fixes htif behavior and instrumentation interface

CMakeLists.txt

@@ -18,6 +18,7 @@ add_subdirectory(softfloat)
 set(LIB_SOURCES
     src/iss/plugin/instruction_count.cpp
     src/iss/arch/tgc5c.cpp
+    src/iss/mmio/memory_if.cpp
     src/vm/interp/vm_tgc5c.cpp
     src/vm/fp_functions.cpp
     src/iss/debugger/csr_names.cpp
@@ -109,16 +110,6 @@ 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

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/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())

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())
@@ -340,6 +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>
-vm_impl<ARCH>::gen_single_inst_behavior(virt_addr_t &pc, unsigned int &inst_cnt, tu_builder& tu) {
+continuation_e
+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/iss/arch/mstatus.h

@@ -0,0 +1,233 @@
+/*******************************************************************************
+ * Copyright (C) 2025 MINRES Technologies GmbH
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ *    this list of conditions and the following disclaimer.
+ *
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ *    this list of conditions and the following disclaimer in the documentation
+ *    and/or other materials provided with the distribution.
+ *
+ * 3. Neither the name of the copyright holder nor the names of its contributors
+ *    may be used to endorse or promote products derived from this software
+ *    without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ *
+ * Contributors:
+ *       eyck@minres.com - initial implementation
+ ******************************************************************************/
+#ifndef _MSTATUS_TYPE
+#define _MSTATUS_TYPE
+
+#include <cstdint>
+#include <type_traits>
+#include <util/bit_field.h>
+#include <util/ities.h>
+
+namespace iss {
+namespace arch {
+
+template <class T, class Enable = void> struct status {};
+// specialization 32bit
+template <typename T> struct status<T, typename std::enable_if<std::is_same<T, uint32_t>::value>::type> {
+    static inline unsigned SD(T v) { return bit_sub<63, 1>(v); }
+    // value of XLEN for S-mode
+    static inline unsigned SXL(T v) { return bit_sub<34, 2>(v); };
+    // value of XLEN for U-mode
+    static inline unsigned UXL(T v) { return bit_sub<32, 2>(v); };
+    // Trap SRET
+    static inline unsigned TSR(T v) { return bit_sub<22, 1>(v); };
+    // Timeout Wait
+    static inline unsigned TW(T v) { return bit_sub<21, 1>(v); };
+    // Trap Virtual Memory
+    static inline unsigned TVM(T v) { return bit_sub<20, 1>(v); };
+    // Make eXecutable Readable
+    static inline unsigned MXR(T v) { return bit_sub<19, 1>(v); };
+    // permit Supervisor User Memory access
+    static inline unsigned SUM(T v) { return bit_sub<18, 1>(v); };
+    // Modify PRiVilege
+    static inline unsigned MPRV(T v) { return bit_sub<17, 1>(v); };
+    // status of additional user-mode extensions and associated state, All off/None dirty or clean, some on/None
+    // dirty, some clean/Some dirty
+    static inline unsigned XS(T v) { return bit_sub<15, 2>(v); };
+    // floating-point unit status Off/Initial/Clean/Dirty
+    static inline unsigned FS(T v) { return bit_sub<13, 2>(v); };
+    // machine previous privilege
+    static inline unsigned MPP(T v) { return bit_sub<11, 2>(v); };
+    // supervisor previous privilege
+    static inline unsigned SPP(T v) { return bit_sub<8, 1>(v); };
+    // previous machine interrupt-enable
+    static inline unsigned MPIE(T v) { return bit_sub<7, 1>(v); };
+    // previous supervisor interrupt-enable
+    static inline unsigned SPIE(T v) { return bit_sub<5, 1>(v); };
+    // previous user interrupt-enable
+    static inline unsigned UPIE(T v) { return bit_sub<4, 1>(v); };
+    // machine interrupt-enable
+    static inline unsigned MIE(T v) { return bit_sub<3, 1>(v); };
+    // supervisor interrupt-enable
+    static inline unsigned SIE(T v) { return bit_sub<1, 1>(v); };
+    // user interrupt-enable
+    static inline unsigned UIE(T v) { return bit_sub<0, 1>(v); };
+};
+
+template <typename T> struct status<T, typename std::enable_if<std::is_same<T, uint64_t>::value>::type> {
+public:
+    // SD bit is read-only and is set when either the FS or XS bits encode a Dirty state (i.e., SD=((FS==11) OR
+    // XS==11)))
+    static inline unsigned SD(T v) { return bit_sub<63, 1>(v); };
+    // value of XLEN for S-mode
+    static inline unsigned SXL(T v) { return bit_sub<34, 2>(v); };
+    // value of XLEN for U-mode
+    static inline unsigned UXL(T v) { return bit_sub<32, 2>(v); };
+    // Trap SRET
+    static inline unsigned TSR(T v) { return bit_sub<22, 1>(v); };
+    // Timeout Wait
+    static inline unsigned TW(T v) { return bit_sub<21, 1>(v); };
+    // Trap Virtual Memory
+    static inline unsigned TVM(T v) { return bit_sub<20, 1>(v); };
+    // Make eXecutable Readable
+    static inline unsigned MXR(T v) { return bit_sub<19, 1>(v); };
+    // permit Supervisor User Memory access
+    static inline unsigned SUM(T v) { return bit_sub<18, 1>(v); };
+    // Modify PRiVilege
+    static inline unsigned MPRV(T v) { return bit_sub<17, 1>(v); };
+    // status of additional user-mode extensions and associated state, All off/None dirty or clean, some on/None
+    // dirty, some clean/Some dirty
+    static inline unsigned XS(T v) { return bit_sub<15, 2>(v); };
+    // floating-point unit status Off/Initial/Clean/Dirty
+    static inline unsigned FS(T v) { return bit_sub<13, 2>(v); };
+    // machine previous privilege
+    static inline unsigned MPP(T v) { return bit_sub<11, 2>(v); };
+    // supervisor previous privilege
+    static inline unsigned SPP(T v) { return bit_sub<8, 1>(v); };
+    // previous machine interrupt-enable
+    static inline unsigned MPIE(T v) { return bit_sub<7, 1>(v); };
+    // previous supervisor interrupt-enable
+    static inline unsigned SPIE(T v) { return bit_sub<5, 1>(v); };
+    // previous user interrupt-enable
+    static inline unsigned UPIE(T v) { return bit_sub<4, 1>(v); };
+    // machine interrupt-enable
+    static inline unsigned MIE(T v) { return bit_sub<3, 1>(v); };
+    // supervisor interrupt-enable
+    static inline unsigned SIE(T v) { return bit_sub<1, 1>(v); };
+    // user interrupt-enable
+    static inline unsigned UIE(T v) { return bit_sub<0, 1>(v); };
+};
+
+// primary template
+template <class T, class Enable = void> struct hart_state {};
+// specialization 32bit
+template <typename T> class hart_state<T, typename std::enable_if<std::is_same<T, uint32_t>::value>::type> {
+public:
+    BEGIN_BF_DECL(mstatus_t, T);
+    // SD bit is read-only and is set when either the FS or XS bits encode a Dirty state (i.e., SD=((FS==11) OR
+    // XS==11)))
+    BF_FIELD(SD, 31, 1);
+    // Trap SRET
+    BF_FIELD(TSR, 22, 1);
+    // Timeout Wait
+    BF_FIELD(TW, 21, 1);
+    // Trap Virtual Memory
+    BF_FIELD(TVM, 20, 1);
+    // Make eXecutable Readable
+    BF_FIELD(MXR, 19, 1);
+    // permit Supervisor User Memory access
+    BF_FIELD(SUM, 18, 1);
+    // Modify PRiVilege
+    BF_FIELD(MPRV, 17, 1);
+    // status of additional user-mode extensions and associated state, All off/None dirty or clean, some on/None
+    // dirty, some clean/Some dirty
+    BF_FIELD(XS, 15, 2);
+    // floating-point unit status Off/Initial/Clean/Dirty
+    BF_FIELD(FS, 13, 2);
+    // machine previous privilege
+    BF_FIELD(MPP, 11, 2);
+    // supervisor previous privilege
+    BF_FIELD(SPP, 8, 1);
+    // previous machine interrupt-enable
+    BF_FIELD(MPIE, 7, 1);
+    // previous supervisor interrupt-enable
+    BF_FIELD(SPIE, 5, 1);
+    // previous user interrupt-enable
+    BF_FIELD(UPIE, 4, 1);
+    // machine interrupt-enable
+    BF_FIELD(MIE, 3, 1);
+    // supervisor interrupt-enable
+    BF_FIELD(SIE, 1, 1);
+    // user interrupt-enable
+    BF_FIELD(UIE, 0, 1);
+    END_BF_DECL();
+
+    mstatus_t mstatus;
+
+    static const T mstatus_reset_val = 0x1800;
+};
+
+// specialization 64bit
+template <typename T> class hart_state<T, typename std::enable_if<std::is_same<T, uint64_t>::value>::type> {
+public:
+    BEGIN_BF_DECL(mstatus_t, T);
+    // SD bit is read-only and is set when either the FS or XS bits encode a Dirty state (i.e., SD=((FS==11) OR
+    // XS==11)))
+    BF_FIELD(SD, 63, 1);
+    // value of XLEN for S-mode
+    BF_FIELD(SXL, 34, 2);
+    // value of XLEN for U-mode
+    BF_FIELD(UXL, 32, 2);
+    // Trap SRET
+    BF_FIELD(TSR, 22, 1);
+    // Timeout Wait
+    BF_FIELD(TW, 21, 1);
+    // Trap Virtual Memory
+    BF_FIELD(TVM, 20, 1);
+    // Make eXecutable Readable
+    BF_FIELD(MXR, 19, 1);
+    // permit Supervisor User Memory access
+    BF_FIELD(SUM, 18, 1);
+    // Modify PRiVilege
+    BF_FIELD(MPRV, 17, 1);
+    // status of additional user-mode extensions and associated state, All off/None dirty or clean, some on/None
+    // dirty, some clean/Some dirty
+    BF_FIELD(XS, 15, 2);
+    // floating-point unit status Off/Initial/Clean/Dirty
+    BF_FIELD(FS, 13, 2);
+    // machine previous privilege
+    BF_FIELD(MPP, 11, 2);
+    // supervisor previous privilege
+    BF_FIELD(SPP, 8, 1);
+    // previous machine interrupt-enable
+    BF_FIELD(MPIE, 7, 1);
+    // previous supervisor interrupt-enable
+    BF_FIELD(SPIE, 5, 1);
+    // previous user interrupt-enable
+    BF_FIELD(UPIE, 4, 1);
+    // machine interrupt-enable
+    BF_FIELD(MIE, 3, 1);
+    // supervisor interrupt-enable
+    BF_FIELD(SIE, 1, 1);
+    // user interrupt-enable
+    BF_FIELD(UIE, 0, 1);
+    END_BF_DECL();
+
+    mstatus_t mstatus;
+
+    static const T mstatus_reset_val = 0x1800;
+};
+} // namespace arch
+} // namespace iss
+#endif // _MSTATUS_TYPE

