sends iss output to logfile if logfile is specified

vm_base.h needs to be included before gdb_session.h as termios.h (via
boost and gdb_server) has a define which clashes with a variable name in
ConstantRange.h (via iss/llvm/vm_base.h)

CMakeLists.txt

@@ -18,9 +18,10 @@ add_subdirectory(softfloat)
 set(LIB_SOURCES
     src/iss/plugin/instruction_count.cpp
     src/iss/arch/tgc5c.cpp
-    src/iss/mmio/memory_if.cpp
+    src/iss/mem/memory_if.cpp
     src/vm/interp/vm_tgc5c.cpp
     src/vm/fp_functions.cpp
+    src/vm/vector_functions.cpp
     src/iss/debugger/csr_names.cpp
     src/iss/semihosting/semihosting.cpp
 )

contrib/instr/TGC5C_instr.yaml

@@ -20,7 +20,7 @@ RVI:
     mask: 0b00000000000000000000000001111111
     size:   32
     branch:   true
-    delay:   1
+    delay:   [1,1]
   JALR:
     index: 3
     encoding: 0b00000000000000000000000001100111

gen_input/templates/CORENAME.h.gtl

@@ -30,11 +30,21 @@
  *
  *******************************************************************************/
 <%
-def nativeTypeSize(int size){
-    if(size<=8) return 8; else if(size<=16) return 16; else if(size<=32) return 32; else return 64;
+def nativeSize(int size){
+    if(size<=8) return 8;
+    if(size<=16) return 16;
+    if(size<=32) return 32;
+    if(size<=64) return 64;
+    if(size<=128) return 128;
+    if(size<=256) return 256;
+    if(size<=512) return 512;
+    if(size<=1024) return 1024;
+    if(size<=2048) return 2048;
+    if(size<=4096) return 4096;
+    throw new IllegalArgumentException("Unsupported size in nativeSize in CORENAME.h.gtl");
 }
 def getRegisterSizes(){
-    def regs = registers.collect{nativeTypeSize(it.size)}
+    def regs = registers.collect{nativeSize(it.size)}
     regs+=[32,32, 64, 64, 64, 32, 32] // append TRAP_STATE, PENDING_TRAP, ICOUNT, CYCLE, INSTRET, INSTRUCTION, LAST_BRANCH
     return regs
 }
@@ -47,13 +57,7 @@ def getRegisterOffsets(){
     }
     return offsets
 }
-def byteSize(int size){
-    if(size<=8) return 8;
-    if(size<=16) return 16;
-    if(size<=32) return 32;
-    if(size<=64) return 64;
-    return 128;
-}
+
 def getCString(def val){
     return val.toString()+'ULL'
 }
@@ -84,6 +88,8 @@ template <> struct traits<${coreDef.name.toLowerCase()}> {
     enum constants {${constants.collect{c -> c.name+"="+getCString(c.value)}.join(', ')}};
 
     constexpr static unsigned FP_REGS_SIZE = ${constants.find {it.name=='FLEN'}?.value?:0};
+    constexpr static unsigned V_REGS_SIZE = ${constants.find {it.name=='VLEN'}?.value?:0};
+
 
     enum reg_e {
         ${registers.collect{it.name}.join(', ')}, NUM_REGS, TRAP_STATE=NUM_REGS, PENDING_TRAP, ICOUNT, CYCLE, INSTRET, INSTRUCTION, LAST_BRANCH
@@ -142,15 +148,17 @@ struct ${coreDef.name.toLowerCase()}: public arch_if {
 
 #pragma pack(push, 1)
     struct ${coreDef.name}_regs {<%
-        registers.each { reg -> if(reg.size>0) {%> 
-        uint${byteSize(reg.size)}_t ${reg.name} = 0;<%
+        registers.each { reg -> if(reg.size>64) {%>
+        uint8_t ${reg.name}[${reg.size/8}] = {0};<%
+        }else if(reg.size>0) {%>
+        uint${nativeSize(reg.size)}_t ${reg.name} = 0;<%
         }}%>
         uint32_t trap_state = 0, pending_trap = 0;
-        uint64_t icount = 0;
-        uint64_t cycle = 0;
-        uint64_t instret = 0;
-        uint32_t instruction = 0;
-        uint32_t last_branch = 0;
+        uint64_t icount = 0;      // counts number of instructions undisturbed
+        uint64_t cycle = 0;       // counts number of cycles, in functional mode equals icount
+        uint64_t instret = 0;     // counts number of instructions, can be reset via CSR write
+        uint32_t instruction = 0; // holds op code of currently executed instruction
+        uint32_t last_branch = 0; // indicates if last branch was taken
     } reg;
 #pragma pack(pop)
     std::array<address_type, 4> addr_mode;
@@ -164,6 +172,31 @@ if(fcsr != null) {%>
 <%} else { %>
     uint32_t get_fcsr(){return 0;}
     void set_fcsr(uint32_t val){}
+<%}
+def vstart = registers.find {it.name=='vstart'}
+def vl = registers.find {it.name=='vl'}
+def vtype = registers.find {it.name=='vtype'}
+def vxsat = registers.find {it.name=='vxsat'}
+def vxrm = registers.find {it.name=='vxrm'}
+
+if(vtype != null) {%>
+    uint${vstart.size}_t get_vstart(){return reg.vstart;}
+    void set_vstart(uint${vstart.size}_t val){reg.vstart = val;}
+    uint${vl.size}_t get_vl(){return reg.vl;}
+    uint${vtype.size}_t get_vtype(){return reg.vtype;}
+    uint${vxsat.size}_t get_vxsat(){return reg.vxsat;}
+    void set_vxsat(uint${vxsat.size}_t val){reg.vxsat = val;}
+    uint${vxrm.size}_t get_vxrm(){return reg.vxrm;}
+    void set_vxrm(uint${vxrm.size}_t val){reg.vxrm = val;}
+<%} else { %>
+    uint32_t get_vstart(){return 0;}
+    void set_vstart(uint32_t val){}
+    uint32_t get_vl(){return 0;}
+    uint32_t get_vtype(){return 0;}
+    uint32_t get_vxsat(){return 0;}
+    void set_vxsat(uint32_t val){}
+    uint32_t get_vxrm(){return 0;}
+    void set_vxrm(uint32_t val){}
 <%}%>
 };
 

gen_input/templates/asmjit/CORENAME.cpp.gtl

@@ -40,7 +40,11 @@
 #include <iss/instruction_decoder.h>
 <%def fcsr = registers.find {it.name=='FCSR'}
 if(fcsr != null) {%>
-#include <vm/fp_functions.h><%}%>
+#include <vm/fp_functions.h><%}
+def aes = functions.find { it.contains('aes') }
+if(aes != null) {%>
+#include <vm/aes_sbox.h>
+<%}%>
 #ifndef FMT_HEADER_ONLY
 #define FMT_HEADER_ONLY
 #endif
@@ -49,6 +53,22 @@ if(fcsr != null) {%>
 #include <array>
 #include <iss/debugger/riscv_target_adapter.h>
 
+#ifndef _MSC_VER
+using int128_t = __int128;
+using uint128_t = unsigned __int128;
+namespace std {
+template <> struct make_unsigned<__int128> { typedef unsigned __int128 type; };
+template <> class __make_unsigned_selector<__int128 unsigned, false, false> {
+public:
+    typedef unsigned __int128 __type;
+};
+template <> struct is_signed<int128_t> { static constexpr bool value = true; };
+template <> struct is_signed<uint128_t> { static constexpr bool value = false; };
+template <> struct is_unsigned<int128_t> { static constexpr bool value = false; };
+template <> struct is_unsigned<uint128_t> { static constexpr bool value = true; };
+} // namespace std
+#endif
+
 namespace iss {
 namespace asmjit {
 
@@ -116,9 +136,14 @@ protected:
         auto sign_mask = 1ULL<<(W-1);
         return (from & mask) | ((from & sign_mask) ? ~mask : 0);
     }
+    inline void raise(uint16_t trap_id, uint16_t cause){
+        auto trap_val =  0x80ULL << 24 | (cause << 16) | trap_id;
+        this->core.reg.trap_state = trap_val;
+    }
+
 <%functions.each{ it.eachLine { %>
-    ${it}<%}%>
-<%}%>
+    ${it}<%}
+}%>
 private:
     /****************************************************************************
      * start opcode definitions
@@ -166,7 +191,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         <%instr.behavior.eachLine{%>${it}
         <%}%>
@@ -197,7 +221,6 @@ private:
         pc = pc + ((instr & 3) == 3 ? 4 : 2);
         mov(cc, jh.next_pc, pc.val);
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         gen_raise(jh, 0, 2);
         gen_sync(jh, POST_SYNC, instr_descr.size());
         gen_instr_epilogue(jh);
@@ -226,8 +249,6 @@ continuation_e vm_impl<ARCH>::gen_single_inst_behavior(virt_addr_t &pc, jit_hold
     code_word_t instr = 0;
     phys_addr_t paddr(pc);
     auto *const data = (uint8_t *)&instr;
-    if(this->core.has_mmu())
-        paddr = this->core.virt2phys(pc);
     auto res = this->core.read(paddr, 4, data);
     if (res != iss::Ok)
         return ILLEGAL_FETCH;
@@ -245,24 +266,25 @@ template <typename ARCH>
 void vm_impl<ARCH>::gen_instr_prologue(jit_holder& jh) {
     auto& cc = jh.cc;
 
-    cc.comment("//gen_instr_prologue");
-
     x86_reg_t current_trap_state = get_reg_for(cc, traits::TRAP_STATE);
     mov(cc, current_trap_state, get_ptr_for(jh, traits::TRAP_STATE));
     mov(cc, get_ptr_for(jh, traits::PENDING_TRAP), current_trap_state);
+    cc.comment("//Instruction prologue end");
 
 }
 template <typename ARCH>
 void vm_impl<ARCH>::gen_instr_epilogue(jit_holder& jh) {
     auto& cc = jh.cc;
 
-    cc.comment("//gen_instr_epilogue");
+    cc.comment("//Instruction epilogue begin");
     x86_reg_t current_trap_state = get_reg_for(cc, traits::TRAP_STATE);
     mov(cc, current_trap_state, get_ptr_for(jh, traits::TRAP_STATE));
     cmp(cc, current_trap_state, 0);
     cc.jne(jh.trap_entry);
     cc.inc(get_ptr_for(jh, traits::ICOUNT));
     cc.inc(get_ptr_for(jh, traits::CYCLE));
+    cc.comment("//Instruction epilogue end");
+
 }
 template <typename ARCH>
 void vm_impl<ARCH>::gen_block_prologue(jit_holder& jh){
@@ -274,7 +296,7 @@ void vm_impl<ARCH>::gen_block_prologue(jit_holder& jh){
 template <typename ARCH>
 void vm_impl<ARCH>::gen_block_epilogue(jit_holder& jh){
     x86::Compiler& cc = jh.cc;
-    cc.comment("//gen_block_epilogue");
+    cc.comment("//block epilogue begin");
     cc.ret(jh.next_pc);
 
     cc.bind(jh.trap_entry);
@@ -286,7 +308,6 @@ void vm_impl<ARCH>::gen_block_epilogue(jit_holder& jh){
     x86::Gp current_pc = get_reg_for_Gp(cc, traits::PC);
     mov(cc, current_pc, get_ptr_for(jh, traits::PC));
 
-    cc.comment("//enter trap call;");
     InvokeNode* call_enter_trap;
     cc.invoke(&call_enter_trap, &enter_trap, FuncSignature::build<uint64_t, void*, uint64_t, uint64_t, uint64_t>());
     call_enter_trap->setArg(0, jh.arch_if_ptr);
@@ -304,7 +325,6 @@ void vm_impl<ARCH>::gen_block_epilogue(jit_holder& jh){
 template <typename ARCH>
 inline void vm_impl<ARCH>::gen_raise(jit_holder& jh, uint16_t trap_id, uint16_t cause) {
     auto& cc = jh.cc;
-    cc.comment("//gen_raise");
     auto tmp1 = get_reg_for(cc, traits::TRAP_STATE);
     mov(cc, tmp1, 0x80ULL << 24 | (cause << 16) | trap_id);
     mov(cc, get_ptr_for(jh, traits::TRAP_STATE), tmp1);

gen_input/templates/interp/CORENAME.cpp.gtl

@@ -30,6 +30,9 @@
  *
  *******************************************************************************/
 <%
+def floating_point = registers.find {it.name=='FCSR'}
+def vector = registers.find {it.name=='vtype'}
+def aes = functions.find { it.contains('aes') }
 def nativeTypeSize(int size){
     if(size<=8) return 8; else if(size<=16) return 16; else if(size<=32) return 32; else return 64;
 }
@@ -41,7 +44,16 @@ def nativeTypeSize(int size){
 #include <iss/debugger/server.h>
 #include <iss/iss.h>
 #include <iss/interp/vm_base.h>
+<%
+if(floating_point != null) {%>
 #include <vm/fp_functions.h>
+<%}
+if(vector != null) {%>
+#include <vm/vector_functions.h>
+<%}
+if(aes != null) {%>
+#include <vm/aes_sbox.h>
+<%}%>
 #include <util/logging.h>
 #include <boost/coroutine2/all.hpp>
 #include <functional>
@@ -101,10 +113,48 @@ protected:
     using compile_func = compile_ret_t (this_class::*)(virt_addr_t &pc, code_word_t instr);
 
     inline const char *name(size_t index){return traits::reg_aliases.at(index);}
-<%
-def fcsr = registers.find {it.name=='FCSR'}
-if(fcsr != null) {%>
+<% 
+if(floating_point != null) {%>
     inline const char *fname(size_t index){return index < 32?name(index+traits::F0):"illegal";}     
+<%}
+if(vector != null) {%>
+    inline const char* vname(size_t index) { return index < 32 ? name(index + traits::V0) : "illegal"; }
+    inline const char* sew_name(size_t bits) {
+        switch(bits) {
+        case 0b000:
+            return "e8";
+        case 0b001:
+            return "e16";
+        case 0b010:
+            return "e32";
+        case 0b011:
+            return "e64";
+        default:
+            return "illegal";
+        }
+    }
+    inline const char* lmul_name(size_t bits) {
+        switch(bits) {
+        case 0b101:
+            return "mf8";
+        case 0b110:
+            return "mf4";
+        case 0b111:
+            return "mf2";
+        case 0b000:
+            return "m1";
+        case 0b001:
+            return "m2";
+        case 0b010:
+            return "m4";
+        case 0b011:
+            return "m8";
+        default:
+            return "illegal";
+        }
+    }
+    inline const char* ma_name(bool ma) { return ma ? "ma" : "mu"; }
+    inline const char* ta_name(bool ta) { return ta ? "ta" : "tu"; }
 <%}%>
 
     virt_addr_t execute_inst(finish_cond_e cond, virt_addr_t start, uint64_t icount_limit) override;
@@ -127,7 +177,860 @@ if(fcsr != null) {%>
     inline void set_tval(uint64_t new_tval){
         tval = new_tval;
     }
+<%if(vector != null) {
+    def xlen = constants.find { it.name == 'XLEN' }?.value ?: 0
+    def vlen = constants.find { it.name == 'VLEN' }?.value ?: 0 %>
+    inline void lower(){
+        this->core.reg.trap_state = 0;
+    }
+    uint64_t vlseg(uint8_t* V, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint64_t rs1_val, uint8_t width_val, uint8_t segment_size){
+        switch(width_val){
+            case 0b000:
+                return softvector::vector_load_store<${vlen}, uint8_t>(this->get_arch(), softvector::softvec_read, V, vl, vstart, vtype, vm, vd, rs1_val, segment_size);
+            case 0b101:
+                return softvector::vector_load_store<${vlen}, uint16_t>(this->get_arch(), softvector::softvec_read, V, vl, vstart, vtype, vm, vd, rs1_val, segment_size);
+            case 0b110:
+                return softvector::vector_load_store<${vlen}, uint32_t>(this->get_arch(), softvector::softvec_read, V, vl, vstart, vtype, vm, vd, rs1_val, segment_size);
+            case 0b111:
+                return softvector::vector_load_store<${vlen}, uint64_t>(this->get_arch(), softvector::softvec_read, V, vl, vstart, vtype, vm, vd, rs1_val, segment_size);
+            default:
+                throw new std::runtime_error("Unsupported width bit value");
+        }
+    }
+    uint64_t vsseg(uint8_t* V, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint64_t rs1_val, uint8_t width_val, uint8_t segment_size){
+        switch(width_val){
+            case 0b000:
+                return softvector::vector_load_store<${vlen}, uint8_t>(this->get_arch(), softvector::softvec_write, V, vl, vstart, vtype, vm, vd, rs1_val, segment_size);
+            case 0b101:
+                return softvector::vector_load_store<${vlen}, uint16_t>(this->get_arch(), softvector::softvec_write, V, vl, vstart, vtype, vm, vd, rs1_val, segment_size);
+            case 0b110:
+                return softvector::vector_load_store<${vlen}, uint32_t>(this->get_arch(), softvector::softvec_write, V, vl, vstart, vtype, vm, vd, rs1_val, segment_size);
+            case 0b111:
+                return softvector::vector_load_store<${vlen}, uint64_t>(this->get_arch(), softvector::softvec_write, V, vl, vstart, vtype, vm, vd, rs1_val, segment_size);
+            default:
+                throw new std::runtime_error("Unsupported width bit value");
+        }
+    }
+    uint64_t vlsseg(uint8_t* V, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint64_t rs1_val, uint8_t width_val, uint8_t segment_size, int64_t stride){
+        switch(width_val){
+            case 0b000:
+                return softvector::vector_load_store<${vlen}, uint8_t>(this->get_arch(), softvector::softvec_read, V, vl, vstart, vtype, vm, vd, rs1_val, segment_size, stride, true);
+            case 0b101:
+                return softvector::vector_load_store<${vlen}, uint16_t>(this->get_arch(), softvector::softvec_read, V, vl, vstart, vtype, vm, vd, rs1_val, segment_size, stride, true);
+            case 0b110:
+                return softvector::vector_load_store<${vlen}, uint32_t>(this->get_arch(), softvector::softvec_read, V, vl, vstart, vtype, vm, vd, rs1_val, segment_size, stride, true);
+            case 0b111:
+                return softvector::vector_load_store<${vlen}, uint64_t>(this->get_arch(), softvector::softvec_read, V, vl, vstart, vtype, vm, vd, rs1_val, segment_size, stride, true);
+            default:
+                throw new std::runtime_error("Unsupported width bit value");
+        }    
+    }
+    uint64_t vssseg(uint8_t* V, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint64_t rs1_val, uint8_t width_val, uint8_t segment_size, int64_t stride){
+        switch(width_val){
+            case 0b000:
+                return softvector::vector_load_store<${vlen}, uint8_t>(this->get_arch(), softvector::softvec_write, V, vl, vstart, vtype, vm, vd, rs1_val, segment_size, stride, true);
+            case 0b101:
+                return softvector::vector_load_store<${vlen}, uint16_t>(this->get_arch(), softvector::softvec_write, V, vl, vstart, vtype, vm, vd, rs1_val, segment_size, stride, true);
+            case 0b110:
+                return softvector::vector_load_store<${vlen}, uint32_t>(this->get_arch(), softvector::softvec_write, V, vl, vstart, vtype, vm, vd, rs1_val, segment_size, stride, true);
+            case 0b111:
+                return softvector::vector_load_store<${vlen}, uint64_t>(this->get_arch(), softvector::softvec_write, V, vl, vstart, vtype, vm, vd, rs1_val, segment_size, stride, true);
+            default:
+                throw new std::runtime_error("Unsupported width bit value");
+        }    
+    }
 
