=0x80000000UL) throw memory_access_exception();
-                                        int16_t res = (int16_t)res_28;
+                                        int16_t res = (int16_t)res_2;
                                         if(rd != 0) {
                                             *(X+rd) = (uint32_t)res;
                                         }
@@ -771,9 +768,9 @@ typename vm_base<ARCH>::virt_addr_t vm_impl<ARCH>::execute_inst(finish_cond_e co
                                     }
                                     else {
                                         uint32_t load_address = (uint32_t)((uint64_t)(*(X+rs1) ) + (uint64_t)((int16_t)sext<12>(imm) ));
-                                        int32_t res_29 = super::template read_mem<int32_t>(traits::MEM, load_address);
+                                        int32_t res_3 = super::template read_mem<int32_t>(traits::MEM, load_address);
                                         if(this->core.reg.trap_state>=0x80000000UL) throw memory_access_exception();
-                                        int32_t res = (int32_t)res_29;
+                                        int32_t res = (int32_t)res_3;
                                         if(rd != 0) {
                                             *(X+rd) = (uint32_t)res;
                                         }
@@ -802,9 +799,9 @@ typename vm_base<ARCH>::virt_addr_t vm_impl<ARCH>::execute_inst(finish_cond_e co
                                     }
                                     else {
                                         uint32_t load_address = (uint32_t)((uint64_t)(*(X+rs1) ) + (uint64_t)((int16_t)sext<12>(imm) ));
-                                        uint8_t res_30 = super::template read_mem<uint8_t>(traits::MEM, load_address);
+                                        uint8_t res_4 = super::template read_mem<uint8_t>(traits::MEM, load_address);
                                         if(this->core.reg.trap_state>=0x80000000UL) throw memory_access_exception();
-                                        uint8_t res = res_30;
+                                        uint8_t res = res_4;
                                         if(rd != 0) {
                                             *(X+rd) = (uint32_t)res;
                                         }
@@ -833,9 +830,9 @@ typename vm_base<ARCH>::virt_addr_t vm_impl<ARCH>::execute_inst(finish_cond_e co
                                     }
                                     else {
                                         uint32_t load_address = (uint32_t)((uint64_t)(*(X+rs1) ) + (uint64_t)((int16_t)sext<12>(imm) ));
-                                        uint16_t res_31 = super::template read_mem<uint16_t>(traits::MEM, load_address);
+                                        uint16_t res_5 = super::template read_mem<uint16_t>(traits::MEM, load_address);
                                         if(this->core.reg.trap_state>=0x80000000UL) throw memory_access_exception();
-                                        uint16_t res = res_31;
+                                        uint16_t res = res_5;
                                         if(rd != 0) {
                                             *(X+rd) = (uint32_t)res;
                                         }
@@ -1541,9 +1538,9 @@ typename vm_base<ARCH>::virt_addr_t vm_impl<ARCH>::execute_inst(finish_cond_e co
                                     else {
                                         uint32_t xrs1 = *(X+rs1);
                                         if(rd != 0) {
-                                            uint32_t res_32 = super::template read_mem<uint32_t>(traits::CSR, csr);
+                                            uint32_t res_6 = super::template read_mem<uint32_t>(traits::CSR, csr);
                                             if(this->core.reg.trap_state>=0x80000000UL) throw memory_access_exception();
-                                            uint32_t xrd = res_32;
+                                            uint32_t xrd = res_6;
                                             super::template write_mem<uint32_t>(traits::CSR, csr, xrs1);
                                             if(this->core.reg.trap_state>=0x80000000UL) throw memory_access_exception();
                                             *(X+rd) = xrd;
@@ -1576,9 +1573,9 @@ typename vm_base<ARCH>::virt_addr_t vm_impl<ARCH>::execute_inst(finish_cond_e co
                                         raise(0, traits::RV_CAUSE_ILLEGAL_INSTRUCTION);
                                     }
                                     else {
-                                        uint32_t res_33 = super::template read_mem<uint32_t>(traits::CSR, csr);
+                                        uint32_t res_7 = super::template read_mem<uint32_t>(traits::CSR, csr);
                                         if(this->core.reg.trap_state>=0x80000000UL) throw memory_access_exception();
-                                        uint32_t xrd = res_33;
+                                        uint32_t xrd = res_7;
                                         uint32_t xrs1 = *(X+rs1);
                                         if(rs1 != 0) {
                                             super::template write_mem<uint32_t>(traits::CSR, csr, xrd | xrs1);
@@ -1611,9 +1608,9 @@ typename vm_base<ARCH>::virt_addr_t vm_impl<ARCH>::execute_inst(finish_cond_e co
                                         raise(0, traits::RV_CAUSE_ILLEGAL_INSTRUCTION);