src/iss/arch/riscv_hart_common.h

@@ -1,5 +1,5 @@
 /*******************************************************************************
- * Copyright (C) 2017, 2018, 2021 MINRES Technologies GmbH
+ * Copyright (C) 2017 - 2025 MINRES Technologies GmbH
  * All rights reserved.
  *
  * Redistribution and use in source and binary forms, with or without
@@ -35,15 +35,25 @@
 #ifndef _RISCV_HART_COMMON
 #define _RISCV_HART_COMMON
 
+#include "iss/arch/traits.h"
+#include "iss/log_categories.h"
+#include "iss/mmio/memory_if.h"
 #include "iss/vm_types.h"
+#include "mstatus.h"
+#include "util/delegate.h"
+#include <array>
 #include <cstdint>
 #include <elfio/elfio.hpp>
 #include <fmt/format.h>
 #include <iss/arch_if.h>
 #include <iss/log_categories.h>
+#include <iss/semihosting/semihosting.h>
+#include <limits>
+#include <sstream>
 #include <string>
 #include <unordered_map>
 #include <util/logging.h>
+#include <util/sparse_array.h>
 
 #if defined(__GNUC__)
 #define likely(x) ::__builtin_expect(!!(x), 1)
@@ -56,9 +66,7 @@
 namespace iss {
 namespace arch {
 
-enum { tohost_dflt = 0xF0001000, fromhost_dflt = 0xF0001040 };
-
-enum features_e { FEAT_NONE, FEAT_PMP = 1, FEAT_EXT_N = 2, FEAT_CLIC = 4, FEAT_DEBUG = 8, FEAT_TCM = 16 };
+enum features_e { FEAT_NONE, FEAT_EXT_N = 1, FEAT_DEBUG = 2 };
 
 enum riscv_csr {
     /* user-level CSR */
@@ -234,10 +242,6 @@ struct vm_info {
 };
 
 struct feature_config {
-    uint64_t clic_base{0xc0000000};
-    unsigned clic_int_ctl_bits{4};
-    unsigned clic_num_irq{16};
-    unsigned clic_num_trigger{0};
     uint64_t tcm_base{0x10000000};
     uint64_t tcm_size{0x8000};
     uint64_t io_address{0xf0000000};
@@ -270,57 +274,144 @@ public:
     : trap_access(15 << 16, badaddr) {}
 };
 
-inline void read_reg_uint32(uint64_t offs, uint32_t& reg, uint8_t* const data, unsigned length) {
-    auto reg_ptr = reinterpret_cast<uint8_t*>(&reg);
-    switch(offs & 0x3) {
-    case 0:
-        for(auto i = 0U; i < length; ++i)
-            *(data + i) = *(reg_ptr + i);
-        break;
-    case 1:
-        for(auto i = 0U; i < length; ++i)
-            *(data + i) = *(reg_ptr + 1 + i);
-        break;
-    case 2:
-        for(auto i = 0U; i < length; ++i)
-            *(data + i) = *(reg_ptr + 2 + i);
-        break;
-    case 3:
-        *data = *(reg_ptr + 3);
-        break;
-    }
-}
+template <typename WORD_TYPE> struct priv_if {
+    using rd_csr_f = std::function<iss::status(unsigned addr, WORD_TYPE&)>;
+    using wr_csr_f = std::function<iss::status(unsigned addr, WORD_TYPE)>;
 
-inline void write_reg_uint32(uint64_t offs, uint32_t& reg, const uint8_t* const data, unsigned length) {
-    auto reg_ptr = reinterpret_cast<uint8_t*>(&reg);
-    switch(offs & 0x3) {
-    case 0:
-        for(auto i = 0U; i < length; ++i)
-            *(reg_ptr + i) = *(data + i);
-        break;
-    case 1:
-        for(auto i = 0U; i < length; ++i)
-            *(reg_ptr + 1 + i) = *(data + i);
-        break;
-    case 2:
-        for(auto i = 0U; i < length; ++i)
-            *(reg_ptr + 2 + i) = *(data + i);
-        break;
-    case 3:
-        *(reg_ptr + 3) = *data;
-        break;
+    std::function<iss::status(unsigned, WORD_TYPE&)> read_csr;
+    std::function<iss::status(unsigned, WORD_TYPE)> write_csr;
+    std::function<iss::status(uint8_t const*)> exec_htif;
+    std::unordered_map<unsigned, rd_csr_f>& csr_rd_cb;
+    std::unordered_map<unsigned, wr_csr_f>& csr_wr_cb;
+    hart_state<WORD_TYPE>& mstatus;
+    uint64_t& tohost;
+    uint64_t& fromhost;
+    unsigned& mcause_max_irq;
+};
+
+template <typename BASE, typename LOGCAT = logging::disass> struct riscv_hart_common : public BASE, public mmio::memory_elem {
+    const std::array<const char, 4> lvl = {{'U', 'S', 'H', 'M'}};
+    const std::array<const char*, 16> trap_str = {{""
+                                                   "Instruction address misaligned", // 0
+                                                   "Instruction access fault",       // 1
+                                                   "Illegal instruction",            // 2
+                                                   "Breakpoint",                     // 3
+                                                   "Load address misaligned",        // 4
+                                                   "Load access fault",              // 5
+                                                   "Store/AMO address misaligned",   // 6
+                                                   "Store/AMO access fault",         // 7
+                                                   "Environment call from U-mode",   // 8
+                                                   "Environment call from S-mode",   // 9
+                                                   "Reserved",                       // a
+                                                   "Environment call from M-mode",   // b
+                                                   "Instruction page fault",         // c
+                                                   "Load page fault",                // d
+                                                   "Reserved",                       // e
+                                                   "Store/AMO page fault"}};
+    const std::array<const char*, 12> irq_str = {{"User software interrupt", "Supervisor software interrupt", "Reserved",
+                                                  "Machine software interrupt", "User timer interrupt", "Supervisor timer interrupt",
+                                                  "Reserved", "Machine timer interrupt", "User external interrupt",
+                                                  "Supervisor external interrupt", "Reserved", "Machine external interrupt"}};
+    constexpr static unsigned MEM = traits<BASE>::MEM;
+
+    using core = BASE;
+    using this_class = riscv_hart_common<BASE, LOGCAT>;
+    using phys_addr_t = typename core::phys_addr_t;
+    using reg_t = typename core::reg_t;
+    using addr_t = typename core::addr_t;
+
+    using rd_csr_f = std::function<iss::status(unsigned addr, reg_t&)>;
+    using wr_csr_f = std::function<iss::status(unsigned addr, reg_t)>;
+
+#define MK_CSR_RD_CB(FCT) [this](unsigned a, reg_t& r) -> iss::status { return this->FCT(a, r); };
+#define MK_CSR_WR_CB(FCT) [this](unsigned a, reg_t r) -> iss::status { return this->FCT(a, r); };
+
+    riscv_hart_common()
+    : state()
+    , instr_if(*this) {
+        // reset values
+        csr[misa] = traits<BASE>::MISA_VAL;
+        csr[mvendorid] = 0x669;
+        csr[marchid] = traits<BASE>::MARCHID_VAL;
+        csr[mimpid] = 1;
+
+        if(traits<BASE>::FLEN > 0) {
+            csr_rd_cb[fcsr] = MK_CSR_RD_CB(read_fcsr);
+            csr_wr_cb[fcsr] = MK_CSR_WR_CB(write_fcsr);
+            csr_rd_cb[fflags] = MK_CSR_RD_CB(read_fcsr);
+            csr_wr_cb[fflags] = MK_CSR_WR_CB(write_fcsr);
+            csr_rd_cb[frm] = MK_CSR_RD_CB(read_fcsr);
+            csr_wr_cb[frm] = MK_CSR_WR_CB(write_fcsr);
+        }
+        for(unsigned addr = mhpmcounter3; addr <= mhpmcounter31; ++addr) {
+            csr_rd_cb[addr] = MK_CSR_RD_CB(read_null);
+            csr_wr_cb[addr] = MK_CSR_WR_CB(write_plain);
+        }
+        if(traits<BASE>::XLEN == 32)
+            for(unsigned addr = mhpmcounter3h; addr <= mhpmcounter31h; ++addr) {
+                csr_rd_cb[addr] = MK_CSR_RD_CB(read_null);
+                csr_wr_cb[addr] = MK_CSR_WR_CB(write_plain);
+            }
+        for(unsigned addr = mhpmevent3; addr <= mhpmevent31; ++addr) {
+            csr_rd_cb[addr] = MK_CSR_RD_CB(read_null);
+            csr_wr_cb[addr] = MK_CSR_WR_CB(write_plain);
+        }
+        for(unsigned addr = hpmcounter3; addr <= hpmcounter31; ++addr) {
+            csr_rd_cb[addr] = MK_CSR_RD_CB(read_null);
+        }
+        if(traits<BASE>::XLEN == 32)
+            for(unsigned addr = hpmcounter3h; addr <= hpmcounter31h; ++addr) {
+                csr_rd_cb[addr] = MK_CSR_RD_CB(read_null);
+            }
+        // common regs
+        const std::array<unsigned, 4> roaddrs{{misa, mvendorid, marchid, mimpid}};
+        for(auto addr : roaddrs) {
+            csr_rd_cb[addr] = MK_CSR_RD_CB(read_plain);
+            csr_wr_cb[addr] = MK_CSR_WR_CB(write_null);
+        }
+        // special handling & overrides
+        csr_rd_cb[time] = MK_CSR_RD_CB(read_time);
+        if(traits<BASE>::XLEN == 32)
+            csr_rd_cb[timeh] = MK_CSR_RD_CB(read_time);
+        csr_rd_cb[cycle] = MK_CSR_RD_CB(read_cycle);
+        if(traits<BASE>::XLEN == 32)
+            csr_rd_cb[cycleh] = MK_CSR_RD_CB(read_cycle);
+        csr_rd_cb[instret] = MK_CSR_RD_CB(read_instret);
+        if(traits<BASE>::XLEN == 32)
+            csr_rd_cb[instreth] = MK_CSR_RD_CB(read_instret);
+
+        csr_rd_cb[mcycle] = MK_CSR_RD_CB(read_cycle);
+        csr_wr_cb[mcycle] = MK_CSR_WR_CB(write_cycle);
+        if(traits<BASE>::XLEN == 32)
+            csr_rd_cb[mcycleh] = MK_CSR_RD_CB(read_cycle);
+        if(traits<BASE>::XLEN == 32)
+            csr_wr_cb[mcycleh] = MK_CSR_WR_CB(write_cycle);
+        csr_rd_cb[minstret] = MK_CSR_RD_CB(read_instret);
+        csr_wr_cb[minstret] = MK_CSR_WR_CB(write_instret);
+        if(traits<BASE>::XLEN == 32)
+            csr_rd_cb[minstreth] = MK_CSR_RD_CB(read_instret);
+        if(traits<BASE>::XLEN == 32)
+            csr_wr_cb[minstreth] = MK_CSR_WR_CB(write_instret);
+        csr_rd_cb[mhartid] = MK_CSR_RD_CB(read_hartid);
+    };
+    ~riscv_hart_common() {
+        if(io_buf.str().length()) {
+            CPPLOG(INFO) << "tohost send '" << io_buf.str() << "'";
+        }
     }
-}
-struct riscv_hart_common {
-    riscv_hart_common(){};
-    ~riscv_hart_common(){};
     std::unordered_map<std::string, uint64_t> symbol_table;
     uint64_t entry_address{0};
-    uint64_t tohost = tohost_dflt;
-    uint64_t fromhost = fromhost_dflt;
+    uint64_t tohost = std::numeric_limits<uint64_t>::max();
+    uint64_t fromhost = std::numeric_limits<uint64_t>::max();
+    std::stringstream io_buf;
 