+    using indexed_load_store_t = std::function<uint64_t(void*, std::function<bool(void*, uint64_t, uint64_t, uint8_t*)>, uint8_t*, uint64_t, uint64_t, softvector::vtype_t, bool, uint8_t, uint64_t, uint8_t, uint8_t)>;
+    template <typename T1, typename T2> indexed_load_store_t getFunction() {
+        return [this](void* core, std::function<uint64_t(void*, uint64_t, uint64_t, uint8_t*)> load_store_fn, uint8_t* V, uint64_t vl,
+                      uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint64_t rs1, uint8_t vs2, uint8_t segment_size) {
+            return softvector::vector_load_store_index<${xlen}, ${vlen}, T1, T2>(core, load_store_fn, V, vl, vstart, vtype, vm, vd, rs1, vs2, segment_size);
+        };
+    }
+
+    const std::array<std::array<indexed_load_store_t, 4>, 4> functionTable = {{
+        {getFunction<uint8_t, uint8_t>(), getFunction<uint8_t, uint16_t>(), getFunction<uint8_t, uint32_t>(), getFunction<uint8_t, uint64_t>()},
+        {getFunction<uint16_t, uint8_t>(), getFunction<uint16_t, uint16_t>(), getFunction<uint16_t, uint32_t>(), getFunction<uint16_t, uint64_t>()},
+        {getFunction<uint32_t, uint8_t>(), getFunction<uint32_t, uint16_t>(), getFunction<uint32_t, uint32_t>(), getFunction<uint32_t, uint64_t>()},
+        {getFunction<uint64_t, uint8_t>(), getFunction<uint64_t, uint16_t>(), getFunction<uint64_t, uint32_t>(), getFunction<uint64_t, uint64_t>()}
+    }};
+    const size_t map_index_size[9] = { 0, 0, 1, 0, 2, 0, 0, 0, 3 }; // translate number of bytes to index in functionTable
+    uint64_t vlxseg(uint8_t* V, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint64_t rs1_val, uint8_t vs2, uint8_t segment_size, uint8_t index_byte_size, uint8_t data_byte_size, bool ordered){
+        return functionTable[map_index_size[index_byte_size]][data_byte_size](this->get_arch(), softvector::softvec_read, V, vl, vstart, vtype, vm, vd, rs1_val, vs2, segment_size);
+    }
+    uint64_t vsxseg(uint8_t* V, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vs3, uint64_t rs1_val, uint8_t vs2, uint8_t segment_size, uint8_t index_byte_size, uint8_t data_byte_size, bool ordered){
+        return functionTable[map_index_size[index_byte_size]][data_byte_size](this->get_arch(), softvector::softvec_write, V, vl, vstart, vtype, vm, vs3, rs1_val, vs2, segment_size);
+    }
+    void vector_vector_op(uint8_t* V, uint8_t funct6, uint8_t funct3, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint8_t vs2, uint8_t vs1, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                return softvector::vector_vector_op<${vlen}, uint8_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b001:
+                return softvector::vector_vector_op<${vlen}, uint16_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b010:
+                return softvector::vector_vector_op<${vlen}, uint32_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b011:
+                return softvector::vector_vector_op<${vlen}, uint64_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1);
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void vector_imm_op(uint8_t* V, uint8_t funct6, uint8_t funct3, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint8_t vs2, int64_t imm, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                return softvector::vector_imm_op<${vlen}, uint8_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b001:
+                return softvector::vector_imm_op<${vlen}, uint16_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b010:
+                return softvector::vector_imm_op<${vlen}, uint32_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b011:
+                return softvector::vector_imm_op<${vlen}, uint64_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, imm);
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void vector_vector_wv(uint8_t* V, uint8_t funct6, uint8_t funct3, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint8_t vs2, uint8_t vs1, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                return softvector::vector_vector_op<${vlen}, uint16_t, uint8_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b001:
+                return softvector::vector_vector_op<${vlen}, uint32_t, uint16_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b010:
+                return softvector::vector_vector_op<${vlen}, uint64_t, uint32_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b011: // would widen to 128 bits
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void vector_imm_wv(uint8_t* V, uint8_t funct6, uint8_t funct3, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint8_t vs2, int64_t imm, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                return softvector::vector_imm_op<${vlen}, uint16_t, uint8_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b001:
+                return softvector::vector_imm_op<${vlen}, uint32_t, uint16_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b010:
+                return softvector::vector_imm_op<${vlen}, uint64_t, uint32_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b011: // would widen to 128 bits
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void vector_vector_ww(uint8_t* V, uint8_t funct6, uint8_t funct3, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint8_t vs2, uint8_t vs1, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                return softvector::vector_vector_op<${vlen}, uint16_t, uint16_t, uint8_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b001:
+                return softvector::vector_vector_op<${vlen}, uint32_t, uint32_t, uint16_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b010:
+                return softvector::vector_vector_op<${vlen}, uint64_t, uint64_t, uint32_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b011: // would widen to 128 bits
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void vector_imm_ww(uint8_t* V, uint8_t funct6, uint8_t funct3, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint8_t vs2, int64_t imm, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                return softvector::vector_imm_op<${vlen}, uint16_t, uint16_t, uint8_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b001:
+                return softvector::vector_imm_op<${vlen}, uint32_t, uint32_t, uint16_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b010:
+                return softvector::vector_imm_op<${vlen}, uint64_t, uint64_t, uint32_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b011: // would widen to 128 bits
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void vector_extend(uint8_t* V, uint8_t unary_op, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint8_t vs2, uint8_t target_sew_pow, uint8_t frac_pow){
+        switch(target_sew_pow){
+            case 4: // uint16_t target
+                if(frac_pow != 1) throw new std::runtime_error("Unsupported frac_pow");
+                return softvector::vector_unary_op<${vlen}, uint16_t, uint8_t>(V, unary_op, vl, vstart, vtype, vm, vd, vs2);
+            case 5: // uint32_t target
+                switch(frac_pow){
+                    case 1:
+                        return softvector::vector_unary_op<${vlen}, uint32_t, uint16_t>(V, unary_op, vl, vstart, vtype, vm, vd, vs2);
+                    case 2:
+                        return softvector::vector_unary_op<${vlen}, uint32_t, uint8_t>(V, unary_op, vl, vstart, vtype, vm, vd, vs2);
+                    default: 
+                        throw new std::runtime_error("Unsupported frac_pow");
+                }
+            case 6: // uint64_t target
+                switch(frac_pow){
+                    case 1:
+                        return softvector::vector_unary_op<${vlen}, uint64_t, uint32_t>(V, unary_op, vl, vstart, vtype, vm, vd, vs2);
+                    case 2:
+                        return softvector::vector_unary_op<${vlen}, uint64_t, uint16_t>(V, unary_op, vl, vstart, vtype, vm, vd, vs2);
+                    case 3:
+                        return softvector::vector_unary_op<${vlen}, uint64_t, uint8_t>(V, unary_op, vl, vstart, vtype, vm, vd, vs2);
+                    default: 
+                        throw new std::runtime_error("Unsupported frac_pow");
+                }
+            default: 
+                throw new std::runtime_error("Unsupported target_sew_pow");
+        }
+    }
+    void vector_vector_carry(uint8_t* V, uint8_t funct6, uint8_t funct3, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, uint8_t vd, uint8_t vs2, uint8_t vs1, uint8_t sew_val, int8_t carry){
+        switch(sew_val){
+            case 0b000:
+                return softvector::vector_vector_carry<${vlen}, uint8_t>(V, funct6, funct3, vl, vstart, vtype, vd, vs2, vs1, carry);
+            case 0b001:
+                return softvector::vector_vector_carry<${vlen}, uint16_t>(V, funct6, funct3, vl, vstart, vtype, vd, vs2, vs1, carry);
+            case 0b010:
+                return softvector::vector_vector_carry<${vlen}, uint32_t>(V, funct6, funct3, vl, vstart, vtype, vd, vs2, vs1, carry);
+            case 0b011:
+                return softvector::vector_vector_carry<${vlen}, uint64_t>(V, funct6, funct3, vl, vstart, vtype, vd, vs2, vs1, carry);
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }    }
+    void vector_imm_carry(uint8_t* V, uint8_t funct6, uint8_t funct3, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, uint8_t vd, uint8_t vs2, int64_t imm, uint8_t sew_val, int8_t carry){
+        switch(sew_val){
+            case 0b000:
+                return softvector::vector_imm_carry<${vlen}, uint8_t>(V, funct6, funct3, vl, vstart, vtype, vd, vs2, imm, carry);
+            case 0b001:
+                return softvector::vector_imm_carry<${vlen}, uint16_t>(V, funct6, funct3, vl, vstart, vtype, vd, vs2, imm, carry);
+            case 0b010:
+                return softvector::vector_imm_carry<${vlen}, uint32_t>(V, funct6, funct3, vl, vstart, vtype, vd, vs2, imm, carry);
+            case 0b011:
+                return softvector::vector_imm_carry<${vlen}, uint64_t>(V, funct6, funct3, vl, vstart, vtype, vd, vs2, imm, carry);
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void carry_vector_vector_op(uint8_t* V, unsigned funct6, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, unsigned vd, unsigned vs2, unsigned vs1, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                return softvector::carry_vector_vector_op<${vlen}, uint8_t>(V, funct6, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b001:
+                return softvector::carry_vector_vector_op<${vlen}, uint16_t>(V, funct6, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b010:
+                return softvector::carry_vector_vector_op<${vlen}, uint32_t>(V, funct6, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b011:
+                return softvector::carry_vector_vector_op<${vlen}, uint64_t>(V, funct6, vl, vstart, vtype, vm, vd, vs2, vs1);
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void carry_vector_imm_op(uint8_t* V, unsigned funct6, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, unsigned vd, unsigned vs2, int64_t imm, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                return softvector::carry_vector_imm_op<${vlen}, uint8_t>(V, funct6, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b001:
+                return softvector::carry_vector_imm_op<${vlen}, uint16_t>(V, funct6, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b010:
+                return softvector::carry_vector_imm_op<${vlen}, uint32_t>(V, funct6, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b011:
+                return softvector::carry_vector_imm_op<${vlen}, uint64_t>(V, funct6, vl, vstart, vtype, vm, vd, vs2, imm);
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void mask_vector_vector_op(uint8_t* V, unsigned funct6, uint8_t funct3, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, unsigned vd, unsigned vs2, unsigned vs1, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                return softvector::mask_vector_vector_op<${vlen}, uint8_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b001:
+                return softvector::mask_vector_vector_op<${vlen}, uint16_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b010:
+                return softvector::mask_vector_vector_op<${vlen}, uint32_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b011:
+                return softvector::mask_vector_vector_op<${vlen}, uint64_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1);
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void mask_vector_imm_op(uint8_t* V, unsigned funct6, uint8_t funct3, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, unsigned vd, unsigned vs2, int64_t imm, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                return softvector::mask_vector_imm_op<${vlen}, uint8_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b001:
+                return softvector::mask_vector_imm_op<${vlen}, uint16_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b010:
+                return softvector::mask_vector_imm_op<${vlen}, uint32_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b011:
+                return softvector::mask_vector_imm_op<${vlen}, uint64_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, imm);
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void vector_vector_vw(uint8_t* V, uint8_t funct6, uint8_t funct3, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint8_t vs2, uint8_t vs1, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                return softvector::vector_vector_op<${vlen}, uint8_t, uint16_t, uint8_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b001:
+                return softvector::vector_vector_op<${vlen}, uint16_t, uint32_t, uint16_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b010:
+                return softvector::vector_vector_op<${vlen}, uint32_t, uint64_t, uint32_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b011: // would require 128 bits vs2 value
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void vector_imm_vw(uint8_t* V, uint8_t funct6, uint8_t funct3, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint8_t vs2, int64_t imm, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                return softvector::vector_imm_op<${vlen}, uint8_t,  uint16_t, uint8_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b001:
+                return softvector::vector_imm_op<${vlen}, uint16_t, uint32_t, uint16_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b010:
+                return softvector::vector_imm_op<${vlen}, uint32_t, uint64_t, uint32_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b011: // would require 128 bits vs2 value
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void vector_vector_merge(uint8_t* V, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint8_t vs2, uint8_t vs1, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                return softvector::vector_vector_merge<${vlen}, uint8_t>(V, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b001:
+                return softvector::vector_vector_merge<${vlen}, uint16_t>(V, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b010:
+                return softvector::vector_vector_merge<${vlen}, uint32_t>(V, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b011:
+                return softvector::vector_vector_merge<${vlen}, uint64_t>(V, vl, vstart, vtype, vm, vd, vs2, vs1);
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void vector_imm_merge(uint8_t* V, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint8_t vs2, int64_t imm, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                return softvector::vector_imm_merge<${vlen}, uint8_t>(V, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b001:
+                return softvector::vector_imm_merge<${vlen}, uint16_t>(V, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b010:
+                return softvector::vector_imm_merge<${vlen}, uint32_t>(V, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b011:
+                return softvector::vector_imm_merge<${vlen}, uint64_t>(V, vl, vstart, vtype, vm, vd, vs2, imm);
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    bool sat_vector_vector_op(uint8_t* V, uint8_t funct6, uint8_t funct3, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, uint64_t vxrm, bool vm, uint8_t vd, uint8_t vs2, uint8_t vs1, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                return softvector::sat_vector_vector_op<${vlen}, uint8_t>(V, funct6, funct3, vl, vstart, vtype, vxrm, vm, vd, vs2, vs1);
+            case 0b001:
+                return softvector::sat_vector_vector_op<${vlen}, uint16_t>(V, funct6, funct3, vl, vstart, vtype, vxrm, vm, vd, vs2, vs1);
+            case 0b010:
+                return softvector::sat_vector_vector_op<${vlen}, uint32_t>(V, funct6, funct3, vl, vstart, vtype, vxrm, vm, vd, vs2, vs1);
+            case 0b011:
+                return softvector::sat_vector_vector_op<${vlen}, uint64_t>(V, funct6, funct3, vl, vstart, vtype, vxrm, vm, vd, vs2, vs1);
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    bool sat_vector_imm_op(uint8_t* V, uint8_t funct6, uint8_t funct3, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, uint64_t vxrm, bool vm, uint8_t vd, uint8_t vs2, int64_t imm, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                return softvector::sat_vector_imm_op<${vlen}, uint8_t>(V, funct6, funct3, vl, vstart, vtype, vxrm, vm, vd, vs2, imm);
+            case 0b001:
+                return softvector::sat_vector_imm_op<${vlen}, uint16_t>(V, funct6, funct3, vl, vstart, vtype, vxrm, vm, vd, vs2, imm);
+            case 0b010:
+                return softvector::sat_vector_imm_op<${vlen}, uint32_t>(V, funct6, funct3, vl, vstart, vtype, vxrm, vm, vd, vs2, imm);
+            case 0b011:
+                return softvector::sat_vector_imm_op<${vlen}, uint64_t>(V, funct6, funct3, vl, vstart, vtype, vxrm, vm, vd, vs2, imm);
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    bool sat_vector_vector_vw(uint8_t* V, uint8_t funct6, uint8_t funct3, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, uint64_t vxrm, bool vm, uint8_t vd, uint8_t vs2, uint8_t vs1, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                return softvector::sat_vector_vector_op<${vlen}, uint8_t, uint16_t>(V, funct6, funct3, vl, vstart, vtype, vxrm, vm, vd, vs2, vs1);
+            case 0b001:
+                return softvector::sat_vector_vector_op<${vlen}, uint16_t, uint32_t>(V, funct6, funct3, vl, vstart, vtype, vxrm, vm, vd, vs2, vs1);
+            case 0b010:
+                return softvector::sat_vector_vector_op<${vlen}, uint32_t, uint64_t>(V, funct6, funct3, vl, vstart, vtype, vxrm, vm, vd, vs2, vs1);
+            case 0b011: // would require 128 bits vs2 value
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    bool sat_vector_imm_vw(uint8_t* V, uint8_t funct6, uint8_t funct3, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, uint64_t vxrm, bool vm, uint8_t vd, uint8_t vs2, int64_t imm, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                return softvector::sat_vector_imm_op<${vlen}, uint8_t, uint16_t>(V, funct6, funct3, vl, vstart, vtype, vxrm, vm, vd, vs2, imm);
+            case 0b001:
+                return softvector::sat_vector_imm_op<${vlen}, uint16_t, uint32_t>(V, funct6, funct3, vl, vstart, vtype, vxrm, vm, vd, vs2, imm);
+            case 0b010:
+                return softvector::sat_vector_imm_op<${vlen}, uint32_t, uint64_t>(V, funct6, funct3, vl, vstart, vtype, vxrm, vm, vd, vs2, imm);
+            case 0b011: // would require 128 bits vs2 value
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void vector_red_op(uint8_t* V, uint8_t funct6, uint8_t funct3, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint8_t vs2, uint8_t vs1, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                return softvector::vector_red_op<${vlen}, uint8_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b001:
+                return softvector::vector_red_op<${vlen}, uint16_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b010:
+                return softvector::vector_red_op<${vlen}, uint32_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b011:
+                return softvector::vector_red_op<${vlen}, uint64_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1);
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void vector_red_wv(uint8_t* V, uint8_t funct6, uint8_t funct3, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint8_t vs2, uint8_t vs1, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                return softvector::vector_red_op<${vlen}, uint16_t, uint8_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b001:
+                return softvector::vector_red_op<${vlen}, uint32_t, uint16_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b010:
+                return softvector::vector_red_op<${vlen}, uint64_t, uint32_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b011: // would require 128 bits vs2 value
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void mask_mask_op(uint8_t* V, unsigned funct6, unsigned funct3, uint64_t vl, uint64_t vstart, unsigned vd, unsigned vs2, unsigned vs1){
+        return softvector::mask_mask_op<${vlen}>(V, funct6, funct3, vl, vstart, vd, vs2, vs1);
+    }
+    uint64_t vcpop(uint8_t* V, uint64_t vl, uint64_t vstart, bool vm, unsigned vs2){
+        return softvector::vcpop<${vlen}>(V, vl, vstart, vm, vs2);
+    }
+    int64_t vfirst(uint8_t* V, uint64_t vl, uint64_t vstart, bool vm, unsigned vs2){
+        return softvector::vfirst<${vlen}>(V, vl, vstart, vm, vs2);
+    }
+    void mask_set_op(uint8_t* V, unsigned enc, uint64_t vl, uint64_t vstart, bool vm, unsigned vd, unsigned vs2){
+        return softvector::mask_set_op<${vlen}>(V, enc, vl, vstart, vm, vd, vs2);
+    }
+    void viota(uint8_t* V, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint8_t vs2, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                return softvector::viota<${vlen}, uint8_t>(V, vl, vstart, vtype, vm, vd, vs2);
+            case 0b001:
+                return softvector::viota<${vlen}, uint16_t>(V, vl, vstart, vtype, vm, vd, vs2);
+            case 0b010:
+                return softvector::viota<${vlen}, uint32_t>(V, vl, vstart, vtype, vm, vd, vs2);
+            case 0b011:
+                return softvector::viota<${vlen}, uint64_t>(V, vl, vstart, vtype, vm, vd, vs2);
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void vid(uint8_t* V, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                return softvector::vid<${vlen}, uint8_t>(V, vl, vstart, vtype, vm, vd);
+            case 0b001:
+                return softvector::vid<${vlen}, uint16_t>(V, vl, vstart, vtype, vm, vd);
+            case 0b010:
+                return softvector::vid<${vlen}, uint32_t>(V, vl, vstart, vtype, vm, vd);
+            case 0b011:
+                return softvector::vid<${vlen}, uint64_t>(V, vl, vstart, vtype, vm, vd);
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void scalar_to_vector(uint8_t* V, softvector::vtype_t vtype, unsigned vd, uint64_t val, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                softvector::scalar_move<${vlen}, uint8_t>(V, vtype, vd, val, true);
+                break;
+            case 0b001:
+                softvector::scalar_move<${vlen}, uint16_t>(V, vtype, vd, val, true);
+                break;
+            case 0b010:
+                softvector::scalar_move<${vlen}, uint32_t>(V, vtype, vd, val, true);
+                break;
+            case 0b011:
+                softvector::scalar_move<${vlen}, uint64_t>(V, vtype, vd, val, true);
+                break;
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    uint64_t scalar_from_vector(uint8_t* V, softvector::vtype_t vtype, unsigned vd, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                return softvector::scalar_move<${vlen}, uint8_t>(V, vtype, vd, 0, false);
+            case 0b001:
+                return softvector::scalar_move<${vlen}, uint16_t>(V, vtype, vd, 0, false);
+            case 0b010:
+                return softvector::scalar_move<${vlen}, uint32_t>(V, vtype, vd, 0, false);
+            case 0b011:
+                return softvector::scalar_move<${vlen}, uint64_t>(V, vtype, vd, 0, false);
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void vector_slideup(uint8_t* V, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, unsigned vd, unsigned vs2, uint64_t imm, uint8_t sew_val) {
+        switch(sew_val){
+            case 0b000:
+                return softvector::vector_slideup<${vlen}, uint8_t>(V, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b001:
+                return softvector::vector_slideup<${vlen}, uint16_t>(V, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b010:
+                return softvector::vector_slideup<${vlen}, uint32_t>(V, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b011:
+                return softvector::vector_slideup<${vlen}, uint64_t>(V, vl, vstart, vtype, vm, vd, vs2, imm);
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void vector_slidedown(uint8_t* V, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, unsigned vd, unsigned vs2, uint64_t imm, uint8_t sew_val) {
+        switch(sew_val){
+            case 0b000:
+                return softvector::vector_slidedown<${vlen}, uint8_t>(V, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b001:
+                return softvector::vector_slidedown<${vlen}, uint16_t>(V, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b010:
+                return softvector::vector_slidedown<${vlen}, uint32_t>(V, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b011:
+                return softvector::vector_slidedown<${vlen}, uint64_t>(V, vl, vstart, vtype, vm, vd, vs2, imm);
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void vector_slide1up(uint8_t* V, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, unsigned vd, unsigned vs2, uint64_t imm, uint8_t sew_val) {
+        switch(sew_val){
+            case 0b000:
+                return softvector::vector_slide1up<${vlen}, uint8_t>(V, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b001:
+                return softvector::vector_slide1up<${vlen}, uint16_t>(V, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b010:
+                return softvector::vector_slide1up<${vlen}, uint32_t>(V, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b011:
+                return softvector::vector_slide1up<${vlen}, uint64_t>(V, vl, vstart, vtype, vm, vd, vs2, imm);
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void vector_slide1down(uint8_t* V, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, unsigned vd, unsigned vs2, uint64_t imm, uint8_t sew_val) {
+        switch(sew_val){
+            case 0b000:
+                return softvector::vector_slide1down<${vlen}, uint8_t>(V, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b001:
+                return softvector::vector_slide1down<${vlen}, uint16_t>(V, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b010:
+                return softvector::vector_slide1down<${vlen}, uint32_t>(V, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b011:
+                return softvector::vector_slide1down<${vlen}, uint64_t>(V, vl, vstart, vtype, vm, vd, vs2, imm);
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void vector_vector_gather(uint8_t* V, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint8_t vs2, uint8_t vs1, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                return softvector::vector_vector_gather<${vlen}, uint8_t>(V, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b001:
+                return softvector::vector_vector_gather<${vlen}, uint16_t>(V, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b010:
+                return softvector::vector_vector_gather<${vlen}, uint32_t>(V, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b011:
+                return softvector::vector_vector_gather<${vlen}, uint64_t>(V, vl, vstart, vtype, vm, vd, vs2, vs1);
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void vector_vector_gatherei16(uint8_t* V, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint8_t vs2, uint8_t vs1, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                return softvector::vector_vector_gather<${vlen}, uint8_t, uint16_t>(V, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b001:
+                return softvector::vector_vector_gather<${vlen}, uint16_t, uint16_t>(V, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b010:
+                return softvector::vector_vector_gather<${vlen}, uint32_t, uint16_t>(V, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b011:
+                return softvector::vector_vector_gather<${vlen}, uint64_t, uint16_t>(V, vl, vstart, vtype, vm, vd, vs2, vs1);
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void vector_imm_gather(uint8_t* V, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint8_t vs2, uint64_t imm, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                return softvector::vector_imm_gather<${vlen}, uint8_t>(V, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b001:
+                return softvector::vector_imm_gather<${vlen}, uint16_t>(V, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b010:
+                return softvector::vector_imm_gather<${vlen}, uint32_t>(V, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b011:
+                return softvector::vector_imm_gather<${vlen}, uint64_t>(V, vl, vstart, vtype, vm, vd, vs2, imm);
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void vector_compress(uint8_t* V, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, uint8_t vd, uint8_t vs2, uint8_t vs1, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                return softvector::vector_compress<${vlen}, uint8_t>(V, vl, vstart, vtype, vd, vs2, vs1);
+            case 0b001:
+                return softvector::vector_compress<${vlen}, uint16_t>(V, vl, vstart, vtype, vd, vs2, vs1);
+            case 0b010:
+                return softvector::vector_compress<${vlen}, uint32_t>(V, vl, vstart, vtype, vd, vs2, vs1);
+            case 0b011:
+                return softvector::vector_compress<${vlen}, uint64_t>(V, vl, vstart, vtype, vd, vs2, vs1);
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void vector_whole_move(uint8_t* V, uint8_t vd, uint8_t vs2, uint8_t count){
+        return softvector::vector_whole_move<${vlen}>(V, vd, vs2, count);
+    }
+    uint64_t fp_scalar_from_vector(uint8_t* V, softvector::vtype_t vtype, unsigned vd, uint8_t sew_val){
+        return scalar_from_vector(V, vtype, vd, sew_val);
+    }
+    void fp_vector_slide1up(uint8_t* V, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, unsigned vd, unsigned vs2, uint64_t imm, uint8_t sew_val) {
+        return vector_slide1up(V, vl, vstart, vtype, vm, vd, vs2, imm, sew_val);
+    }
+    void fp_vector_slide1down(uint8_t* V, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, unsigned vd, unsigned vs2, uint64_t imm, uint8_t sew_val) {
+        return vector_slide1down(V, vl, vstart, vtype, vm, vd, vs2, imm, sew_val);
+    }
+    void fp_vector_red_op(uint8_t* V, uint8_t funct6, uint8_t funct3, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint8_t vs2, uint8_t vs1, uint8_t rm, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                throw new std::runtime_error("Unsupported sew bit value");
+            case 0b001:
+                return softvector::fp_vector_red_op<${vlen}, uint16_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1, rm);
+            case 0b010:
+                return softvector::fp_vector_red_op<${vlen}, uint32_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1, rm);
+            case 0b011:
+                return softvector::fp_vector_red_op<${vlen}, uint64_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1, rm);
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void fp_vector_red_wv(uint8_t* V, uint8_t funct6, uint8_t funct3, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint8_t vs2, uint8_t vs1, uint8_t rm, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                return softvector::fp_vector_red_op<${vlen}, uint16_t, uint8_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1, rm);
+            case 0b001:
+                return softvector::fp_vector_red_op<${vlen}, uint32_t, uint16_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1, rm);
+            case 0b010:
+                return softvector::fp_vector_red_op<${vlen}, uint64_t, uint32_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1, rm);
+            case 0b011: // would require 128 bits vs2 value
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void fp_vector_vector_op(uint8_t* V, uint8_t funct6, uint8_t funct3, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint8_t vs2, uint8_t vs1, uint8_t rm, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                throw new std::runtime_error("Unsupported sew bit value");
+            case 0b001:
+                return softvector::fp_vector_vector_op<${vlen}, uint16_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1, rm);
+            case 0b010:
+                return softvector::fp_vector_vector_op<${vlen}, uint32_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1, rm);
+            case 0b011:
+                return softvector::fp_vector_vector_op<${vlen}, uint64_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1, rm);
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void fp_vector_imm_op(uint8_t* V, uint8_t funct6, uint8_t funct3, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint8_t vs2, uint64_t imm, uint8_t rm, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                throw new std::runtime_error("Unsupported sew bit value");
+            case 0b001:
+                return softvector::fp_vector_imm_op<${vlen}, uint16_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, imm, rm);
+            case 0b010:
+                return softvector::fp_vector_imm_op<${vlen}, uint32_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, imm, rm);
+            case 0b011:
+                return softvector::fp_vector_imm_op<${vlen}, uint64_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, imm, rm);
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void fp_vector_vector_wv(uint8_t* V, uint8_t funct6, uint8_t funct3, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint8_t vs2, uint8_t vs1, uint8_t rm, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                throw new std::runtime_error("Unsupported sew bit value");
+            case 0b001:
+                return softvector::fp_vector_vector_op<${vlen}, uint32_t, uint16_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1, rm);
+            case 0b010:
+                return softvector::fp_vector_vector_op<${vlen}, uint64_t, uint32_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1, rm);
+            case 0b011: // would widen to 128 bits
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void fp_vector_imm_wv(uint8_t* V, uint8_t funct6, uint8_t funct3, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint8_t vs2, uint64_t imm, uint8_t rm, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                throw new std::runtime_error("Unsupported sew bit value");
+            case 0b001:
+                return softvector::fp_vector_imm_op<${vlen}, uint32_t, uint16_t, uint16_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, imm, rm);
+            case 0b010:
+                return softvector::fp_vector_imm_op<${vlen}, uint64_t, uint32_t, uint32_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, imm, rm);
+            case 0b011: // would widen to 128 bits
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void fp_vector_vector_ww(uint8_t* V, uint8_t funct6, uint8_t funct3, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint8_t vs2, uint8_t vs1, uint8_t rm, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                throw new std::runtime_error("Unsupported sew bit value");
+            case 0b001:
+                return softvector::fp_vector_vector_op<${vlen}, uint32_t, uint32_t, uint16_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1, rm);
+            case 0b010:
+                return softvector::fp_vector_vector_op<${vlen}, uint64_t, uint64_t, uint32_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, vs1, rm);
+            case 0b011: // would widen to 128 bits
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void fp_vector_imm_ww(uint8_t* V, uint8_t funct6, uint8_t funct3, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint8_t vs2, uint64_t imm, uint8_t rm, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                throw new std::runtime_error("Unsupported sew bit value");
+            case 0b001:
+                return softvector::fp_vector_imm_op<${vlen}, uint32_t, uint32_t, uint16_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, imm, rm);
+            case 0b010:
+                return softvector::fp_vector_imm_op<${vlen}, uint64_t, uint64_t, uint32_t>(V, funct6, funct3, vl, vstart, vtype, vm, vd, vs2, imm, rm);
+            case 0b011: // would widen to 128 bits
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void fp_vector_unary_op(uint8_t* V, uint8_t encoding_space, uint8_t unary_op, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint8_t vs2, uint8_t rm, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                throw new std::runtime_error("Unsupported sew bit value");
+            case 0b001:
+                return softvector::fp_vector_unary_op<${vlen}, uint16_t>(V, encoding_space, unary_op, vl, vstart, vtype, vm, vd, vs2, rm);
+            case 0b010:
+                return softvector::fp_vector_unary_op<${vlen}, uint32_t>(V, encoding_space, unary_op, vl, vstart, vtype, vm, vd, vs2, rm);
+            case 0b011:
+                return softvector::fp_vector_unary_op<${vlen}, uint64_t>(V, encoding_space, unary_op, vl, vstart, vtype, vm, vd, vs2, rm);
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void mask_fp_vector_vector_op(uint8_t* V, uint8_t funct6, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint8_t vs2, uint8_t vs1, uint8_t rm, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                throw new std::runtime_error("Unsupported sew bit value");
+            case 0b001:
+                return softvector::mask_fp_vector_vector_op<${vlen}, uint16_t>(V, funct6, vl, vstart, vtype, vm, vd, vs2, vs1, rm);
+            case 0b010:
+                return softvector::mask_fp_vector_vector_op<${vlen}, uint32_t>(V, funct6, vl, vstart, vtype, vm, vd, vs2, vs1, rm);
+            case 0b011:
+                return softvector::mask_fp_vector_vector_op<${vlen}, uint64_t>(V, funct6, vl, vstart, vtype, vm, vd, vs2, vs1, rm);
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void mask_fp_vector_imm_op(uint8_t* V, uint8_t funct6, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint8_t vs2, uint64_t imm, uint8_t rm, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                throw new std::runtime_error("Unsupported sew bit value");
+            case 0b001:
+                return softvector::mask_fp_vector_imm_op<${vlen}, uint16_t>(V, funct6, vl, vstart, vtype, vm, vd, vs2, imm, rm);
+            case 0b010:
+                return softvector::mask_fp_vector_imm_op<${vlen}, uint32_t>(V, funct6, vl, vstart, vtype, vm, vd, vs2, imm, rm);
+            case 0b011:
+                return softvector::mask_fp_vector_imm_op<${vlen}, uint64_t>(V, funct6, vl, vstart, vtype, vm, vd, vs2, imm, rm);
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void fp_vector_imm_merge(uint8_t* V, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint8_t vs2, uint64_t imm, uint8_t sew_val){
+        vector_imm_merge(V, vl, vstart, vtype, vm, vd, vs2, imm, sew_val);
+    }
+    void fp_vector_unary_w(uint8_t* V, uint8_t unary_op, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint8_t vs2, uint8_t rm, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                return softvector::fp_vector_unary_w<${vlen}, uint16_t, uint8_t>(V, unary_op, vl, vstart, vtype, vm, vd, vs2, rm);
+            case 0b001:
+                return softvector::fp_vector_unary_w<${vlen}, uint32_t, uint16_t>(V, unary_op, vl, vstart, vtype, vm, vd, vs2, rm);
+            case 0b010:
+                return softvector::fp_vector_unary_w<${vlen}, uint64_t, uint32_t>(V, unary_op, vl, vstart, vtype, vm, vd, vs2, rm);
+            case 0b011: // would widen to 128 bits
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void fp_vector_unary_n(uint8_t* V, uint8_t unary_op, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint8_t vs2, uint8_t rm, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                return softvector::fp_vector_unary_n<${vlen}, uint8_t, uint16_t>(V, unary_op, vl, vstart, vtype, vm, vd, vs2, rm);
+            case 0b001:
+                return softvector::fp_vector_unary_n<${vlen}, uint16_t, uint32_t>(V, unary_op, vl, vstart, vtype, vm, vd, vs2, rm);
+            case 0b010:
+                return softvector::fp_vector_unary_n<${vlen}, uint32_t, uint64_t>(V, unary_op, vl, vstart, vtype, vm, vd, vs2, rm);
+            case 0b011: // would require 128 bit value to narrow
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+    void vector_unary_op(uint8_t* V, uint8_t unary_op, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint8_t vs2, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                return softvector::vector_unary_op<${vlen}, uint8_t>(V, unary_op, vl, vstart, vtype, vm, vd, vs2);
+            case 0b001:
+                return softvector::vector_unary_op<${vlen}, uint16_t>(V, unary_op, vl, vstart, vtype, vm, vd, vs2);
+            case 0b010:
+                return softvector::vector_unary_op<${vlen}, uint32_t>(V, unary_op, vl, vstart, vtype, vm, vd, vs2);
+            case 0b011:
+                return softvector::vector_unary_op<${vlen}, uint64_t>(V, unary_op, vl, vstart, vtype, vm, vd, vs2);
+            default: 
+                throw new std::runtime_error("Unsupported sew_val");
+        }
+    }
+    void vector_vector_crypto(uint8_t* V, uint8_t funct6, uint64_t eg_len, uint64_t eg_start, softvector::vtype_t vtype, uint8_t vd, uint8_t vs2, uint8_t vs1, uint8_t egs){
+        switch(egs){
+            case 4:
+                return softvector::vector_vector_crypto<${vlen}, 4>(V, funct6, eg_len, eg_start, vtype, vd, vs2, vs1);
+            case 8:
+                return softvector::vector_vector_crypto<${vlen}, 8>(V, funct6, eg_len, eg_start, vtype, vd, vs2, vs1);
+            default: 
+                throw new std::runtime_error("Unsupported egs");
+        } 
+    }
+    void vector_scalar_crypto(uint8_t* V, uint8_t funct6, uint64_t eg_len, uint64_t eg_start, softvector::vtype_t vtype, uint8_t vd, uint8_t vs2, uint8_t vs1, uint8_t egs){
+        switch(egs){
+            case 4:
+                return softvector::vector_scalar_crypto<${vlen}, 4>(V, funct6, eg_len, eg_start, vtype, vd, vs2, vs1);
+            case 8:
+                return softvector::vector_scalar_crypto<${vlen}, 8>(V, funct6, eg_len, eg_start, vtype, vd, vs2, vs1);
+            default: 
+                throw new std::runtime_error("Unsupported egs");
+        } 
+    }
+    void vector_imm_crypto(uint8_t* V, uint8_t funct6, uint64_t eg_len, uint64_t eg_start, softvector::vtype_t vtype, uint8_t vd, uint8_t vs2, uint8_t imm, uint8_t egs){
+        switch(egs){
+            case 4:
+                return softvector::vector_imm_crypto<${vlen}, 4>(V, funct6, eg_len, eg_start, vtype, vd, vs2, imm);
+            case 8:
+                return softvector::vector_imm_crypto<${vlen}, 8>(V, funct6, eg_len, eg_start, vtype, vd, vs2, imm);
+            default: 
+                throw new std::runtime_error("Unsupported egs");
+        } 
+    }
+    void vector_crypto(uint8_t* V, uint8_t funct6, uint64_t eg_len, uint64_t eg_start, softvector::vtype_t vtype, uint8_t vd, uint8_t vs2, uint8_t vs1, uint8_t egs, uint8_t sew){
+        switch(egs){
+            case 4:
+                switch(sew){
+                    case 32:
+                        return softvector::vector_crypto<${vlen}, 4, uint32_t>(V, funct6, eg_len, eg_start, vtype, vd, vs2, vs1);
+                    case 64:
+                        return softvector::vector_crypto<${vlen}, 4, uint64_t>(V, funct6, eg_len, eg_start, vtype, vd, vs2, vs1);
+                    default: 
+                        throw new std::runtime_error("Unsupported sew");
+                }
+            case 8:
+                switch(sew){
+                    case 32:
+                        return softvector::vector_crypto<${vlen}, 8, uint32_t>(V, funct6, eg_len, eg_start, vtype, vd, vs2, vs1);
+                    case 64:
+                        return softvector::vector_crypto<${vlen}, 8, uint64_t>(V, funct6, eg_len, eg_start, vtype, vd, vs2, vs1);
+                    default: 
+                        throw new std::runtime_error("Unsupported sew");
+                }            
+            default: 
+                throw new std::runtime_error("Unsupported egs");
+        } 
+    }
+<%}%>
     uint64_t fetch_count{0};
     uint64_t tval{0};
 
@@ -175,22 +1078,8 @@ private:
     decoder instr_decoder;
 
     iss::status fetch_ins(virt_addr_t pc, uint8_t * data){
-        if(this->core.has_mmu()) {
-            auto phys_pc = this->core.virt2phys(pc);
-//            if ((pc.val & upper_bits) != ((pc.val + 2) & upper_bits)) { // we may cross a page boundary
-//                if (this->core.read(phys_pc, 2, data) != iss::Ok) return iss::Err;
-//                if ((data[0] & 0x3) == 0x3) // this is a 32bit instruction
-//                    if (this->core.read(this->core.v2p(pc + 2), 2, data + 2) != iss::Ok)
-//                        return iss::Err;
-//            } else {
-                if (this->core.read(phys_pc, 4, data) != iss::Ok)
-                    return iss::Err;
-//            }
-        } else {
-            if (this->core.read(phys_addr_t(pc.access, pc.space, pc.val), 4, data) != iss::Ok)
-                return iss::Err;
-
-        }
+        if (this->core.read(iss::address_type::PHYSICAL, pc.access, pc.space, pc.val, 4, data) != iss::Ok)
+            return iss::Err;
         return iss::Ok;
     }
 };
@@ -285,8 +1174,8 @@ typename vm_base<ARCH>::virt_addr_t vm_impl<ARCH>::execute_inst(finish_cond_e co
                     // used registers<%instr.usedVariables.each{ k,v->
                     if(v.isArray) {%>
                     auto* ${k} = reinterpret_cast<uint${nativeTypeSize(v.type.size)}_t*>(this->regs_base_ptr+arch::traits<ARCH>::reg_byte_offsets[arch::traits<ARCH>::${k}0]);<% }else{ %> 
-                    auto* ${k} = reinterpret_cast<uint${nativeTypeSize(v.type.size)}_t*>(this->regs_base_ptr+arch::traits<ARCH>::reg_byte_offsets[arch::traits<ARCH>::${k}]);
-                    <%}}%>// calculate next pc value
+                    auto* ${k} = reinterpret_cast<uint${nativeTypeSize(v.type.size)}_t*>(this->regs_base_ptr+arch::traits<ARCH>::reg_byte_offsets[arch::traits<ARCH>::${k}]);<%}}%>
+                    // calculate next pc value
                     *NEXT_PC = *PC + ${instr.length/8};
                     // execute instruction<%instr.behavior.eachLine{%>
                     ${it}<%}%>

gen_input/templates/llvm/CORENAME.cpp.gtl

@@ -31,12 +31,14 @@
  *******************************************************************************/
 // clang-format off
 #include <iss/arch/${coreDef.name.toLowerCase()}.h>
+// vm_base needs to be included before gdb_session as termios.h (via boost and gdb_server) has a define which clashes with a variable
+// name in ConstantRange.h
+#include <iss/llvm/vm_base.h>
+#include <iss/iss.h>
 #include <iss/debugger/gdb_session.h>
 #include <iss/debugger/server.h>
-#include <iss/iss.h>
-#include <iss/llvm/vm_base.h>
-#include <util/logging.h>
 #include <iss/instruction_decoder.h>
+#include <util/logging.h>
 <%def fcsr = registers.find {it.name=='FCSR'}
 if(fcsr != null) {%>
 #include <vm/fp_functions.h><%}%>
@@ -136,7 +138,28 @@ protected:
         return (from & mask) | ((from & sign_mask) ? ~mask : 0);
     }
 <%functions.each{ it.eachLine { %>
-    ${it}<%}%>
+    ${it}<%}
+}
+if(fcsr != null) {%>
+    Value* NaNBox16(BasicBlock* bb, Value* NaNBox16_val){
+        if(static_cast<uint32_t>(traits::FLEN) == 16)
+            return this->gen_ext(NaNBox16_val, traits::FLEN, false);
+        auto box = this->builder.CreateNot((this->gen_ext(0, 32, false)));
+        return this->gen_ext((this->builder.CreateOr(this->builder.CreateShl(this->gen_ext(box, traits::FLEN), 16), this->gen_ext(NaNBox16_val, traits::FLEN))), traits::FLEN, false);
+    }
+    Value* NaNBox32(BasicBlock* bb, Value* NaNBox32_val){
+        if(static_cast<uint32_t>(traits::FLEN) == 32)
+            return this->gen_ext(NaNBox32_val, traits::FLEN, false);
+        auto box = this->builder.CreateNot((this->gen_ext(0, 64, false)));
+        return this->gen_ext((this->builder.CreateOr(this->builder.CreateShl(this->gen_ext(box, traits::FLEN), 32), this->gen_ext(NaNBox32_val, traits::FLEN))), traits::FLEN, false);
+    }
+    Value* NaNBox64(BasicBlock* bb, Value* NaNBox64_val){
+        if(static_cast<uint32_t>(traits::FLEN) == 64)
+            return this->gen_ext(NaNBox64_val, traits::FLEN, false);
+        auto box = this->builder.CreateNot((this->gen_ext(0, 128, false)));
+        return this->gen_ext((this->builder.CreateOr(this->builder.CreateShl(this->gen_ext(box, traits::FLEN), 64), this->gen_ext(NaNBox64_val, traits::FLEN))), traits::FLEN, false);
+    }
+
 <%}%>
 private:
     /****************************************************************************
@@ -251,8 +274,6 @@ vm_impl<ARCH>::gen_single_inst_behavior(virt_addr_t &pc, BasicBlock *this_block)
     // const typename traits::addr_t upper_bits = ~traits::PGMASK;
     phys_addr_t paddr(pc);
     auto *const data = (uint8_t *)&instr;
-    if(this->core.has_mmu())
-        paddr = this->core.virt2phys(pc);
     auto res = this->core.read(paddr, 4, data);
     if (res != iss::Ok) 
         return std::make_tuple(ILLEGAL_FETCH, nullptr);