                                     }
                                     else {
-                                        uint32_t res_34 = super::template read_mem<uint32_t>(traits::CSR, csr);
+                                        uint32_t res_8 = super::template read_mem<uint32_t>(traits::CSR, csr);
                                         if(this->core.reg.trap_state>=0x80000000UL) throw memory_access_exception();
-                                        uint32_t xrd = res_34;
+                                        uint32_t xrd = res_8;
                                         uint32_t xrs1 = *(X+rs1);
                                         if(rs1 != 0) {
                                             super::template write_mem<uint32_t>(traits::CSR, csr, xrd & ~ xrs1);
@@ -1646,9 +1643,9 @@ typename vm_base<ARCH>::virt_addr_t vm_impl<ARCH>::execute_inst(finish_cond_e co
                                         raise(0, traits::RV_CAUSE_ILLEGAL_INSTRUCTION);
                                     }
                                     else {
-                                        uint32_t res_35 = super::template read_mem<uint32_t>(traits::CSR, csr);
+                                        uint32_t res_9 = super::template read_mem<uint32_t>(traits::CSR, csr);
                                         if(this->core.reg.trap_state>=0x80000000UL) throw memory_access_exception();
-                                        uint32_t xrd = res_35;
+                                        uint32_t xrd = res_9;
                                         super::template write_mem<uint32_t>(traits::CSR, csr, (uint32_t)zimm);
                                         if(this->core.reg.trap_state>=0x80000000UL) throw memory_access_exception();
                                         if(rd != 0) {
@@ -1678,9 +1675,9 @@ typename vm_base<ARCH>::virt_addr_t vm_impl<ARCH>::execute_inst(finish_cond_e co
                                         raise(0, traits::RV_CAUSE_ILLEGAL_INSTRUCTION);
                                     }
                                     else {
-                                        uint32_t res_36 = super::template read_mem<uint32_t>(traits::CSR, csr);
+                                        uint32_t res_10 = super::template read_mem<uint32_t>(traits::CSR, csr);
                                         if(this->core.reg.trap_state>=0x80000000UL) throw memory_access_exception();
-                                        uint32_t xrd = res_36;
+                                        uint32_t xrd = res_10;
                                         if(zimm != 0) {
                                             super::template write_mem<uint32_t>(traits::CSR, csr, xrd | (uint32_t)zimm);
                                             if(this->core.reg.trap_state>=0x80000000UL) throw memory_access_exception();
@@ -1712,9 +1709,9 @@ typename vm_base<ARCH>::virt_addr_t vm_impl<ARCH>::execute_inst(finish_cond_e co
                                         raise(0, traits::RV_CAUSE_ILLEGAL_INSTRUCTION);
                                     }
                                     else {
-                                        uint32_t res_37 = super::template read_mem<uint32_t>(traits::CSR, csr);
+                                        uint32_t res_11 = super::template read_mem<uint32_t>(traits::CSR, csr);
                                         if(this->core.reg.trap_state>=0x80000000UL) throw memory_access_exception();
-                                        uint32_t xrd = res_37;
+                                        uint32_t xrd = res_11;
                                         if(zimm != 0) {
                                             super::template write_mem<uint32_t>(traits::CSR, csr, xrd & ~ ((uint32_t)zimm));
                                             if(this->core.reg.trap_state>=0x80000000UL) throw memory_access_exception();
@@ -2049,9 +2046,9 @@ typename vm_base<ARCH>::virt_addr_t vm_impl<ARCH>::execute_inst(finish_cond_e co
                     // execute instruction
                     {
                         uint32_t offs = (uint32_t)((uint64_t)(*(X+rs1 + 8) ) + (uint64_t)(uimm ));
-                        int32_t res_38 = super::template read_mem<int32_t>(traits::MEM, offs);
+                        int32_t res_12 = super::template read_mem<int32_t>(traits::MEM, offs);
                         if(this->core.reg.trap_state>=0x80000000UL) throw memory_access_exception();
-                        *(X+rd + 8) = (uint32_t)(int32_t)res_38;
+                        *(X+rd + 8) = (uint32_t)(int32_t)res_12;
                     }
                     break;
                 }// @suppress("No break at end of case")
@@ -2475,9 +2472,9 @@ typename vm_base<ARCH>::virt_addr_t vm_impl<ARCH>::execute_inst(finish_cond_e co
                         }
                         else {
                             uint32_t offs = (uint32_t)((uint64_t)(*(X+2) ) + (uint64_t)(uimm ));
-                            int32_t res_39 = super::template read_mem<int32_t>(traits::MEM, offs);
+                            int32_t res_13 = super::template read_mem<int32_t>(traits::MEM, offs);
                             if(this->core.reg.trap_state>=0x80000000UL) throw memory_access_exception();
-                            *(X+rd) = (uint32_t)(int32_t)res_39;
+                            *(X+rd) = (uint32_t)(int32_t)res_13;
                         }
                     }
                     break;