-    bool read_elf_file(std::string name, uint8_t expected_elf_class,
-                       std::function<iss::status(uint64_t, uint64_t, const uint8_t* const)> cb) {
+    void set_semihosting_callback(semihosting_cb_t<reg_t> cb) { semihosting_cb = cb; };
+
+    std::pair<uint64_t, bool> load_file(std::string name, int type) {
+        return std::make_pair(entry_address, read_elf_file(name, sizeof(reg_t) == 4 ? ELFIO::ELFCLASS32 : ELFIO::ELFCLASS64));
+    }
+
+    bool read_elf_file(std::string name, uint8_t expected_elf_class) {
         // Create elfio reader
         ELFIO::elfio reader;
         // Load ELF data
@@ -337,8 +428,9 @@ struct riscv_hart_common {
                 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(type == ELFIO::PT_LOAD && 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();
                 }
@@ -365,11 +457,10 @@ struct riscv_hart_common {
                 }
                 try {
                     tohost = symbol_table.at("tohost");
-                    try {
-                        fromhost = symbol_table.at("fromhost");
-                    } catch(std::out_of_range& e) {
-                        fromhost = tohost + 0x40;
-                    }
+                } catch(std::out_of_range& e) {
+                }
+                try {
+                    fromhost = symbol_table.at("fromhost");
                 } catch(std::out_of_range& e) {
                 }
             }
@@ -377,6 +468,385 @@ struct riscv_hart_common {
         }
         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;
+    }
+
+    constexpr bool has_compressed() { return traits<BASE>::MISA_VAL & 0b0100; }
+
+    constexpr reg_t get_pc_mask() { return has_compressed() ? (reg_t)~1 : (reg_t)~3; }
+
+    void disass_output(uint64_t pc, const std::string instr) override {
+        // NSCLOG(INFO, LOGCAT) << fmt::format("0x{:016x}    {:40} [p:{};s:0x{:x};c:{}]", pc, instr, lvl[this->reg.PRIV],
+        // (reg_t)state.mstatus,
+        //                                     this->reg.cycle + cycle_offset);
+        NSCLOG(INFO, LOGCAT) << fmt::format("0x{:016x}    {:40} [p:{};c:{}]", pc, instr, lvl[this->reg.PRIV],
+                                            this->reg.cycle + cycle_offset);
+    };
+
+    void register_csr(unsigned addr, rd_csr_f f) { csr_rd_cb[addr] = f; }
+    void register_csr(unsigned addr, wr_csr_f f) { csr_wr_cb[addr] = f; }
+    void register_csr(unsigned addr, rd_csr_f rdf, wr_csr_f wrf) {
+        csr_rd_cb[addr] = rdf;
+        csr_wr_cb[addr] = wrf;
+    }
+    void unregister_csr_rd(unsigned addr) { csr_rd_cb.erase(addr); }
+    void unregister_csr_wr(unsigned addr) { csr_wr_cb.erase(addr); }
+
+    bool debug_mode_active() { return this->reg.PRIV & 0x4; }
+
+    const reg_t& get_mhartid() const { return mhartid_reg; }
+    void set_mhartid(reg_t mhartid) { mhartid_reg = mhartid; };
+
+    iss::status read_csr(unsigned addr, reg_t& val) {
+        if(addr >= csr.size())
+            return iss::Err;
+        auto req_priv_lvl = (addr >> 8) & 0x3;
+        if(this->reg.PRIV < req_priv_lvl) // not having required privileges
+            throw illegal_instruction_fault(this->fault_data);
+        auto it = csr_rd_cb.find(addr);
+        if(it == csr_rd_cb.end() || !it->second) // non existent register
+            throw illegal_instruction_fault(this->fault_data);
+        return it->second(addr, val);
+    }
+
+    iss::status write_csr(unsigned addr, reg_t val) {
+        if(addr >= csr.size())
+            return iss::Err;
+        auto req_priv_lvl = (addr >> 8) & 0x3;
+        if(this->reg.PRIV < req_priv_lvl) // not having required privileges
+            throw illegal_instruction_fault(this->fault_data);
+        if((addr & 0xc00) == 0xc00) // writing to read-only region
+            throw illegal_instruction_fault(this->fault_data);
+        auto it = csr_wr_cb.find(addr);
+        if(it == csr_wr_cb.end() || !it->second) // non existent register
+            throw illegal_instruction_fault(this->fault_data);
+        return it->second(addr, val);
+    }
+
+    iss::status read_null(unsigned addr, reg_t& val) {
+        val = 0;
+        return iss::Ok;
+    }
+
+    iss::status write_null(unsigned addr, reg_t val) { return iss::status::Ok; }
+
+    iss::status read_plain(unsigned addr, reg_t& val) {
+        val = csr[addr];
+        return iss::Ok;
+    }
+
+    iss::status write_plain(unsigned addr, reg_t val) {
+        csr[addr] = val;
+        return iss::Ok;
+    }
+
+    iss::status read_cycle(unsigned addr, reg_t& val) {
+        auto cycle_val = this->reg.cycle + cycle_offset;
+        if(addr == mcycle) {
+            val = static_cast<reg_t>(cycle_val);
+        } else if(addr == mcycleh) {
+            val = static_cast<reg_t>(cycle_val >> 32);
+        }
+        return iss::Ok;
+    }
+
+    iss::status write_cycle(unsigned addr, reg_t val) {
+        if(sizeof(typename traits<BASE>::reg_t) != 4) {
+            mcycle_csr = static_cast<uint64_t>(val);
+        } else {
+            if(addr == mcycle) {
+                mcycle_csr = (mcycle_csr & 0xffffffff00000000) + val;
+            } else {
+                mcycle_csr = (static_cast<uint64_t>(val) << 32) + (mcycle_csr & 0xffffffff);
+            }
+        }
+        cycle_offset = mcycle_csr - this->reg.cycle; // TODO: relying on wrap-around
+        return iss::Ok;
+    }
+
+    iss::status read_instret(unsigned addr, reg_t& val) {
+        if((addr & 0xff) == (minstret & 0xff)) {
+            val = static_cast<reg_t>(this->reg.instret);
+        } else if((addr & 0xff) == (minstreth & 0xff)) {
+            val = static_cast<reg_t>(this->reg.instret >> 32);
+        }
+        return iss::Ok;
+    }
+
+    iss::status write_instret(unsigned addr, reg_t val) {
+        if(sizeof(typename traits<BASE>::reg_t) != 4) {
+            this->reg.instret = static_cast<uint64_t>(val);
+        } else {
+            if((addr & 0xff) == (minstret & 0xff)) {
+                this->reg.instret = (this->reg.instret & 0xffffffff00000000) + val;
+            } else {
+                this->reg.instret = (static_cast<uint64_t>(val) << 32) + (this->reg.instret & 0xffffffff);
+            }
+        }
+        this->reg.instret--;
+        return iss::Ok;
+    }
+
+    iss::status read_time(unsigned addr, reg_t& val) {
+        uint64_t time_val = this->reg.cycle / (100000000 / 32768 - 1); //-> ~3052;
+        if(addr == time) {
+            val = static_cast<reg_t>(time_val);
+        } else if(addr == timeh) {
+            if(sizeof(typename traits<BASE>::reg_t) != 4)
+                return iss::Err;
+            val = static_cast<reg_t>(time_val >> 32);
+        }
+        return iss::Ok;
+    }
+
+    iss::status read_tvec(unsigned addr, reg_t& val) {
+        val = csr[addr] & ~2;
+        return iss::Ok;
+    }
+
+    iss::status read_hartid(unsigned addr, reg_t& val) {
+        val = mhartid_reg;
+        return iss::Ok;
+    }
+
+    iss::status write_epc(unsigned addr, reg_t val) {
+        csr[addr] = val & get_pc_mask();
+        return iss::Ok;
+    }
+
+    iss::status write_dcsr(unsigned addr, reg_t val) {
+        if(!debug_mode_active())
+            throw illegal_instruction_fault(this->fault_data);
+        //                  +-------------- ebreakm
+        //                  |   +---------- stepi
+        //                  |   |  +++----- cause
+        //                  |   |  |||   +- step
+        csr[addr] = val & 0b1000100111000100U;
+        return iss::Ok;
+    }
+