gen_input/templates/tcc/CORENAME.cpp.gtl

@@ -139,7 +139,34 @@ if(fcsr != null) {%>
     }
 
 <%functions.each{ it.eachLine { %>
-    ${it}<%}%>
+    ${it}<%}
+}
+if(fcsr != null) {%>
+    value NaNBox16(tu_builder& tu, value NaNBox16_val){
+        if(static_cast<uint32_t>(traits::FLEN) == 16)
+            return tu.ext(NaNBox16_val, traits::FLEN, false);
+        else {         
+            auto box = tu.assignment(tu.logical_neg((tu.ext(0, 32, false))), traits::FLEN) ;
+            return tu.ext((tu.bitwise_or(tu.shl(box, 16), NaNBox16_val)), traits::FLEN, false);
+        } 
+    }
+    value NaNBox32(tu_builder& tu, value NaNBox32_val){
+        if(static_cast<uint32_t>(traits::FLEN) == 32)
+            return tu.ext(NaNBox32_val, traits::FLEN, false);
+        else {         
+            auto box = tu.assignment(tu.logical_neg((tu.ext(0, 64, false))), traits::FLEN) ;
+            return tu.ext((tu.bitwise_or(tu.shl(box, 32), NaNBox32_val)), traits::FLEN, false);
+        } 
+    }
+    value NaNBox64(tu_builder& tu, value NaNBox64_val){
+        if(static_cast<uint32_t>(traits::FLEN) == 64)
+            return tu.ext(NaNBox64_val, traits::FLEN, false);
+        else {
+            throw new std::runtime_error("tcc does not support Registers wider than 64 bits");     
+            auto box = tu.assignment(tu.logical_neg((tu.ext(0, 128, false))), traits::FLEN) ;
+            return tu.ext((tu.bitwise_or(tu.shl(box, 64), NaNBox64_val)), traits::FLEN, false);
+        }
+    }
 <%}%>
 private:
     /****************************************************************************
@@ -181,6 +208,8 @@ private:
         this->gen_set_tval(tu, instr);
         <%instr.behavior.eachLine{%>${it}
         <%}%>
+        tu("(*icount)++;");
+        tu("(*instret)++;");
         tu.close_scope();
         vm_base<ARCH>::gen_sync(tu, POST_SYNC,${idx});
         gen_trap_check(tu);        
@@ -232,8 +261,6 @@ vm_impl<ARCH>::gen_single_inst_behavior(virt_addr_t &pc, tu_builder& tu) {
     enum {TRAP_ID=1<<16};
     code_word_t instr = 0;
     phys_addr_t paddr(pc);
-    if(this->core.has_mmu())
-        paddr = this->core.virt2phys(pc);
     auto res = this->core.read(paddr, 4, reinterpret_cast<uint8_t*>(&instr));
     if (res != iss::Ok)
         return ILLEGAL_FETCH;
@@ -278,34 +305,75 @@ template <typename ARCH> void vm_impl<ARCH>::add_prologue(tu_builder& tu){
     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) {%>
+<%if(fcsr != null) {
+    def flen = constants.find { it.name == 'FLEN' }?.value ?: 0
+    %>
     os << "uint32_t (*fget_flags)()=" << (uintptr_t)&fget_flags << ";\\n";
+
+    os << "uint16_t (*fadd_h)(uint16_t v1, uint16_t v2, uint8_t mode)=" << (uintptr_t)&fadd_h << ";\\n";
+    os << "uint16_t (*fsub_h)(uint16_t v1, uint16_t v2, uint8_t mode)=" << (uintptr_t)&fsub_h << ";\\n";
+    os << "uint16_t (*fmul_h)(uint16_t v1, uint16_t v2, uint8_t mode)=" << (uintptr_t)&fmul_h << ";\\n";
+    os << "uint16_t (*fdiv_h)(uint16_t v1, uint16_t v2, uint8_t mode)=" << (uintptr_t)&fdiv_h << ";\\n";
+    os << "uint16_t (*fsqrt_h)(uint16_t v1, uint8_t mode)=" << (uintptr_t)&fsqrt_h << ";\\n";
+    os << "uint16_t (*fcmp_h)(uint16_t v1, uint16_t v2, uint16_t op)=" << (uintptr_t)&fcmp_h << ";\\n";
+    os << "uint16_t (*fmadd_h)(uint16_t v1, uint16_t v2, uint16_t v3, uint16_t op, uint8_t mode)=" << (uintptr_t)&fmadd_h << ";\\n";
+    os << "uint16_t (*fsel_h)(uint16_t v1, uint16_t v2, uint16_t op)=" << (uintptr_t)&fsel_h << ";\\n";
+    os << "uint16_t (*fclass_h)(uint16_t v1)=" << (uintptr_t)&fclass_h << ";\\n";
+    os << "uint16_t (*unbox_h)(uint8_t FLEN, uint64_t v)=" << (uintptr_t)&unbox_h << ";\\n";
+
+    os << "uint32_t (*f16toi32)(uint32_t v1, uint8_t mode)=" << (uintptr_t)&f16toi32 << ";\\n";
+    os << "uint32_t (*f16toui32)(uint32_t v1, uint8_t mode)=" << (uintptr_t)&f16toui32 << ";\\n";
+    os << "uint16_t (*i32tof16)(uint32_t v1, uint8_t mode)=" << (uintptr_t)&i32tof16 << ";\\n";
+    os << "uint16_t (*ui32tof16)(uint32_t v1, uint8_t mode)=" << (uintptr_t)&ui32tof16 << ";\\n";
+    os << "uint64_t (*f16toi64)(uint32_t v1, uint8_t mode)=" << (uintptr_t)&f16toi64 <<";\\n";
+    os << "uint64_t (*f16toui64)(uint32_t v1, uint8_t mode)=" << (uintptr_t)&f16toui64 <<";\\n";
+    os << "uint16_t (*i64tof16)(uint64_t v1, uint8_t mode)=" << (uintptr_t)&i64tof16 <<";\\n";
+    os << "uint16_t (*ui64tof16)(uint64_t v1, uint8_t mode)=" << (uintptr_t)&ui64tof16 <<";\\n";
+
     os << "uint32_t (*fadd_s)(uint32_t v1, uint32_t v2, uint8_t mode)=" << (uintptr_t)&fadd_s << ";\\n";
     os << "uint32_t (*fsub_s)(uint32_t v1, uint32_t v2, uint8_t mode)=" << (uintptr_t)&fsub_s << ";\\n";
     os << "uint32_t (*fmul_s)(uint32_t v1, uint32_t v2, uint8_t mode)=" << (uintptr_t)&fmul_s << ";\\n";
     os << "uint32_t (*fdiv_s)(uint32_t v1, uint32_t v2, uint8_t mode)=" << (uintptr_t)&fdiv_s << ";\\n";
     os << "uint32_t (*fsqrt_s)(uint32_t v1, uint8_t mode)=" << (uintptr_t)&fsqrt_s << ";\\n";
     os << "uint32_t (*fcmp_s)(uint32_t v1, uint32_t v2, uint32_t op)=" << (uintptr_t)&fcmp_s << ";\\n";
-    os << "uint32_t (*fcvt_s)(uint32_t v1, uint32_t op, uint8_t mode)=" << (uintptr_t)&fcvt_s << ";\\n";
     os << "uint32_t (*fmadd_s)(uint32_t v1, uint32_t v2, uint32_t v3, uint32_t op, uint8_t mode)=" << (uintptr_t)&fmadd_s << ";\\n";
     os << "uint32_t (*fsel_s)(uint32_t v1, uint32_t v2, uint32_t op)=" << (uintptr_t)&fsel_s << ";\\n";
     os << "uint32_t (*fclass_s)( uint32_t v1 )=" << (uintptr_t)&fclass_s << ";\\n";
-    os << "uint32_t (*fconv_d2f)(uint64_t v1, uint8_t mode)=" << (uintptr_t)&fconv_d2f << ";\\n";
-    os << "uint64_t (*fconv_f2d)(uint32_t v1, uint8_t mode)=" << (uintptr_t)&fconv_f2d << ";\\n";
+    os << "uint32_t (*unbox_s)(uint8_t FLEN, uint64_t v)=" << (uintptr_t)&unbox_s << ";\\n";
+
+    os << "uint32_t (*f32toi32)(uint32_t v1, uint8_t mode)=" << (uintptr_t)&f32toi32 << ";\\n";
+    os << "uint32_t (*f32toui32)(uint32_t v1, uint8_t mode)=" << (uintptr_t)&f32toui32 << ";\\n";
+    os << "uint32_t (*i32tof32)(uint32_t v1, uint8_t mode)=" << (uintptr_t)&i32tof32 << ";\\n";
+    os << "uint32_t (*ui32tof32)(uint32_t v1, uint8_t mode)=" << (uintptr_t)&ui32tof32 << ";\\n";
+    os << "uint64_t (*f32toi64)(uint32_t v1, uint8_t mode)=" << (uintptr_t)&f32toi64 <<";\\n";
+    os << "uint64_t (*f32toui64)(uint32_t v1, uint8_t mode)=" << (uintptr_t)&f32toui64 <<";\\n";
+    os << "uint32_t (*i64tof32)(uint64_t v1, uint8_t mode)=" << (uintptr_t)&i64tof32 <<";\\n";
+    os << "uint32_t (*ui64tof32)(uint64_t v1, uint8_t mode)=" << (uintptr_t)&ui64tof32 <<";\\n";
+    <%if(flen > 32) {%>
+
     os << "uint64_t (*fadd_d)(uint64_t v1, uint64_t v2, uint8_t mode)=" << (uintptr_t)&fadd_d << ";\\n";
     os << "uint64_t (*fsub_d)(uint64_t v1, uint64_t v2, uint8_t mode)=" << (uintptr_t)&fsub_d << ";\\n";
     os << "uint64_t (*fmul_d)(uint64_t v1, uint64_t v2, uint8_t mode)=" << (uintptr_t)&fmul_d << ";\\n";
     os << "uint64_t (*fdiv_d)(uint64_t v1, uint64_t v2, uint8_t mode)=" << (uintptr_t)&fdiv_d << ";\\n";
     os << "uint64_t (*fsqrt_d)(uint64_t v1, uint8_t mode)=" << (uintptr_t)&fsqrt_d << ";\\n";
     os << "uint64_t (*fcmp_d)(uint64_t v1, uint64_t v2, uint32_t op)=" << (uintptr_t)&fcmp_d << ";\\n";
-    os << "uint64_t (*fcvt_d)(uint64_t v1, uint32_t op, uint8_t mode)=" << (uintptr_t)&fcvt_d << ";\\n";
     os << "uint64_t (*fmadd_d)(uint64_t v1, uint64_t v2, uint64_t v3, uint32_t op, uint8_t mode)=" << (uintptr_t)&fmadd_d << ";\\n";
     os << "uint64_t (*fsel_d)(uint64_t v1, uint64_t v2, uint32_t op)=" << (uintptr_t)&fsel_d << ";\\n";
     os << "uint64_t (*fclass_d)(uint64_t v1  )=" << (uintptr_t)&fclass_d << ";\\n";
-    os << "uint64_t (*fcvt_32_64)(uint32_t v1, uint32_t op, uint8_t mode)=" << (uintptr_t)&fcvt_32_64 << ";\\n";
-    os << "uint32_t (*fcvt_64_32)(uint64_t v1, uint32_t op, uint8_t mode)=" << (uintptr_t)&fcvt_64_32 << ";\\n";
-    os << "uint32_t (*unbox_s)(uint64_t v)=" << (uintptr_t)&unbox_s << ";\\n";
-    <%}%>
+    os << "uint64_t (*unbox_d)(uint8_t FLEN, uint64_t v)=" << (uintptr_t)&unbox_d << ";\\n";
+
+    os << "uint32_t (*f64tof32)(uint64_t v1, uint8_t mode)=" << (uintptr_t)&f64tof32 << ";\\n";
+    os << "uint64_t (*f32tof64)(uint32_t v1, uint8_t mode)=" << (uintptr_t)&f32tof64 << ";\\n";
+    os << "uint64_t (*f64toi64)(uint64_t v1, uint8_t mode)=" << (uintptr_t)&f64toi64 <<";\\n";
+    os << "uint64_t (*f64toui64)(uint64_t v1, uint8_t mode)=" << (uintptr_t)&f64toui64 <<";\\n";
+    os << "uint64_t (*i64tof64)(uint64_t v1, uint8_t mode)=" << (uintptr_t)&i64tof64 <<";\\n";
+    os << "uint64_t (*ui64tof64)(uint64_t v1, uint8_t mode)=" << (uintptr_t)&ui64tof64 <<";\\n";
+    os << "uint64_t (*i32tof64)(uint32_t v1, uint8_t mode)=" << (uintptr_t)&i32tof64 <<";\\n";
+    os << "uint64_t (*ui32tof64)(uint32_t v1, uint8_t mode)=" << (uintptr_t)&ui32tof64 <<";\\n";
+    os << "uint32_t (*f64toi32)(uint64_t v1, uint8_t mode)=" << (uintptr_t)&f64toi32 <<";\\n";
+    os << "uint32_t (*f64toui32)(uint64_t v1, uint8_t mode)=" << (uintptr_t)&f64toui32 <<";\\n";
+    <%}
+    }%>
     tu.add_prologue(os.str());
 }
 

src/elfio.cpp

@@ -1,3 +1,37 @@
+/*******************************************************************************
+ * Copyright (C) 2024 MINRES Technologies GmbH
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ *    this list of conditions and the following disclaimer.
+ *
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ *    this list of conditions and the following disclaimer in the documentation
+ *    and/or other materials provided with the distribution.
+ *
+ * 3. Neither the name of the copyright holder nor the names of its contributors
+ *    may be used to endorse or promote products derived from this software
+ *    without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ *
+ * Contributors:
+ *       eyck@minres.com - initial implementation
+ ******************************************************************************/
+
 #ifdef _MSC_VER
 #define _SCL_SECURE_NO_WARNINGS
 #define ELFIO_NO_INTTYPES
@@ -32,4 +66,4 @@ int main(int argc, char** argv) {
     dump::segment_datas(std::cout, reader);
 
     return 0;
-}
+}

src/iss/arch/riscv_hart_common.h

@@ -35,19 +35,18 @@
 #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/traits.h>
 #include <iss/arch_if.h>
 #include <iss/log_categories.h>
+#include <iss/mem/memory_if.h>
 #include <iss/semihosting/semihosting.h>
+#include <iss/vm_types.h>
 #include <limits>
 #include <sstream>
 #include <string>
@@ -195,26 +194,19 @@ enum riscv_csr {
     dcsr = 0x7B0,
     dpc = 0x7B1,
     dscratch0 = 0x7B2,
-    dscratch1 = 0x7B3
+    dscratch1 = 0x7B3,
+    // vector CSR
+    //  URW
+    vstart = 0x008,
+    vxsat = 0x009,
+    vxrm = 0x00A,
+    vcsr = 0x00F,
+    // URO
+    vl = 0xC20,
+    vtype = 0xC21,
+    vlenb = 0xC22,
 };
 
-enum {
-    PGSHIFT = 12,
-    PTE_PPN_SHIFT = 10,
-    // page table entry (PTE) fields
-    PTE_V = 0x001,   // Valid
-    PTE_R = 0x002,   // Read
-    PTE_W = 0x004,   // Write
-    PTE_X = 0x008,   // Execute
-    PTE_U = 0x010,   // User
-    PTE_G = 0x020,   // Global
-    PTE_A = 0x040,   // Accessed
-    PTE_D = 0x080,   // Dirty
-    PTE_SOFT = 0x300 // Reserved for Software
-};
-
-template <typename T> inline bool PTE_TABLE(T PTE) { return (((PTE) & (PTE_V | PTE_R | PTE_W | PTE_X)) == PTE_V); }
-
 enum { PRIV_U = 0, PRIV_S = 1, PRIV_M = 3, PRIV_D = 4 };
 
 enum {
@@ -233,31 +225,11 @@ enum {
     ISA_U = 1 << 20
 };
 
-struct vm_info {
-    int levels;
-    int idxbits;
-    int ptesize;
-    uint64_t ptbase;
-    bool is_active() { return levels; }
-};
-
-struct feature_config {
-    uint64_t tcm_base{0x10000000};
-    uint64_t tcm_size{0x8000};
-    uint64_t io_address{0xf0000000};
-    uint64_t io_addr_mask{0xf0000000};
-};
-
 class trap_load_access_fault : public trap_access {
 public:
     trap_load_access_fault(uint64_t badaddr)
     : trap_access(5 << 16, badaddr) {}
 };
-class illegal_instruction_fault : public trap_access {
-public:
-    illegal_instruction_fault(uint64_t badaddr)
-    : trap_access(2 << 16, badaddr) {}
-};
 class trap_instruction_page_fault : public trap_access {
 public:
     trap_instruction_page_fault(uint64_t badaddr)
@@ -281,15 +253,18 @@ template <typename WORD_TYPE> struct priv_if {
     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::function<void(uint16_t, uint16_t, WORD_TYPE)> raise_trap; // trap_id, cause, fault_data
     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;
+    hart_state<WORD_TYPE>& state;
+    uint8_t& PRIV;
+    WORD_TYPE& PC;
     uint64_t& tohost;
     uint64_t& fromhost;
-    unsigned& mcause_max_irq;
+    unsigned& max_irq;
 };
 
-template <typename BASE, typename LOGCAT = logging::disass> struct riscv_hart_common : public BASE, public mmio::memory_elem {
+template <typename BASE, typename LOGCAT = logging::disass> struct riscv_hart_common : public BASE, public mem::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
@@ -343,6 +318,19 @@ template <typename BASE, typename LOGCAT = logging::disass> struct riscv_hart_co
             csr_rd_cb[frm] = MK_CSR_RD_CB(read_fcsr);
             csr_wr_cb[frm] = MK_CSR_WR_CB(write_fcsr);
         }
+        if(traits<BASE>::V_REGS_SIZE > 0) {
+            csr_rd_cb[vstart] = MK_CSR_RD_CB(read_vstart);
+            csr_wr_cb[vstart] = MK_CSR_WR_CB(write_vstart);
+            csr_rd_cb[vxsat] = MK_CSR_RD_CB(read_vxsat);
+            csr_wr_cb[vxsat] = MK_CSR_WR_CB(write_vxsat);
+            csr_rd_cb[vxrm] = MK_CSR_RD_CB(read_vxrm);
+            csr_wr_cb[vxrm] = MK_CSR_WR_CB(write_vxrm);
+            csr_rd_cb[vcsr] = MK_CSR_RD_CB(read_vcsr);
+            csr_wr_cb[vcsr] = MK_CSR_WR_CB(write_vcsr);
+            csr_rd_cb[vl] = MK_CSR_RD_CB(read_vl);
+            csr_rd_cb[vtype] = MK_CSR_RD_CB(read_vtype);
+            csr_rd_cb[vlenb] = MK_CSR_RD_CB(read_vlenb);
+        }
         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);
@@ -394,6 +382,7 @@ template <typename BASE, typename LOGCAT = logging::disass> struct riscv_hart_co
             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() << "'";
@@ -418,8 +407,10 @@ template <typename BASE, typename LOGCAT = logging::disass> struct riscv_hart_co
         // Load ELF data
         if(reader.load(name)) {
             // check elf properties
-            if(reader.get_class() != expected_elf_class)
+            if(reader.get_class() != expected_elf_class) {
+                CPPLOG(ERR) << "ISA missmatch, selected XLEN does not match supplied file ";
                 return false;
+            }
             if(reader.get_type() != ELFIO::ET_EXEC)
                 return false;
             if(reader.get_machine() != ELFIO::EM_RISCV)
@@ -456,14 +447,12 @@ template <typename BASE, typename LOGCAT = logging::disass> struct riscv_hart_co
 #endif
                     }
                 }
-                try {
-                    tohost = symbol_table.at("tohost");
-                } catch(std::out_of_range& e) {
-                }
-                try {
-                    fromhost = symbol_table.at("fromhost");
-                } catch(std::out_of_range& e) {
-                }
+                auto to_it = symbol_table.find("tohost");
+                if(to_it != std::end(symbol_table))
+                    tohost = to_it->second;
+                auto from_it = symbol_table.find("tohost");
+                if(from_it != std::end(symbol_table))
+                    tohost = from_it->second;
             }
             return true;
         }
@@ -529,10 +518,10 @@ template <typename BASE, typename LOGCAT = logging::disass> struct riscv_hart_co
             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);
+            return iss::Err;
         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 iss::Err;
         return it->second(addr, val);
     }
 
@@ -541,12 +530,12 @@ template <typename BASE, typename LOGCAT = logging::disass> struct riscv_hart_co
             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);
+            return iss::Err;
         if((addr & 0xc00) == 0xc00) // writing to read-only region
-            throw illegal_instruction_fault(this->fault_data);
+            return iss::Err;
         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 iss::Err;
         return it->second(addr, val);
     }
 
@@ -643,7 +632,7 @@ template <typename BASE, typename LOGCAT = logging::disass> struct riscv_hart_co
 
     iss::status write_dcsr(unsigned addr, reg_t val) {
         if(!debug_mode_active())
-            throw illegal_instruction_fault(this->fault_data);
+            return iss::Err;
         //                  +-------------- ebreakm
         //                  |   +---------- stepi
         //                  |   |  +++----- cause
@@ -654,28 +643,28 @@ template <typename BASE, typename LOGCAT = logging::disass> struct riscv_hart_co
 
     iss::status read_debug(unsigned addr, reg_t& val) {
         if(!debug_mode_active())
-            throw illegal_instruction_fault(this->fault_data);
+            return iss::Err;
         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);
+            return iss::Err;
         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);
+            return iss::Err;
         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);
+            return iss::Err;
         this->reg.DPC = val;
         return iss::Ok;
     }
@@ -714,16 +703,80 @@ template <typename BASE, typename LOGCAT = logging::disass> struct riscv_hart_co
         return iss::Ok;
     }
 
+    iss::status read_vstart(unsigned addr, reg_t& val) {
+        val = this->get_vstart();
+        return iss::Ok;
+    }
+
+    iss::status write_vstart(unsigned addr, reg_t val) {
+        this->set_vstart(val);
+        return iss::Ok;
+    }
+
+    iss::status read_vxsat(unsigned addr, reg_t& val) {
+        val = this->get_vxsat();
+        return iss::Ok;
+    }
+
+    iss::status write_vxsat(unsigned addr, reg_t val) {
+        this->set_vxsat(val & 1);
+        csr[vcsr] = (~1ULL & csr[vcsr]) | (val & 1);
+        return iss::Ok;
+    }
+
+    iss::status read_vxrm(unsigned addr, reg_t& val) {
+        val = this->get_vxrm();
+        return iss::Ok;
+    }
+
+    iss::status write_vxrm(unsigned addr, reg_t val) {
+        this->set_vxrm(val & 0b11);
+        csr[vcsr] = (~0b110ULL & csr[vcsr]) | ((val & 0b11) << 1);
+        return iss::Ok;
+    }
+
+    iss::status read_vcsr(unsigned addr, reg_t& val) {
+        val = csr[vcsr];
+        return iss::Ok;
+    }
+
+    iss::status write_vcsr(unsigned addr, reg_t val) {
+        csr[vcsr] = val;
+        return iss::Ok;
+    }
+
+    iss::status read_vl(unsigned addr, reg_t& val) {
+        val = this->get_vl();
+        return iss::Ok;
+    }
+
+    iss::status read_vtype(unsigned addr, reg_t& val) {
+        val = this->get_vtype();
+        return iss::Ok;
+    }
+
+    iss::status read_vlenb(unsigned addr, reg_t& val) {
+        val = csr[vlenb];
+        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); },
+                              .raise_trap =
+                                  [this](uint16_t trap_id, uint16_t cause, reg_t fault_data) {
+                                      this->reg.trap_state = 0x80ULL << 24 | (cause << 16) | trap_id;
+                                      this->fault_data = fault_data;
+                                  },
                               .csr_rd_cb{this->csr_rd_cb},
                               .csr_wr_cb{csr_wr_cb},
-                              .mstatus{this->state},
+                              .state{this->state},
+                              .PRIV{this->reg.PRIV},
+                              .PC{this->reg.PC},
                               .tohost{this->tohost},
                               .fromhost{this->fromhost},
-                              .mcause_max_irq{mcause_max_irq}};
+                              .max_irq{mcause_max_irq}};
     }
 
     iss::status execute_htif(uint8_t const* data) {
@@ -763,14 +816,14 @@ template <typename BASE, typename LOGCAT = logging::disass> struct riscv_hart_co
         }
     }
 
-    mmio::memory_hierarchy memories;
+    mem::memory_hierarchy memories;
 
-    virtual mmio::memory_if get_mem_if() override {
-        assert(false || "This function should nevver be called");
-        return mmio::memory_if{};
+    mem::memory_if get_mem_if() override {
+        assert(false || "This function should never be called");
+        return mem::memory_if{};
     }
 
-    virtual void set_next(mmio::memory_if mem_if) { memory = mem_if; };
+    void set_next(mem::memory_if mem_if) override { memory = mem_if; };
 
     void set_irq_num(unsigned i) { mcause_max_irq = 1 << util::ilog2(i); }
 
@@ -789,7 +842,7 @@ protected:
         }
     }
 
-    mmio::memory_if memory;
+    mem::memory_if memory;
     struct riscv_instrumentation_if : public iss::instrumentation_if {
 
         riscv_instrumentation_if(riscv_hart_common<BASE, LOGCAT>& arch)
@@ -831,8 +884,7 @@ protected:
 
     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;
+    using csr_type = std::array<typename traits<BASE>::reg_t, 1ULL << 12>;
     csr_type csr;
 
     std::unordered_map<unsigned, rd_csr_f> csr_rd_cb;
@@ -846,7 +898,6 @@ protected:
     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};
 };
 

src/iss/arch/riscv_hart_m_p.h

@@ -48,7 +48,7 @@
 #include <array>
 #include <elfio/elfio.hpp>
 #include <fmt/format.h>
-#include <iss/mmio/memory_with_htif.h>
+#include <iss/mem/memory_with_htif.h>
 #include <unordered_map>
 
 namespace iss {
@@ -111,7 +111,7 @@ public:
         return 0b100010001000; // only machine mode is supported
     }
 
-    riscv_hart_m_p(feature_config cfg = feature_config{});
+    riscv_hart_m_p();
 
     virtual ~riscv_hart_m_p() = default;
 
@@ -128,15 +128,12 @@ public:
 
     void set_csr(unsigned addr, reg_t val) { this->csr[addr & this->csr.page_addr_mask] = val; }
 
-    void set_num_of_irq(unsigned i) { this->mcause_max_irq = 1 << util::ilog2(i); }
-
 protected:
     using mem_read_f = iss::status(phys_addr_t addr, unsigned, uint8_t* const);
     using mem_write_f = iss::status(phys_addr_t addr, unsigned, uint8_t const* const);
 
     hart_state<reg_t> state;
 
-    std::unordered_map<reg_t, uint64_t> ptw;
     std::unordered_map<uint64_t, uint8_t> atomic_reservation;
 
     iss::status read_status(unsigned addr, reg_t& val);
@@ -154,18 +151,15 @@ protected:
     iss::status write_dpc(unsigned addr, reg_t val);
 
     void check_interrupt();
-    feature_config cfg;
-    mmio::memory_with_htif<reg_t> default_mem;
+    mem::memory_with_htif<reg_t> default_mem;
 };
 
 template <typename BASE, features_e FEAT, typename LOGCAT>
-riscv_hart_m_p<BASE, FEAT, LOGCAT>::riscv_hart_m_p(feature_config cfg)
-: cfg(cfg)
-, default_mem(base::get_priv_if()) {
+riscv_hart_m_p<BASE, FEAT, LOGCAT>::riscv_hart_m_p()
+: default_mem(base::get_priv_if()) {
     const std::array<unsigned, 4> rwaddrs{{mepc, mtvec, mscratch, mtval}};
     for(auto addr : rwaddrs) {
         this->csr_rd_cb[addr] = MK_CSR_RD_CB(read_plain);
-        // MK_CSR_RD_CB(read_plain(a,r);};
         this->csr_wr_cb[addr] = MK_CSR_WR_CB(write_plain);
     }
     this->csr_rd_cb[mstatus] = MK_CSR_RD_CB(read_status);
@@ -195,7 +189,7 @@ riscv_hart_m_p<BASE, FEAT, LOGCAT>::riscv_hart_m_p(feature_config cfg)
     }
     this->rd_func = util::delegate<arch_if::rd_func_sig>::from<this_class, &this_class::read>(this);
     this->wr_func = util::delegate<arch_if::wr_func_sig>::from<this_class, &this_class::write>(this);
-    this->memories.prepend(*this);
+    this->memories.root(*this);
     this->memories.append(default_mem);
 }
 
@@ -300,7 +294,7 @@ iss::status riscv_hart_m_p<BASE, FEAT, LOGCAT>::write(const address_type type, c
     try {
         switch(space) {
         case traits<BASE>::MEM: {
-            if(unlikely((access && iss::access_type::FETCH) && (addr & 0x1) == 1)) {
+            if(unlikely(is_fetch(access) && (addr & 0x1) == 1)) {
                 this->fault_data = addr;
                 if(access && iss::access_type::DEBUG)
                     throw trap_access(0, addr);
@@ -403,15 +397,13 @@ template <typename BASE, features_e FEAT, typename LOGCAT> inline void riscv_har
 }
 
 template <typename BASE, features_e FEAT, typename LOGCAT> void riscv_hart_m_p<BASE, FEAT, LOGCAT>::check_interrupt() {
-    // TODO: Implement CLIC functionality
-    // auto ideleg = csr[mideleg];
     // Multiple simultaneous interrupts and traps at the same privilege level are
     // handled in the following decreasing priority order:
     // external interrupts, software interrupts, timer interrupts, then finally
     // any synchronous traps.
     auto ena_irq = this->csr[mip] & this->csr[mie];
 
-    bool mstatus_mie = this->state.mstatus.MIE;
+    bool mstatus_mie = state.mstatus.MIE;
     auto m_enabled = this->reg.PRIV < PRIV_M || mstatus_mie;
     auto enabled_interrupts = m_enabled ? ena_irq : 0;
 
@@ -525,9 +517,6 @@ uint64_t riscv_hart_m_p<BASE, FEAT, LOGCAT>::enter_trap(uint64_t flags, uint64_t
         if((xtvec & 0x1) == 1 && trap_id != 0)
             this->reg.NEXT_PC += 4 * cause;
     }
-    // reset trap state
-    this->reg.PRIV = new_priv;
-    this->reg.trap_state = 0;
     std::array<char, 32> buffer;
 #if defined(_MSC_VER)
     sprintf(buffer.data(), "0x%016llx", addr);
@@ -538,6 +527,9 @@ uint64_t riscv_hart_m_p<BASE, FEAT, LOGCAT>::enter_trap(uint64_t flags, uint64_t
         NSCLOG(INFO, LOGCAT) << (trap_id ? "Interrupt" : "Trap") << " with cause '"
                              << (trap_id ? this->irq_str[cause] : this->trap_str[cause]) << "' (" << cause << ")"
                              << " at address " << buffer.data() << " occurred";
+    // reset trap state
+    this->reg.PRIV = new_priv;
+    this->reg.trap_state = 0;
     return this->reg.NEXT_PC;
 }
 

src/iss/arch/riscv_hart_msu_vp.h

@@ -40,23 +40,25 @@
 #include "riscv_hart_common.h"
 #include "util/logging.h"
 #include <algorithm>
+#include <array>
 #include <cstdint>
 #include <elfio/elf_types.hpp>
+#include <elfio/elfio.hpp>
 #include <limits>
 #ifndef FMT_HEADER_ONLY
 #define FMT_HEADER_ONLY
 #endif
-#include <array>
-#include <elfio/elfio.hpp>
+#include <iss/mem/memory_with_htif.h>
+#include <iss/mem/mmu.h>
 #include <fmt/format.h>
 #include <type_traits>
 #include <unordered_map>
-#include <util/bit_field.h>
 
 namespace iss {
 namespace arch {
 
-template <typename BASE> class riscv_hart_msu_vp : public riscv_hart_common<BASE> {
+template <typename BASE, features_e FEAT = FEAT_NONE, typename LOGCAT = logging::disass>
+class riscv_hart_msu_vp : public riscv_hart_common<BASE> {
 public:
     using core = BASE;
     using base = riscv_hart_common<BASE>;
@@ -65,10 +67,6 @@ public:
     using reg_t = typename core::reg_t;
     using addr_t = typename core::addr_t;
 
-    static constexpr reg_t get_misa() {
-        return (sizeof(reg_t) == 4 ? (1UL << 30) : (2ULL << 62)) | ISA_I | ISA_M | ISA_A | ISA_U | ISA_S | ISA_M;
-    }
-
     static constexpr reg_t get_mstatus_mask(unsigned priv_lvl) {
         if(sizeof(reg_t) == 4) {
 #if __cplusplus < 201402L
@@ -76,21 +74,21 @@ public:
 #else
             switch(priv_lvl) {
             case PRIV_U:
-                return 0x80000011UL; // 0b1000 0000 0000 0000 0000 0000 0001 0001
+                return 0x80000000UL; // 0b1000 0000 0000 0000 0000 0000 0001 0001
             case PRIV_S:
-                return 0x800de133UL; // 0b1000 0000 0000 1101 1110 0001 0011 0011
+                return 0x800de122UL; // 0b1000 0000 0000 1101 1110 0001 0011 0011
             default:
-                return 0x807ff9ddUL; // 0b1000 0000 0111 1111 1111 1001 1011 1011
+                return 0x807ff9aaUL; // 0b1000 0000 0111 1111 1111 1001 1011 1011
             }
 #endif
         } else if(sizeof(reg_t) == 8) {
             switch(priv_lvl) {
             case PRIV_U:
-                return 0x8000000f00000011ULL; // 0b1...0 1111 0000 0000 0111 1111 1111 1001 1011 1011
+                return 0x8000000f00000000ULL; // 0b1...0 1111 0000 0000 0111 1111 1111 1001 1011 1011
             case PRIV_S:
-                return 0x8000000f000de133ULL; // 0b1...0 0011 0000 0000 0000 1101 1110 0001 0011 0011
+                return 0x8000000f000de122ULL; // 0b1...0 0011 0000 0000 0000 1101 1110 0001 0011 0011
             default:
-                return 0x8000000f007ff9ddULL; // 0b1...0 1111 0000 0000 0111 1111 1111 1001 1011 1011
+                return 0x8000000f007ff9aaULL; // 0b1...0 1111 0000 0000 0111 1111 1111 1001 1011 1011
             }
         } else
             assert(false && "Unsupported XLEN value");
@@ -117,48 +115,6 @@ public:
         }
         this->state.mstatus = new_val;
     }
-    reg_t satp;
-
-    static inline vm_info decode_vm_info(uint32_t state, uint32_t sptbr) {
-        if(state == PRIV_M)
-            return {0, 0, 0, 0};
-        if(state <= PRIV_S)
-            switch(bit_sub<31, 1>(sptbr)) {
-            case 0:
-                return {0, 0, 0, 0}; // off
-            case 1:
-                return {2, 10, 4, bit_sub<0, 22>(sptbr) << PGSHIFT}; // SV32
-            default:
-                abort();
-            }
-        abort();
-        return {0, 0, 0, 0}; // dummy
-    }
-
-    static inline vm_info decode_vm_info(uint32_t state, uint64_t sptbr) {
-        if(state == PRIV_M)
-            return {0, 0, 0, 0};
-        if(state <= PRIV_S)
-            switch(bit_sub<60, 4>(sptbr)) {
-            case 0:
-                return {0, 0, 0, 0}; // off
-            case 8:
-                return {3, 9, 8, bit_sub<0, 44>(sptbr) << PGSHIFT}; // SV39
-            case 9:
-                return {4, 9, 8, bit_sub<0, 44>(sptbr) << PGSHIFT}; // SV48
-            case 10:
-                return {5, 9, 8, bit_sub<0, 44>(sptbr) << PGSHIFT}; // SV57
-            case 11:
-                return {6, 9, 8, bit_sub<0, 44>(sptbr) << PGSHIFT}; // SV64
-            default:
-                abort();
-            }
-        abort();
-        return {0, 0, 0, 0}; // dummy
-    }
-
-    const typename core::reg_t PGSIZE = 1 << PGSHIFT;
-    const typename core::reg_t PGMASK = PGSIZE - 1;
 
     constexpr reg_t get_irq_mask(size_t mode) {
         std::array<const reg_t, 4> m = {{
@@ -176,8 +132,6 @@ public:
 
     void reset(uint64_t address) override;
 
-    phys_addr_t virt2phys(const iss::addr_t& addr) override;
-
     iss::status read(const address_type type, const access_type access, const uint32_t space, const uint64_t addr, const unsigned length,
                      uint8_t* const data);
     iss::status write(const address_type type, const access_type access, const uint32_t space, const uint64_t addr, const unsigned length,
@@ -190,22 +144,14 @@ public:
 
     void set_csr(unsigned addr, reg_t val) { this->csr[addr & this->csr.page_addr_mask] = val; }
 
-    void set_irq_num(unsigned i) { this->mcause_max_irq = 1 << util::ilog2(i); }
-
 protected:
-    virtual iss::status read_mem(phys_addr_t addr, unsigned length, uint8_t* const data);
-    virtual iss::status write_mem(phys_addr_t addr, unsigned length, const uint8_t* const data);
+    using mem_read_f = iss::status(phys_addr_t addr, unsigned, uint8_t* const);
+    using mem_write_f = iss::status(phys_addr_t addr, unsigned, uint8_t const* const);
 
     hart_state<reg_t> state;
 
-    using mem_type = util::sparse_array<uint8_t, 1ULL << 32>;
-    mem_type mem;
-    void update_vm_info();
-    std::unordered_map<reg_t, uint64_t> ptw;
     std::unordered_map<uint64_t, uint8_t> atomic_reservation;
 
-    std::vector<uint8_t> tcm;
-
     iss::status read_status(unsigned addr, reg_t& val);
     iss::status write_status(unsigned addr, reg_t val);
     iss::status write_cause(unsigned addr, reg_t val);
@@ -214,26 +160,17 @@ protected:
     iss::status read_ip(unsigned addr, reg_t& val);
     iss::status write_ideleg(unsigned addr, reg_t val);
     iss::status write_edeleg(unsigned addr, reg_t val);
-    iss::status read_satp(unsigned addr, reg_t& val);
-    iss::status write_satp(unsigned addr, reg_t val);
-
-    virtual iss::status read_custom_csr_reg(unsigned addr, reg_t& val) { return iss::status::Err; };
-    virtual iss::status write_custom_csr_reg(unsigned addr, reg_t val) { return iss::status::Err; };
-
-    void register_custom_csr_rd(unsigned addr) { this->csr_rd_cb[addr] = MK_CSR_RD_CB(read_custom_csr_reg); }
-    void register_custom_csr_wr(unsigned addr) { this->csr_wr_cb[addr] = MK_CSR_WR_CB(write_custom_csr_reg); }
-
-    reg_t mhartid_reg{0x0};
 
     void check_interrupt();
+    mem::mmu<reg_t> mmu;
+    mem::memory_with_htif<reg_t> default_mem;
 };
 