src/vm/llvm/vm_tgc5c.cpp

@@ -97,7 +97,7 @@ protected:
         return super::gen_cond_assign(cond, this->gen_ext(trueVal, size), this->gen_ext(falseVal, size));
     }
 
-    std::tuple<continuation_e, BasicBlock *> gen_single_inst_behavior(virt_addr_t &, unsigned int &, BasicBlock *) override;
+    std::tuple<continuation_e, BasicBlock *> gen_single_inst_behavior(virt_addr_t &, BasicBlock *) override;
 
     void gen_leave_behavior(BasicBlock *leave_blk) override;
     void gen_raise_trap(uint16_t trap_id, uint16_t cause);
@@ -1490,7 +1490,7 @@ private:
                 ),
                 this->gen_const(8,1),
                 this->gen_const(8,0),
-                1), 32),
+                8), 32),
                 get_reg_ptr(rd + traits::X0), false);
             }
         }
@@ -1543,7 +1543,7 @@ private:
                 ),
                 this->gen_const(8,1),
                 this->gen_const(8,0),
-                1), 32),
+                8), 32),
                 get_reg_ptr(rd + traits::X0), false);
             }
         }
@@ -2057,7 +2057,7 @@ private:
                 ,
                 this->gen_const(8,1),
                 this->gen_const(8,0),
-                1), 32),
+                8), 32),
                 get_reg_ptr(rd + traits::X0), false);
             }
         }
@@ -2110,7 +2110,7 @@ private:
                 ,
                 this->gen_const(8,1),
                 this->gen_const(8,0),
-                1), 32),
+                8), 32),
                 get_reg_ptr(rd + traits::X0), false);
             }
         }
@@ -2553,11 +2553,10 @@ private:
         
         this->gen_instr_prologue();
         /*generate behavior*/
-        auto wait_arg0 = this->gen_const(8,1);
+        std::vector<Value*> wait_231_args{
-        std::vector<Value*> wait_args{
+            this->gen_ext(this->gen_const(8,1), 32)
-            wait_arg0
         };
-        this->builder.CreateCall(this->mod->getFunction("wait"), wait_args);
+        this->builder.CreateCall(this->mod->getFunction("wait"), wait_231_args);
         bb = BasicBlock::Create(this->mod->getContext(), "entry", this->func, this->leave_blk);
         auto returnValue = std::make_tuple(CONT,bb);
         
@@ -2719,7 +2718,7 @@ private:
                 csr,
                 this->builder.CreateAnd(
                    xrd,
-                   this->builder.CreateNeg(xrs1))
+                   this->builder.CreateNot(xrs1))
                 );
             }
             if(rd!=0) {
@@ -4938,7 +4937,7 @@ vm_impl<ARCH>::vm_impl(ARCH &core, unsigned core_id, unsigned cluster_id)
 
 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) {
+vm_impl<ARCH>::gen_single_inst_behavior(virt_addr_t &pc, BasicBlock *this_block) {
     // we fetch at max 4 byte, alignment is 2
     enum {TRAP_ID=1<<16};
     code_word_t instr = 0;
@@ -4950,9 +4949,10 @@ vm_impl<ARCH>::gen_single_inst_behavior(virt_addr_t &pc, unsigned int &inst_cnt,
     auto res = this->core.read(paddr, 4, data);
     if (res != iss::Ok) 
         return std::make_tuple(ILLEGAL_FETCH, nullptr);
-    if (instr == 0x0000006f || (instr&0xffff)==0xa001)
+    if (instr == 0x0000006f || (instr&0xffff)==0xa001){
+        this->builder.CreateBr(this->leave_blk);
         return std::make_tuple(JUMP_TO_SELF, nullptr);
-    ++inst_cnt;
+        }
     uint32_t inst_index = instr_decoder.decode_instr(instr);
     compile_func f = nullptr;
     if(inst_index < instr_descr.size())
@@ -4973,6 +4973,7 @@ 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::TRAP_STATE), true);
+    this->builder.CreateBr(this->trap_blk);
 }
 
 template <typename ARCH>
@@ -5033,6 +5034,10 @@ void vm_impl<ARCH>::gen_instr_epilogue(BasicBlock *bb) {
     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);
+    //increment cyclecount
+    auto* cycle_val = this->builder.CreateAdd(
+        this->builder.CreateLoad(this->get_typeptr(arch::traits<ARCH>::CYCLE), get_reg_ptr(arch::traits<ARCH>::CYCLE)), this->gen_const(64U, 1));
+    this->builder.CreateStore(cycle_val, get_reg_ptr(arch::traits<ARCH>::CYCLE), false);
 }
 
 } // namespace tgc5c