+    iss::status read_debug(unsigned addr, reg_t& val) {
+        if(!debug_mode_active())
+            throw illegal_instruction_fault(this->fault_data);
+        val = csr[addr];
+        return iss::Ok;
+    }
+
+    iss::status write_dscratch(unsigned addr, reg_t val) {
+        if(!debug_mode_active())
+            throw illegal_instruction_fault(this->fault_data);
+        csr[addr] = val;
+        return iss::Ok;
+    }
+
+    iss::status read_dpc(unsigned addr, reg_t& val) {
+        if(!debug_mode_active())
+            throw illegal_instruction_fault(this->fault_data);
+        val = this->reg.DPC;
+        return iss::Ok;
+    }
+
+    iss::status write_dpc(unsigned addr, reg_t val) {
+        if(!debug_mode_active())
+            throw illegal_instruction_fault(this->fault_data);
+        this->reg.DPC = val;
+        return iss::Ok;
+    }
+
+    iss::status read_fcsr(unsigned addr, reg_t& val) {
+        switch(addr) {
+        case 1: // fflags, 4:0
+            val = bit_sub<0, 5>(this->get_fcsr());
+            break;
+        case 2: // frm, 7:5
+            val = bit_sub<5, 3>(this->get_fcsr());
+            break;
+        case 3: // fcsr
+            val = this->get_fcsr();
+            break;
+        default:
+            return iss::Err;
+        }
+        return iss::Ok;
+    }
+
+    iss::status write_fcsr(unsigned addr, reg_t val) {
+        switch(addr) {
+        case 1: // fflags, 4:0
+            this->set_fcsr((this->get_fcsr() & 0xffffffe0) | (val & 0x1f));
+            break;
+        case 2: // frm, 7:5
+            this->set_fcsr((this->get_fcsr() & 0xffffff1f) | ((val & 0x7) << 5));
+            break;
+        case 3: // fcsr
+            this->set_fcsr(val & 0xff);
+            break;
+        default:
+            return iss::Err;
+        }
+        return iss::Ok;
+    }
+
+    priv_if<reg_t> get_priv_if() {
+        return priv_if<reg_t>{.read_csr = [this](unsigned addr, reg_t& val) -> iss::status { return read_csr(addr, val); },
+                              .write_csr = [this](unsigned addr, reg_t val) -> iss::status { return write_csr(addr, val); },
+                              .exec_htif = [this](uint8_t const* data) -> iss::status { return execute_htif(data); },
+                              .csr_rd_cb{this->csr_rd_cb},
+                              .csr_wr_cb{csr_wr_cb},
+                              .mstatus{this->state},
+                              .tohost{this->tohost},
+                              .fromhost{this->fromhost},
+                              .mcause_max_irq{mcause_max_irq}};
+    }
+
+    iss::status execute_htif(uint8_t const* data) {
+        reg_t cur_data = *reinterpret_cast<const reg_t*>(data);
+        // Extract Device (bits 63:56)
+        uint8_t device = traits<BASE>::XLEN == 32 ? 0 : (cur_data >> 56) & 0xFF;
+        // Extract Command (bits 55:48)
+        uint8_t command = traits<BASE>::XLEN == 32 ? 0 : (cur_data >> 48) & 0xFF;
+        // Extract payload (bits 47:0)
+        uint64_t payload_addr = cur_data & 0xFFFFFFFFFFFFULL;
+        if(payload_addr & 1) {
+            CPPLOG(FATAL) << "this->tohost value is 0x" << std::hex << payload_addr << std::dec << " (" << payload_addr
+                          << "), stopping simulation";
+            this->reg.trap_state = std::numeric_limits<uint32_t>::max();
+            this->interrupt_sim = payload_addr;
+            return iss::Ok;
+        } else if(device == 0 && command == 0) {
+            std::array<uint64_t, 8> loaded_payload;
+            if(memory.rd_mem(access_type::DEBUG_READ, payload_addr, 8 * sizeof(uint64_t),
+                             reinterpret_cast<uint8_t*>(loaded_payload.data())) == iss::Err)
+                CPPLOG(ERR) << "Syscall read went wrong";
+            uint64_t syscall_num = loaded_payload.at(0);
+            if(syscall_num == 64) { // SYS_WRITE
+                return this->execute_sys_write(this, loaded_payload, traits<BASE>::MEM);
+            } else {
+                CPPLOG(ERR) << "this->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;
+            }
+        } else {
+            CPPLOG(ERR) << "this->tohost functionality not implemented for device " << device << " and command " << command;
+            this->reg.trap_state = std::numeric_limits<uint32_t>::max();
+            this->interrupt_sim = payload_addr;
+            return iss::Ok;
+        }
+    }
+
+    mmio::memory_hierarchy memories;
+
+    virtual mmio::memory_if get_mem_if() override {
+        assert(false || "This function should nevver be called");
+        return mmio::memory_if{};
+    }
+
+    virtual void set_next(mmio::memory_if mem_if) { memory = mem_if; };
+
+    void set_irq_num(unsigned i) { mcause_max_irq = 1 << util::ilog2(i); }
+
+protected:
+    hart_state<reg_t> state;
+
+    static constexpr reg_t get_mstatus_mask_t(unsigned priv_lvl = PRIV_M) {
+        if(sizeof(reg_t) == 4) {
+            return priv_lvl == PRIV_U ? 0x80000011UL :   // 0b1...0 0001 0001
+                       priv_lvl == PRIV_S ? 0x800de133UL // 0b0...0 0001 1000 1001 1001;
+                                          : 0x807ff9ddUL;
+        } else {
+            return priv_lvl == PRIV_U ? 0x011ULL : // 0b1...0 0001 0001
+                       priv_lvl == PRIV_S ? 0x000de133ULL
+                                          : 0x007ff9ddULL;
+        }
+    }
+
+    mmio::memory_if memory;
+    struct riscv_instrumentation_if : public iss::instrumentation_if {
+
+        riscv_instrumentation_if(riscv_hart_common<BASE, LOGCAT>& arch)
+        : arch(arch) {}
+        /**
+         * get the name of this architecture
+         *
+         * @return the name of this architecture
+         */
+        const std::string core_type_name() const override { return traits<BASE>::core_type; }
+
+        uint64_t get_pc() override { return arch.reg.PC; }
+
+        uint64_t get_next_pc() override { return arch.reg.NEXT_PC; }
+
+        uint64_t get_instr_word() override { return arch.reg.instruction; }
+
+        uint64_t get_instr_count() override { return arch.reg.icount; }
+
+        uint64_t get_pendig_traps() override { return arch.reg.trap_state; }
+
+        uint64_t get_total_cycles() override { return arch.reg.cycle + arch.cycle_offset; }
+
+        void update_last_instr_cycles(unsigned cycles) override { arch.cycle_offset += cycles - 1; }
+
+        bool is_branch_taken() override { return arch.reg.last_branch; }
+
+        unsigned get_reg_num() override { return traits<BASE>::NUM_REGS; }
+
+        unsigned get_reg_size(unsigned num) override { return traits<BASE>::reg_bit_widths[num]; }
+
+        std::unordered_map<std::string, uint64_t> const& get_symbol_table(std::string name) override { return arch.symbol_table; }
+
+        riscv_hart_common<BASE, LOGCAT>& arch;
+    };
+
+    friend struct riscv_instrumentation_if;
+    riscv_instrumentation_if instr_if;
+
+    instrumentation_if* get_instrumentation_if() override { return &instr_if; };
+
+    using csr_type = util::sparse_array<typename traits<BASE>::reg_t, 1ULL << 12, 12>;
+    using csr_page_type = typename csr_type::page_type;
+    csr_type csr;
+
+    std::unordered_map<unsigned, rd_csr_f> csr_rd_cb;
+    std::unordered_map<unsigned, wr_csr_f> csr_wr_cb;
+
+    reg_t mhartid_reg{0x0};
+    uint64_t mcycle_csr{0};
+    uint64_t minstret_csr{0};
+    reg_t fault_data;
+
+    int64_t cycle_offset{0};
+    int64_t instret_offset{0};
+    semihosting_cb_t<reg_t> semihosting_cb;
+    std::array<vm_info, 2> vm;
+    unsigned mcause_max_irq{16U};
 };
 