-template <typename BASE>
-riscv_hart_msu_vp<BASE>::riscv_hart_msu_vp()
-: state() {
-    this->mmu = true;
-    this->rd_func = util::delegate<arch_if::rd_func_sig>::from<this_class, &this_class::read>(this);
-    this->rd_func = util::delegate<arch_if::wr_func_sig>::from<this_class, &this_class::write>(this);
+template <typename BASE, features_e FEAT, typename LOGCAT>
+riscv_hart_msu_vp<BASE, FEAT, LOGCAT>::riscv_hart_msu_vp()
+: state()
+, mmu(base::get_priv_if())
+, default_mem(base::get_priv_if()) {
     // common regs
     const std::array<unsigned, 17> rwaddrs{{mepc, mtvec, mscratch, mtval, mscratch, sepc, stvec, sscratch, scause, stval, sscratch, uepc,
                                             utvec, uscratch, ucause, utval, uscratch}};
@@ -272,17 +209,32 @@ riscv_hart_msu_vp<BASE>::riscv_hart_msu_vp()
     this->csr_wr_cb[mvendorid] = MK_CSR_WR_CB(write_null);
     this->csr_wr_cb[marchid] = MK_CSR_WR_CB(write_null);
     this->csr_wr_cb[mimpid] = MK_CSR_WR_CB(write_null);
-    this->csr_rd_cb[satp] = MK_CSR_RD_CB(read_satp);
-    this->csr_wr_cb[satp] = MK_CSR_WR_CB(write_satp);
+    this->rd_func = util::delegate<arch_if::rd_func_sig>::from<this_class, &this_class::read>(this);
+    this->wr_func = util::delegate<arch_if::wr_func_sig>::from<this_class, &this_class::write>(this);
+    if(FEAT & FEAT_DEBUG) {
+        this->csr_wr_cb[dscratch0] = MK_CSR_WR_CB(write_dscratch);
+        this->csr_rd_cb[dscratch0] = MK_CSR_RD_CB(read_debug);
+        this->csr_wr_cb[dscratch1] = MK_CSR_WR_CB(write_dscratch);
+        this->csr_rd_cb[dscratch1] = MK_CSR_RD_CB(read_debug);
+        this->csr_wr_cb[dpc] = MK_CSR_WR_CB(write_dpc);
+        this->csr_rd_cb[dpc] = MK_CSR_RD_CB(read_dpc);
+        this->csr_wr_cb[dcsr] = MK_CSR_WR_CB(write_dcsr);
+        this->csr_rd_cb[dcsr] = MK_CSR_RD_CB(read_debug);
+    }
+    this->rd_func = util::delegate<arch_if::rd_func_sig>::from<this_class, &this_class::read>(this);
+    this->wr_func = util::delegate<arch_if::wr_func_sig>::from<this_class, &this_class::write>(this);
+    this->set_next(mmu.get_mem_if());
+    this->memories.root(mmu);
+    this->memories.append(default_mem);
 }
 
-template <typename BASE>
-iss::status riscv_hart_msu_vp<BASE>::read(const address_type type, const access_type access, const uint32_t space, const uint64_t addr,
-                                          const unsigned length, uint8_t* const data) {
+template <typename BASE, features_e FEAT, typename LOGCAT>
+iss::status riscv_hart_msu_vp<BASE, FEAT, LOGCAT>::read(const address_type type, const access_type access, const uint32_t space,
+                                                        const uint64_t addr, const unsigned length, uint8_t* const data) {
 #ifndef NDEBUG
     if(access && iss::access_type::DEBUG) {
         CPPLOG(TRACEALL) << "debug read of " << length << " bytes @addr 0x" << std::hex << addr;
-    } else if(access && iss::access_type::FETCH) {
+    } else if(is_fetch(access)) {
         CPPLOG(TRACEALL) << "fetch of " << length << " bytes  @addr 0x" << std::hex << addr;
     } else {
         CPPLOG(TRACE) << "read of " << length << " bytes  @addr 0x" << std::hex << addr;
@@ -294,31 +246,20 @@ iss::status riscv_hart_msu_vp<BASE>::read(const address_type type, const access_
             auto alignment = is_fetch(access) ? (this->has_compressed() ? 2 : 4) : std::min<unsigned>(length, sizeof(reg_t));
             if(unlikely(is_fetch(access) && (addr & (alignment - 1)))) {
                 this->fault_data = addr;
-                if(access && iss::access_type::DEBUG)
+                if(is_debug(access))
                     throw trap_access(0, addr);
-                this->reg.trap_state = (1 << 31); // issue trap 0
+                this->reg.trap_state = (1UL << 31); // issue trap 0
                 return iss::Err;
             }
             try {
                 if(!is_debug(access) && (addr & (alignment - 1))) {
-                    this->reg.trap_state = 1 << 31 | 4 << 16;
+                    this->reg.trap_state = (1UL << 31) | 4 << 16;
                     this->fault_data = addr;
                     return iss::Err;
                 }
-                if(unlikely((addr & ~PGMASK) != ((addr + length - 1) & ~PGMASK))) { // we may cross a page boundary
-                    vm_info vm = decode_vm_info(this->reg.PRIV, satp);
-                    if(vm.levels != 0) { // VM is active
-                        auto split_addr = (addr + length) & ~PGMASK;
-                        auto len1 = split_addr - addr;
-                        auto res = read(type, access, space, addr, len1, data);
-                        if(res == iss::Ok)
-                            res = read(type, access, space, split_addr, length - len1, data + len1);
-                        return res;
-                    }
-                }
-                auto res = read_mem(virt2phys(iss::addr_t{access, type, space, addr}), length, data);
+                auto res = this->memory.rd_mem(access, addr, length, data);
                 if(unlikely(res != iss::Ok && (access & access_type::DEBUG) == 0)) {
-                    this->reg.trap_state = (1 << 31) | (5 << 16); // issue trap 5 (load access fault
+                    this->reg.trap_state = (1UL << 31) | (5 << 16); // issue trap 5 (load access fault
                     this->fault_data = addr;
                 }
                 return res;
@@ -336,8 +277,6 @@ iss::status riscv_hart_msu_vp<BASE>::read(const address_type type, const access_
             return this->read_csr(addr, *reinterpret_cast<reg_t* const>(data));
         } break;
         case traits<BASE>::FENCE: {
-            if((addr + length) > mem.size())
-                return iss::Err;
             switch(addr) {
             case 2:   // SFENCE:VMA lower
             case 3: { // SFENCE:VMA upper
@@ -372,9 +311,9 @@ iss::status riscv_hart_msu_vp<BASE>::read(const address_type type, const access_
     }
 }
 
-template <typename BASE>
-iss::status riscv_hart_msu_vp<BASE>::write(const address_type type, const access_type access, const uint32_t space, const uint64_t addr,
-                                           const unsigned length, const uint8_t* const data) {
+template <typename BASE, features_e FEAT, typename LOGCAT>
+iss::status riscv_hart_msu_vp<BASE, FEAT, LOGCAT>::write(const address_type type, const access_type access, const uint32_t space,
+                                                         const uint64_t addr, const unsigned length, const uint8_t* const data) {
 #ifndef NDEBUG
     const char* prefix = (access && iss::access_type::DEBUG) ? "debug " : "";
     switch(length) {
@@ -401,29 +340,22 @@ iss::status riscv_hart_msu_vp<BASE>::write(const address_type type, const access
     try {
         switch(space) {
         case traits<BASE>::MEM: {
-            if(unlikely((access && iss::access_type::FETCH) && (addr & 0x1) == 1)) {
+            if(unlikely(is_fetch(access) && (addr & 0x1) == 1)) {
                 this->fault_data = addr;
                 if(access && iss::access_type::DEBUG)
                     throw trap_access(0, addr);
-                this->reg.trap_state = (1 << 31); // issue trap 0
+                this->reg.trap_state = (1UL << 31); // issue trap 0
                 return iss::Err;
             }
-            phys_addr_t paddr = virt2phys(iss::addr_t{access, type, space, addr});
             try {
-                // TODO: There is no check for alignment
-                if(unlikely((addr & ~PGMASK) != ((addr + length - 1) & ~PGMASK))) { // we may cross a page boundary
-                    vm_info vm = decode_vm_info(this->reg.PRIV, satp);
-                    if(vm.levels != 0) { // VM is active
-                        auto split_addr = (addr + length) & ~PGMASK;
-                        auto len1 = split_addr - addr;
-                        auto res = write(type, access, space, addr, len1, data);
-                        if(res == iss::Ok)
-                            res = write(type, access, space, split_addr, length - len1, data + len1);
-                        return res;
-                    }
+                auto alignment = std::min<unsigned>(length, sizeof(reg_t));
+                if(length > 1 && (addr & (alignment - 1)) && !is_debug(access)) {
+                    this->reg.trap_state = (1UL << 31) | 6 << 16;
+                    this->fault_data = addr;
+                    return iss::Err;
                 }
-                auto res = write_mem(paddr, length, data);
-                if(unlikely(res != iss::Ok && (access & access_type::DEBUG) == 0)) {
+                auto res = this->memory.wr_mem(access, addr, length, data);
+                if(unlikely(res != iss::Ok && !is_debug(access))) {
                     this->reg.trap_state = (1UL << 31) | (7UL << 16); // issue trap 7 (Store/AMO access fault)
                     this->fault_data = addr;
                 }
@@ -440,12 +372,9 @@ iss::status riscv_hart_msu_vp<BASE>::write(const address_type type, const access
             return this->write_csr(addr, *reinterpret_cast<const reg_t*>(data));
         } break;
         case traits<BASE>::FENCE: {
-            if((addr + length) > mem.size())
-                return iss::Err;
             switch(addr) {
             case 2:
             case 3: {
-                ptw.clear();
                 auto tvm = state.mstatus.TVM;
                 if(this->reg.PRIV == PRIV_S & tvm != 0) {
                     this->reg.trap_state = (1 << 31) | (2 << 16);
@@ -472,26 +401,29 @@ iss::status riscv_hart_msu_vp<BASE>::write(const address_type type, const access
     }
 }
 
-template <typename BASE> iss::status riscv_hart_msu_vp<BASE>::read_status(unsigned addr, reg_t& val) {
+template <typename BASE, features_e FEAT, typename LOGCAT>
+iss::status riscv_hart_msu_vp<BASE, FEAT, LOGCAT>::read_status(unsigned addr, reg_t& val) {
     auto req_priv_lvl = (addr >> 8) & 0x3;
     val = state.mstatus & get_mstatus_mask(req_priv_lvl);
     return iss::Ok;
 }
 
-template <typename BASE> iss::status riscv_hart_msu_vp<BASE>::write_status(unsigned addr, reg_t val) {
+template <typename BASE, features_e FEAT, typename LOGCAT>
+iss::status riscv_hart_msu_vp<BASE, FEAT, LOGCAT>::write_status(unsigned addr, reg_t val) {
     auto req_priv_lvl = (addr >> 8) & 0x3;
     write_mstatus(val, req_priv_lvl);
     check_interrupt();
-    update_vm_info();
     return iss::Ok;
 }
 
-template <typename BASE> iss::status riscv_hart_msu_vp<BASE>::write_cause(unsigned addr, reg_t val) {
+template <typename BASE, features_e FEAT, typename LOGCAT>
+iss::status riscv_hart_msu_vp<BASE, FEAT, LOGCAT>::write_cause(unsigned addr, reg_t val) {
     this->csr[addr] = val & ((1UL << (traits<BASE>::XLEN - 1)) | 0xf); // TODO: make exception code size configurable
     return iss::Ok;
 }
 
-template <typename BASE> iss::status riscv_hart_msu_vp<BASE>::read_ie(unsigned addr, reg_t& val) {
+template <typename BASE, features_e FEAT, typename LOGCAT>
+iss::status riscv_hart_msu_vp<BASE, FEAT, LOGCAT>::read_ie(unsigned addr, reg_t& val) {
     val = this->csr[mie];
     if(addr < mie)
         val &= this->csr[mideleg];
@@ -500,7 +432,8 @@ template <typename BASE> iss::status riscv_hart_msu_vp<BASE>::read_ie(unsigned a
     return iss::Ok;
 }
 
-template <typename BASE> iss::status riscv_hart_msu_vp<BASE>::write_ie(unsigned addr, reg_t val) {
+template <typename BASE, features_e FEAT, typename LOGCAT>
+iss::status riscv_hart_msu_vp<BASE, FEAT, LOGCAT>::write_ie(unsigned addr, reg_t val) {
     auto req_priv_lvl = (addr >> 8) & 0x3;
     auto mask = get_irq_mask(req_priv_lvl);
     this->csr[mie] = (this->csr[mie] & ~mask) | (val & mask);
@@ -508,7 +441,8 @@ template <typename BASE> iss::status riscv_hart_msu_vp<BASE>::write_ie(unsigned
     return iss::Ok;
 }
 
-template <typename BASE> iss::status riscv_hart_msu_vp<BASE>::read_ip(unsigned addr, reg_t& val) {
+template <typename BASE, features_e FEAT, typename LOGCAT>
+iss::status riscv_hart_msu_vp<BASE, FEAT, LOGCAT>::read_ip(unsigned addr, reg_t& val) {
     val = this->csr[mip];
     if(addr < mip)
         val &= this->csr[mideleg];
@@ -517,111 +451,12 @@ template <typename BASE> iss::status riscv_hart_msu_vp<BASE>::read_ip(unsigned a
     return iss::Ok;
 }
 
-template <typename BASE> iss::status riscv_hart_msu_vp<BASE>::read_satp(unsigned addr, reg_t& val) {
-    reg_t tvm = state.mstatus.TVM;
-    if(this->reg.PRIV == PRIV_S & tvm != 0) {
-        this->reg.trap_state = (1 << 31) | (2 << 16);
-        this->fault_data = this->reg.PC;
-        return iss::Err;
-    }
-    val = satp;
-    return iss::Ok;
-}
-
-template <typename BASE> iss::status riscv_hart_msu_vp<BASE>::write_satp(unsigned addr, reg_t val) {
-    reg_t tvm = state.mstatus.TVM;
-    if(this->reg.PRIV == PRIV_S & tvm != 0) {
-        this->reg.trap_state = (1 << 31) | (2 << 16);
-        this->fault_data = this->reg.PC;
-        return iss::Err;
-    }
-    satp = val;
-    update_vm_info();
-    return iss::Ok;
-}
-template <typename BASE> iss::status riscv_hart_msu_vp<BASE>::read_mem(phys_addr_t paddr, unsigned length, uint8_t* const data) {
-    switch(paddr.val) {
-    default: {
-        for(auto offs = 0U; offs < length; ++offs) {
-            *(data + offs) = mem[(paddr.val + offs) % mem.size()];
-        }
-    }
-    }
-    return iss::Ok;
-}
-
-template <typename BASE> iss::status riscv_hart_msu_vp<BASE>::write_mem(phys_addr_t paddr, unsigned length, const uint8_t* const data) {
-    mem_type::page_type& p = mem(paddr.val / mem.page_size);
-    std::copy(data, data + length, p.data() + (paddr.val & mem.page_addr_mask));
-    // tohost handling in case of riscv-test
-    // according to https://github.com/riscv-software-src/riscv-isa-sim/issues/364#issuecomment-607657754:
-    if(paddr.access && iss::access_type::FUNC) {
-        if(paddr.val == this->tohost) {
-            reg_t cur_data = *reinterpret_cast<const reg_t*>(data);
-            // Extract Device (bits 63:56)
-            uint8_t device = traits<BASE>::XLEN == 32
-                                 ? *reinterpret_cast<uint32_t*>(p.data() + ((this->tohost + 4) & mem.page_addr_mask)) >> 24
-                                 : (cur_data >> 56) & 0xFF;
-            // Extract Command (bits 55:48)
-            uint8_t command = traits<BASE>::XLEN == 32
-                                  ? *reinterpret_cast<uint32_t*>(p.data() + ((this->tohost + 4) & mem.page_addr_mask)) >> 16
-                                  : (cur_data >> 48) & 0xFF;
-            // Extract payload (bits 47:0)
-            uint64_t payload_addr = cur_data & 0xFFFFFFFFFFFFULL;
-            if(payload_addr & 1) {
-                CPPLOG(FATAL) << "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(read(address_type::PHYSICAL, access_type::DEBUG_READ, traits<BASE>::MEM, 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) << "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) << "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;
-            }
-        }
-        if((traits<BASE>::XLEN == 32 && paddr.val == this->fromhost + 4) || (traits<BASE>::XLEN == 64 && paddr.val == this->fromhost)) {
-            uint64_t fhostvar = *reinterpret_cast<uint64_t*>(p.data() + (this->fromhost & mem.page_addr_mask));
-            *reinterpret_cast<uint64_t*>(p.data() + (this->tohost & mem.page_addr_mask)) = fhostvar;
-        }
-    }
-    return iss::Ok;
-}
-
-template <typename BASE> inline void riscv_hart_msu_vp<BASE>::reset(uint64_t address) {
+template <typename BASE, features_e FEAT, typename LOGCAT> inline void riscv_hart_msu_vp<BASE, FEAT, LOGCAT>::reset(uint64_t address) {
     BASE::reset(address);
     state.mstatus = hart_state<reg_t>::mstatus_reset_val;
-    update_vm_info();
 }
 
-template <typename BASE> inline void riscv_hart_msu_vp<BASE>::update_vm_info() {
-    this->vm[1] = decode_vm_info(this->reg.PRIV, satp);
-    BASE::addr_mode[3] = BASE::addr_mode[2] = this->vm[1].is_active() ? iss::address_type::VIRTUAL : iss::address_type::PHYSICAL;
-    if(state.mstatus.MPRV)
-        this->vm[0] = decode_vm_info(state.mstatus.MPP, satp);
-    else
-        this->vm[0] = this->vm[1];
-    BASE::addr_mode[1] = BASE::addr_mode[0] = this->vm[0].is_active() ? iss::address_type::VIRTUAL : iss::address_type::PHYSICAL;
-    ptw.clear();
-}
-
-template <typename BASE> void riscv_hart_msu_vp<BASE>::check_interrupt() {
+template <typename BASE, features_e FEAT, typename LOGCAT> void riscv_hart_msu_vp<BASE, FEAT, LOGCAT>::check_interrupt() {
     auto status = state.mstatus;
     auto ip = this->csr[mip];
     auto ie = this->csr[mie];
@@ -649,101 +484,8 @@ template <typename BASE> void riscv_hart_msu_vp<BASE>::check_interrupt() {
     }
 }
 
-template <typename BASE> typename riscv_hart_msu_vp<BASE>::phys_addr_t riscv_hart_msu_vp<BASE>::virt2phys(const iss::addr_t& addr) {
-    const auto type = addr.access & iss::access_type::FUNC;
-    auto it = ptw.find(addr.val >> PGSHIFT);
-    if(it != ptw.end()) {
-        const reg_t pte = it->second;
-        const reg_t ad = PTE_A | (type == iss::access_type::WRITE) * PTE_D;
-#ifdef RISCV_ENABLE_DIRTY
-        // set accessed and possibly dirty bits.
-        *(uint32_t*)ppte |= ad;
-        return {addr.getAccessType(), addr.space, (pte & (~PGMASK)) | (addr.val & PGMASK)};
-#else
-        // take exception if access or possibly dirty bit is not set.
-        if((pte & ad) == ad)
-            return {addr.access, addr.space, (pte & (~PGMASK)) | (addr.val & PGMASK)};
-        else
-            ptw.erase(it); // throw an exception
-#endif
-    } else {
-        uint32_t mode = type != iss::access_type::FETCH && state.mstatus.MPRV ? // MPRV
-                            state.mstatus.MPP
-                                                                              : this->reg.PRIV;
-
-        const vm_info& vm = this->vm[static_cast<uint16_t>(type) / 2];
-
-        const bool s_mode = mode == PRIV_S;
-        const bool sum = state.mstatus.SUM;
-        const bool mxr = state.mstatus.MXR;
-
-        // verify bits xlen-1:va_bits-1 are all equal
-        const int va_bits = PGSHIFT + vm.levels * vm.idxbits;
-        const reg_t mask = (reg_t(1) << (traits<BASE>::XLEN > -(va_bits - 1))) - 1;
-        const reg_t masked_msbs = (addr.val >> (va_bits - 1)) & mask;
-        const int levels = (masked_msbs != 0 && masked_msbs != mask) ? 0 : vm.levels;
-
-        reg_t base = vm.ptbase;
-        for(int i = levels - 1; i >= 0; i--) {
-            const int ptshift = i * vm.idxbits;
-            const reg_t idx = (addr.val >> (PGSHIFT + ptshift)) & ((1 << vm.idxbits) - 1);
-
-            // check that physical address of PTE is legal
-            reg_t pte = 0;
-            const uint8_t res = this->read(iss::address_type::PHYSICAL, addr.access, traits<BASE>::MEM, base + idx * vm.ptesize, vm.ptesize,
-                                           (uint8_t*)&pte);
-            if(res != 0)
-                throw trap_load_access_fault(addr.val);
-            const reg_t ppn = pte >> PTE_PPN_SHIFT;
-
-            if(PTE_TABLE(pte)) { // next level of page table
-                base = ppn << PGSHIFT;
-            } else if((pte & PTE_U) ? s_mode && (type == iss::access_type::FETCH || !sum) : !s_mode) {
-                break;
-            } else if(!(pte & PTE_V) || (!(pte & PTE_R) && (pte & PTE_W))) {
-                break;
-            } else if(type == (type == iss::access_type::FETCH  ? !(pte & PTE_X)
-                               : type == iss::access_type::READ ? !(pte & PTE_R) && !(mxr && (pte & PTE_X))
-                                                                : !((pte & PTE_R) && (pte & PTE_W)))) {
-                break;
-            } else if((ppn & ((reg_t(1) << ptshift) - 1)) != 0) {
-                break;
-            } else {
-                const reg_t ad = PTE_A | ((type == iss::access_type::WRITE) * PTE_D);
-#ifdef RISCV_ENABLE_DIRTY
-                // set accessed and possibly dirty bits.
-                *(uint32_t*)ppte |= ad;
-#else
-                // take exception if access or possibly dirty bit is not set.
-                if((pte & ad) != ad)
-                    break;
-#endif
-                // for superpage mappings, make a fake leaf PTE for the TLB's benefit.
-                const reg_t vpn = addr.val >> PGSHIFT;
-                const reg_t value = (ppn | (vpn & ((reg_t(1) << ptshift) - 1))) << PGSHIFT;
-                const reg_t offset = addr.val & PGMASK;
-                ptw[vpn] = value | (pte & 0xff);
-                return {addr.access, addr.space, value | offset};
-            }
-        }
-    }
-    switch(type) {
-    case access_type::FETCH:
-        this->fault_data = addr.val;
-        throw trap_instruction_page_fault(addr.val);
-    case access_type::READ:
-        this->fault_data = addr.val;
-        throw trap_load_page_fault(addr.val);
-    case access_type::WRITE:
-        this->fault_data = addr.val;
-        throw trap_store_page_fault(addr.val);
-    default:
-        abort();
-    }
-}
-
-template <typename BASE> uint64_t riscv_hart_msu_vp<BASE>::enter_trap(uint64_t flags, uint64_t addr, uint64_t instr) {
-    auto cur_priv = this->reg.PRIV;
+template <typename BASE, features_e FEAT, typename LOGCAT>
+uint64_t riscv_hart_msu_vp<BASE, FEAT, LOGCAT>::enter_trap(uint64_t flags, uint64_t addr, uint64_t instr) {
     // flags are ACTIVE[31:31], CAUSE[30:16], TRAPID[15:0]
     // calculate and write mcause val
     if(flags == std::numeric_limits<uint64_t>::max())
@@ -751,11 +493,11 @@ template <typename BASE> uint64_t riscv_hart_msu_vp<BASE>::enter_trap(uint64_t f
     auto trap_id = bit_sub<0, 16>(flags);
     auto cause = bit_sub<16, 15>(flags);
     if(trap_id == 0 && cause == 11)
-        cause = 0x8 + cur_priv; // adjust environment call cause
+        cause = 0x8 + this->reg.PRIV; // adjust environment call cause
     // calculate effective privilege level
-    auto new_priv = PRIV_M;
+    unsigned new_priv = PRIV_M;
     if(trap_id == 0) { // exception
-        if(cur_priv != PRIV_M && ((this->csr[medeleg] >> cause) & 0x1) != 0)
+        if(this->reg.PRIV != PRIV_M && ((this->csr[medeleg] >> cause) & 0x1) != 0)
             new_priv = (this->csr[sedeleg] >> cause) & 0x1 ? PRIV_U : PRIV_S;
         // store ret addr in xepc register
         this->csr[uepc | (new_priv << 8)] = static_cast<reg_t>(addr); // store actual address instruction of exception
@@ -771,21 +513,23 @@ template <typename BASE> uint64_t riscv_hart_msu_vp<BASE>::enter_trap(uint64_t f
             this->csr[utval | (new_priv << 8)] = static_cast<reg_t>(addr);
             break;
         case 2:
-            this->csr[utval | (new_priv << 8)] = (instr & 0x3) == 3 ? instr : instr & 0xffff;
+            this->csr[utval | (new_priv << 8)] = (!this->has_compressed() || (instr & 0x3) == 3) ? instr : instr & 0xffff;
             break;
         case 3:
-            // TODO: implement debug mode behavior
-            // csr[dpc] = addr;
-            // csr[dcsr] = (csr[dcsr] & ~0x1c3) | (1<<6) | PRIV_M; //FIXME: cause should not be 4 (stepi)
-            this->csr[utval | (new_priv << 8)] = addr;
+            if((FEAT & FEAT_DEBUG) && (this->csr[dcsr] & 0x8000)) {
+                this->reg.DPC = addr;
+                this->csr[dcsr] = (this->csr[dcsr] & ~0x1c3) | (1 << 6) | PRIV_M; // FIXME: cause should not be 4 (stepi)
+                new_priv = this->reg.PRIV | PRIV_D;
+            } else {
+                this->csr[utval | (new_priv << 8)] = addr;
+            }
             if(this->semihosting_cb) {
                 // Check for semihosting call
-                phys_addr_t p_addr(access_type::DEBUG_READ, traits<BASE>::MEM, addr - 4);
                 std::array<uint8_t, 8> data;
                 // check for SLLI_X0_X0_0X1F and SRAI_X0_X0_0X07
-                this->read_mem(p_addr, 4, data.data());
-                p_addr.val += 8;
-                this->read_mem(p_addr, 4, data.data() + 4);
+                this->memory.rd_mem(iss::access_type::DEBUG_READ, addr - 4, 4, data.data());
+                addr += 8;
+                this->memory.rd_mem(iss::access_type::DEBUG_READ, addr - 4, 4, data.data() + 4);
 
                 const std::array<uint8_t, 8> ref_data = {0x13, 0x10, 0xf0, 0x01, 0x13, 0x50, 0x70, 0x40};
                 if(data == ref_data) {
@@ -797,7 +541,7 @@ template <typename BASE> uint64_t riscv_hart_msu_vp<BASE>::enter_trap(uint64_t f
 #else
                     sprintf(buffer.data(), "0x%016lx", addr);
 #endif
-                    CLOG(INFO, disass) << "Semihosting call at address " << buffer.data() << " occurred ";
+                    NSCLOG(INFO, LOGCAT) << "Semihosting call at address " << buffer.data() << " occurred ";
 
                     this->semihosting_cb(this, &(this->reg.X10) /*a0*/, &(this->reg.X11) /*a1*/);
                     return this->reg.NEXT_PC;
@@ -814,7 +558,7 @@ template <typename BASE> uint64_t riscv_hart_msu_vp<BASE>::enter_trap(uint64_t f
         }
         this->fault_data = 0;
     } else {
-        if(cur_priv != PRIV_M && ((this->csr[mideleg] >> cause) & 0x1) != 0)
+        if(this->reg.PRIV != PRIV_M && ((this->csr[mideleg] >> cause) & 0x1) != 0)
             new_priv = (this->csr[sideleg] >> cause) & 0x1 ? PRIV_U : PRIV_S;
         this->csr[uepc | (new_priv << 8)] = this->reg.NEXT_PC; // store next address if interrupt
         this->reg.pending_trap = 0;
@@ -830,12 +574,12 @@ template <typename BASE> uint64_t riscv_hart_msu_vp<BASE>::enter_trap(uint64_t f
     // store the actual privilege level in yPP and store interrupt enable flags
     switch(new_priv) {
     case PRIV_M:
-        state.mstatus.MPP = cur_priv;
+        state.mstatus.MPP = this->reg.PRIV;
         state.mstatus.MPIE = state.mstatus.MIE;
         state.mstatus.MIE = false;
         break;
     case PRIV_S:
-        state.mstatus.SPP = cur_priv;
+        state.mstatus.SPP = this->reg.PRIV;
         state.mstatus.SPIE = state.mstatus.SIE;
         state.mstatus.SIE = false;
         break;
@@ -848,71 +592,82 @@ template <typename BASE> uint64_t riscv_hart_msu_vp<BASE>::enter_trap(uint64_t f
     }
 
     // get trap vector
-    auto ivec = this->csr[utvec | (new_priv << 8)];
+    auto xtvec = this->csr[utvec | (new_priv << 8)];
     // calculate addr// set NEXT_PC to trap addressess to jump to based on MODE
     // bits in mtvec
-    this->reg.NEXT_PC = ivec & ~0x3UL;
-    if((ivec & 0x1) == 1 && trap_id != 0)
-        this->reg.NEXT_PC += 4 * cause;
+    this->reg.NEXT_PC = xtvec & ~0x3UL;
+    if(trap_id != 0) {
+        if((xtvec & 0x3UL) == 3UL) {
+            reg_t data;
+            auto ret = read(address_type::LOGICAL, access_type::READ, 0, this->csr[mtvt], sizeof(reg_t), reinterpret_cast<uint8_t*>(&data));
+            if(ret == iss::Err)
+                return this->reg.PC;
+            this->reg.NEXT_PC = data;
+        } else if((xtvec & 0x3UL) == 1UL)
+            this->reg.NEXT_PC += 4 * cause;
+    }
     std::array<char, 32> buffer;
+#if defined(_MSC_VER)
+    sprintf(buffer.data(), "0x%016llx", addr);
+#else
     sprintf(buffer.data(), "0x%016lx", addr);
+#endif
     if((flags & 0xffffffff) != 0xffffffff)
-        CLOG(INFO, disass) << (trap_id ? "Interrupt" : "Trap") << " with cause '"
-                           << (trap_id ? this->irq_str[cause] : this->trap_str[cause]) << "' (" << cause << ")"
-                           << " at address " << buffer.data() << " occurred, changing privilege level from " << this->lvl[cur_priv]
-                           << " to " << this->lvl[new_priv];
+        NSCLOG(INFO, LOGCAT) << (trap_id ? "Interrupt" : "Trap") << " with cause '"
+                             << (trap_id ? this->irq_str[cause] : this->trap_str[cause]) << "' (" << cause << ")"
+                             << " at address " << buffer.data() << " occurred, changing privilege level from " << this->lvl[this->reg.PRIV]
+                             << " to " << this->lvl[new_priv];
     // reset trap state
     this->reg.PRIV = new_priv;
     this->reg.trap_state = 0;
-    update_vm_info();
     return this->reg.NEXT_PC;
 }
 
-template <typename BASE> uint64_t riscv_hart_msu_vp<BASE>::leave_trap(uint64_t flags) {
+template <typename BASE, features_e FEAT, typename LOGCAT> uint64_t riscv_hart_msu_vp<BASE, FEAT, LOGCAT>::leave_trap(uint64_t flags) {
     auto cur_priv = this->reg.PRIV;
     auto inst_priv = flags & 0x3;
-    auto status = state.mstatus;
 
     auto tsr = state.mstatus.TSR;
     if(cur_priv == PRIV_S && inst_priv == PRIV_S && tsr != 0) {
-        this->reg.trap_state = (1 << 31) | (2 << 16);
-        this->fault_data = this->reg.PC;
-        return this->reg.PC;
+        this->reg.trap_state = 0x80ULL << 24 | (2 << 16); // illegal instruction
+        this->reg.NEXT_PC = std::numeric_limits<uint32_t>::max();
+    } else {
+        auto status = state.mstatus;
+        // pop the relevant lower-privilege interrupt enable and privilege mode stack
+        // clear respective yIE
+        switch(inst_priv) {
+        case PRIV_M:
+            this->reg.PRIV = state.mstatus.MPP;
+            state.mstatus.MPP = 0; // clear mpp to U mode
+            state.mstatus.MIE = state.mstatus.MPIE;
+            state.mstatus.MPIE = 1;
+            break;
+        case PRIV_S:
+            this->reg.PRIV = state.mstatus.SPP;
+            state.mstatus.SPP = 0; // clear spp to U mode
+            state.mstatus.SIE = state.mstatus.SPIE;
+            state.mstatus.SPIE = 1;
+            break;
+        case PRIV_U:
+            this->reg.PRIV = 0;
+            state.mstatus.UIE = state.mstatus.UPIE;
+            state.mstatus.UPIE = 1;
+            break;
+        }
+        // sets the pc to the value stored in the x epc register.
+        this->reg.NEXT_PC = this->csr[uepc | inst_priv << 8];
+        NSCLOG(INFO, LOGCAT) << "Executing xRET , changing privilege level from " << this->lvl[cur_priv] << " to "
+                             << this->lvl[this->reg.PRIV];
+        check_interrupt();
     }
-
-    // pop the relevant lower-privilege interrupt enable and privilege mode stack
-    // clear respective yIE
-    switch(inst_priv) {
-    case PRIV_M:
-        this->reg.PRIV = state.mstatus.MPP;
-        state.mstatus.MPP = 0; // clear mpp to U mode
-        state.mstatus.MIE = state.mstatus.MPIE;
-        state.mstatus.MPIE = 1;
-        break;
-    case PRIV_S:
-        this->reg.PRIV = state.mstatus.SPP;
-        state.mstatus.SPP = 0; // clear spp to U mode
-        state.mstatus.SIE = state.mstatus.SPIE;
-        state.mstatus.SPIE = 1;
-        break;
-    case PRIV_U:
-        this->reg.PRIV = 0;
-        state.mstatus.UIE = state.mstatus.UPIE;
-        state.mstatus.UPIE = 1;
-        break;
-    }
-    // sets the pc to the value stored in the x epc register.
-    this->reg.NEXT_PC = this->csr[uepc | inst_priv << 8];
-    CLOG(INFO, disass) << "Executing xRET , changing privilege level from " << this->lvl[cur_priv] << " to " << this->lvl[this->reg.PRIV];
-    update_vm_info();
-    check_interrupt();
+    this->reg.trap_state = this->reg.pending_trap;
     return this->reg.NEXT_PC;
 }
 