 } // namespace arch

src/iss/mmio/clic.h

@@ -0,0 +1,252 @@
+
+#include "iss/arch/riscv_hart_common.h"
+#include "iss/vm_types.h"
+#include "memory_if.h"
+#include <util/logging.h>
+
+namespace iss {
+namespace mmio {
+struct clic_config {
+    uint64_t clic_base{0xc0000000};
+    unsigned clic_int_ctl_bits{4};
+    unsigned clic_num_irq{16};
+    unsigned clic_num_trigger{0};
+    bool nmode{false};
+};
+
+inline void read_reg_with_offset(uint32_t reg, uint8_t offs, uint8_t* const data, unsigned length) {
+    auto reg_ptr = reinterpret_cast<uint8_t*>(&reg);
+    switch(offs) {
+    default:
+        for(auto i = 0U; i < length; ++i)
+            *(data + i) = *(reg_ptr + i);
+        break;
+    case 1:
+        for(auto i = 0U; i < length; ++i)
+            *(data + i) = *(reg_ptr + 1 + i);
+        break;
+    case 2:
+        for(auto i = 0U; i < length; ++i)
+            *(data + i) = *(reg_ptr + 2 + i);
+        break;
+    case 3:
+        *data = *(reg_ptr + 3);
+        break;
+    }
+}
+
+inline void write_reg_with_offset(uint32_t& reg, uint8_t offs, const uint8_t* const data, unsigned length) {
+    auto reg_ptr = reinterpret_cast<uint8_t*>(&reg);
+    switch(offs) {
+    default:
+        for(auto i = 0U; i < length; ++i)
+            *(reg_ptr + i) = *(data + i);
+        break;
+    case 1:
+        for(auto i = 0U; i < length; ++i)
+            *(reg_ptr + 1 + i) = *(data + i);
+        break;
+    case 2:
+        for(auto i = 0U; i < length; ++i)
+            *(reg_ptr + 2 + i) = *(data + i);
+        break;
+    case 3:
+        *(reg_ptr + 3) = *data;
+        break;
+    }
+}
+
+template <typename WORD_TYPE> struct clic : public memory_elem {
+    using this_class = clic<WORD_TYPE>;
+    using reg_t = WORD_TYPE;
+    constexpr static unsigned WORD_LEN = sizeof(WORD_TYPE) * 8;
+
+    clic(arch::priv_if<WORD_TYPE> hart_if, clic_config cfg)
+    : hart_if(hart_if)
+    , cfg(cfg) {
+        clic_int_reg.resize(cfg.clic_num_irq, clic_int_reg_t{.raw = 0});
+        clic_cfg_reg = 0x30;
+        clic_mact_lvl = clic_mprev_lvl = (1 << (cfg.clic_int_ctl_bits)) - 1;
+        clic_uact_lvl = clic_uprev_lvl = (1 << (cfg.clic_int_ctl_bits)) - 1;
+        hart_if.csr_rd_cb[arch::mtvt] = MK_CSR_RD_CB(read_plain);
+        hart_if.csr_wr_cb[arch::mtvt] = MK_CSR_WR_CB(write_xtvt);
+        //        hart_if.csr_rd_cb[mxnti] = MK_CSR_RD_CB(read_plain(a,r);};
+        //        hart_if.csr_wr_cb[mxnti] = MK_CSR_WR_CB(write_plain(a,r);};
+        hart_if.csr_rd_cb[arch::mintstatus] = MK_CSR_RD_CB(read_intstatus);
+        hart_if.csr_wr_cb[arch::mintstatus] = MK_CSR_WR_CB(write_null);
+        //        hart_if.csr_rd_cb[mscratchcsw] = MK_CSR_RD_CB(read_plain(a,r);};
+        //        hart_if.csr_wr_cb[mscratchcsw] = MK_CSR_WR_CB(write_plain(a,r);};
+        //        hart_if.csr_rd_cb[mscratchcswl] = MK_CSR_RD_CB(read_plain(a,r);};
+        //        hart_if.csr_wr_cb[mscratchcswl] = MK_CSR_WR_CB(write_plain(a,r);};
+        hart_if.csr_rd_cb[arch::mintthresh] = MK_CSR_RD_CB(read_plain);
+        hart_if.csr_wr_cb[arch::mintthresh] = MK_CSR_WR_CB(write_intthresh);
+        if(cfg.nmode) {
+            hart_if.csr_rd_cb[arch::utvt] = MK_CSR_RD_CB(read_plain);
+            hart_if.csr_wr_cb[arch::utvt] = MK_CSR_WR_CB(write_xtvt);
+            hart_if.csr_rd_cb[arch::uintstatus] = MK_CSR_RD_CB(read_intstatus);
+            hart_if.csr_wr_cb[arch::uintstatus] = MK_CSR_WR_CB(write_null);
+            hart_if.csr_rd_cb[arch::uintthresh] = MK_CSR_RD_CB(read_plain);
+            hart_if.csr_wr_cb[arch::uintthresh] = MK_CSR_WR_CB(write_intthresh);
+        }
+        hart_if.csr[arch::mintthresh] = (1 << (cfg.clic_int_ctl_bits)) - 1;
+        hart_if.csr[arch::uintthresh] = (1 << (cfg.clic_int_ctl_bits)) - 1;
+    }
+
+    ~clic() = default;
+
+    memory_if get_mem_if() override {
+        return memory_if{.rd_mem{util::delegate<rd_mem_func_sig>::from<this_class, &this_class::read_mem>(this)},
+                         .wr_mem{util::delegate<wr_mem_func_sig>::from<this_class, &this_class::write_mem>(this)}};
+    }
+
+    void set_next(memory_if mem) override { down_stream_mem = mem; }
+
+    std::tuple<uint64_t, uint64_t> get_range() override { return {cfg.clic_base, cfg.clic_base + 0x7fff}; }
+
+private:
+    iss::status read_mem(iss::access_type access, uint64_t addr, unsigned length, uint8_t* data) {
+        if(addr >= cfg.clic_base && (addr + length) < (cfg.clic_base + 0x8000))
+            return read_clic(addr, length, data);
+        return down_stream_mem.rd_mem(access, addr, length, data);
+    }
+
+    iss::status write_mem(iss::access_type access, uint64_t addr, unsigned length, uint8_t const* data) {
+        if(addr >= cfg.clic_base && (addr + length) < (cfg.clic_base + 0x8000))
+            return write_clic(addr, length, data);
+        return down_stream_mem.wr_mem(access, addr, length, data);
+    }
+
+    iss::status read_clic(uint64_t addr, unsigned length, uint8_t* data);
+
+    iss::status write_clic(uint64_t addr, unsigned length, uint8_t const* data);
+
+    iss::status write_null(unsigned addr, reg_t val) { return iss::status::Ok; }
+
+    iss::status read_plain(unsigned addr, reg_t& val) {
+        val = hart_if.csr[addr];
+        return iss::Ok;
+    }
+
+    iss::status write_xtvt(unsigned addr, reg_t val) {
+        hart_if.csr[addr] = val & ~0x3fULL;
+        return iss::Ok;
+    }
+
+    iss::status read_cause(unsigned addr, reg_t& val);
+    iss::status write_cause(unsigned addr, reg_t val);
+
+    iss::status read_intstatus(unsigned addr, reg_t& val);
+    iss::status write_intthresh(unsigned addr, reg_t val);
+
+protected:
+    arch::priv_if<WORD_TYPE> hart_if;
+    memory_if down_stream_mem;
+    clic_config cfg;
+    uint8_t clic_cfg_reg{0};
+    std::array<uint32_t, 32> clic_inttrig_reg;
+    union clic_int_reg_t {
+        struct {
+            uint8_t ip;
+            uint8_t ie;
+            uint8_t attr;
+            uint8_t ctl;
+        };
+        uint32_t raw;
+    };
+    std::vector<clic_int_reg_t> clic_int_reg;
+    uint8_t clic_mprev_lvl{0}, clic_uprev_lvl{0};
+    uint8_t clic_mact_lvl{0}, clic_uact_lvl{0};
+};
+
+template <typename WORD_TYPE> iss::status clic<WORD_TYPE>::read_clic(uint64_t addr, unsigned length, uint8_t* const data) {
+    if(addr == cfg.clic_base) { // cliccfg
+        *data = clic_cfg_reg;
+        for(auto i = 1; i < length; ++i)
+            *(data + i) = 0;
+    } else if(addr >= (cfg.clic_base + 0x40) && (addr + length) <= (cfg.clic_base + 0x40 + cfg.clic_num_trigger * 4)) { // clicinttrig
+        auto offset = ((addr & 0x7fff) - 0x40) / 4;
+        read_reg_with_offset(clic_inttrig_reg[offset], addr & 0x3, data, length);
+    } else if(addr >= (cfg.clic_base + 0x1000) &&
+              (addr + length) <= (cfg.clic_base + 0x1000 + cfg.clic_num_irq * 4)) { // clicintip/clicintie/clicintattr/clicintctl
+        auto offset = ((addr & 0x7fff) - 0x1000) / 4;
+        read_reg_with_offset(clic_int_reg[offset].raw, addr & 0x3, data, length);
+    } else {
+        for(auto i = 0U; i < length; ++i)
+            *(data + i) = 0;
+    }
+    return iss::Ok;
+}
+
+template <typename WORD_TYPE> iss::status clic<WORD_TYPE>::write_clic(uint64_t addr, unsigned length, const uint8_t* const data) {
+    if(addr == cfg.clic_base) { // cliccfg
+        clic_cfg_reg = (clic_cfg_reg & ~0x1e) | (*data & 0x1e);
+    } else if(addr >= (cfg.clic_base + 0x40) && (addr + length) <= (cfg.clic_base + 0x40 + cfg.clic_num_trigger * 4)) { // clicinttrig
+        auto offset = ((addr & 0x7fff) - 0x40) / 4;
+        write_reg_with_offset(clic_inttrig_reg[offset], addr & 0x3, data, length);
+    } else if(addr >= (cfg.clic_base + 0x1000) &&
+              (addr + length) <= (cfg.clic_base + 0x1000 + cfg.clic_num_irq * 4)) { // clicintip/clicintie/clicintattr/clicintctl
+        auto offset = ((addr & 0x7fff) - 0x1000) / 4;
+        write_reg_with_offset(clic_int_reg[offset].raw, addr & 0x3, data, length);
+        clic_int_reg[offset].raw &= 0xf0c70101; // clicIntCtlBits->0xf0, clicintattr->0xc7, clicintie->0x1, clicintip->0x1
+    }
+    return iss::Ok;
+}
+
+template <typename WORD_TYPE> iss::status clic<WORD_TYPE>::read_cause(unsigned addr, reg_t& val) {
+    if((hart_if.csr[arch::mtvec] & 0x3) == 3) {
+        val = hart_if.csr[addr] & (1UL << (sizeof(reg_t) * 8) | (hart_if.mcause_max_irq - 1) | (0xfUL << 16));
+        auto mode = (addr >> 8) & 0x3;
+        switch(mode) {
+        case 0:
+            val |= clic_uprev_lvl << 16;
+            val |= hart_if.mstatus.UPIE << 27;
+            break;
+        default:
+            val |= clic_mprev_lvl << 16;
+            val |= hart_if.mstatus.MPIE << 27;
+            val |= hart_if.mstatus.MPP << 28;
+            break;
+        }
+    } else
+        val = hart_if.csr[addr] & ((1UL << (sizeof(WORD_TYPE) * 8 - 1)) | (hart_if.mcause_max_irq - 1));
+    return iss::Ok;
+}
+
+template <typename WORD_TYPE> iss::status clic<WORD_TYPE>::write_cause(unsigned addr, reg_t val) {
+    if((hart_if.csr[arch::mtvec] & 0x3) == 3) {
+        auto mask = ((1UL << (sizeof(WORD_TYPE) * 8 - 1)) | (hart_if.mcause_max_irq - 1) | (0xfUL << 16));
+        hart_if.csr[addr] = (val & mask) | (hart_if.csr[addr] & ~mask);
+        auto mode = (addr >> 8) & 0x3;
+        switch(mode) {
+        case 0:
+            clic_uprev_lvl = ((val >> 16) & 0xff) | (1 << (8 - cfg.clic_int_ctl_bits)) - 1;
+            hart_if.mstatus.UPIE = (val >> 27) & 0x1;
+            break;
+        default:
+            clic_mprev_lvl = ((val >> 16) & 0xff) | (1 << (8 - cfg.clic_int_ctl_bits)) - 1;
+            hart_if.mstatus.MPIE = (val >> 27) & 0x1;
+            hart_if.mstatus.MPP = (val >> 28) & 0x3;
+            break;
+        }
+    } else {
+        auto mask = ((1UL << (sizeof(WORD_TYPE) * 8 - 1)) | (hart_if.mcause_max_irq - 1));
+        hart_if.csr[addr] = (val & mask) | (hart_if.csr[addr] & ~mask);
+    }
+    return iss::Ok;
+}
+
+template <typename WORD_TYPE> iss::status clic<WORD_TYPE>::read_intstatus(unsigned addr, reg_t& val) {
+    auto mode = (addr >> 8) & 0x3;
+    val = clic_uact_lvl & 0xff;
+    if(mode == 0x3)
+        val += (clic_mact_lvl & 0xff) << 24;
+    return iss::Ok;
+}
+
+template <typename WORD_TYPE> iss::status clic<WORD_TYPE>::write_intthresh(unsigned addr, reg_t val) {
+    hart_if.csr[addr] = (val & 0xff) | (1 << (cfg.clic_int_ctl_bits)) - 1;
+    return iss::Ok;
+}
+
+} // namespace mmio
+} // namespace iss