-template <typename BASE> void riscv_hart_msu_vp<BASE>::wait_until(uint64_t flags) {
+template <typename BASE, features_e FEAT, typename LOGCAT> void riscv_hart_msu_vp<BASE, FEAT, LOGCAT>::wait_until(uint64_t flags) {
     auto status = state.mstatus;
     auto tw = status.TW;
-    if(this->reg.PRIV == PRIV_S && tw != 0) {
+    if(this->reg.PRIV < PRIV_M && tw != 0) {
         this->reg.trap_state = (1 << 31) | (2 << 16);
         this->fault_data = this->reg.PC;
     }

src/iss/arch/riscv_hart_mu_p.h

@@ -40,15 +40,15 @@
 #include "riscv_hart_common.h"
 #include "util/logging.h"
 #include <algorithm>
+#include <array>
 #include <cstdint>
 #include <elfio/elf_types.hpp>
+#include <elfio/elfio.hpp>
 #ifndef FMT_HEADER_ONLY
 #define FMT_HEADER_ONLY
 #endif
-#include <array>
-#include <elfio/elfio.hpp>
 #include <fmt/format.h>
-#include <iss/mmio/memory_with_htif.h>
+#include <iss/mem/memory_with_htif.h>
 #include <unordered_map>
 
 namespace iss {
@@ -71,7 +71,7 @@ public:
 #else
             switch(priv_lvl) {
             case PRIV_U:
-                return 0x00000011UL; // 0b1...0 0001 0001
+                return FEAT & features_e::FEAT_EXT_N ? 0x00000011UL : 0UL; // 0b1...0 0001 0001
             default:
                 //       +-SD
                 //       |        +-TSR
@@ -88,7 +88,7 @@ public:
                 //       |        |||||| | | |  ||  +-UPIE
                 //       |        ||||||/|/|/|  ||  |+-MIE
                 //       |        ||||||/|/|/|  ||  ||  +-UIE
-                return 0b00000000000000000001100010011001;
+                return 0b10000000001000000001100010011001;
             }
 #endif
         } else if(sizeof(reg_t) == 8) {
@@ -97,24 +97,23 @@ public:
 #else
             switch(priv_lvl) {
             case PRIV_U:
-                return 0x00000011UL; // 0b1...0 0001 0001
+                return FEAT & features_e::FEAT_EXT_N ? 0x8000000000000011ULL : 0ULL; // 0b1...0 0001 0001
             default:
-                //       +-SD
-                //       |        +-TSR
-                //       |        |+-TW
-                //       |        ||+-TVM
-                //       |        |||+-MXR
-                //       |        ||||+-SUM
-                //       |        |||||+-MPRV
-                //       |        |||||| +-XS
-                //       |        |||||| | +-FS
-                //       |        |||||| | | +-MPP
-                //       |        |||||| | | |  +-SPP
-                //       |        |||||| | | |  |+-MPIE
-                //       |        |||||| | | |  ||  +-UPIE
-                //       |        ||||||/|/|/|  ||  |+-MIE
-                //       |        ||||||/|/|/|  ||  ||  +-UIE
-                return 0b00000000000000000001100010011001;
+                //                +-TSR
+                //                |+-TW
+                //                ||+-TVM
+                //                |||+-MXR
+                //                ||||+-SUM
+                //                |||||+-MPRV
+                //                |||||| +-XS
+                //                |||||| | +-FS
+                //                |||||| | | +-MPP
+                //                |||||| | | |  +-SPP
+                //                |||||| | | |  |+-MPIE
+                //                |||||| | | |  ||  +-UPIE
+                //                ||||||/|/|/|  ||  |+-MIE
+                //                ||||||/|/|/|  ||  ||  +-UIE
+                return 0b00000000001000000001100010011001 | 0x8000000000000000ULL;
             }
 #endif
         } else
@@ -137,7 +136,7 @@ public:
         return m[mode];
     }
 
-    riscv_hart_mu_p(feature_config cfg = feature_config{});
+    riscv_hart_mu_p();
 
     virtual ~riscv_hart_mu_p() = default;
 
@@ -151,6 +150,7 @@ public:
     uint64_t enter_trap(uint64_t flags) override { return riscv_hart_mu_p::enter_trap(flags, this->fault_data, this->fault_data); }
     uint64_t enter_trap(uint64_t flags, uint64_t addr, uint64_t instr) override;
     uint64_t leave_trap(uint64_t flags) override;
+    void wait_until(uint64_t flags) override;
 
     void set_csr(unsigned addr, reg_t val) { this->csr[addr & this->csr.page_addr_mask] = val; }
 
@@ -160,7 +160,6 @@ protected:
 
     hart_state<reg_t> state;
 
-    std::unordered_map<reg_t, uint64_t> ptw;
     std::unordered_map<uint64_t, uint8_t> atomic_reservation;
 
     iss::status read_status(unsigned addr, reg_t& val);
@@ -180,14 +179,12 @@ protected:
     iss::status write_edeleg(unsigned addr, reg_t val);
 
     void check_interrupt();
-    feature_config cfg;
-    mmio::memory_with_htif<reg_t> default_mem;
+    mem::memory_with_htif<reg_t> default_mem;
 };
 
 template <typename BASE, features_e FEAT, typename LOGCAT>
-riscv_hart_mu_p<BASE, FEAT, LOGCAT>::riscv_hart_mu_p(feature_config cfg)
+riscv_hart_mu_p<BASE, FEAT, LOGCAT>::riscv_hart_mu_p()
 : state()
-, cfg(cfg)
 , default_mem(base::get_priv_if()) {
     const std::array<unsigned, 8> rwaddrs{{
         mepc,
@@ -248,7 +245,7 @@ riscv_hart_mu_p<BASE, FEAT, LOGCAT>::riscv_hart_mu_p(feature_config cfg)
     }
     this->rd_func = util::delegate<arch_if::rd_func_sig>::from<this_class, &this_class::read>(this);
     this->wr_func = util::delegate<arch_if::wr_func_sig>::from<this_class, &this_class::write>(this);
-    this->memories.prepend(*this);
+    this->memories.root(*this);
     this->memories.append(default_mem);
 }
 
@@ -384,16 +381,8 @@ iss::status riscv_hart_mu_p<BASE, FEAT, LOGCAT>::write(const address_type type,
                 return iss::Err;
             return this->write_csr(addr, *reinterpret_cast<const reg_t*>(data));
         } break;
-        case traits<BASE>::FENCE: {
-            switch(addr) {
-            case 2:
-            case 3: {
-                ptw.clear();
-                auto tvm = state.mstatus.TVM;
-                return iss::Ok;
-            }
-            }
-        } break;
+        case traits<BASE>::FENCE:
+            break;
         case traits<BASE>::RES: {
             atomic_reservation[addr] = data[0];
         } break;
@@ -505,12 +494,12 @@ template <typename BASE, features_e FEAT, typename LOGCAT> void riscv_hart_mu_p<
 }
 
 template <typename BASE, features_e FEAT, typename LOGCAT>
-uint64_t riscv_hart_mu_p<BASE, FEAT, LOGCAT>::enter_trap(uint64_t flags, uint64_t addr, uint64_t instr) {
+uint64_t riscv_hart_mu_p<BASE, FEAT, LOGCAT>::enter_trap(uint64_t flags, uint64_t addr, uint64_t tval) {
     // flags are ACTIVE[31:31], CAUSE[30:16], TRAPID[15:0]
     // calculate and write mcause val
     if(flags == std::numeric_limits<uint64_t>::max())
         flags = this->reg.trap_state;
-    auto trap_id = bit_sub<0, 16>(flags);
+    auto const trap_id = bit_sub<0, 16>(flags);
     auto cause = bit_sub<16, 15>(flags);
     if(trap_id == 0 && cause == 11)
         cause = 0x8 + this->reg.PRIV; // adjust environment call cause
@@ -533,7 +522,7 @@ uint64_t riscv_hart_mu_p<BASE, FEAT, LOGCAT>::enter_trap(uint64_t flags, uint64_
             this->csr[utval | (new_priv << 8)] = static_cast<reg_t>(addr);
             break;
         case 2:
-            this->csr[utval | (new_priv << 8)] = (!this->has_compressed() || (instr & 0x3) == 3) ? instr : instr & 0xffff;
+            this->csr[utval | (new_priv << 8)] = (!this->has_compressed() || (tval & 0x3) == 3) ? tval : tval & 0xffff;
             break;
         case 3:
             if((FEAT & FEAT_DEBUG) && (this->csr[dcsr] & 0x8000)) {
@@ -545,7 +534,6 @@ uint64_t riscv_hart_mu_p<BASE, FEAT, LOGCAT>::enter_trap(uint64_t flags, uint64_
             }
             if(this->semihosting_cb) {
                 // Check for semihosting call
-                phys_addr_t p_addr(access_type::DEBUG_READ, traits<BASE>::MEM, addr - 4);
                 std::array<uint8_t, 8> data;
                 // check for SLLI_X0_X0_0X1F and SRAI_X0_X0_0X07
                 this->memory.rd_mem(iss::access_type::DEBUG_READ, addr - 4, 4, data.data());
@@ -562,7 +550,7 @@ uint64_t riscv_hart_mu_p<BASE, FEAT, LOGCAT>::enter_trap(uint64_t flags, uint64_
 #else
                     sprintf(buffer.data(), "0x%016lx", addr);
 #endif
-                    CLOG(INFO, disass) << "Semihosting call at address " << buffer.data() << " occurred ";
+                    NSCLOG(INFO, LOGCAT) << "Semihosting call at address " << buffer.data() << " occurred ";
 
                     this->semihosting_cb(this, &(this->reg.X10) /*a0*/, &(this->reg.X11) /*a1*/);
                     return this->reg.NEXT_PC;
@@ -611,15 +599,15 @@ uint64_t riscv_hart_mu_p<BASE, FEAT, LOGCAT>::enter_trap(uint64_t flags, uint64_
     auto xtvec = this->csr[utvec | (new_priv << 8)];
     // calculate addr// set NEXT_PC to trap addressess to jump to based on MODE
     // bits in mtvec
-    if(trap_id != 0 && (xtvec & 0x3UL) == 3UL) {
-        reg_t data;
-        auto ret = read(address_type::LOGICAL, access_type::READ, 0, this->csr[mtvt], sizeof(reg_t), reinterpret_cast<uint8_t*>(&data));
-        if(ret == iss::Err)
-            return this->reg.PC;
-        this->reg.NEXT_PC = data;
-    } else {
-        this->reg.NEXT_PC = xtvec & ~0x3UL;
-        if((xtvec & 0x1) == 1 && trap_id != 0)
+    this->reg.NEXT_PC = xtvec & ~0x3UL;
+    if(trap_id != 0) {
+        if((xtvec & 0x3UL) == 3UL) {
+            reg_t data;
+            auto ret = read(address_type::LOGICAL, access_type::READ, 0, this->csr[mtvt], sizeof(reg_t), reinterpret_cast<uint8_t*>(&data));
+            if(ret == iss::Err)
+                return this->reg.PC;
+            this->reg.NEXT_PC = data;
+        } else if((xtvec & 0x3UL) == 1UL)
             this->reg.NEXT_PC += 4 * cause;
     }
     std::array<char, 32> buffer;
@@ -629,10 +617,10 @@ uint64_t riscv_hart_mu_p<BASE, FEAT, LOGCAT>::enter_trap(uint64_t flags, uint64_
     sprintf(buffer.data(), "0x%016lx", addr);
 #endif
     if((flags & 0xffffffff) != 0xffffffff)
-        CLOG(INFO, disass) << (trap_id ? "Interrupt" : "Trap") << " with cause '"
-                           << (trap_id ? this->irq_str[cause] : this->trap_str[cause]) << "' (" << cause << ")"
-                           << " at address " << buffer.data() << " occurred, changing privilege level from " << this->lvl[this->reg.PRIV]
-                           << " to " << this->lvl[new_priv];
+        NSCLOG(INFO, LOGCAT) << (trap_id ? "Interrupt" : "Trap") << " with cause '"
+                             << (trap_id ? this->irq_str[cause] : this->trap_str[cause]) << "' (" << cause << ")"
+                             << " at address " << buffer.data() << " occurred, changing privilege level from " << this->lvl[this->reg.PRIV]
+                             << " to " << this->lvl[new_priv];
     // reset trap state
     this->reg.PRIV = new_priv;
     this->reg.trap_state = 0;
@@ -643,8 +631,7 @@ template <typename BASE, features_e FEAT, typename LOGCAT> uint64_t riscv_hart_m
     auto cur_priv = this->reg.PRIV;
     auto inst_priv = (flags & 0x3) ? 3 : 0;
     if(inst_priv > cur_priv) {
-        auto trap_val = 0x80ULL << 24 | (2 << 16); // illegal instruction
-        this->reg.trap_state = trap_val;
+        this->reg.trap_state = 0x80ULL << 24 | (2 << 16); // illegal instruction
         this->reg.NEXT_PC = std::numeric_limits<uint32_t>::max();
     } else {
         auto status = state.mstatus;
@@ -665,13 +652,22 @@ template <typename BASE, features_e FEAT, typename LOGCAT> uint64_t riscv_hart_m
         }
         // sets the pc to the value stored in the x epc register.
         this->reg.NEXT_PC = this->csr[uepc | inst_priv << 8];
-        CLOG(INFO, disass) << "Executing xRET , changing privilege level from " << this->lvl[cur_priv] << " to "
-                           << this->lvl[this->reg.PRIV];
+        NSCLOG(INFO, LOGCAT) << "Executing xRET , changing privilege level from " << this->lvl[cur_priv] << " to "
+                             << this->lvl[this->reg.PRIV];
         check_interrupt();
     }
+    this->reg.trap_state = this->reg.pending_trap;
     return this->reg.NEXT_PC;
 }
 
+template <typename BASE, features_e FEAT, typename LOGCAT> void riscv_hart_mu_p<BASE, FEAT, LOGCAT>::wait_until(uint64_t flags) {
+    auto status = state.mstatus;
+    auto tw = status.TW;
+    if(this->reg.PRIV == PRIV_S && tw != 0) {
+        this->reg.trap_state = (1 << 31) | (2 << 16);
+        this->fault_data = this->reg.PC;
+    }
+}
 } // namespace arch
 } // namespace iss
 

src/iss/arch/tgc_mapper.h

@@ -1,3 +1,37 @@
+/*******************************************************************************
+ * Copyright (C) 2023 - 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 _ISS_ARCH_TGC_MAPPER_H
 #define _ISS_ARCH_TGC_MAPPER_H
 
@@ -23,35 +57,29 @@ using tgc5c_xrb_nn_plat_type = iss::arch::hwl<iss::arch::riscv_hart_m_p<iss::arc
 #ifdef CORE_TGC5D
 #include "riscv_hart_mu_p.h"
 #include <iss/arch/tgc5d.h>
-using tgc5d_plat_type = iss::arch::riscv_hart_mu_p<iss::arch::tgc5d, (iss::arch::features_e)(iss::arch::FEAT_PMP | iss::arch::FEAT_CLIC |
-                                                                                             iss::arch::FEAT_EXT_N)>;
+using tgc5d_plat_type = iss::arch::riscv_hart_mu_p<iss::arch::tgc5d, (iss::arch::features_e)(iss::arch::FEAT_EXT_N)>;
 #endif
 #ifdef CORE_TGC5D_XRB_MAC
 #include "riscv_hart_mu_p.h"
 #include <iss/arch/tgc5d_xrb_mac.h>
-using tgc5d_xrb_mac_plat_type =
-    iss::arch::riscv_hart_mu_p<iss::arch::tgc5d_xrb_mac,
-                               (iss::arch::features_e)(iss::arch::FEAT_PMP | iss::arch::FEAT_CLIC | iss::arch::FEAT_EXT_N)>;
+using tgc5d_xrb_mac_plat_type = iss::arch::riscv_hart_mu_p<iss::arch::tgc5d_xrb_mac(iss::arch::features_e)(iss::arch::FEAT_EXT_N)>;
 #endif
 #ifdef CORE_TGC5D_XRB_NN
 #include "hwl.h"
 #include "riscv_hart_mu_p.h"
 #include <iss/arch/tgc5d_xrb_nn.h>
 using tgc5d_xrb_nn_plat_type =
-    iss::arch::hwl<iss::arch::riscv_hart_mu_p<iss::arch::tgc5d_xrb_nn,
-                                              (iss::arch::features_e)(iss::arch::FEAT_PMP | iss::arch::FEAT_CLIC | iss::arch::FEAT_EXT_N)>>;
+    iss::arch::hwl<iss::arch::riscv_hart_mu_p<iss::arch::tgc5d_xrb_nn, (iss::arch::features_e)(iss::arch::FEAT_EXT_N)>>;
 #endif
 #ifdef CORE_TGC5E
 #include "riscv_hart_mu_p.h"
 #include <iss/arch/tgc5e.h>
-using tgc5e_plat_type = iss::arch::riscv_hart_mu_p<iss::arch::tgc5e, (iss::arch::features_e)(iss::arch::FEAT_PMP | iss::arch::FEAT_CLIC |
-                                                                                             iss::arch::FEAT_EXT_N)>;
+using tgc5e_plat_type = iss::arch::riscv_hart_mu_p<iss::arch::tgc5e, (iss::arch::features_e)(iss::arch::FEAT_EXT_N)>;
 #endif
 #ifdef CORE_TGC5X
 #include "riscv_hart_mu_p.h"
 #include <iss/arch/tgc5x.h>
-using tgc5x_plat_type = iss::arch::riscv_hart_mu_p<iss::arch::tgc5x, (iss::arch::features_e)(iss::arch::FEAT_PMP | iss::arch::FEAT_CLIC |
-                                                                                             iss::arch::FEAT_EXT_N | iss::arch::FEAT_TCM)>;
+using tgc5x_plat_type = iss::arch::riscv_hart_mu_p<iss::arch::tgc5x, (iss::arch::features_e)(iss::arch::FEAT_EXT_N)>;
 #endif
 
 #endif

src/iss/debugger/csr_names.cpp

@@ -1,3 +1,37 @@
+/*******************************************************************************
+ * 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
+ ******************************************************************************/
+
 #include <array>
 // generated from:
 // * /scratch/eyck/workarea/Other/riscv-opcodes/csrs.csv

src/iss/debugger/riscv_target_adapter.h

@@ -39,7 +39,6 @@
 #include <iss/iss.h>
 
 #include <array>
-#include <memory>
 #ifndef FMT_HEADER_ONLY
 #define FMT_HEADER_ONLY
 #endif

src/iss/mem/clic.h

@@ -1,11 +1,44 @@
+/*******************************************************************************
+ * 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
+ ******************************************************************************/
 
+#include "memory_if.h"
 #include "iss/arch/riscv_hart_common.h"
 #include "iss/vm_types.h"
-#include "memory_if.h"
 #include <util/logging.h>
 
 namespace iss {
-namespace mmio {
+namespace mem {
 struct clic_config {
     uint64_t clic_base{0xc0000000};
     unsigned clic_int_ctl_bits{4};
@@ -199,12 +232,12 @@ template <typename WORD_TYPE> iss::status clic<WORD_TYPE>::read_cause(unsigned a
         switch(mode) {
         case 0:
             val |= clic_uprev_lvl << 16;
-            val |= hart_if.mstatus.UPIE << 27;
+            val |= hart_if.state.mstatus.UPIE << 27;
             break;
         default:
             val |= clic_mprev_lvl << 16;
-            val |= hart_if.mstatus.MPIE << 27;
-            val |= hart_if.mstatus.MPP << 28;
+            val |= hart_if.state.mstatus.MPIE << 27;
+            val |= hart_if.state.mstatus.MPP << 28;
             break;
         }
     } else
@@ -220,12 +253,12 @@ template <typename WORD_TYPE> iss::status clic<WORD_TYPE>::write_cause(unsigned
         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;
+            hart_if.state.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;
+            hart_if.state.mstatus.MPIE = (val >> 27) & 0x1;
+            hart_if.state.mstatus.MPP = (val >> 28) & 0x3;
             break;
         }
     } else {
@@ -248,5 +281,5 @@ template <typename WORD_TYPE> iss::status clic<WORD_TYPE>::write_intthresh(unsig
     return iss::Ok;
 }
 
-} // namespace mmio
+} // namespace mem
 } // namespace iss

src/iss/mem/memory_if.cpp

@@ -0,0 +1,101 @@
+/*******************************************************************************
+ * 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
+ ******************************************************************************/
+
+#include "memory_if.h"
+#include <algorithm>
+
+namespace iss {
+namespace mem {
+void memory_hierarchy::root(memory_elem& e) {
+    hierarchy.push_front(&e);
+    root_set = true;
+    update_chain();
+}
+void memory_hierarchy::prepend(memory_elem& e) {
+    if(root_set)
+        hierarchy.insert(hierarchy.begin() + 1, &e);
+    else
+        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& e) {
+    auto old = hierarchy.back();
+    auto it = std::find_if(std::begin(owned_elems), std::end(owned_elems),
+                           [old](std::unique_ptr<memory_elem> const& p) { return p.get() == old; });
+    hierarchy.pop_back();
+    if(it != std::end(owned_elems))
+        owned_elems.erase(it);
+    hierarchy.push_back(&e);
+    update_chain();
+}
+void memory_hierarchy::update_chain() {
+    bool tail = false;
+    for(size_t i = 1; i < hierarchy.size(); ++i) {
+        hierarchy[i - 1]->set_next(hierarchy[i]->get_mem_if());
+    }
+}
+
+void memory_hierarchy::prepend(std::unique_ptr<memory_elem>&& p) {
+    prepend(*p);
+    owned_elems.push_back(std::move(p));
+}
+
+void memory_hierarchy::append(std::unique_ptr<memory_elem>&& p) {
+    append(*p);
+    owned_elems.push_back(std::move(p));
+}
+
+void memory_hierarchy::insert_before(std::unique_ptr<memory_elem>&& p) {
+    insert_before(*p);
+    owned_elems.push_back(std::move(p));
+}
+
+void memory_hierarchy::insert_after(std::unique_ptr<memory_elem>&& p) {
+    insert_after(*p);
+    owned_elems.push_back(std::move(p));
+}
+
+void memory_hierarchy::replace_last(std::unique_ptr<memory_elem>&& p) {
+    replace_last(*p);
+    owned_elems.push_back(std::move(p));
+}
+
+} // namespace mem
+} // namespace iss

src/iss/mem/memory_if.h

@@ -39,10 +39,12 @@
 #include <deque>
 #include <functional>
 #include <limits>
+#include <memory>
 #include <util/delegate.h>
+#include <vector>
 
 namespace iss {
-namespace mmio {
+namespace mem {
 
 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*);
@@ -53,24 +55,32 @@ struct memory_if {
 };
 
 struct memory_elem {
+    virtual ~memory_elem() = default;
     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 root(memory_elem&);
     void prepend(memory_elem&);
     void append(memory_elem&);
     void insert_before(memory_elem&);
     void insert_after(memory_elem&);
     void replace_last(memory_elem&);
+    void prepend(std::unique_ptr<memory_elem>&&);
+    void append(std::unique_ptr<memory_elem>&&);
+    void insert_before(std::unique_ptr<memory_elem>&&);
+    void insert_after(std::unique_ptr<memory_elem>&&);
+    void replace_last(std::unique_ptr<memory_elem>&&);
 
 protected:
     void update_chain();
-    std::deque<std::reference_wrapper<memory_elem>> hierarchy;
+    std::deque<memory_elem*> hierarchy;
+    std::vector<std::unique_ptr<memory_elem>> owned_elems;
+    bool root_set{false};
 };
 
-} // namespace mmio
+} // namespace mem
 } // namespace iss
-#endif
+#endif

src/iss/mem/memory_with_htif.h

@@ -1,14 +1,48 @@
+/*******************************************************************************
+ * 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_WITH_HTIF_
 #define _MEMORY_WITH_HTIF_
 
+#include "memory_if.h"
 #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 {
+namespace mem {
 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;
@@ -51,6 +85,6 @@ protected:
     mem_type mem;
     arch::priv_if<WORD_TYPE> hart_if;
 };
-} // namespace mmio
+} // namespace mem
 } // namespace iss
 #endif // _MEMORY_WITH_HTIF_

src/iss/mem/mmu.h

@@ -0,0 +1,353 @@
+/*******************************************************************************
+ * 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
+ ******************************************************************************/
+
+#include "memory_if.h"
+#include "iss/arch/riscv_hart_common.h"
+#include "iss/vm_types.h"
+#include <util/logging.h>
+
+namespace iss {
+namespace mem {
+enum {
+    PGSHIFT = 12,
+    PTE_PPN_SHIFT = 10,
+    // page table entry (PTE) fields
+    PTE_V = 0x001,   // Valid
+    PTE_R = 0x002,   // Read
+    PTE_W = 0x004,   // Write
+    PTE_X = 0x008,   // Execute
+    PTE_U = 0x010,   // User
+    PTE_G = 0x020,   // Global
+    PTE_A = 0x040,   // Accessed
+    PTE_D = 0x080,   // Dirty
+    PTE_SOFT = 0x300 // Reserved for Software
+};
+
+template <typename T> inline bool PTE_TABLE(T PTE) { return (((PTE) & (PTE_V | PTE_R | PTE_W | PTE_X)) == PTE_V); }
+
+struct vm_info {
+    int levels;
+    int idxbits;
+    int ptesize;
+    uint64_t ptbase;
+    bool is_active() { return levels; }
+};
+
+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;
+    }
+}
+// TODO: update vminfo on trap enter and leave as well as mstatus write, reset
+template <typename WORD_TYPE> struct mmu : public memory_elem {
+    using this_class = mmu<WORD_TYPE>;
+    using reg_t = WORD_TYPE;
+    constexpr static unsigned WORD_LEN = sizeof(WORD_TYPE) * 8;
+
+    constexpr static reg_t PGSIZE = 1 << PGSHIFT;
+    constexpr static reg_t PGMASK = PGSIZE - 1;
+
+    mmu(arch::priv_if<WORD_TYPE> hart_if)
+    : hart_if(hart_if) {
+        hart_if.csr_rd_cb[satp] = MK_CSR_RD_CB(read_satp);
+        hart_if.csr_wr_cb[satp] = MK_CSR_WR_CB(write_satp);
+    }
+
+    virtual ~mmu() = 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(unlikely((addr & ~PGMASK) != ((addr + length - 1) & ~PGMASK))) { // we may cross a page boundary
+            vm_info vm = decode_vm_info(hart_if.PRIV, satp);
+            if(vm.levels != 0) { // VM is active
+                auto split_addr = (addr + length) & ~PGMASK;
+                auto len1 = split_addr - addr;
+                auto res = down_stream_mem.rd_mem(access, addr, len1, data);
+                if(res == iss::Ok)
+                    res = down_stream_mem.rd_mem(access, split_addr, length - len1, data + len1);
+                return res;
+            }
+        }
+        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(unlikely((addr & ~PGMASK) != ((addr + length - 1) & ~PGMASK))) { // we may cross a page boundary
+            vm_info vm = decode_vm_info(hart_if.PRIV, satp);
+            if(vm.levels != 0) { // VM is active
+                auto split_addr = (addr + length) & ~PGMASK;
+                auto len1 = split_addr - addr;
+                auto res = down_stream_mem.wr_mem(access, addr, len1, data);
+                if(res == iss::Ok)
+                    res = down_stream_mem.wr_mem(access, split_addr, length - len1, data + len1);
+                return res;
+            }
+        }
+        return down_stream_mem.wr_mem(access, virt2phys(access, addr), length, data);
+    }
+    void update_vm_info();
+
+    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 read_satp(unsigned addr, reg_t& val) {
+        auto tvm = bit_sub<20, 1>(hart_if.state.mstatus());
+        if(hart_if.PRIV == arch::PRIV_S & tvm != 0) {
+            hart_if.raise_trap(2, 0, hart_if.PC);
+            //            hart_if.reg.trap_state = (1 << 31) | (2 << 16);
+            //            hart_if.fault_data = hart_if.reg.PC;
+            return iss::Err;
+        }
+        val = satp;
+        return iss::Ok;
+    }
+
+    iss::status write_satp(unsigned addr, reg_t val) {
+        reg_t tvm = hart_if.state.mstatus.TVM;
+        if(hart_if.PRIV == arch::PRIV_S & tvm != 0) {
+            hart_if.raise_trap(2, 0, hart_if.PC);
+            //            hart_if.reg.trap_state = (1 << 31) | (2 << 16);
+            //            hart_if.fault_data = hart_if.reg.PC;
+            return iss::Err;
+        }
+        satp = val;
+        update_vm_info();
+        return iss::Ok;
+    }
+
+    uint64_t virt2phys(iss::access_type access, uint64_t addr);
+
+    static inline vm_info decode_vm_info(uint32_t state, uint32_t sptbr) {
+        if(state == arch::PRIV_M)
+            return {0, 0, 0, 0};
+        if(state <= arch::PRIV_S)
+            switch(bit_sub<31, 1>(sptbr)) {
+            case 0:
+                return {0, 0, 0, 0}; // off
+            case 1:
+                return {2, 10, 4, bit_sub<0, 22>(sptbr) << PGSHIFT}; // SV32
+            default:
+                abort();
+            }
+        abort();
+        return {0, 0, 0, 0}; // dummy
+    }
+
+    static inline vm_info decode_vm_info(uint32_t state, uint64_t sptbr) {
+        if(state == arch::PRIV_M)
+            return {0, 0, 0, 0};
+        if(state <= arch::PRIV_S)
+            switch(bit_sub<60, 4>(sptbr)) {
+            case 0:
+                return {0, 0, 0, 0}; // off
+            case 8:
+                return {3, 9, 8, bit_sub<0, 44>(sptbr) << PGSHIFT}; // SV39
+            case 9:
+                return {4, 9, 8, bit_sub<0, 44>(sptbr) << PGSHIFT}; // SV48
+            case 10:
+                return {5, 9, 8, bit_sub<0, 44>(sptbr) << PGSHIFT}; // SV57
+            case 11:
+                return {6, 9, 8, bit_sub<0, 44>(sptbr) << PGSHIFT}; // SV64
+            default:
+                abort();
+            }
+        abort();
+        return {0, 0, 0, 0}; // dummy
+    }
+
+protected:
+    reg_t satp;
+    std::unordered_map<reg_t, uint64_t> ptw;
+    std::array<vm_info, 2> vmt;
+    std::array<address_type, 4> addr_mode;
+
+    arch::priv_if<WORD_TYPE> hart_if;
+    memory_if down_stream_mem;
+};
+
+template <typename WORD_TYPE> uint64_t mmu<WORD_TYPE>::virt2phys(iss::access_type access, uint64_t addr) {
+    const auto type = access & iss::access_type::FUNC;
+    auto it = ptw.find(addr >> PGSHIFT);
+    if(it != ptw.end()) {
+        const reg_t pte = it->second;
+        const reg_t ad = PTE_A | (type == iss::access_type::WRITE) * PTE_D;
+#ifdef RISCV_ENABLE_DIRTY
+        // set accessed and possibly dirty bits.
+        *(uint32_t*)ppte |= ad;
+        return {addr.getAccessType(), addr.space, (pte & (~PGMASK)) | (addr.val & PGMASK)};
+#else
+        // take exception if access or possibly dirty bit is not set.
+        if((pte & ad) == ad)
+            return {(pte & (~PGMASK)) | (addr & PGMASK)};
+        else
+            ptw.erase(it); // throw an exception
+#endif
+    } else {
+        uint32_t mode = type != iss::access_type::FETCH && hart_if.state.mstatus.MPRV ? // MPRV
+                            hart_if.state.mstatus.MPP
+                                                                                      : hart_if.PRIV;
+
+        const vm_info& vm = vmt[static_cast<uint16_t>(type) / 2];
+
+        const bool s_mode = mode == arch::PRIV_S;
+        const bool sum = hart_if.state.mstatus.SUM;
+        const bool mxr = hart_if.state.mstatus.MXR;
+
+        // verify bits xlen-1:va_bits-1 are all equal
+        const int va_bits = PGSHIFT + vm.levels * vm.idxbits;
+        const reg_t mask = (reg_t(1) << (sizeof(reg_t) * 8 - (va_bits - 1))) - 1;
+        const reg_t masked_msbs = (addr >> (va_bits - 1)) & mask;
+        const int levels = (masked_msbs != 0 && masked_msbs != mask) ? 0 : vm.levels;
+
+        reg_t base = vm.ptbase;
+        for(int i = levels - 1; i >= 0; i--) {
+            const int ptshift = i * vm.idxbits;
+            const reg_t idx = (addr >> (PGSHIFT + ptshift)) & ((1 << vm.idxbits) - 1);
+
+            // check that physical address of PTE is legal
+            reg_t pte = 0;
+            const uint8_t res = down_stream_mem.rd_mem(iss::access_type::READ, base + idx * vm.ptesize, vm.ptesize, (uint8_t*)&pte);
+            if(res != 0)
+                throw arch::trap_load_access_fault(addr);
+            const reg_t ppn = pte >> PTE_PPN_SHIFT;
+
+            if(PTE_TABLE(pte)) { // next level of page table
+                base = ppn << PGSHIFT;
+            } else if((pte & PTE_U) ? s_mode && (type == iss::access_type::FETCH || !sum) : !s_mode) {
+                break;
+            } else if(!(pte & PTE_V) || (!(pte & PTE_R) && (pte & PTE_W))) {
+                break;
+            } else if(type == (type == iss::access_type::FETCH  ? !(pte & PTE_X)
+                               : type == iss::access_type::READ ? !(pte & PTE_R) && !(mxr && (pte & PTE_X))
+                                                                : !((pte & PTE_R) && (pte & PTE_W)))) {
+                break;
+            } else if((ppn & ((reg_t(1) << ptshift) - 1)) != 0) {
+                break;
+            } else {
+                const reg_t ad = PTE_A | ((type == iss::access_type::WRITE) * PTE_D);
+#ifdef RISCV_ENABLE_DIRTY
+                // set accessed and possibly dirty bits.
+                *(uint32_t*)ppte |= ad;
+#else
+                // take exception if access or possibly dirty bit is not set.
+                if((pte & ad) != ad)
+                    break;
+#endif
+                // for superpage mappings, make a fake leaf PTE for the TLB's benefit.
+                const reg_t vpn = addr >> PGSHIFT;
+                const reg_t value = (ppn | (vpn & ((reg_t(1) << ptshift) - 1))) << PGSHIFT;
+                const reg_t offset = addr & PGMASK;
+                ptw[vpn] = value | (pte & 0xff);
+                return value | offset;
+            }
+        }
+    }
+    switch(type) {
+    case access_type::FETCH:
+        hart_if.raise_trap(12, 0, addr);
+        throw arch::trap_instruction_page_fault(addr);
+    case access_type::READ:
+        hart_if.raise_trap(13, 0, addr);
+        throw arch::trap_load_page_fault(addr);
+    case access_type::WRITE:
+        hart_if.raise_trap(15, 0, addr);
+        throw arch::trap_store_page_fault(addr);
+    default:
+        abort();
+    }
+}
+
+template <typename WORD_TYPE> inline void mmu<WORD_TYPE>::update_vm_info() {
+    vmt[1] = decode_vm_info(hart_if.PRIV, satp);
+    addr_mode[3] = addr_mode[2] = vmt[1].is_active() ? iss::address_type::VIRTUAL : iss::address_type::PHYSICAL;
+    if(hart_if.state.mstatus.MPRV)
+        vmt[0] = decode_vm_info(hart_if.state.mstatus.MPP, satp);
+    else
+        vmt[0] = vmt[1];
+    addr_mode[1] = addr_mode[0] = vmt[0].is_active() ? iss::address_type::VIRTUAL : iss::address_type::PHYSICAL;
+    ptw.clear();
+}
+
+} // namespace mem
+} // namespace iss

src/iss/mem/pmp.h

@@ -1,11 +1,44 @@
+/*******************************************************************************
+ * 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
+ ******************************************************************************/
 
+#include "memory_if.h"
 #include "iss/arch/riscv_hart_common.h"
 #include "iss/vm_types.h"
-#include "memory_if.h"
 #include <util/logging.h>
 
 namespace iss {
-namespace mmio {
+namespace mem {
 struct clic_config {
     uint64_t clic_base{0xc0000000};
     unsigned clic_int_ctl_bits{4};
@@ -61,9 +94,8 @@ template <typename WORD_TYPE> struct pmp : public memory_elem {
     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) {
+    pmp(arch::priv_if<WORD_TYPE> hart_if)
+    : hart_if(hart_if) {
         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);
@@ -74,7 +106,7 @@ template <typename WORD_TYPE> struct pmp : public memory_elem {
         }
     }
 
-    ~pmp() = default;
+    virtual ~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)},
@@ -208,5 +240,5 @@ template <typename WORD_TYPE> bool pmp<WORD_TYPE>::pmp_check(const access_type t
     return !any_active || hart_if.reg.PRIV == arch::PRIV_M;
 }
 
-} // namespace mmio
+} // namespace mem
 } // namespace iss

src/iss/mmio/memory_if.cpp

@@ -1,26 +0,0 @@
-#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/semihosting/semihosting.cpp

@@ -1,3 +1,37 @@
+/*******************************************************************************
+ * 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
+ ******************************************************************************/
+
 #include "semihosting.h"
 #include <chrono>
 #include <cstdint>
@@ -216,7 +250,7 @@ template <typename T> void semihosting_callback<T>::operator()(iss::arch_if* arc
         T cmd_addr = sh_read_field<T>(arch_if_ptr, *parameter);
         T cmd_len = sh_read_field<T>(arch_if_ptr, (*parameter) + 1);
         std::string cmd = sh_read_string<T>(arch_if_ptr, cmd_addr, cmd_len);
-        system(cmd.c_str());
+        auto val = system(cmd.c_str());
         break;
     }
     case semihosting_syscalls::SYS_TICKFREQ: {

src/iss/semihosting/semihosting.h

@@ -1,3 +1,37 @@
+/*******************************************************************************
+ * 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 _SEMIHOSTING_H_
 #define _SEMIHOSTING_H_
 #include <chrono>
@@ -58,4 +92,4 @@ template <typename T> struct semihosting_callback {
 };
 
 template <typename T> using semihosting_cb_t = std::function<void(iss::arch_if*, T*, T*)>;
-#endif
+#endif

src/main.cpp

@@ -40,6 +40,7 @@
 #include <vector>
 
 #include "iss/arch/tgc_mapper.h"
+#include "util/logging.h"
 #include <boost/lexical_cast.hpp>
 #include <boost/program_options.hpp>
 #ifdef WITH_LLVM
@@ -98,14 +99,18 @@ int main(int argc, char* argv[]) {
 
     LOGGER(DEFAULT)::print_time() = false;
     LOGGER(connection)::print_time() = false;
+    LOGGER(dbt_rise_iss)::print_time() = false;
+
     auto l = logging::as_log_level(clim["verbose"].as<int>());
     LOGGER(DEFAULT)::reporting_level() = l;
     LOGGER(connection)::reporting_level() = l;
+    LOGGER(dbt_rise_iss)::reporting_level() = l;
     if(clim.count("logfile")) {
         // configure the connection logger
         auto f = fopen(clim["logfile"].as<std::string>().c_str(), "w");
         LOG_OUTPUT(DEFAULT)::stream() = f;
         LOG_OUTPUT(connection)::stream() = f;
+        LOG_OUTPUT(dbt_rise_iss)::stream() = f;
     }
 
     std::vector<iss::vm_plugin*> plugin_list;
@@ -258,6 +263,7 @@ int main(int argc, char* argv[]) {
                 LOG(ERR) << "Error opening file " << filename << std::endl;
                 return 1;
             }
+            LOGGER(DEFAULT)::reporting_level() = logging::ERR;
             for(auto addr = start_addr; addr < end_addr; addr += data.size()) {
                 vm->get_arch()->read(iss::address_type::PHYSICAL, iss::access_type::DEBUG_READ, 0 /*MEM*/, addr, data.size(),
                                      data.data()); // FIXME: get space from iss::arch::traits<ARCH>::mem_type_e::MEM

src/sysc/core_complex.cpp

@@ -1,5 +1,5 @@
 /*******************************************************************************
- * Copyright (C) 2017, 2018 MINRES Technologies GmbH
+ * Copyright (C) 2017 - 2025 MINRES Technologies GmbH
  * All rights reserved.
  *
  * Redistribution and use in source and binary forms, with or without
@@ -45,11 +45,10 @@
 #include <scc/report.h>
 #include <util/ities.h>
 #include <iostream>
-#include <sstream>
 #include <array>
-#include <numeric>
 #include <iss/plugin/cycle_estimate.h>
 #include <iss/plugin/instruction_count.h>
+#include <util/ities.h>
 
 // clang-format on
 
@@ -267,7 +266,10 @@ template <unsigned int BUSWIDTH> void core_complex<BUSWIDTH>::before_end_of_elab
     cpu = new core_wrapper(this);
     cpu->create_cpu(GET_PROP_VALUE(core_type), GET_PROP_VALUE(backend), GET_PROP_VALUE(gdb_server_port), GET_PROP_VALUE(mhartid));
     sc_assert(cpu->vm != nullptr);
-    cpu->vm->setDisassEnabled(GET_PROP_VALUE(enable_disass) || trc->m_db != nullptr);
+    auto disass = GET_PROP_VALUE(enable_disass);
+    if(disass && trc->m_db)
+        SCCINFO(SCMOD) << "Disasssembly will only be in transaction trace database!";
+    cpu->vm->setDisassEnabled(disass || trc->m_db != nullptr);
     if(GET_PROP_VALUE(plugins).length()) {
         auto p = util::split(GET_PROP_VALUE(plugins), ';');
         for(std::string const& opt_val : p) {
@@ -305,17 +307,20 @@ template <unsigned int BUSWIDTH> void core_complex<BUSWIDTH>::before_end_of_elab
 template <unsigned int BUSWIDTH> void core_complex<BUSWIDTH>::start_of_simulation() {
     // quantum_keeper.reset();
     if(GET_PROP_VALUE(elf_file).size() > 0) {
-        istringstream is(GET_PROP_VALUE(elf_file));
-        string s;
-        while(getline(is, s, ',')) {
-            std::pair<uint64_t, bool> start_addr = cpu->load_file(s);
+        auto file_names = util::split(GET_PROP_VALUE(elf_file), ',');
+        for(auto& s : file_names) {
+            std::pair<uint64_t, bool> load_result = cpu->load_file(s);
+            if(!std::get<1>(load_result)) {
+                SCCWARN(SCMOD) << "Could not load FW file " << s;
+            } else {
 #ifndef CWR_SYSTEMC
-            if(reset_address.is_default_value() && start_addr.second == true)
-                reset_address.set_value(start_addr.first);
+                if(reset_address.is_default_value())
+                    reset_address.set_value(load_result.first);
 #else
-            if(start_addr.second == true)
-                reset_address = start_addr.first;
+                if(start_addr.second == true)
+                    reset_address = start_addr.first;
 #endif
+            }
         }
     }
     if(trc->m_db != nullptr && trc->stream_handle == nullptr) {
@@ -419,7 +424,7 @@ template <unsigned int BUSWIDTH> bool core_complex<BUSWIDTH>::read_mem(uint64_t
             gp.set_extension(preExt);
         }
         auto pre_delay = delay;
-        dbus->b_transport(gp, delay);
+        sckt->b_transport(gp, delay);
         if(pre_delay > delay) {
             quantum_keeper.reset();
         } else {

src/sysc/core_complex.h

@@ -1,5 +1,5 @@
 /*******************************************************************************
- * Copyright (C) 2017-2021 MINRES Technologies GmbH
+ * Copyright (C) 2017 - 2025 MINRES Technologies GmbH
  * All rights reserved.
  *
  * Redistribution and use in source and binary forms, with or without

src/sysc/sc_core_adapter.h

@@ -1,9 +1,36 @@
-/*
- * sc_core_adapter.h
+/*******************************************************************************
+ * Copyright (C) 2023 - 2025 MINRES Technologies GmbH
+ * All rights reserved.
  *
- *  Created on: Jul 5, 2023
- *      Author: eyck
- */
+ * 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 _SYSC_SC_CORE_ADAPTER_H_
 #define _SYSC_SC_CORE_ADAPTER_H_
@@ -11,6 +38,7 @@
 #include "sc_core_adapter_if.h"
 #include <iostream>
 #include <iss/iss.h>
+#include <iss/mem/memory_if.h>
 #include <iss/vm_types.h>
 #include <scc/report.h>
 #include <util/ities.h>
@@ -18,11 +46,16 @@
 namespace sysc {
 template <typename PLAT> class sc_core_adapter : public PLAT, public sc_core_adapter_if {
 public:
+    using this_class = sc_core_adapter<PLAT>;
     using reg_t = typename iss::arch::traits<typename PLAT::core>::reg_t;
     using phys_addr_t = typename iss::arch::traits<typename PLAT::core>::phys_addr_t;
-    using heart_state_t = typename PLAT::hart_state_type;
     sc_core_adapter(sysc::tgfs::core_complex_if* owner)
-    : owner(owner) {}
+    : owner(owner) {
+        this->csr_rd_cb[iss::arch::time] = MK_CSR_RD_CB(read_time);
+        if(sizeof(reg_t) == 4)
+            this->csr_rd_cb[iss::arch::timeh] = MK_CSR_RD_CB(read_time);
+        this->memories.replace_last(*this);
+    }
 
     iss::arch_if* get_arch_if() override { return this; }
 
@@ -60,18 +93,23 @@ public:
         }
     };
 
-    iss::status read_mem(phys_addr_t addr, unsigned length, uint8_t* const data) override {
-        if(addr.access && iss::access_type::DEBUG)
-            return owner->read_mem_dbg(addr.val, length, data) ? iss::Ok : iss::Err;
+    iss::mem::memory_if get_mem_if() override {
+        return iss::mem::memory_if{.rd_mem{util::delegate<iss::mem::rd_mem_func_sig>::from<this_class, &this_class::read_mem>(this)},
+                                   .wr_mem{util::delegate<iss::mem::wr_mem_func_sig>::from<this_class, &this_class::write_mem>(this)}};
+    }
+
+    iss::status read_mem(iss::access_type access, uint64_t addr, unsigned length, uint8_t* data) {
+        if(access && iss::access_type::DEBUG)
+            return owner->read_mem_dbg(addr, length, data) ? iss::Ok : iss::Err;
         else {
-            return owner->read_mem(addr.val, length, data, is_fetch(addr.access)) ? iss::Ok : iss::Err;
+            return owner->read_mem(addr, length, data, is_fetch(access)) ? iss::Ok : iss::Err;
         }
     }
 
-    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;
-        if(addr.val == this->tohost) {
+    iss::status write_mem(iss::access_type access, uint64_t addr, unsigned length, uint8_t const* data) {
+        if(access && iss::access_type::DEBUG)
+            return owner->write_mem_dbg(addr, length, data) ? iss::Ok : iss::Err;
+        if(addr == 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;
@@ -89,9 +127,6 @@ public:
                 }
                 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) {
@@ -117,39 +152,28 @@ public:
             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));
-        }
+        auto res = owner->write_mem(addr, length, data) ? iss::Ok : iss::Err;
         return res;
     }
 
-    iss::status read_csr(unsigned addr, reg_t& val) override {
-        if((addr == iss::arch::time || addr == iss::arch::timeh)) {
-            uint64_t time_val = owner->mtime_i.get_interface() ? owner->mtime_i.read() : 0;
-            if(addr == iss::arch::time) {
-                val = static_cast<reg_t>(time_val);
-            } else if(addr == iss::arch::timeh) {
-                if(sizeof(reg_t) != 4)
-                    return iss::Err;
-                val = static_cast<reg_t>(time_val >> 32);
-            }
-            return iss::Ok;
-        } else {
-            return PLAT::read_csr(addr, val);
+    iss::status read_time(unsigned addr, reg_t& val) {
+        uint64_t time_val = owner->mtime_i.get_interface() ? owner->mtime_i.read() : 0;
+        if(addr == iss::arch::time) {
+            val = static_cast<reg_t>(time_val);
+        } else if(addr == iss::arch::timeh) {
+            if(sizeof(reg_t) != 4)
+                return iss::Err;
+            val = static_cast<reg_t>(time_val >> 32);
         }
+        return iss::Ok;
     }
 
     void wait_until(uint64_t flags) override {
         SCCDEBUG(owner->hier_name()) << "Sleeping until interrupt";
+        PLAT::wait_until(flags);
         while(this->reg.pending_trap == 0 && (this->csr[iss::arch::mip] & this->csr[iss::arch::mie]) == 0) {
             sc_core::wait(wfi_evt);
         }
-        PLAT::wait_until(flags);
     }
 
     void local_irq(short id, bool value) override {

src/sysc/sc_core_adapter_if.h

@@ -1,9 +1,36 @@
-/*
- * sc_core_adapter.h
+/*******************************************************************************
+ * Copyright (C) 2023 MINRES Technologies GmbH
+ * All rights reserved.
  *
- *  Created on: Jul 5, 2023
- *      Author: eyck
- */
+ * 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 _SYSC_SC_CORE_ADAPTER_IF_H_
 #define _SYSC_SC_CORE_ADAPTER_IF_H_

src/vm/aes_sbox.h

@@ -0,0 +1,70 @@
+////////////////////////////////////////////////////////////////////////////////
+// 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:
+//       ales@minres.com - initial API and implementation
+////////////////////////////////////////////////////////////////////////////////
+
+#ifndef _VM_AES_SBOX_H_
+#define _VM_AES_SBOX_H_
+#include <cstdint>
+extern "C" {
+const uint8_t AES_ENC_SBOX[] = {
+    0x63, 0x7C, 0x77, 0x7B, 0xF2, 0x6B, 0x6F, 0xC5, 0x30, 0x01, 0x67, 0x2B, 0xFE, 0xD7, 0xAB, 0x76, 0xCA, 0x82, 0xC9, 0x7D, 0xFA, 0x59,
+    0x47, 0xF0, 0xAD, 0xD4, 0xA2, 0xAF, 0x9C, 0xA4, 0x72, 0xC0, 0xB7, 0xFD, 0x93, 0x26, 0x36, 0x3F, 0xF7, 0xCC, 0x34, 0xA5, 0xE5, 0xF1,
+    0x71, 0xD8, 0x31, 0x15, 0x04, 0xC7, 0x23, 0xC3, 0x18, 0x96, 0x05, 0x9A, 0x07, 0x12, 0x80, 0xE2, 0xEB, 0x27, 0xB2, 0x75, 0x09, 0x83,
+    0x2C, 0x1A, 0x1B, 0x6E, 0x5A, 0xA0, 0x52, 0x3B, 0xD6, 0xB3, 0x29, 0xE3, 0x2F, 0x84, 0x53, 0xD1, 0x00, 0xED, 0x20, 0xFC, 0xB1, 0x5B,
+    0x6A, 0xCB, 0xBE, 0x39, 0x4A, 0x4C, 0x58, 0xCF, 0xD0, 0xEF, 0xAA, 0xFB, 0x43, 0x4D, 0x33, 0x85, 0x45, 0xF9, 0x02, 0x7F, 0x50, 0x3C,
+    0x9F, 0xA8, 0x51, 0xA3, 0x40, 0x8F, 0x92, 0x9D, 0x38, 0xF5, 0xBC, 0xB6, 0xDA, 0x21, 0x10, 0xFF, 0xF3, 0xD2, 0xCD, 0x0C, 0x13, 0xEC,
+    0x5F, 0x97, 0x44, 0x17, 0xC4, 0xA7, 0x7E, 0x3D, 0x64, 0x5D, 0x19, 0x73, 0x60, 0x81, 0x4F, 0xDC, 0x22, 0x2A, 0x90, 0x88, 0x46, 0xEE,
+    0xB8, 0x14, 0xDE, 0x5E, 0x0B, 0xDB, 0xE0, 0x32, 0x3A, 0x0A, 0x49, 0x06, 0x24, 0x5C, 0xC2, 0xD3, 0xAC, 0x62, 0x91, 0x95, 0xE4, 0x79,
+    0xE7, 0xC8, 0x37, 0x6D, 0x8D, 0xD5, 0x4E, 0xA9, 0x6C, 0x56, 0xF4, 0xEA, 0x65, 0x7A, 0xAE, 0x08, 0xBA, 0x78, 0x25, 0x2E, 0x1C, 0xA6,
+    0xB4, 0xC6, 0xE8, 0xDD, 0x74, 0x1F, 0x4B, 0xBD, 0x8B, 0x8A, 0x70, 0x3E, 0xB5, 0x66, 0x48, 0x03, 0xF6, 0x0E, 0x61, 0x35, 0x57, 0xB9,
+    0x86, 0xC1, 0x1D, 0x9E, 0xE1, 0xF8, 0x98, 0x11, 0x69, 0xD9, 0x8E, 0x94, 0x9B, 0x1E, 0x87, 0xE9, 0xCE, 0x55, 0x28, 0xDF, 0x8C, 0xA1,
+    0x89, 0x0D, 0xBF, 0xE6, 0x42, 0x68, 0x41, 0x99, 0x2D, 0x0F, 0xB0, 0x54, 0xBB, 0x16};
+uint8_t inline aes_sbox_fwd(uint8_t index) { return AES_ENC_SBOX[index]; }
+
+const uint8_t AES_DEC_SBOX[] = {
+    0x52, 0x09, 0x6A, 0xD5, 0x30, 0x36, 0xA5, 0x38, 0xBF, 0x40, 0xA3, 0x9E, 0x81, 0xF3, 0xD7, 0xFB, 0x7C, 0xE3, 0x39, 0x82, 0x9B, 0x2F,
+    0xFF, 0x87, 0x34, 0x8E, 0x43, 0x44, 0xC4, 0xDE, 0xE9, 0xCB, 0x54, 0x7B, 0x94, 0x32, 0xA6, 0xC2, 0x23, 0x3D, 0xEE, 0x4C, 0x95, 0x0B,
+    0x42, 0xFA, 0xC3, 0x4E, 0x08, 0x2E, 0xA1, 0x66, 0x28, 0xD9, 0x24, 0xB2, 0x76, 0x5B, 0xA2, 0x49, 0x6D, 0x8B, 0xD1, 0x25, 0x72, 0xF8,
+    0xF6, 0x64, 0x86, 0x68, 0x98, 0x16, 0xD4, 0xA4, 0x5C, 0xCC, 0x5D, 0x65, 0xB6, 0x92, 0x6C, 0x70, 0x48, 0x50, 0xFD, 0xED, 0xB9, 0xDA,
+    0x5E, 0x15, 0x46, 0x57, 0xA7, 0x8D, 0x9D, 0x84, 0x90, 0xD8, 0xAB, 0x00, 0x8C, 0xBC, 0xD3, 0x0A, 0xF7, 0xE4, 0x58, 0x05, 0xB8, 0xB3,
+    0x45, 0x06, 0xD0, 0x2C, 0x1E, 0x8F, 0xCA, 0x3F, 0x0F, 0x02, 0xC1, 0xAF, 0xBD, 0x03, 0x01, 0x13, 0x8A, 0x6B, 0x3A, 0x91, 0x11, 0x41,
+    0x4F, 0x67, 0xDC, 0xEA, 0x97, 0xF2, 0xCF, 0xCE, 0xF0, 0xB4, 0xE6, 0x73, 0x96, 0xAC, 0x74, 0x22, 0xE7, 0xAD, 0x35, 0x85, 0xE2, 0xF9,
+    0x37, 0xE8, 0x1C, 0x75, 0xDF, 0x6E, 0x47, 0xF1, 0x1A, 0x71, 0x1D, 0x29, 0xC5, 0x89, 0x6F, 0xB7, 0x62, 0x0E, 0xAA, 0x18, 0xBE, 0x1B,
+    0xFC, 0x56, 0x3E, 0x4B, 0xC6, 0xD2, 0x79, 0x20, 0x9A, 0xDB, 0xC0, 0xFE, 0x78, 0xCD, 0x5A, 0xF4, 0x1F, 0xDD, 0xA8, 0x33, 0x88, 0x07,
+    0xC7, 0x31, 0xB1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xEC, 0x5F, 0x60, 0x51, 0x7F, 0xA9, 0x19, 0xB5, 0x4A, 0x0D, 0x2D, 0xE5, 0x7A, 0x9F,
+    0x93, 0xC9, 0x9C, 0xEF, 0xA0, 0xE0, 0x3B, 0x4D, 0xAE, 0x2A, 0xF5, 0xB0, 0xC8, 0xEB, 0xBB, 0x3C, 0x83, 0x53, 0x99, 0x61, 0x17, 0x2B,
+    0x04, 0x7E, 0xBA, 0x77, 0xD6, 0x26, 0xE1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0C, 0x7D};
+uint8_t inline aes_sbox_inv(uint8_t index) { return AES_DEC_SBOX[index]; }
+}
+
+#endif /* _VM_AES_SBOX_H_ */

src/vm/fp_functions.cpp

@@ -33,7 +33,9 @@
 ////////////////////////////////////////////////////////////////////////////////
 
 #include "fp_functions.h"
+#include "softfloat_types.h"
 #include <array>
+#include <cstdint>
 
 extern "C" {
 #include "internals.h"
@@ -44,21 +46,375 @@ extern "C" {
 #include <limits>
 
 using this_t = uint8_t*;
-// this does not inlcude any reserved rm or the DYN rm, as DYN rm should be taken care of in the vm_impl
-const std::array<uint8_t, 5> rmm_map = {
-    softfloat_round_near_even /*RNE*/, softfloat_round_minMag /*RTZ*/, softfloat_round_min /*RDN*/, softfloat_round_max /*RUP?*/,
-    softfloat_round_near_maxMag /*RMM*/
-};
+template <typename T> T constexpr defaultNaN();
+template <> uint16_t constexpr defaultNaN<uint16_t>() { return defaultNaNF16UI; }
+template <> uint32_t constexpr defaultNaN<uint32_t>() { return defaultNaNF32UI; }
+template <> uint64_t constexpr defaultNaN<uint64_t>() { return defaultNaNF64UI; }
+template <typename T> T constexpr posInf();
+template <> uint16_t constexpr posInf<uint16_t>() { return 0x7C00; }
+template <> uint32_t constexpr posInf<uint32_t>() { return 0x7F800000; }
+template <> uint64_t constexpr posInf<uint64_t>() { return 0x7FF0000000000000; }
+template <typename T> T constexpr negInf();
+template <> uint16_t constexpr negInf<uint16_t>() { return 0xFC00; }
+template <> uint32_t constexpr negInf<uint32_t>() { return 0xFF800000; }
+template <> uint64_t constexpr negInf<uint64_t>() { return 0xFFF0000000000000; }
+template <typename T> T constexpr negZero();
+template <> uint16_t constexpr negZero<uint16_t>() { return 0x8000; }
+template <> uint32_t constexpr negZero<uint32_t>() { return 0x80000000; }
+template <> uint64_t constexpr negZero<uint64_t>() { return 0x8000000000000000; }
 
-const uint32_t quiet_nan32 = 0x7fC00000;
+template <typename T> bool rsqrt_check(T fclass_val, bool& subnormal, T& ret_val) {
+    softfloat_exceptionFlags = 0;
+    switch(fclass_val) {
+    case 0x0001: {
+        softfloat_exceptionFlags |= softfloat_flag_invalid;
+        ret_val = defaultNaN<T>();
+        return true;
+    }
+    case 0x0002: {
+        softfloat_exceptionFlags |= softfloat_flag_invalid;
+        ret_val = defaultNaN<T>();
+        return true;
+    }
+    case 0x0004: {
+        softfloat_exceptionFlags |= softfloat_flag_invalid;
+        ret_val = defaultNaN<T>();
+        return true;
+    }
+    case 0x0100: {
+        softfloat_exceptionFlags |= softfloat_flag_invalid;
+        ret_val = defaultNaN<T>();
+        return true;
+    }
+    case 0x0200: {
+        ret_val = defaultNaN<T>();
+        return true;
+    }
+    case 0x0008: {
+        softfloat_exceptionFlags |= softfloat_flag_infinite;
+        ret_val = negInf<T>();
+        return true;
+    }
+    case 0x0010: {
+        softfloat_exceptionFlags |= softfloat_flag_infinite;
+        ret_val = posInf<T>();
+        return true;
+    }
+    case 0x0080: {
+        ret_val = 0;
+        return true;
+    }
+    case 0x0020: {
+        subnormal = true;
+    }
+    default:
+        return false;
+    }
+}
+static constexpr std::array<std::array<uint64_t, 64>, 2> rsqrt_table{
+    {{
+         52, 51, 50, 48, 47, 46, 44, 43, 42, 41, 40, 39, 38, 36, 35, 34, 33, 32, 31, 30, 30, 29, 28, 27, 26, 25, 24, 23, 23, 22, 21, 20,
+         19, 19, 18, 17, 16, 16, 15, 14, 14, 13, 12, 12, 11, 10, 10, 9,  9,  8,  7,  7,  6,  6,  5,  4,  4,  3,  3,  2,  2,  1,  1,  0,
+     },
+     {127, 125, 123, 121, 119, 118, 116, 114, 113, 111, 109, 108, 106, 105, 103, 102, 100, 99, 97, 96, 95, 93,
+      92,  91,  90,  88,  87,  86,  85,  84,  83,  82,  80,  79,  78,  77,  76,  75,  74,  73, 72, 71, 70, 70,
+      69,  68,  67,  66,  65,  64,  63,  63,  62,  61,  60,  59,  59,  58,  57,  56,  56,  55, 54, 53}}};
+
+uint64_t constexpr frsqrt7_general(const unsigned s, const unsigned e, const uint64_t sign, const int64_t exp, const uint64_t sig,
+                                   const bool subnormal) {
+    int64_t normalized_exp = exp;
+    uint64_t normalized_sig = sig;
+    if(subnormal) {
+        signed nr_leadingzeros = __builtin_clzll(sig) - (64 - s);
+        normalized_exp = -nr_leadingzeros;
+        normalized_sig = (sig << (1 + nr_leadingzeros)) & ((1ULL << s) - 1);
+    }
+    unsigned exp_idx = normalized_exp & 1;
+    unsigned sig_idx = (normalized_sig >> (s - 6)) & 0x3f;
+    // The output of the table becomes the seven high bits of the result significand (after the leading one); the remainder of the
+    // result significand is zero.
+    uint64_t out_sig = rsqrt_table[exp_idx][sig_idx] << (s - 7);
+    // The output exponent equals floor((3*B - 1 - the normalized input exponent) / 2), where B is the exponent bias.
+    unsigned bias = (1UL << (e - 1)) - 1;
+    uint64_t out_exp = (3 * bias - 1 - normalized_exp) / 2;
+    // The output sign equals the input sign.
+    return (sign << (s + e)) | (out_exp << s) | out_sig;
+}
+template <typename T> bool recip_check(T fclass_val, bool& subnormal, uint64_t& ret_val) {
+    softfloat_exceptionFlags = 0;
+    switch(fclass_val) {
+    case 0x0001: {
+        ret_val = negZero<T>();
+        return true;
+    }
+    case 0x0080: {
+        ret_val = 0;
+        return true;
+    }
+    case 0x0008: {
+        softfloat_exceptionFlags |= softfloat_flag_infinite;
+        ret_val = negInf<T>();
+        return true;
+    }
+    case 0x0010: {
+        softfloat_exceptionFlags |= softfloat_flag_infinite;
+        ret_val = posInf<T>();
+        return true;
+    }
+    case 0x0100: {
+        softfloat_exceptionFlags |= softfloat_flag_invalid;
+        ret_val = defaultNaN<T>();
+        return true;
+    }
+    case 0x0200: {
+        ret_val = defaultNaN<T>();
+        return true;
+    }
+    case 0x0004: {
+        subnormal = true;
+        return false;
+    }
+    case 0x0020: {
+        subnormal = true;
+        return false;
+    }
+    default: {
+        subnormal = false;
+        return false;
+    }
+    }
+}
+static constexpr std::array<uint64_t, 128> rec_table{
+    {127, 125, 123, 121, 119, 117, 116, 114, 112, 110, 109, 107, 105, 104, 102, 100, 99, 97, 96, 94, 93, 91, 90, 88, 87, 85,
+     84,  83,  81,  80,  79,  77,  76,  75,  74,  72,  71,  70,  69,  68,  66,  65,  64, 63, 62, 61, 60, 59, 58, 57, 56, 55,
+     54,  53,  52,  51,  50,  49,  48,  47,  46,  45,  44,  43,  42,  41,  40,  40,  39, 38, 37, 36, 35, 35, 34, 33, 32, 31,
+     31,  30,  29,  28,  28,  27,  26,  25,  25,  24,  23,  23,  22,  21,  21,  20,  19, 19, 18, 17, 17, 16, 15, 15, 14, 14,
+     13,  12,  12,  11,  11,  10,  9,   9,   8,   8,   7,   7,   6,   5,   5,   4,   4,  3,  3,  2,  2,  1,  1,  0}};
+bool frec_general(uint64_t& res, const unsigned s, const unsigned e, const uint64_t sign, const int64_t exp, const uint64_t sig,
+                  const bool subnormal, uint8_t mode) {
+    int nr_leadingzeros = __builtin_clzll(sig) - (64 - s);
+    int64_t normalized_exp = subnormal ? -nr_leadingzeros : exp;
+    uint64_t normalized_sig = subnormal ? ((sig << (1 + nr_leadingzeros)) & ((1ULL << s) - 1)) : sig;
+    unsigned idx = (normalized_sig >> (s - 7)) & 0x7f;
+    unsigned bias = (1UL << (e - 1)) - 1;
+    uint64_t mid_exp = 2 * (bias)-1 - normalized_exp;
+    uint64_t mid_sig = rec_table[idx] << (s - 7);
+
+    uint64_t out_exp = mid_exp;
+    uint64_t out_sig = mid_sig;
+    if(mid_exp == 0) {
+        out_exp = mid_exp;
+        out_sig = (mid_sig >> 1) | (1ULL << (s - 1));
+    } else if(mid_exp == (1ULL << e) - 1) {
+        out_exp = 0;
+        out_sig = (mid_sig >> 2) | (1ULL << (s - 2));
+    }
+    if(subnormal && nr_leadingzeros > 1) {
+        if((mode == 0b001) || (mode == 0b010 && sign == 0b0) || (mode == 0b011 && sign == 0b1)) {
+            res = (sign << (s + e)) | ((1ULL << (e - 1)) - 1) << s | ((1ULL << s) - 1);
+            return true;
+        } else {
+            res = (sign << (s + e)) | ((1ULL << e) - 1) << s;
+            return true;
+        }
+    }
+    res = (sign << (s + e)) | (out_exp << s) | out_sig;
+    return false;
+}
 