src/iss/mmio/memory_if.cpp

@@ -0,0 +1,26 @@
+#include "memory_if.h"
+
+namespace iss {
+namespace mmio {
+void memory_hierarchy::prepend(memory_elem& e) {
+    hierarchy.push_front(e);
+    update_chain();
+}
+void memory_hierarchy::append(memory_elem& e) {
+    hierarchy.push_back(e);
+    update_chain();
+}
+void memory_hierarchy::insert_before(memory_elem&) {}
+void memory_hierarchy::insert_after(memory_elem&) {}
+void memory_hierarchy::replace_last(memory_elem&) {}
+void memory_hierarchy::update_chain() {
+    bool tail = false;
+    for(size_t i = 0; i < hierarchy.size(); ++i) {
+        hierarchy[i].get().register_csrs();
+        if(i)
+            hierarchy[i - 1].get().set_next(hierarchy[i].get().get_mem_if());
+    }
+}
+
+} // namespace mmio
+} // namespace iss

src/iss/mmio/memory_if.h

@@ -0,0 +1,76 @@
+/*******************************************************************************
+ * Copyright (C) 2025 MINRES Technologies GmbH
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ *    this list of conditions and the following disclaimer.
+ *
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ *    this list of conditions and the following disclaimer in the documentation
+ *    and/or other materials provided with the distribution.
+ *
+ * 3. Neither the name of the copyright holder nor the names of its contributors
+ *    may be used to endorse or promote products derived from this software
+ *    without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ *
+ * Contributors:
+ *       eyck@minres.com - initial implementation
+ ******************************************************************************/
+
+#ifndef _MEMORY_MEMORY_IF_
+#define _MEMORY_MEMORY_IF_
+
+#include "iss/vm_types.h"
+#include <deque>
+#include <functional>
+#include <limits>
+#include <util/delegate.h>
+
+namespace iss {
+namespace mmio {
+
+using rd_mem_func_sig = iss::status(iss::access_type, uint64_t, unsigned, uint8_t*);
+using wr_mem_func_sig = iss::status(iss::access_type, uint64_t, unsigned, uint8_t const*);
+
+struct memory_if {
+    util::delegate<iss::status(access_type, uint64_t, unsigned, uint8_t*)> rd_mem;
+    util::delegate<iss::status(access_type, uint64_t, unsigned, uint8_t const*)> wr_mem;
+};
+
+struct memory_elem {
+    virtual memory_if get_mem_if() = 0;
+    virtual void set_next(memory_if) = 0;
+    virtual void register_csrs() {}
+    virtual std::tuple<uint64_t, uint64_t> get_range() { return {0, std::numeric_limits<uint64_t>::max()}; }
+};
+
+struct memory_hierarchy {
+    void prepend(memory_elem&);
+    void append(memory_elem&);
+    void insert_before(memory_elem&);
+    void insert_after(memory_elem&);
+    void replace_last(memory_elem&);
+
+protected:
+    void update_chain();
+    std::deque<std::reference_wrapper<memory_elem>> hierarchy;
+};
+
+} // namespace mmio
+} // namespace iss
+#endif

src/iss/mmio/memory_with_htif.h

@@ -0,0 +1,62 @@
+#ifndef _MEMORY_WITH_HTIF_
+#define _MEMORY_WITH_HTIF_
+
+#include "iss/arch/riscv_hart_common.h"
+#include "iss/vm_types.h"
+#include "memory_if.h"
+#include <util/logging.h>
+#include <util/sparse_array.h>
+
+namespace iss {
+namespace mmio {
+template <typename WORD_TYPE> struct memory_with_htif : public memory_elem {
+    using this_class = memory_with_htif<WORD_TYPE>;
+    constexpr static unsigned WORD_LEN = sizeof(WORD_TYPE) * 8;
+
+    memory_with_htif(arch::priv_if<WORD_TYPE> hart_if)
+    : hart_if(hart_if) {}
+
+    ~memory_with_htif() = default;
+
+    memory_if get_mem_if() override {
+        return memory_if{.rd_mem{util::delegate<rd_mem_func_sig>::from<this_class, &this_class::read_mem>(this)},
+                         .wr_mem{util::delegate<wr_mem_func_sig>::from<this_class, &this_class::write_mem>(this)}};
+    }
+
+    void set_next(memory_if) override {
+        // intenrionally left empty, leaf element
+    }
+
+private:
+    iss::status read_mem(iss::access_type access, uint64_t addr, unsigned length, uint8_t* data) {
+        for(auto offs = 0U; offs < length; ++offs) {
+            *(data + offs) = mem[(addr + offs) % mem.size()];
+        }
+        return iss::Ok;
+    }
+
+    iss::status write_mem(iss::access_type access, uint64_t addr, unsigned length, uint8_t const* data) {
+        mem_type::page_type& p = mem(addr / mem.page_size);
+        std::copy(data, data + length, p.data() + (addr & mem.page_addr_mask));
+        // this->tohost handling in case of riscv-test
+        // according to https://github.com/riscv-software-src/riscv-isa-sim/issues/364#issuecomment-607657754:
+        if(access && iss::access_type::FUNC) {
+            if(addr == hart_if.tohost) {
+                return hart_if.exec_htif(data);
+            }
+            if((WORD_LEN == 32 && addr == hart_if.fromhost + 4) || (WORD_LEN == 64 && addr == hart_if.fromhost)) {
+                uint64_t fhostvar = *reinterpret_cast<uint64_t*>(p.data() + (hart_if.fromhost & mem.page_addr_mask));
+                *reinterpret_cast<uint64_t*>(p.data() + (hart_if.tohost & mem.page_addr_mask)) = fhostvar;
+            }
+        }
+        return iss::Ok;
+    }
+
+protected:
+    using mem_type = util::sparse_array<uint8_t, 1ULL << 32>;
+    mem_type mem;
+    arch::priv_if<WORD_TYPE> hart_if;
+};
+} // namespace mmio
+} // namespace iss
+#endif // _MEMORY_WITH_HTIF_

src/iss/mmio/pmp.h

@@ -0,0 +1,212 @@
+
+#include "iss/arch/riscv_hart_common.h"
+#include "iss/vm_types.h"
+#include "memory_if.h"
+#include <util/logging.h>
+
+namespace iss {
+namespace mmio {
+struct clic_config {
+    uint64_t clic_base{0xc0000000};
+    unsigned clic_int_ctl_bits{4};
+    unsigned clic_num_irq{16};
+    unsigned clic_num_trigger{0};
+    bool nmode{false};
+};
+
+inline void read_reg_with_offset(uint32_t reg, uint8_t offs, uint8_t* const data, unsigned length) {
+    auto reg_ptr = reinterpret_cast<uint8_t*>(&reg);
+    switch(offs) {
+    default:
+        for(auto i = 0U; i < length; ++i)
+            *(data + i) = *(reg_ptr + i);
+        break;
+    case 1:
+        for(auto i = 0U; i < length; ++i)
+            *(data + i) = *(reg_ptr + 1 + i);
+        break;
+    case 2:
+        for(auto i = 0U; i < length; ++i)
+            *(data + i) = *(reg_ptr + 2 + i);
+        break;
+    case 3:
+        *data = *(reg_ptr + 3);
+        break;
+    }
+}
+
+inline void write_reg_with_offset(uint32_t& reg, uint8_t offs, const uint8_t* const data, unsigned length) {
+    auto reg_ptr = reinterpret_cast<uint8_t*>(&reg);
+    switch(offs) {
+    default:
+        for(auto i = 0U; i < length; ++i)
+            *(reg_ptr + i) = *(data + i);
+        break;
+    case 1:
+        for(auto i = 0U; i < length; ++i)
+            *(reg_ptr + 1 + i) = *(data + i);
+        break;
+    case 2:
+        for(auto i = 0U; i < length; ++i)
+            *(reg_ptr + 2 + i) = *(data + i);
+        break;
+    case 3:
+        *(reg_ptr + 3) = *data;
+        break;
+    }
+}
+
+template <typename WORD_TYPE> struct pmp : public memory_elem {
+    using this_class = pmp<WORD_TYPE>;
+    using reg_t = WORD_TYPE;
+    constexpr static unsigned WORD_LEN = sizeof(WORD_TYPE) * 8;
+
+    pmp(arch::priv_if<WORD_TYPE> hart_if, clic_config cfg)
+    : hart_if(hart_if)
+    , cfg(cfg) {
+        for(size_t i = arch::pmpaddr0; i <= arch::pmpaddr15; ++i) {
+            hart_if.csr_rd_cb[i] = MK_CSR_RD_CB(read_plain);
+            hart_if.csr_wr_cb[i] = MK_CSR_WR_CB(write_plain);
+        }
+        for(size_t i = arch::pmpcfg0; i < arch::pmpcfg0 + 16 / sizeof(reg_t); ++i) {
+            hart_if.csr_rd_cb[i] = MK_CSR_RD_CB(read_plain);
+            hart_if.csr_wr_cb[i] = MK_CSR_WR_CB(write_pmpcfg);
+        }
+    }
+
+    ~pmp() = default;
+
+    memory_if get_mem_if() override {
+        return memory_if{.rd_mem{util::delegate<rd_mem_func_sig>::from<this_class, &this_class::read_mem>(this)},
+                         .wr_mem{util::delegate<wr_mem_func_sig>::from<this_class, &this_class::write_mem>(this)}};
+    }
+
+    void set_next(memory_if mem) override { down_stream_mem = mem; }
+
+private:
+    iss::status read_mem(iss::access_type access, uint64_t addr, unsigned length, uint8_t* data) {
+        if(!pmp_check(access, addr, length) && !is_debug(access)) {
+            hart_if.fault_data = addr;
+            if(is_debug(access))
+                throw trap_access(0, addr);
+            hart_if.reg.trap_state = (1UL << 31) | ((access == access_type::FETCH ? 1 : 5) << 16); // issue trap 1
+            return iss::Err;
+        }
+        return down_stream_mem.rd_mem(access, addr, length, data);
+    }
+
+    iss::status write_mem(iss::access_type access, uint64_t addr, unsigned length, uint8_t const* data) {
+        if(!pmp_check(access, addr, length) && !is_debug(access)) {
+            hart_if.fault_data = addr;
+            if(is_debug(access))
+                throw trap_access(0, addr);
+            hart_if.reg.trap_state = (1UL << 31) | (7 << 16); // issue trap 1
+            return iss::Err;
+        }
+        return down_stream_mem.wr_mem(access, addr, length, data);
+    }
+
+    iss::status read_plain(unsigned addr, reg_t& val) {
+        val = hart_if.csr[addr];
+        return iss::Ok;
+    }
+
+    iss::status write_plain(unsigned addr, reg_t const& val) {
+        hart_if.csr[addr] = val;
+        return iss::Ok;
+    }
+
+    iss::status write_pmpcfg(unsigned addr, reg_t val) {
+        hart_if.csr[addr] = val & 0x9f9f9f9f;
+        return iss::Ok;
+    }
+
+    bool pmp_check(const access_type type, const uint64_t addr, const unsigned len);
+
+protected:
+    arch::priv_if<WORD_TYPE> hart_if;
+    memory_if down_stream_mem;
+};
+
+template <typename WORD_TYPE> bool pmp<WORD_TYPE>::pmp_check(const access_type type, const uint64_t addr, const unsigned len) {
+    constexpr auto PMP_SHIFT = 2U;
+    constexpr auto PMP_R = 0x1U;
+    constexpr auto PMP_W = 0x2U;
+    constexpr auto PMP_X = 0x4U;
+    constexpr auto PMP_A = 0x18U;
+    constexpr auto PMP_L = 0x80U;
+    constexpr auto PMP_TOR = 0x1U;
+    constexpr auto PMP_NA4 = 0x2U;
+    constexpr auto PMP_NAPOT = 0x3U;
+    reg_t base = 0;
+    auto any_active = false;
+    auto const cfg_reg_size = sizeof(reg_t);
+    for(size_t i = 0; i < 16; i++) {
+        reg_t tor = hart_if.csr[arch::pmpaddr0 + i] << PMP_SHIFT;
+        uint8_t cfg = hart_if.csr[arch::pmpcfg0 + (i / cfg_reg_size)] >> (i % cfg_reg_size);
+        if(cfg & PMP_A) {
+            any_active = true;
+            auto pmp_a = (cfg & PMP_A) >> 3;
+            auto is_tor = pmp_a == PMP_TOR;
+            auto is_na4 = pmp_a == PMP_NA4;
+
+            reg_t mask = (hart_if.csr[arch::pmpaddr0 + i] << 1) | (!is_na4);
+            mask = ~(mask & ~(mask + 1)) << PMP_SHIFT;
+
+            // Check each 4-byte sector of the access
+            auto any_match = false;
+            auto all_match = true;
+            for(reg_t offset = 0; offset < len; offset += 1 << PMP_SHIFT) {
+                reg_t cur_addr = addr + offset;
+                auto napot_match = ((cur_addr ^ tor) & mask) == 0;
+                auto tor_match = base <= (cur_addr + len - 1) && cur_addr < tor;
+                auto match = is_tor ? tor_match : napot_match;
+                any_match |= match;
+                all_match &= match;
+            }
+            if(any_match) {
+                // If the PMP matches only a strict subset of the access, fail it
+                if(!all_match)
+                    return false;
+                return (hart_if.reg.PRIV == arch::PRIV_M && !(cfg & PMP_L)) || (type == access_type::READ && (cfg & PMP_R)) ||
+                       (type == access_type::WRITE && (cfg & PMP_W)) || (type == access_type::FETCH && (cfg & PMP_X));
+            }
+        }
+        base = tor;
+    }
+    //    constexpr auto pmp_num_regs = 16;
+    //    reg_t tor_base = 0;
+    //    auto any_active = false;
+    //    auto lower_addr = addr >>2;
+    //    auto upper_addr = (addr+len-1)>>2;
+    //    for (size_t i = 0; i < pmp_num_regs; i++) {
+    //        uint8_t cfg = csr[pmpcfg0+(i/4)]>>(i%4);
+    //        uint8_t cfg_next = i==(pmp_num_regs-1)? 0 : csr[pmpcfg0+((i+1)/4)]>>((i+1)%4);
+    //        auto pmpaddr = csr[pmpaddr0+i];
+    //        if (cfg & PMP_A) {
+    //            any_active=true;
+    //            auto is_tor = bit_sub<3, 2>(cfg) == PMP_TOR;
+    //            auto is_napot = bit_sub<4, 1>(cfg) && bit_sub<3, 2>(cfg_next)!= PMP_TOR;
+    //            if(is_napot) {
+    //                reg_t mask = bit_sub<3, 1>(cfg)?~( pmpaddr & ~(pmpaddr + 1)): 0x3fffffff;
+    //                auto mpmpaddr = pmpaddr & mask;
+    //                if((lower_addr&mask) == mpmpaddr && (upper_addr&mask)==mpmpaddr)
+    //                    return  (hart_if.reg.PRIV == PRIV_M && !(cfg & PMP_L)) ||
+    //                            (type == access_type::READ && (cfg & PMP_R)) ||
+    //                            (type == access_type::WRITE && (cfg & PMP_W)) ||
+    //                            (type == access_type::FETCH && (cfg & PMP_X));
+    //            } else if(is_tor) {
+    //                if(lower_addr>=tor_base && upper_addr<=pmpaddr)
+    //                    return  (hart_if.reg.PRIV == PRIV_M && !(cfg & PMP_L)) ||
+    //                            (type == access_type::READ && (cfg & PMP_R)) ||
+    //                            (type == access_type::WRITE && (cfg & PMP_W)) ||
+    //                            (type == access_type::FETCH && (cfg & PMP_X));
+    //            }
+    //        }
+    //        tor_base = pmpaddr;
+    //    }
+    return !any_active || hart_if.reg.PRIV == arch::PRIV_M;
+}
+
+} // namespace mmio
+} // namespace iss