 extern "C" {
 
 uint32_t fget_flags() { return softfloat_exceptionFlags & 0x1f; }
+uint16_t fadd_h(uint16_t v1, uint16_t v2, uint8_t mode) {
+    float16_t v1f{v1}, v2f{v2};
+    softfloat_roundingMode = mode;
+    softfloat_exceptionFlags = 0;
+    float16_t r = f16_add(v1f, v2f);
+    return r.v;
+}
+
+uint16_t fsub_h(uint16_t v1, uint16_t v2, uint8_t mode) {
+    float16_t v1f{v1}, v2f{v2};
+    softfloat_roundingMode = mode;
+    softfloat_exceptionFlags = 0;
+    float16_t r = f16_sub(v1f, v2f);
+    return r.v;
+}
+
+uint16_t fmul_h(uint16_t v1, uint16_t v2, uint8_t mode) {
+    float16_t v1f{v1}, v2f{v2};
+    softfloat_roundingMode = mode;
+    softfloat_exceptionFlags = 0;
+    float16_t r = f16_mul(v1f, v2f);
+    return r.v;
+}
+
+uint16_t fdiv_h(uint16_t v1, uint16_t v2, uint8_t mode) {
+    float16_t v1f{v1}, v2f{v2};
+    softfloat_roundingMode = mode;
+    softfloat_exceptionFlags = 0;
+    float16_t r = f16_div(v1f, v2f);
+    return r.v;
+}
+
+uint16_t fsqrt_h(uint16_t v1, uint8_t mode) {
+    float16_t v1f{v1};
+    softfloat_roundingMode = mode;
+    softfloat_exceptionFlags = 0;
+    float16_t r = f16_sqrt(v1f);
+    return r.v;
+}
+
+uint16_t fcmp_h(uint16_t v1, uint16_t v2, uint16_t op) {
+    float16_t v1f{v1}, v2f{v2};
+    softfloat_exceptionFlags = 0;
+    bool nan = v1 == defaultNaNF16UI || v2 & defaultNaNF16UI;
+    bool snan = softfloat_isSigNaNF16UI(v1) || softfloat_isSigNaNF16UI(v2);
+    switch(op) {
+    case 0:
+        if(nan | snan) {
+            if(snan)
+                softfloat_raiseFlags(softfloat_flag_invalid);
+            return 0;
+        } else
+            return f16_eq(v1f, v2f) ? 1 : 0;
+    case 1:
+        if(nan | snan) {
+            softfloat_raiseFlags(softfloat_flag_invalid);
+            return 0;
+        } else
+            return f16_le(v1f, v2f) ? 1 : 0;
+    case 2:
+        if(nan | snan) {
+            softfloat_raiseFlags(softfloat_flag_invalid);
+            return 0;
+        } else
+            return f16_lt(v1f, v2f) ? 1 : 0;
+    default:
+        break;
+    }
+    return -1;
+}
+
+uint16_t fmadd_h(uint16_t v1, uint16_t v2, uint16_t v3, uint16_t op, uint8_t mode) {
+    uint16_t F16_SIGN = 1UL << 15;
+    switch(op) {
+    case 0: // FMADD_S
+        break;
+    case 1: // FMSUB_S
+        v3 ^= F16_SIGN;
+        break;
+    case 2: // FNMADD_S
+        v1 ^= F16_SIGN;
+        v3 ^= F16_SIGN;
+        break;
+    case 3: // FNMSUB_S
+        v1 ^= F16_SIGN;
+        break;
+    }
+    softfloat_roundingMode = mode;
+    softfloat_exceptionFlags = 0;
+    float16_t res = softfloat_mulAddF16(v1, v2, v3, 0);
+    return res.v;
+}
+
+uint16_t fsel_h(uint16_t v1, uint16_t v2, uint16_t op) {
+    softfloat_exceptionFlags = 0;
+    bool v1_nan = (v1 & defaultNaNF16UI) == defaultNaNF16UI;
+    bool v2_nan = (v2 & defaultNaNF16UI) == defaultNaNF16UI;
+    bool v1_snan = softfloat_isSigNaNF16UI(v1);
+    bool v2_snan = softfloat_isSigNaNF16UI(v2);
+    if(v1_snan || v2_snan)
+        softfloat_raiseFlags(softfloat_flag_invalid);
+    if(v1_nan || v1_snan)
+        return (v2_nan || v2_snan) ? defaultNaNF16UI : v2;
+    else if(v2_nan || v2_snan)
+        return v1;
+    else {
+        if((v1 & 0x7fff) == 0 && (v2 & 0x7fff) == 0) {
+            return op == 0 ? ((v1 & 0x8000) ? v1 : v2) : ((v1 & 0x8000) ? v2 : v1);
+        } else {
+            float16_t v1f{v1}, v2f{v2};
+            return op == 0 ? (f16_lt(v1f, v2f) ? v1 : v2) : (f16_lt(v1f, v2f) ? v2 : v1);
+        }
+    }
+}
+
+uint16_t fclass_h(uint16_t v1) {
+
+    float16_t a{v1};
+    union ui16_f16 uA;
+    uint_fast16_t uiA;
+
+    uA.f = a;
+    uiA = uA.ui;
+
+    bool infOrNaN = expF16UI(uiA) == 0x1F;
+    bool subnormalOrZero = expF16UI(uiA) == 0;
+    bool sign = signF16UI(uiA);
+    bool fracZero = fracF16UI(uiA) == 0;
+    bool isNaN = isNaNF16UI(uiA);
+    bool isSNaN = softfloat_isSigNaNF16UI(uiA);
+
+    return (sign && infOrNaN && fracZero) << 0 | (sign && !infOrNaN && !subnormalOrZero) << 1 |
+           (sign && subnormalOrZero && !fracZero) << 2 | (sign && subnormalOrZero && fracZero) << 3 | (!sign && infOrNaN && fracZero) << 7 |
+           (!sign && !infOrNaN && !subnormalOrZero) << 6 | (!sign && subnormalOrZero && !fracZero) << 5 |
+           (!sign && subnormalOrZero && fracZero) << 4 | (isNaN && isSNaN) << 8 | (isNaN && !isSNaN) << 9;
+}
+
+uint16_t frsqrt7_h(uint16_t v) {
+    bool subnormal = false;
+    uint16_t ret_val = 0;
+    if(rsqrt_check(fclass_h(v), subnormal, ret_val)) {
+        return ret_val;
+    }
+    uint16_t sig = fracF64UI(v);
+    int16_t exp = expF64UI(v);
+    uint16_t sign = signF64UI(v);
+    unsigned constexpr e = 5;
+    unsigned constexpr s = 10;
+    return frsqrt7_general(s, e, sign, exp, sig, subnormal);
+}
+
+uint16_t frec7_h(uint16_t v, uint8_t mode) {
+    bool subnormal = false;
+    uint64_t ret_val = 0;
+    if(recip_check(fclass_h(v), subnormal, ret_val)) {
+        return ret_val;
+    }
+    uint16_t sig = fracF16UI(v);
+    int exp = expF16UI(v);
+    uint16_t sign = signF16UI(v);
+    unsigned constexpr e = 5;
+    unsigned constexpr s = 10;
+    if(frec_general(ret_val, s, e, sign, exp, sig, subnormal, mode))
+        softfloat_exceptionFlags |= (softfloat_flag_inexact | softfloat_flag_overflow);
+    return ret_val;
+}
+
+uint16_t unbox_h(uint8_t FLEN, uint64_t v) {
+    uint64_t mask = 0;
+    switch(FLEN) {
+    case 32: {
+        mask = std::numeric_limits<uint32_t>::max() & ~((uint64_t)std::numeric_limits<uint16_t>::max());
+        break;
+    }
+    case 64: {
+        mask = std::numeric_limits<uint64_t>::max() & ~((uint64_t)std::numeric_limits<uint16_t>::max());
+        break;
+    }
+    default:
+        break;
+    }
+    if((v & mask) != mask)
+        return defaultNaNF16UI;
+    else
+        return v & std::numeric_limits<uint32_t>::max();
+}
 
 uint32_t fadd_s(uint32_t v1, uint32_t v2, uint8_t mode) {
     float32_t v1f{v1}, v2f{v2};
-    softfloat_roundingMode = rmm_map.at(mode);
+    softfloat_roundingMode = mode;
     softfloat_exceptionFlags = 0;
     float32_t r = f32_add(v1f, v2f);
     return r.v;
@@ -66,7 +422,7 @@ uint32_t fadd_s(uint32_t v1, uint32_t v2, uint8_t mode) {
 
 uint32_t fsub_s(uint32_t v1, uint32_t v2, uint8_t mode) {
     float32_t v1f{v1}, v2f{v2};
-    softfloat_roundingMode = rmm_map.at(mode);
+    softfloat_roundingMode = mode;
     softfloat_exceptionFlags = 0;
     float32_t r = f32_sub(v1f, v2f);
     return r.v;
@@ -74,7 +430,7 @@ uint32_t fsub_s(uint32_t v1, uint32_t v2, uint8_t mode) {
 
 uint32_t fmul_s(uint32_t v1, uint32_t v2, uint8_t mode) {
     float32_t v1f{v1}, v2f{v2};
-    softfloat_roundingMode = rmm_map.at(mode);
+    softfloat_roundingMode = mode;
     softfloat_exceptionFlags = 0;
     float32_t r = f32_mul(v1f, v2f);
     return r.v;
@@ -82,7 +438,7 @@ uint32_t fmul_s(uint32_t v1, uint32_t v2, uint8_t mode) {
 
 uint32_t fdiv_s(uint32_t v1, uint32_t v2, uint8_t mode) {
     float32_t v1f{v1}, v2f{v2};
-    softfloat_roundingMode = rmm_map.at(mode);
+    softfloat_roundingMode = mode;
     softfloat_exceptionFlags = 0;
     float32_t r = f32_div(v1f, v2f);
     return r.v;
@@ -90,7 +446,7 @@ uint32_t fdiv_s(uint32_t v1, uint32_t v2, uint8_t mode) {
 
 uint32_t fsqrt_s(uint32_t v1, uint8_t mode) {
     float32_t v1f{v1};
-    softfloat_roundingMode = rmm_map.at(mode);
+    softfloat_roundingMode = mode;
     softfloat_exceptionFlags = 0;
     float32_t r = f32_sqrt(v1f);
     return r.v;
@@ -99,7 +455,7 @@ uint32_t fsqrt_s(uint32_t v1, uint8_t mode) {
 uint32_t fcmp_s(uint32_t v1, uint32_t v2, uint32_t op) {
     float32_t v1f{v1}, v2f{v2};
     softfloat_exceptionFlags = 0;
-    bool nan = (v1 & defaultNaNF32UI) == quiet_nan32 || (v2 & defaultNaNF32UI) == quiet_nan32;
+    bool nan = v1 == defaultNaNF32UI || v2 == defaultNaNF32UI;
     bool snan = softfloat_isSigNaNF32UI(v1) || softfloat_isSigNaNF32UI(v2);
     switch(op) {
     case 0:
@@ -127,29 +483,6 @@ uint32_t fcmp_s(uint32_t v1, uint32_t v2, uint32_t op) {
     return -1;
 }
 
-uint32_t fcvt_s(uint32_t v1, uint32_t op, uint8_t mode) {
-    float32_t v1f{v1};
-    softfloat_exceptionFlags = 0;
-    float32_t r;
-    switch(op) {
-    case 0: { // FCVT__W__S
-        uint_fast32_t res = f32_to_i32(v1f, rmm_map.at(mode), true);
-        return (uint32_t)res;
-    }
-    case 1: { // FCVT__WU__S
-        uint_fast32_t res = f32_to_ui32(v1f, rmm_map.at(mode), true);
-        return (uint32_t)res;
-    }
-    case 2: // FCVT__S__W
-        r = i32_to_f32((int32_t)v1);
-        return r.v;
-    case 3: // FCVT__S__WU
-        r = ui32_to_f32(v1);
-        return r.v;
-    }
-    return 0;
-}
-
 uint32_t fmadd_s(uint32_t v1, uint32_t v2, uint32_t v3, uint32_t op, uint8_t mode) {
     uint32_t F32_SIGN = 1UL << 31;
     switch(op) {
@@ -166,7 +499,7 @@ uint32_t fmadd_s(uint32_t v1, uint32_t v2, uint32_t v3, uint32_t op, uint8_t mod
         v1 ^= F32_SIGN;
         break;
     }
-    softfloat_roundingMode = rmm_map.at(mode);
+    softfloat_roundingMode = mode;
     softfloat_exceptionFlags = 0;
     float32_t res = softfloat_mulAddF32(v1, v2, v3, 0);
     return res.v;
@@ -216,37 +549,60 @@ uint32_t fclass_s(uint32_t v1) {
            (!sign && subnormalOrZero && fracZero) << 4 | (isNaN && isSNaN) << 8 | (isNaN && !isSNaN) << 9;
 }
 
-uint32_t fconv_d2f(uint64_t v1, uint8_t mode) {
-    bool isNan = isNaNF64UI(v1);
-    bool isSNaN = softfloat_isSigNaNF64UI(v1);
-    softfloat_roundingMode = rmm_map.at(mode);
-    softfloat_exceptionFlags = 0;
-    if(isNan) {
-        if(isSNaN)
-            softfloat_raiseFlags(softfloat_flag_invalid);
-        return defaultNaNF32UI;
-    } else {
-        float32_t res = f64_to_f32(float64_t{v1});
-        return res.v;
+uint32_t frsqrt7_s(uint32_t v) {
+    bool subnormal = false;
+    uint32_t ret_val = 0;
+    if(rsqrt_check(fclass_s(v), subnormal, ret_val)) {
+        return ret_val;
     }
+    uint32_t sig = fracF32UI(v);
+    int exp = expF32UI(v);
+    uint32_t sign = signF32UI(v);
+    unsigned constexpr e = 8;
+    unsigned constexpr s = 23;
+    return frsqrt7_general(s, e, sign, exp, sig, subnormal);
 }
 
-uint64_t fconv_f2d(uint32_t v1, uint8_t mode) {
-    bool infOrNaN = expF32UI(v1) == 0xFF;
-    bool subnormalOrZero = expF32UI(v1) == 0;
-    if(infOrNaN || subnormalOrZero) {
-        return defaultNaNF64UI;
-    } else {
-        float64_t res = f32_to_f64(float32_t{v1});
-        return res.v;
+uint32_t frec7_s(uint32_t v, uint8_t mode) {
+    bool subnormal = false;
+    uint64_t ret_val = 0;
+    if(recip_check(fclass_s(v), subnormal, ret_val)) {
+        return ret_val;
     }
+    uint32_t sig = fracF32UI(v);
+    int exp = expF32UI(v);
+    uint32_t sign = signF32UI(v);
+    unsigned constexpr e = 8;
+    unsigned constexpr s = 23;
+    if(frec_general(ret_val, s, e, sign, exp, sig, subnormal, mode))
+        softfloat_exceptionFlags |= (softfloat_flag_inexact | softfloat_flag_overflow);
+    return ret_val;
+}
+
+uint32_t unbox_s(uint8_t FLEN, uint64_t v) {
+    uint64_t mask = 0;
+    switch(FLEN) {
+    case 32: {
+        return v;
+    }
+    case 64: {
+        mask = std::numeric_limits<uint64_t>::max() & ~((uint64_t)std::numeric_limits<uint32_t>::max());
+        break;
+    }
+    default:
+        break;
+    }
+    if((v & mask) != mask)
+        return defaultNaNF32UI;
+    else
+        return v & std::numeric_limits<uint32_t>::max();
 }
 
 uint64_t fadd_d(uint64_t v1, uint64_t v2, uint8_t mode) {
-    bool nan = (v1 & defaultNaNF32UI) == quiet_nan32;
+    bool nan = v1 == defaultNaNF32UI;
     bool snan = softfloat_isSigNaNF32UI(v1);
     float64_t v1f{v1}, v2f{v2};
-    softfloat_roundingMode = rmm_map.at(mode);
+    softfloat_roundingMode = mode;
     softfloat_exceptionFlags = 0;
     float64_t r = f64_add(v1f, v2f);
     return r.v;
@@ -254,7 +610,7 @@ uint64_t fadd_d(uint64_t v1, uint64_t v2, uint8_t mode) {
 
 uint64_t fsub_d(uint64_t v1, uint64_t v2, uint8_t mode) {
     float64_t v1f{v1}, v2f{v2};
-    softfloat_roundingMode = rmm_map.at(mode);
+    softfloat_roundingMode = mode;
     softfloat_exceptionFlags = 0;
     float64_t r = f64_sub(v1f, v2f);
     return r.v;
@@ -262,7 +618,7 @@ uint64_t fsub_d(uint64_t v1, uint64_t v2, uint8_t mode) {
 
 uint64_t fmul_d(uint64_t v1, uint64_t v2, uint8_t mode) {
     float64_t v1f{v1}, v2f{v2};
-    softfloat_roundingMode = rmm_map.at(mode);
+    softfloat_roundingMode = mode;
     softfloat_exceptionFlags = 0;
     float64_t r = f64_mul(v1f, v2f);
     return r.v;
@@ -270,7 +626,7 @@ uint64_t fmul_d(uint64_t v1, uint64_t v2, uint8_t mode) {
 
 uint64_t fdiv_d(uint64_t v1, uint64_t v2, uint8_t mode) {
     float64_t v1f{v1}, v2f{v2};
-    softfloat_roundingMode = rmm_map.at(mode);
+    softfloat_roundingMode = mode;
     softfloat_exceptionFlags = 0;
     float64_t r = f64_div(v1f, v2f);
     return r.v;
@@ -278,7 +634,7 @@ uint64_t fdiv_d(uint64_t v1, uint64_t v2, uint8_t mode) {
 
 uint64_t fsqrt_d(uint64_t v1, uint8_t mode) {
     float64_t v1f{v1};
-    softfloat_roundingMode = rmm_map.at(mode);
+    softfloat_roundingMode = mode;
     softfloat_exceptionFlags = 0;
     float64_t r = f64_sqrt(v1f);
     return r.v;
@@ -287,7 +643,7 @@ uint64_t fsqrt_d(uint64_t v1, uint8_t mode) {
 uint64_t fcmp_d(uint64_t v1, uint64_t v2, uint32_t op) {
     float64_t v1f{v1}, v2f{v2};
     softfloat_exceptionFlags = 0;
-    bool nan = (v1 & defaultNaNF64UI) == quiet_nan32 || (v2 & defaultNaNF64UI) == quiet_nan32;
+    bool nan = v1 == defaultNaNF64UI || v2 == defaultNaNF64UI;
     bool snan = softfloat_isSigNaNF64UI(v1) || softfloat_isSigNaNF64UI(v2);
     switch(op) {
     case 0:
@@ -315,30 +671,6 @@ uint64_t fcmp_d(uint64_t v1, uint64_t v2, uint32_t op) {
     return -1;
 }
 
-uint64_t fcvt_d(uint64_t v1, uint32_t op, uint8_t mode) {
-
-    float64_t v1f{v1};
-    softfloat_exceptionFlags = 0;
-    float64_t r;
-    switch(op) {
-    case 0: { // l from d
-        int64_t res = f64_to_i64(v1f, rmm_map.at(mode), true);
-        return (uint64_t)res;
-    }
-    case 1: { // lu from d
-        uint64_t res = f64_to_ui64(v1f, rmm_map.at(mode), true);
-        return res;
-    }
-    case 2: // d from l
-        r = i64_to_f64(v1);
-        return r.v;
-    case 3: // d from lu
-        r = ui64_to_f64(v1);
-        return r.v;
-    }
-    return 0;
-}
-
 uint64_t fmadd_d(uint64_t v1, uint64_t v2, uint64_t v3, uint32_t op, uint8_t mode) {
     uint64_t F64_SIGN = 1ULL << 63;
     switch(op) {
@@ -355,7 +687,7 @@ uint64_t fmadd_d(uint64_t v1, uint64_t v2, uint64_t v3, uint32_t op, uint8_t mod
         v1 ^= F64_SIGN;
         break;
     }
-    softfloat_roundingMode = rmm_map.at(mode);
+    softfloat_roundingMode = mode;
     softfloat_exceptionFlags = 0;
     float64_t res = softfloat_mulAddF64(v1, v2, v3, 0);
     return res.v;
@@ -406,52 +738,211 @@ uint64_t fclass_d(uint64_t v1) {
            (!sign && subnormalOrZero && fracZero) << 4 | (isNaN && isSNaN) << 8 | (isNaN && !isSNaN) << 9;
 }
 
-uint64_t fcvt_32_64(uint32_t v1, uint32_t op, uint8_t mode) {
-    float32_t v1f{v1};
-    softfloat_exceptionFlags = 0;
-    float64_t r;
-    switch(op) {
-    case 0: // l->s, fp to int32
-        return f32_to_i64(v1f, rmm_map.at(mode), true);
-    case 1: // wu->s
-        return f32_to_ui64(v1f, rmm_map.at(mode), true);
-    case 2: // s->w
-        r = i32_to_f64(v1);
-        return r.v;
-    case 3: // s->wu
-        r = ui32_to_f64(v1);
-        return r.v;
+uint64_t frsqrt7_d(uint64_t v) {
+    bool subnormal = false;
+    uint64_t ret_val = 0;
+    if(rsqrt_check(fclass_d(v), subnormal, ret_val)) {
+        return ret_val;
     }
-    return 0;
+    uint64_t sig = fracF64UI(v);
+    int exp = expF64UI(v);
+    uint64_t sign = signF64UI(v);
+    unsigned constexpr e = 11;
+    unsigned constexpr s = 52;
+    return frsqrt7_general(s, e, sign, exp, sig, subnormal);
 }
 
-uint32_t fcvt_64_32(uint64_t v1, uint32_t op, uint8_t mode) {
-    softfloat_exceptionFlags = 0;
-    float32_t r;
-    switch(op) {
-    case 0: { // wu->s
-        int32_t r = f64_to_i32(float64_t{v1}, rmm_map.at(mode), true);
-        return r;
+uint64_t frec7_d(uint64_t v, uint8_t mode) {
+    bool subnormal = false;
+    uint64_t ret_val = 0;
+    if(recip_check(fclass_d(v), subnormal, ret_val)) {
+        return ret_val;
     }
-    case 1: { // wu->s
-        uint32_t r = f64_to_ui32(float64_t{v1}, rmm_map.at(mode), true);
-        return r;
-    }
-    case 2: // l->s, fp to int32
-        r = i64_to_f32(v1);
-        return r.v;
-    case 3: // wu->s
-        r = ui64_to_f32(v1);
-        return r.v;
-    }
-    return 0;
+    uint64_t sig = fracF64UI(v);
+    int exp = expF64UI(v);
+    uint64_t sign = signF64UI(v);
+    unsigned constexpr e = 11;
+    unsigned constexpr s = 52;
+    if(frec_general(ret_val, s, e, sign, exp, sig, subnormal, mode))
+        softfloat_exceptionFlags |= (softfloat_flag_inexact | softfloat_flag_overflow);
+    return ret_val;
 }
 
-uint32_t unbox_s(uint64_t v) {
-    constexpr uint64_t mask = std::numeric_limits<uint64_t>::max() & ~((uint64_t)std::numeric_limits<uint32_t>::max());
+uint64_t unbox_d(uint8_t FLEN, uint64_t v) {
+    uint64_t mask = 0;
+    switch(FLEN) {
+    case 64: {
+        return v;
+        break;
+    }
+    default:
+        break;
+    }
     if((v & mask) != mask)
-        return 0x7fc00000;
+        return defaultNaNF64UI;
     else
         return v & std::numeric_limits<uint32_t>::max();
 }
+
+// conversion: float to float
+uint32_t f16tof32(uint16_t val, uint8_t rm) {
+    softfloat_exceptionFlags = 0;
+    softfloat_roundingMode = rm;
+    return f16_to_f32(float16_t{val}).v;
 }
+uint64_t f16tof64(uint16_t val, uint8_t rm) {
+    softfloat_exceptionFlags = 0;
+    softfloat_roundingMode = rm;
+    return f16_to_f64(float16_t{val}).v;
+}
+
+uint16_t f32tof16(uint32_t val, uint8_t rm) {
+    softfloat_exceptionFlags = 0;
+    softfloat_roundingMode = rm;
+    return f32_to_f16(float32_t{val}).v;
+}
+uint64_t f32tof64(uint32_t val, uint8_t rm) {
+    softfloat_exceptionFlags = 0;
+    softfloat_roundingMode = rm;
+    return f32_to_f64(float32_t{val}).v;
+}
+
+uint16_t f64tof16(uint64_t val, uint8_t rm) {
+    softfloat_exceptionFlags = 0;
+    softfloat_roundingMode = rm;
+    return f64_to_f16(float64_t{val}).v;
+}
+uint32_t f64tof32(uint64_t val, uint8_t rm) {
+    softfloat_exceptionFlags = 0;
+    softfloat_roundingMode = rm;
+    return f64_to_f32(float64_t{val}).v;
+}
+
+// conversions: float to unsigned
+uint32_t f16toui32(uint16_t v, uint8_t rm) {
+    softfloat_exceptionFlags = 0;
+    softfloat_roundingMode = rm;
+    return f16_to_ui32(float16_t{v}, rm, true);
+}
+uint64_t f16toui64(uint16_t v, uint8_t rm) {
+    softfloat_exceptionFlags = 0;
+    softfloat_roundingMode = rm;
+    return f16_to_ui64(float16_t{v}, rm, true);
+}
+uint32_t f32toui32(uint32_t v, uint8_t rm) {
+    softfloat_exceptionFlags = 0;
+    softfloat_roundingMode = rm;
+    return f32_to_ui32(float32_t{v}, rm, true);
+}
+uint64_t f32toui64(uint32_t v, uint8_t rm) {
+    softfloat_exceptionFlags = 0;
+    softfloat_roundingMode = rm;
+    return f32_to_ui64(float32_t{v}, rm, true);
+}
+uint32_t f64toui32(uint64_t v, uint8_t rm) {
+    softfloat_exceptionFlags = 0;
+    softfloat_roundingMode = rm;
+    return f64_to_ui32(float64_t{v}, rm, true);
+}
+uint64_t f64toui64(uint64_t v, uint8_t rm) {
+    softfloat_exceptionFlags = 0;
+    softfloat_roundingMode = rm;
+    return f64_to_ui64(float64_t{v}, rm, true);
+}
+
+// conversions: float to signed
+uint32_t f16toi32(uint16_t v, uint8_t rm) {
+    softfloat_exceptionFlags = 0;
+    softfloat_roundingMode = rm;
+    return f16_to_i32(float16_t{v}, rm, true);
+}
+uint64_t f16toi64(uint16_t v, uint8_t rm) {
+    softfloat_exceptionFlags = 0;
+    softfloat_roundingMode = rm;
+    return f16_to_i64(float16_t{v}, rm, true);
+}
+uint32_t f32toi32(uint32_t v, uint8_t rm) {
+    softfloat_exceptionFlags = 0;
+    softfloat_roundingMode = rm;
+    return f32_to_i32(float32_t{v}, rm, true);
+}
+uint64_t f32toi64(uint32_t v, uint8_t rm) {
+    softfloat_exceptionFlags = 0;
+    softfloat_roundingMode = rm;
+    return f32_to_i64(float32_t{v}, rm, true);
+}
+uint32_t f64toi32(uint64_t v, uint8_t rm) {
+    softfloat_exceptionFlags = 0;
+    softfloat_roundingMode = rm;
+    return f64_to_i32(float64_t{v}, rm, true);
+}
+uint64_t f64toi64(uint64_t v, uint8_t rm) {
+    softfloat_exceptionFlags = 0;
+    softfloat_roundingMode = rm;
+    return f64_to_i64(float64_t{v}, rm, true);
+}
+
+// conversions: unsigned to float
+uint16_t ui32tof16(uint32_t v, uint8_t rm) {
+    softfloat_exceptionFlags = 0;
+    softfloat_roundingMode = rm;
+    return ui32_to_f16(v).v;
+}
+uint16_t ui64tof16(uint64_t v, uint8_t rm) {
+    softfloat_exceptionFlags = 0;
+    softfloat_roundingMode = rm;
+    return ui64_to_f16(v).v;
+}
+uint32_t ui32tof32(uint32_t v, uint8_t rm) {
+    softfloat_exceptionFlags = 0;
+    softfloat_roundingMode = rm;
+    return ui32_to_f32(v).v;
+}
+uint32_t ui64tof32(uint64_t v, uint8_t rm) {
+    softfloat_exceptionFlags = 0;
+    softfloat_roundingMode = rm;
+    return ui64_to_f32(v).v;
+}
+uint64_t ui32tof64(uint32_t v, uint8_t rm) {
+    softfloat_exceptionFlags = 0;
+    softfloat_roundingMode = rm;
+    return ui32_to_f64(v).v;
+}
+uint64_t ui64tof64(uint64_t v, uint8_t rm) {
+    softfloat_exceptionFlags = 0;
+    softfloat_roundingMode = rm;
+    return ui64_to_f64(v).v;
+}
+
+// conversions: signed to float
+uint16_t i32tof16(uint32_t v, uint8_t rm) {
+    softfloat_exceptionFlags = 0;
+    softfloat_roundingMode = rm;
+    return i32_to_f16(v).v;
+}
+uint16_t i64tof16(uint64_t v, uint8_t rm) {
+    softfloat_exceptionFlags = 0;
+    softfloat_roundingMode = rm;
+    return i64_to_f16(v).v;
+}
+uint32_t i32tof32(uint32_t v, uint8_t rm) {
+    softfloat_exceptionFlags = 0;
+    softfloat_roundingMode = rm;
+    return i32_to_f32(v).v;
+}
+uint32_t i64tof32(uint64_t v, uint8_t rm) {
+    softfloat_exceptionFlags = 0;
+    softfloat_roundingMode = rm;
+    return i64_to_f32(v).v;
+}
+uint64_t i32tof64(uint32_t v, uint8_t rm) {
+    softfloat_exceptionFlags = 0;
+    softfloat_roundingMode = rm;
+    return i32_to_f64(v).v;
+}
+uint64_t i64tof64(uint64_t v, uint8_t rm) {
+    softfloat_exceptionFlags = 0;
+    softfloat_roundingMode = rm;
+    return i64_to_f64(v).v;
+}
+}

src/vm/fp_functions.h

@@ -39,30 +39,87 @@
 
 extern "C" {
 uint32_t fget_flags();
+
+// half precision
+uint16_t fadd_h(uint16_t v1, uint16_t v2, uint8_t mode);
+uint16_t fsub_h(uint16_t v1, uint16_t v2, uint8_t mode);
+uint16_t fmul_h(uint16_t v1, uint16_t v2, uint8_t mode);
+uint16_t fdiv_h(uint16_t v1, uint16_t v2, uint8_t mode);
+uint16_t fsqrt_h(uint16_t v1, uint8_t mode);
+uint16_t fcmp_h(uint16_t v1, uint16_t v2, uint16_t op);
+uint16_t fmadd_h(uint16_t v1, uint16_t v2, uint16_t v3, uint16_t op, uint8_t mode);
+uint16_t fsel_h(uint16_t v1, uint16_t v2, uint16_t op);
+uint16_t fclass_h(uint16_t v1);
+uint16_t frsqrt7_h(uint16_t v);
+uint16_t frec7_h(uint16_t v, uint8_t mode);
+uint16_t unbox_h(uint8_t FLEN, uint64_t v);
+
+// single precision
 uint32_t fadd_s(uint32_t v1, uint32_t v2, uint8_t mode);
 uint32_t fsub_s(uint32_t v1, uint32_t v2, uint8_t mode);
 uint32_t fmul_s(uint32_t v1, uint32_t v2, uint8_t mode);
 uint32_t fdiv_s(uint32_t v1, uint32_t v2, uint8_t mode);
 uint32_t fsqrt_s(uint32_t v1, uint8_t mode);
 uint32_t fcmp_s(uint32_t v1, uint32_t v2, uint32_t op);
-uint32_t fcvt_s(uint32_t v1, uint32_t op, uint8_t mode);
 uint32_t fmadd_s(uint32_t v1, uint32_t v2, uint32_t v3, uint32_t op, uint8_t mode);
 uint32_t fsel_s(uint32_t v1, uint32_t v2, uint32_t op);
 uint32_t fclass_s(uint32_t v1);
-uint32_t fconv_d2f(uint64_t v1, uint8_t mode);
-uint64_t fconv_f2d(uint32_t v1, uint8_t mode);
+uint32_t frsqrt7_s(uint32_t v);
+uint32_t frec7_s(uint32_t v, uint8_t mode);
+uint32_t unbox_s(uint8_t FLEN, uint64_t v);
+
+// double precision
 uint64_t fadd_d(uint64_t v1, uint64_t v2, uint8_t mode);
 uint64_t fsub_d(uint64_t v1, uint64_t v2, uint8_t mode);
 uint64_t fmul_d(uint64_t v1, uint64_t v2, uint8_t mode);
 uint64_t fdiv_d(uint64_t v1, uint64_t v2, uint8_t mode);
 uint64_t fsqrt_d(uint64_t v1, uint8_t mode);
 uint64_t fcmp_d(uint64_t v1, uint64_t v2, uint32_t op);
-uint64_t fcvt_d(uint64_t v1, uint32_t op, uint8_t mode);
 uint64_t fmadd_d(uint64_t v1, uint64_t v2, uint64_t v3, uint32_t op, uint8_t mode);
 uint64_t fsel_d(uint64_t v1, uint64_t v2, uint32_t op);
 uint64_t fclass_d(uint64_t v1);
-uint64_t fcvt_32_64(uint32_t v1, uint32_t op, uint8_t mode);
-uint32_t fcvt_64_32(uint64_t v1, uint32_t op, uint8_t mode);
-uint32_t unbox_s(uint64_t v);
+uint64_t frsqrt7_d(uint64_t v);
+uint64_t frec7_d(uint64_t v, uint8_t mode);
+uint64_t unbox_d(uint8_t FLEN, uint64_t v);
+
+// conversion: float to float
+uint32_t f16tof32(uint16_t val, uint8_t rm);
+uint64_t f16tof64(uint16_t val, uint8_t rm);
+uint16_t f32tof16(uint32_t val, uint8_t rm);
+uint64_t f32tof64(uint32_t val, uint8_t rm);
+uint16_t f64tof16(uint64_t val, uint8_t rm);
+uint32_t f64tof32(uint64_t val, uint8_t rm);
+
+// conversions: float to unsigned
+uint32_t f16toui32(uint16_t v, uint8_t rm);
+uint64_t f16toui64(uint16_t v, uint8_t rm);
+uint32_t f32toui32(uint32_t v, uint8_t rm);
+uint64_t f32toui64(uint32_t v, uint8_t rm);
+uint32_t f64toui32(uint64_t v, uint8_t rm);
+uint64_t f64toui64(uint64_t v, uint8_t rm);
+
+// conversions: float to signed
+uint32_t f16toi32(uint16_t v, uint8_t rm);
+uint64_t f16toi64(uint16_t v, uint8_t rm);
+uint32_t f32toi32(uint32_t v, uint8_t rm);
+uint64_t f32toi64(uint32_t v, uint8_t rm);
+uint32_t f64toi32(uint64_t v, uint8_t rm);
+uint64_t f64toi64(uint64_t v, uint8_t rm);
+
+// conversions: unsigned to float
+uint16_t ui32tof16(uint32_t v, uint8_t rm);
+uint16_t ui64tof16(uint64_t v, uint8_t rm);
+uint32_t ui32tof32(uint32_t v, uint8_t rm);
+uint32_t ui64tof32(uint64_t v, uint8_t rm);
+uint64_t ui32tof64(uint32_t v, uint8_t rm);
+uint64_t ui64tof64(uint64_t v, uint8_t rm);
+
+// conversions: signed to float
+uint16_t i32tof16(uint32_t v, uint8_t rm);
+uint16_t i64tof16(uint64_t v, uint8_t rm);
+uint32_t i32tof32(uint32_t v, uint8_t rm);
+uint32_t i64tof32(uint64_t v, uint8_t rm);
+uint64_t i32tof64(uint32_t v, uint8_t rm);
+uint64_t i64tof64(uint64_t v, uint8_t rm);
 }
-#endif /* RISCV_SRC_VM_FP_FUNCTIONS_H_ */
+#endif /* _VM_FP_FUNCTIONS_H_ */