src/sysc/core_complex.cpp

@@ -387,7 +387,7 @@ template <unsigned int BUSWIDTH> void core_complex<BUSWIDTH>::run() {
         quantum_keeper.reset();
         cpu->set_interrupt_execution(false);
         cpu->start(dump_ir);
-    } while(cpu->get_interrupt_execution());
+    } while(!cpu->get_interrupt_execution());
     sc_stop();
 }
 

src/sysc/sc_core_adapter.h

@@ -71,44 +71,61 @@ public:
     iss::status write_mem(phys_addr_t addr, unsigned length, const uint8_t* const data) override {
         if(addr.access && iss::access_type::DEBUG)
             return owner->write_mem_dbg(addr.val, length, data) ? iss::Ok : iss::Err;
-        else {
-            auto tohost_upper = (sizeof(reg_t) == 4 && addr.val == (this->tohost + 4)) || (sizeof(reg_t) == 8 && addr.val == this->tohost);
-            auto tohost_lower = (sizeof(reg_t) == 4 && addr.val == this->tohost) || (sizeof(reg_t) == 64 && addr.val == this->tohost);
-            if(tohost_lower || tohost_upper) {
-                if(tohost_upper || (tohost_lower && to_host_wr_cnt > 0)) {
-                    switch(hostvar >> 48) {
-                    case 0:
-                        if(hostvar != 0x1) {
-                            SCCINFO(owner->hier_name())
-                                << "tohost value is 0x" << std::hex << hostvar << std::dec << " (" << hostvar << "), stopping simulation";
-                        } else {
-                            SCCINFO(owner->hier_name())
-                                << "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;
-#ifndef WITH_TCC
-                        throw(iss::simulation_stopped(hostvar));
-#endif
-                        break;
-                    default:
-                        break;
-                    }
-                } else if(tohost_lower)
-                    to_host_wr_cnt++;
-                return iss::Ok;
-            } else {
-                auto res = owner->write_mem(addr.val, length, data) ? iss::Ok : iss::Err;
-                // clear MTIP on mtimecmp write
-                if(addr.val == 0x2004000) {
-                    reg_t val;
-                    this->read_csr(iss::arch::mip, val);
-                    if(val & (1ULL << 7))
-                        this->write_csr(iss::arch::mip, val & ~(1ULL << 7));
+        if(addr.val == this->tohost) {
+            reg_t cur_data = *reinterpret_cast<const reg_t*>(data);
+            // Extract Device (bits 63:56)
+            uint8_t device = sizeof(reg_t) == 4 ? 0 : (cur_data >> 56) & 0xFF;
+            // Extract Command (bits 55:48)
+            uint8_t command = sizeof(reg_t) == 4 ? 0 : (cur_data >> 48) & 0xFF;
+            // Extract payload (bits 47:0)
+            uint64_t payload_addr = cur_data & 0xFFFFFFFFFFFFULL; // 24bits
+            if(payload_addr & 1) {
+                if(payload_addr != 0x1) {
+                    SCCERR(owner->hier_name()) << "tohost value is 0x" << std::hex << payload_addr << std::dec << " (" << payload_addr
+                                               << "), stopping simulation";
+                } else {
+                    SCCINFO(owner->hier_name())
+                        << "tohost value is 0x" << std::hex << payload_addr << std::dec << " (" << payload_addr << "), stopping simulation";
                 }
-                return res;
+                this->reg.trap_state = std::numeric_limits<uint32_t>::max();
+                this->interrupt_sim = payload_addr;
+#ifndef WITH_TCC
+                throw(iss::simulation_stopped(payload_addr));
+#endif
+                return iss::Ok;
             }
+            if(device == 0 && command == 0) {
+                std::array<uint64_t, 8> loaded_payload;
+                auto res = owner->read_mem(payload_addr, 8 * sizeof(uint64_t), reinterpret_cast<uint8_t*>(loaded_payload.data()), false)
+                               ? iss::Ok
+                               : iss::Err;
+                if(res == iss::Err) {
+                    SCCERR(owner->hier_name()) << "Syscall read went wrong";
+                    return iss::Ok;
+                }
+                uint64_t syscall_num = loaded_payload.at(0);
+                if(syscall_num == 64) // SYS_WRITE
+                    return this->execute_sys_write(this, loaded_payload, PLAT::MEM);
+                SCCERR(owner->hier_name()) << "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;
+            }
+            SCCERR(owner->hier_name()) << "tohost functionality not implemented for device " << device << " and command " << command;
+            this->reg.trap_state = std::numeric_limits<uint32_t>::max();
+            this->interrupt_sim = payload_addr;
+            return iss::Ok;
         }
+        auto res = owner->write_mem(addr.val, length, data) ? iss::Ok : iss::Err;
+        // clear MTIP on mtimecmp write
+        if(addr.val == 0x2004000) {
+            reg_t val;
+            this->read_csr(iss::arch::mip, val);
+            if(val & (1ULL << 7))
+                this->write_csr(iss::arch::mip, val & ~(1ULL << 7));
+        }
+        return res;
     }
 
     iss::status read_csr(unsigned addr, reg_t& val) override {
@@ -165,7 +182,6 @@ public:
 private:
     sysc::tgfs::core_complex_if* const owner{nullptr};
     sc_core::sc_event wfi_evt;
-    uint64_t hostvar{std::numeric_limits<uint64_t>::max()};
     unsigned to_host_wr_cnt = 0;
     bool first{true};
 };

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;
@@ -4780,7 +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);
@@ -4792,7 +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())

src/vm/interp/vm_tgc5c.cpp

@@ -263,7 +263,7 @@ private:
                     return iss::Err;
 //            }
         } else {
-            if (this->core.read(phys_addr_t(pc.access, pc.space, pc.val), 4, data) != iss::Ok)
+            if (this->core.read(iss::address_type::PHYSICAL, pc.access, pc.space, pc.val, 4, data) != iss::Ok)
                 return iss::Err;
 
         }
@@ -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;
@@ -2698,11 +2695,12 @@ std::unique_ptr<vm_if> create<arch::tgc5c>(arch::tgc5c *core, unsigned short por
 } // namespace iss
 
 #include <iss/arch/riscv_hart_m_p.h>
+#include <iss/arch/riscv_hart_msu_vp.h>
 #include <iss/arch/riscv_hart_mu_p.h>
 #include <iss/factory.h>
 namespace iss {
 namespace {
-volatile std::array<bool, 2> dummy = {
+volatile std::array<bool, 3> dummy = {
         core_factory::instance().register_creator("tgc5c|m_p|interp", [](unsigned port, void* init_data) -> std::tuple<cpu_ptr, vm_ptr>{
             auto* cpu = new iss::arch::riscv_hart_m_p<iss::arch::tgc5c>();
 		    auto vm = new interp::tgc5c::vm_impl<arch::tgc5c>(*cpu, false);
@@ -2722,8 +2720,18 @@ volatile std::array<bool, 2> dummy = {
                 cpu->set_semihosting_callback(*cb);
             }
             return {cpu_ptr{cpu}, vm_ptr{vm}};
+        }),
+        core_factory::instance().register_creator("tgc5c|mus_vp|interp", [](unsigned port, void* init_data) -> std::tuple<cpu_ptr, vm_ptr>{
+            auto* cpu = new iss::arch::riscv_hart_msu_vp<iss::arch::tgc5c>();
+		    auto vm = new interp::tgc5c::vm_impl<arch::tgc5c>(*cpu, false);
+		    if (port != 0) debugger::server<debugger::gdb_session>::run_server(vm, port);
+            if(init_data){
+                auto* cb = reinterpret_cast<semihosting_cb_t<arch::traits<arch::tgc5c>::reg_t>*>(init_data);
+                cpu->set_semihosting_callback(*cb);
+            }
+            return {cpu_ptr{cpu}, vm_ptr{vm}};
         })
 };
 }
 }
-// clang-format on
+// clang-format on

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);
@@ -4937,7 +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;
@@ -4949,9 +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())
@@ -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