src/vm/llvm/fp_impl.cpp

@@ -72,32 +72,70 @@ using namespace ::llvm;
 void add_fp_functions_2_module(Module* mod, uint32_t flen, uint32_t xlen) {
     if(flen) {
         FDECL(fget_flags, INT_TYPE(32));
+
+        FDECL(fadd_h, INT_TYPE(16), INT_TYPE(16), INT_TYPE(16), INT_TYPE(8));
+        FDECL(fsub_h, INT_TYPE(16), INT_TYPE(16), INT_TYPE(16), INT_TYPE(8));
+        FDECL(fmul_h, INT_TYPE(16), INT_TYPE(16), INT_TYPE(16), INT_TYPE(8));
+        FDECL(fdiv_h, INT_TYPE(16), INT_TYPE(16), INT_TYPE(16), INT_TYPE(8));
+        FDECL(fsqrt_h, INT_TYPE(16), INT_TYPE(16), INT_TYPE(8));
+        FDECL(fcmp_h, INT_TYPE(16), INT_TYPE(16), INT_TYPE(16), INT_TYPE(16));
+        FDECL(fmadd_h, INT_TYPE(16), INT_TYPE(16), INT_TYPE(16), INT_TYPE(16), INT_TYPE(16), INT_TYPE(8));
+        FDECL(fsel_h, INT_TYPE(16), INT_TYPE(16), INT_TYPE(16), INT_TYPE(16));
+        FDECL(fclass_h, INT_TYPE(16), INT_TYPE(16));
+        FDECL(unbox_h, INT_TYPE(16), INT_TYPE(32), INT_TYPE(64)); // technically the first arg is only 8 bits
+
+        FDECL(f16toi32, INT_TYPE(32), INT_TYPE(32), INT_TYPE(8))
+        FDECL(f16toui32, INT_TYPE(32), INT_TYPE(32), INT_TYPE(8))
+        FDECL(i32tof16, INT_TYPE(16), INT_TYPE(32), INT_TYPE(8))
+        FDECL(ui32tof16, INT_TYPE(16), INT_TYPE(32), INT_TYPE(8))
+        FDECL(f16toi64, INT_TYPE(64), INT_TYPE(32), INT_TYPE(8))
+        FDECL(f16toui64, INT_TYPE(64), INT_TYPE(32), INT_TYPE(8))
+        FDECL(i64tof16, INT_TYPE(16), INT_TYPE(64), INT_TYPE(8))
+        FDECL(ui64tof16, INT_TYPE(16), INT_TYPE(64), INT_TYPE(8))
+
         FDECL(fadd_s, INT_TYPE(32), INT_TYPE(32), INT_TYPE(32), INT_TYPE(8));
         FDECL(fsub_s, INT_TYPE(32), INT_TYPE(32), INT_TYPE(32), INT_TYPE(8));
         FDECL(fmul_s, INT_TYPE(32), INT_TYPE(32), INT_TYPE(32), INT_TYPE(8));
         FDECL(fdiv_s, INT_TYPE(32), INT_TYPE(32), INT_TYPE(32), INT_TYPE(8));
         FDECL(fsqrt_s, INT_TYPE(32), INT_TYPE(32), INT_TYPE(8));
         FDECL(fcmp_s, INT_TYPE(32), INT_TYPE(32), INT_TYPE(32), INT_TYPE(32));
-        FDECL(fcvt_s, INT_TYPE(32), INT_TYPE(32), INT_TYPE(32), INT_TYPE(8));
         FDECL(fmadd_s, INT_TYPE(32), INT_TYPE(32), INT_TYPE(32), INT_TYPE(32), INT_TYPE(32), INT_TYPE(8));
         FDECL(fsel_s, INT_TYPE(32), INT_TYPE(32), INT_TYPE(32), INT_TYPE(32));
         FDECL(fclass_s, INT_TYPE(32), INT_TYPE(32));
-        FDECL(fcvt_32_64, INT_TYPE(64), INT_TYPE(32), INT_TYPE(32), INT_TYPE(8));
-        FDECL(fcvt_64_32, INT_TYPE(32), INT_TYPE(64), INT_TYPE(32), INT_TYPE(8));
+        FDECL(unbox_s, INT_TYPE(32), INT_TYPE(32), INT_TYPE(64)); // technically the first arg is only 8 bits
+
+        FDECL(f32toi32, INT_TYPE(32), INT_TYPE(32), INT_TYPE(8));
+        FDECL(f32toui32, INT_TYPE(32), INT_TYPE(32), INT_TYPE(8));
+        FDECL(i32tof32, INT_TYPE(32), INT_TYPE(32), INT_TYPE(8));
+        FDECL(ui32tof32, INT_TYPE(32), INT_TYPE(32), INT_TYPE(8));
+        FDECL(f32toi64, INT_TYPE(64), INT_TYPE(32), INT_TYPE(8));
+        FDECL(f32toui64, INT_TYPE(64), INT_TYPE(32), INT_TYPE(8));
+        FDECL(i64tof32, INT_TYPE(32), INT_TYPE(64), INT_TYPE(8));
+        FDECL(ui64tof32, INT_TYPE(32), INT_TYPE(64), INT_TYPE(8));
         if(flen > 32) {
-            FDECL(fconv_d2f, INT_TYPE(32), INT_TYPE(64), INT_TYPE(8));
-            FDECL(fconv_f2d, INT_TYPE(64), INT_TYPE(32), INT_TYPE(8));
+
             FDECL(fadd_d, INT_TYPE(64), INT_TYPE(64), INT_TYPE(64), INT_TYPE(8));
             FDECL(fsub_d, INT_TYPE(64), INT_TYPE(64), INT_TYPE(64), INT_TYPE(8));
             FDECL(fmul_d, INT_TYPE(64), INT_TYPE(64), INT_TYPE(64), INT_TYPE(8));
             FDECL(fdiv_d, INT_TYPE(64), INT_TYPE(64), INT_TYPE(64), INT_TYPE(8));
             FDECL(fsqrt_d, INT_TYPE(64), INT_TYPE(64), INT_TYPE(8));
             FDECL(fcmp_d, INT_TYPE(64), INT_TYPE(64), INT_TYPE(64), INT_TYPE(32));
-            FDECL(fcvt_d, INT_TYPE(64), INT_TYPE(64), INT_TYPE(32), INT_TYPE(8));
             FDECL(fmadd_d, INT_TYPE(64), INT_TYPE(64), INT_TYPE(64), INT_TYPE(64), INT_TYPE(32), INT_TYPE(8));
             FDECL(fsel_d, INT_TYPE(64), INT_TYPE(64), INT_TYPE(64), INT_TYPE(32));
             FDECL(fclass_d, INT_TYPE(64), INT_TYPE(64));
-            FDECL(unbox_s, INT_TYPE(32), INT_TYPE(64));
+
+            FDECL(f64tof32, INT_TYPE(32), INT_TYPE(64), INT_TYPE(8));
+            FDECL(f32tof64, INT_TYPE(64), INT_TYPE(32), INT_TYPE(8));
+            FDECL(f64toi64, INT_TYPE(64), INT_TYPE(64), INT_TYPE(8));
+            FDECL(f64toui64, INT_TYPE(64), INT_TYPE(64), INT_TYPE(8));
+            FDECL(i64tof64, INT_TYPE(64), INT_TYPE(64), INT_TYPE(8));
+            FDECL(ui64tof64, INT_TYPE(64), INT_TYPE(64), INT_TYPE(8));
+            FDECL(i32tof64, INT_TYPE(64), INT_TYPE(32), INT_TYPE(8));
+            FDECL(ui32tof64, INT_TYPE(64), INT_TYPE(32), INT_TYPE(8));
+            FDECL(f64toi32, INT_TYPE(32), INT_TYPE(64), INT_TYPE(8));
+            FDECL(f64toui32, INT_TYPE(32), INT_TYPE(64), INT_TYPE(8));
+
+            FDECL(unbox_d, INT_TYPE(64), INT_TYPE(32), INT_TYPE(64)); // technically the first arg is only 8 bits
         }
     }
 }

src/vm/llvm/vm_tgc5c.cpp

@@ -31,10 +31,12 @@
  *******************************************************************************/
 // clang-format off
 #include <iss/arch/tgc5c.h>
+// vm_base needs to be included before gdb_session as termios.h (via boost and gdb_server) has a define which clashes with a variable
+// name in ConstantRange.h
+#include <iss/llvm/vm_base.h>
 #include <iss/debugger/gdb_session.h>
 #include <iss/debugger/server.h>
 #include <iss/iss.h>
-#include <iss/llvm/vm_base.h>
 #include <util/logging.h>
 #include <iss/instruction_decoder.h>
 
@@ -353,7 +355,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             if(rd!=0) {
@@ -399,7 +401,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             if(rd!=0) {
@@ -446,7 +448,7 @@ private:
         /*generate behavior*/
         this->builder.CreateStore(this->gen_const(32U, static_cast<int>(NO_JUMP)), get_reg_ptr(traits::LAST_BRANCH), false);
         if(rd>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             auto new_pc =(uint32_t)(PC+(int32_t)sext<21>(imm));
@@ -503,7 +505,7 @@ private:
         /*generate behavior*/
         this->builder.CreateStore(this->gen_const(32U, static_cast<int>(NO_JUMP)), get_reg_ptr(traits::LAST_BRANCH), false);
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             auto addr_mask =(uint32_t)- 2;
@@ -584,7 +586,7 @@ private:
         /*generate behavior*/
         this->builder.CreateStore(this->gen_const(32U, static_cast<int>(NO_JUMP)), get_reg_ptr(traits::LAST_BRANCH), false);
         if(rs2>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             {
@@ -649,7 +651,7 @@ private:
         /*generate behavior*/
         this->builder.CreateStore(this->gen_const(32U, static_cast<int>(NO_JUMP)), get_reg_ptr(traits::LAST_BRANCH), false);
         if(rs2>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             {
@@ -714,7 +716,7 @@ private:
         /*generate behavior*/
         this->builder.CreateStore(this->gen_const(32U, static_cast<int>(NO_JUMP)), get_reg_ptr(traits::LAST_BRANCH), false);
         if(rs2>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             {
@@ -783,7 +785,7 @@ private:
         /*generate behavior*/
         this->builder.CreateStore(this->gen_const(32U, static_cast<int>(NO_JUMP)), get_reg_ptr(traits::LAST_BRANCH), false);
         if(rs2>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             {
@@ -852,7 +854,7 @@ private:
         /*generate behavior*/
         this->builder.CreateStore(this->gen_const(32U, static_cast<int>(NO_JUMP)), get_reg_ptr(traits::LAST_BRANCH), false);
         if(rs2>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             {
@@ -917,7 +919,7 @@ private:
         /*generate behavior*/
         this->builder.CreateStore(this->gen_const(32U, static_cast<int>(NO_JUMP)), get_reg_ptr(traits::LAST_BRANCH), false);
         if(rs2>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             {
@@ -981,7 +983,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             auto load_address =this->gen_ext(
@@ -1039,7 +1041,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             auto load_address =this->gen_ext(
@@ -1097,7 +1099,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             auto load_address =this->gen_ext(
@@ -1155,7 +1157,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             auto load_address =this->gen_ext(
@@ -1211,7 +1213,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             auto load_address =this->gen_ext(
@@ -1267,7 +1269,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rs2>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             auto store_address =this->gen_ext(
@@ -1320,7 +1322,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rs2>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             auto store_address =this->gen_ext(
@@ -1373,7 +1375,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rs2>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             auto store_address =this->gen_ext(
@@ -1426,7 +1428,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             if(rd!=0) {
@@ -1478,7 +1480,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             if(rd!=0) {
@@ -1532,7 +1534,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             if(rd!=0) {
@@ -1585,7 +1587,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             if(rd!=0) {
@@ -1635,7 +1637,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             if(rd!=0) {
@@ -1685,7 +1687,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             if(rd!=0) {
@@ -1735,7 +1737,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             if(rd!=0) {
@@ -1785,7 +1787,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             if(rd!=0) {
@@ -1835,7 +1837,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             if(rd!=0) {
@@ -1887,7 +1889,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             if(rd!=0) {
@@ -1939,7 +1941,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             if(rd!=0) {
@@ -1991,7 +1993,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             if(rd!=0) {
@@ -2044,7 +2046,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             if(rd!=0) {
@@ -2099,7 +2101,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             if(rd!=0) {
@@ -2152,7 +2154,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             if(rd!=0) {
@@ -2202,7 +2204,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             if(rd!=0) {
@@ -2255,7 +2257,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             if(rd!=0) {
@@ -2311,7 +2313,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             if(rd!=0) {
@@ -2361,7 +2363,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             if(rd!=0) {
@@ -2553,10 +2555,10 @@ private:
         
         this->gen_instr_prologue();
         /*generate behavior*/
-        std::vector<Value*> wait_231_args{
+        std::vector<Value*> wait_57_args{
             this->gen_ext(this->gen_const(8,1), 32)
         };
-        this->builder.CreateCall(this->mod->getFunction("wait"), wait_231_args);
+        this->builder.CreateCall(this->mod->getFunction("wait"), wait_57_args);
         bb = BasicBlock::Create(this->mod->getContext(), "entry", this->func, this->leave_blk);
         auto returnValue = std::make_tuple(CONT,bb);
         
@@ -2595,7 +2597,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             auto xrs1 =this->gen_reg_load(traits::X0+ rs1);
@@ -2651,7 +2653,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             auto xrd =this->gen_read_mem(traits::CSR, csr, 4);
@@ -2708,7 +2710,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             auto xrd =this->gen_read_mem(traits::CSR, csr, 4);
@@ -2765,7 +2767,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             auto xrd =this->gen_read_mem(traits::CSR, csr, 4);
@@ -2816,7 +2818,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             auto xrd =this->gen_read_mem(traits::CSR, csr, 4);
@@ -2872,7 +2874,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             auto xrd =this->gen_read_mem(traits::CSR, csr, 4);
@@ -2968,7 +2970,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             auto res =this->builder.CreateMul(
@@ -3023,7 +3025,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             auto res =this->builder.CreateMul(
@@ -3081,7 +3083,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             auto res =this->builder.CreateMul(
@@ -3138,7 +3140,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             auto res =this->builder.CreateMul(
@@ -3194,7 +3196,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             auto dividend =this->gen_ext(
@@ -3298,7 +3300,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             {
@@ -3371,7 +3373,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             {
@@ -3479,7 +3481,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             {
@@ -3561,7 +3563,7 @@ private:
             get_reg_ptr(rd+8 + traits::X0), false);
         }
         else{
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         bb = BasicBlock::Create(this->mod->getContext(), "entry", this->func, this->leave_blk);
         auto returnValue = std::make_tuple(CONT,bb);
@@ -3698,7 +3700,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rs1>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             if(rs1!=0) {
@@ -3825,7 +3827,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             if(rd!=0) {
@@ -3871,7 +3873,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(imm==0||rd>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         if(rd!=0) {
             this->builder.CreateStore(
@@ -3924,7 +3926,7 @@ private:
             get_reg_ptr(2 + traits::X0), false);
         }
         else{
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         bb = BasicBlock::Create(this->mod->getContext(), "entry", this->func, this->leave_blk);
         auto returnValue = std::make_tuple(CONT,bb);
@@ -3960,7 +3962,7 @@ private:
         
         this->gen_instr_prologue();
         /*generate behavior*/
-        this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+        this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         bb = BasicBlock::Create(this->mod->getContext(), "entry", this->func, this->leave_blk);
         auto returnValue = std::make_tuple(CONT,bb);
         
@@ -4459,7 +4461,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rs1>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             if(rs1!=0) {
@@ -4508,7 +4510,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rd==0) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             auto offs =this->gen_ext(
@@ -4562,7 +4564,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             if(rd!=0) {
@@ -4689,7 +4691,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             if(rd!=0) {
@@ -4740,7 +4742,7 @@ private:
         /*generate behavior*/
         this->builder.CreateStore(this->gen_const(32U, static_cast<int>(NO_JUMP)), get_reg_ptr(traits::LAST_BRANCH), false);
         if(rs1>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             auto addr_mask =(uint32_t)- 2;
@@ -4826,7 +4828,7 @@ private:
         this->gen_instr_prologue();
         /*generate behavior*/
         if(rs2>=static_cast<uint32_t>(traits::RFS)) {
-            this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
             auto offs =this->gen_ext(
@@ -4874,7 +4876,7 @@ private:
         
         this->gen_instr_prologue();
         /*generate behavior*/
-        this->gen_raise_trap(0, static_cast<int32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
+        this->gen_raise_trap(0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         bb = BasicBlock::Create(this->mod->getContext(), "entry", this->func, this->leave_blk);
         auto returnValue = std::make_tuple(CONT,bb);
         
@@ -4944,8 +4946,6 @@ vm_impl<ARCH>::gen_single_inst_behavior(virt_addr_t &pc, BasicBlock *this_block)
     // const typename traits::addr_t upper_bits = ~traits::PGMASK;
     phys_addr_t paddr(pc);
     auto *const data = (uint8_t *)&instr;
-    if(this->core.has_mmu())
-        paddr = this->core.virt2phys(pc);
     auto res = this->core.read(paddr, 4, data);
     if (res != iss::Ok) 
         return std::make_tuple(ILLEGAL_FETCH, nullptr);

src/vm/tcc/vm_tgc5c.cpp

@@ -348,7 +348,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	if(rd!=0) {
@@ -384,7 +384,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	if(rd!=0) {
@@ -421,7 +421,7 @@ private:
         this->gen_set_tval(tu, instr);
         tu.store(traits::LAST_BRANCH, tu.constant(static_cast<int>(NO_JUMP),32));
         if(rd>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	auto new_pc = (uint32_t)(PC+(int32_t)sext<21>(imm));
@@ -468,7 +468,7 @@ private:
         this->gen_set_tval(tu, instr);
         tu.store(traits::LAST_BRANCH, tu.constant(static_cast<int>(NO_JUMP),32));
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	auto addr_mask = (uint32_t)- 2;
@@ -526,7 +526,7 @@ private:
         this->gen_set_tval(tu, instr);
         tu.store(traits::LAST_BRANCH, tu.constant(static_cast<int>(NO_JUMP),32));
         if(rs2>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	tu.open_if(tu.icmp(ICmpInst::ICMP_EQ,
@@ -576,7 +576,7 @@ private:
         this->gen_set_tval(tu, instr);
         tu.store(traits::LAST_BRANCH, tu.constant(static_cast<int>(NO_JUMP),32));
         if(rs2>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	tu.open_if(tu.icmp(ICmpInst::ICMP_NE,
@@ -626,7 +626,7 @@ private:
         this->gen_set_tval(tu, instr);
         tu.store(traits::LAST_BRANCH, tu.constant(static_cast<int>(NO_JUMP),32));
         if(rs2>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	tu.open_if(tu.icmp(ICmpInst::ICMP_SLT,
@@ -676,7 +676,7 @@ private:
         this->gen_set_tval(tu, instr);
         tu.store(traits::LAST_BRANCH, tu.constant(static_cast<int>(NO_JUMP),32));
         if(rs2>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	tu.open_if(tu.icmp(ICmpInst::ICMP_SGE,
@@ -726,7 +726,7 @@ private:
         this->gen_set_tval(tu, instr);
         tu.store(traits::LAST_BRANCH, tu.constant(static_cast<int>(NO_JUMP),32));
         if(rs2>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	tu.open_if(tu.icmp(ICmpInst::ICMP_ULT,
@@ -776,7 +776,7 @@ private:
         this->gen_set_tval(tu, instr);
         tu.store(traits::LAST_BRANCH, tu.constant(static_cast<int>(NO_JUMP),32));
         if(rs2>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	tu.open_if(tu.icmp(ICmpInst::ICMP_UGE,
@@ -825,7 +825,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	auto load_address = tu.assignment(tu.ext((tu.add(
@@ -866,7 +866,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	auto load_address = tu.assignment(tu.ext((tu.add(
@@ -907,7 +907,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	auto load_address = tu.assignment(tu.ext((tu.add(
@@ -948,7 +948,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	auto load_address = tu.assignment(tu.ext((tu.add(
@@ -989,7 +989,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	auto load_address = tu.assignment(tu.ext((tu.add(
@@ -1030,7 +1030,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rs2>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	auto store_address = tu.assignment(tu.ext((tu.add(
@@ -1068,7 +1068,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rs2>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	auto store_address = tu.assignment(tu.ext((tu.add(
@@ -1106,7 +1106,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rs2>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	auto store_address = tu.assignment(tu.ext((tu.add(
@@ -1144,7 +1144,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	if(rd!=0) {
@@ -1183,7 +1183,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	if(rd!=0) {
@@ -1222,7 +1222,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	if(rd!=0) {
@@ -1261,7 +1261,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	if(rd!=0) {
@@ -1300,7 +1300,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	if(rd!=0) {
@@ -1339,7 +1339,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	if(rd!=0) {
@@ -1378,7 +1378,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	if(rd!=0) {
@@ -1417,7 +1417,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	if(rd!=0) {
@@ -1456,7 +1456,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	if(rd!=0) {
@@ -1495,7 +1495,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)||rs2>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	if(rd!=0) {
@@ -1534,7 +1534,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)||rs2>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	if(rd!=0) {
@@ -1573,7 +1573,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)||rs2>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	if(rd!=0) {
@@ -1614,7 +1614,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)||rs2>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	if(rd!=0) {
@@ -1653,7 +1653,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)||rs2>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	if(rd!=0) {
@@ -1692,7 +1692,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)||rs2>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	if(rd!=0) {
@@ -1731,7 +1731,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)||rs2>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	if(rd!=0) {
@@ -1772,7 +1772,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)||rs2>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	if(rd!=0) {
@@ -1813,7 +1813,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)||rs2>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	if(rd!=0) {
@@ -1852,7 +1852,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)||rs2>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	if(rd!=0) {
@@ -2030,7 +2030,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	auto xrs1 = tu.assignment(tu.load(rs1 + traits::X0, 0),32);
@@ -2072,7 +2072,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	auto xrd = tu.assignment(tu.read_mem(traits::CSR, csr, 32),32);
@@ -2116,7 +2116,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	auto xrd = tu.assignment(tu.read_mem(traits::CSR, csr, 32),32);
@@ -2160,7 +2160,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	auto xrd = tu.assignment(tu.read_mem(traits::CSR, csr, 32),32);
@@ -2199,7 +2199,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	auto xrd = tu.assignment(tu.read_mem(traits::CSR, csr, 32),32);
@@ -2242,7 +2242,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	auto xrd = tu.assignment(tu.read_mem(traits::CSR, csr, 32),32);
@@ -2315,7 +2315,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)||rs2>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	auto res = tu.assignment(tu.mul(
@@ -2355,7 +2355,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)||rs2>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	auto res = tu.assignment(tu.mul(
@@ -2397,7 +2397,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)||rs2>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	auto res = tu.assignment(tu.mul(
@@ -2439,7 +2439,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)||rs2>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	auto res = tu.assignment(tu.mul(
@@ -2481,7 +2481,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)||rs2>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	auto dividend = tu.assignment(tu.ext(tu.load(rs1 + traits::X0, 0),32,true),32);
@@ -2546,7 +2546,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)||rs2>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	tu.open_if(tu.icmp(ICmpInst::ICMP_NE,
@@ -2597,7 +2597,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)||rs2>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	tu.open_if(tu.icmp(ICmpInst::ICMP_NE,
@@ -2665,7 +2665,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rs1>=static_cast<uint32_t>(traits:: RFS)||rs2>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	tu.open_if(tu.icmp(ICmpInst::ICMP_NE,
@@ -2720,7 +2720,7 @@ private:
                tu.constant(imm,16))),32,false));
         }
         else{
-        	this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+        	this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         auto returnValue = CONT;
         
@@ -2817,7 +2817,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rs1>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	if(rs1!=0) {
@@ -2913,7 +2913,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	if(rd!=0) {
@@ -2949,7 +2949,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(imm==0||rd>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         if(rd!=0) {
             tu.store(rd + traits::X0, tu.constant((uint32_t)((int32_t)sext<18>(imm)),32));
@@ -2987,7 +2987,7 @@ private:
                tu.constant((int16_t)sext<10>(nzimm),16))),32,false));
         }
         else{
-        	this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+        	this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         auto returnValue = CONT;
         
@@ -3015,7 +3015,7 @@ private:
         tu("(*cycle)++;");
         tu.open_scope();
         this->gen_set_tval(tu, instr);
-        this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+        this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         auto returnValue = CONT;
         
         tu.close_scope();
@@ -3376,7 +3376,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rs1>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	if(rs1!=0) {
@@ -3414,7 +3414,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)||rd==0) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	auto offs = tu.assignment(tu.ext((tu.add(
@@ -3451,7 +3451,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	if(rd!=0) {
@@ -3551,7 +3551,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rd>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	if(rd!=0) {
@@ -3589,7 +3589,7 @@ private:
         this->gen_set_tval(tu, instr);
         tu.store(traits::LAST_BRANCH, tu.constant(static_cast<int>(NO_JUMP),32));
         if(rs1>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	auto addr_mask = (uint32_t)- 2;
@@ -3656,7 +3656,7 @@ private:
         tu.open_scope();
         this->gen_set_tval(tu, instr);
         if(rs2>=static_cast<uint32_t>(traits:: RFS)) {
-            this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+            this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         }
         else{
         	auto offs = tu.assignment(tu.ext((tu.add(
@@ -3689,7 +3689,7 @@ private:
         tu("(*cycle)++;");
         tu.open_scope();
         this->gen_set_tval(tu, instr);
-        this->gen_raise_trap(tu, 0, static_cast<int32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
+        this->gen_raise_trap(tu, 0, static_cast<uint32_t>(traits:: RV_CAUSE_ILLEGAL_INSTRUCTION));
         auto returnValue = CONT;
         
         tu.close_scope();
@@ -3743,8 +3743,6 @@ vm_impl<ARCH>::gen_single_inst_behavior(virt_addr_t &pc, tu_builder& tu) {
     enum {TRAP_ID=1<<16};
     code_word_t instr = 0;
     phys_addr_t paddr(pc);
-    if(this->core.has_mmu())
-        paddr = this->core.virt2phys(pc);
     auto res = this->core.read(paddr, 4, reinterpret_cast<uint8_t*>(&instr));
     if (res != iss::Ok)
         return ILLEGAL_FETCH;

src/vm/vector_functions.cpp

@@ -0,0 +1,407 @@
+////////////////////////////////////////////////////////////////////////////////
+// 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:
+//       alex@minres.com - initial API and implementation
+////////////////////////////////////////////////////////////////////////////////
+
+#include "vector_functions.h"
+#include "iss/vm_types.h"
+#include "vm/aes_sbox.h"
+#include <algorithm>
+#include <cassert>
+#include <cstddef>
+#include <cstdint>
+#include <functional>
+#include <limits>
+#include <math.h>
+#include <stdexcept>
+#include <vector>
+
+namespace softvector {
+
+bool softvec_read(void* core, uint64_t addr, uint64_t length, uint8_t* data) {
+    // Read length bytes from addr into *data
+    iss::status status = static_cast<iss::arch_if*>(core)->read(iss::address_type::PHYSICAL, iss::access_type::READ,
+                                                                0 /*traits<ARCH>::MEM*/, addr, length, data);
+    return status == iss::Ok;
+}
+bool softvec_write(void* core, uint64_t addr, uint64_t length, uint8_t* data) {
+    // Write length bytes from addr into *data
+    iss::status status = static_cast<iss::arch_if*>(core)->write(iss::address_type::PHYSICAL, iss::access_type::READ,
+                                                                 0 /*traits<ARCH>::MEM*/, addr, length, data);
+    return status == iss::Ok;
+}
+
+vtype_t::vtype_t(uint32_t vtype_val) { underlying = (vtype_val & 0x8000) << 32 | (vtype_val & ~0x8000); }
+vtype_t::vtype_t(uint64_t vtype_val) { underlying = vtype_val; }
+bool vtype_t::vill() { return underlying >> 63; }
+bool vtype_t::vma() { return (underlying >> 7) & 1; }
+bool vtype_t::vta() { return (underlying >> 6) & 1; }
+unsigned vtype_t::sew() {
+    uint8_t vsew = (underlying >> 3) & 0b111;
+    // pow(2, 3 + vsew);
+    return 1 << (3 + vsew);
+}
+double vtype_t::lmul() {
+    uint8_t vlmul = underlying & 0b111;
+    assert(vlmul != 0b100); // reserved encoding
+    int8_t signed_vlmul = (vlmul >> 2) ? 0b11111000 | vlmul : vlmul;
+    return pow(2, signed_vlmul);
+}
+
+mask_bit_reference& mask_bit_reference::operator=(const bool new_value) {
+    *start = *start & ~(1U << pos) | static_cast<unsigned>(new_value) << pos;
+    return *this;
+}
+
+mask_bit_reference::mask_bit_reference(uint8_t* start, uint8_t pos)
+: start(start)
+, pos(pos) {
+    assert(pos < 8 && "Bit reference can only be initialized for bytes");
+};
+mask_bit_reference::operator bool() const { return *(start) & (1U << (pos)); }
+
+mask_bit_reference vmask_view::operator[](size_t idx) const {
+    assert(idx < elem_count);
+    return {start + idx / 8, static_cast<uint8_t>(idx % 8)};
+}
+
+vmask_view read_vmask(uint8_t* V, uint16_t VLEN, uint16_t elem_count, uint8_t reg_idx) {
+    uint8_t* mask_start = V + VLEN / 8 * reg_idx;
+    assert(mask_start + elem_count / 8 <= V + VLEN * RFS / 8);
+    return {mask_start, elem_count};
+}
+uint8_t xt2(uint8_t x) { return (x << 1) ^ (bit_sub<7, 1>(x) ? 27 : 0); }
+
+uint8_t xt3(uint8_t x) { return x ^ xt2(x); }
+
+uint8_t gfmul(uint8_t x, uint8_t y) {
+    return (bit_sub<0, 1>(y) ? x : 0) ^ (bit_sub<1, 1>(y) ? xt2(x) : 0) ^ (bit_sub<2, 1>(y) ? xt2(xt2(x)) : 0) ^
+           (bit_sub<3, 1>(y) ? xt2(xt2(xt2(x))) : 0);
+}
+
+uint32_t aes_mixcolumn_byte_fwd(uint8_t so) {
+    return ((uint32_t)gfmul(so, 3) << 24) | ((uint32_t)so << 16) | ((uint32_t)so << 8) | gfmul(so, 2);
+}
+
+uint32_t aes_mixcolumn_byte_inv(uint8_t so) {
+    return ((uint32_t)gfmul(so, 11) << 24) | ((uint32_t)gfmul(so, 13) << 16) | ((uint32_t)gfmul(so, 9) << 8) | gfmul(so, 14);
+}
+
+uint32_t aes_mixcolumn_fwd(uint32_t x) {
+    uint8_t s0 = bit_sub<0, 7 - 0 + 1>(x);
+    uint8_t s1 = bit_sub<8, 15 - 8 + 1>(x);
+    uint8_t s2 = bit_sub<16, 23 - 16 + 1>(x);
+    uint8_t s3 = bit_sub<24, 31 - 24 + 1>(x);
+    uint8_t b0 = xt2(s0) ^ xt3(s1) ^ (s2) ^ (s3);
+    uint8_t b1 = (s0) ^ xt2(s1) ^ xt3(s2) ^ (s3);
+    uint8_t b2 = (s0) ^ (s1) ^ xt2(s2) ^ xt3(s3);
+    uint8_t b3 = xt3(s0) ^ (s1) ^ (s2) ^ xt2(s3);
+    return ((uint32_t)b3 << 24) | ((uint32_t)b2 << 16) | ((uint32_t)b1 << 8) | b0;
+}
+
+uint32_t aes_mixcolumn_inv(uint32_t x) {
+    uint8_t s0 = bit_sub<0, 7 - 0 + 1>(x);
+    uint8_t s1 = bit_sub<8, 15 - 8 + 1>(x);
+    uint8_t s2 = bit_sub<16, 23 - 16 + 1>(x);
+    uint8_t s3 = bit_sub<24, 31 - 24 + 1>(x);
+    uint8_t b0 = gfmul(s0, 14) ^ gfmul(s1, 11) ^ gfmul(s2, 13) ^ gfmul(s3, 9);
+    uint8_t b1 = gfmul(s0, 9) ^ gfmul(s1, 14) ^ gfmul(s2, 11) ^ gfmul(s3, 13);
+    uint8_t b2 = gfmul(s0, 13) ^ gfmul(s1, 9) ^ gfmul(s2, 14) ^ gfmul(s3, 11);
+    uint8_t b3 = gfmul(s0, 11) ^ gfmul(s1, 13) ^ gfmul(s2, 9) ^ gfmul(s3, 14);
+    return ((uint32_t)b3 << 24) | ((uint32_t)b2 << 16) | ((uint32_t)b1 << 8) | b0;
+}
+
+uint32_t aes_decode_rcon(uint8_t r) {
+    switch(r) {
+    case 0:
+        return 1;
+    case 1:
+        return 2;
+    case 2:
+        return 4;
+    case 3:
+        return 8;
+    case 4:
+        return 16;
+    case 5:
+        return 32;
+    case 6:
+        return 64;
+    case 7:
+        return 128;
+    case 8:
+        return 27;
+    case 9:
+        return 54;
+    }
+    return 0;
+}
+
+uint32_t aes_subword_fwd(uint32_t x) {
+    return ((uint32_t)aes_sbox_fwd(bit_sub<24, 31 - 24 + 1>(x)) << 24) | ((uint32_t)aes_sbox_fwd(bit_sub<16, 23 - 16 + 1>(x)) << 16) |
+           ((uint32_t)aes_sbox_fwd(bit_sub<8, 15 - 8 + 1>(x)) << 8) | aes_sbox_fwd(bit_sub<0, 7 - 0 + 1>(x));
+}
+
+uint32_t aes_subword_inv(uint32_t x) {
+    return ((uint32_t)aes_sbox_inv(bit_sub<24, 31 - 24 + 1>(x)) << 24) | ((uint32_t)aes_sbox_inv(bit_sub<16, 23 - 16 + 1>(x)) << 16) |
+           ((uint32_t)aes_sbox_inv(bit_sub<8, 15 - 8 + 1>(x)) << 8) | aes_sbox_inv(bit_sub<0, 7 - 0 + 1>(x));
+}
+
+uint32_t aes_get_column(__uint128_t state, unsigned c) {
+    assert(c < 4);
+    return static_cast<uint32_t>(state >> (32 * c));
+};
+
+uint64_t aes_apply_fwd_sbox_to_each_byte(uint64_t x) {
+    return ((uint64_t)aes_sbox_fwd(bit_sub<56, 63 - 56 + 1>(x)) << 56) | ((uint64_t)aes_sbox_fwd(bit_sub<48, 55 - 48 + 1>(x)) << 48) |
+           ((uint64_t)aes_sbox_fwd(bit_sub<40, 47 - 40 + 1>(x)) << 40) | ((uint64_t)aes_sbox_fwd(bit_sub<32, 39 - 32 + 1>(x)) << 32) |
+           ((uint64_t)aes_sbox_fwd(bit_sub<24, 31 - 24 + 1>(x)) << 24) | ((uint64_t)aes_sbox_fwd(bit_sub<16, 23 - 16 + 1>(x)) << 16) |
+           ((uint64_t)aes_sbox_fwd(bit_sub<8, 15 - 8 + 1>(x)) << 8) | aes_sbox_fwd(bit_sub<0, 7 - 0 + 1>(x));
+}
+
+uint64_t aes_apply_inv_sbox_to_each_byte(uint64_t x) {
+    return ((uint64_t)aes_sbox_inv(bit_sub<56, 63 - 56 + 1>(x)) << 56) | ((uint64_t)aes_sbox_inv(bit_sub<48, 55 - 48 + 1>(x)) << 48) |
+           ((uint64_t)aes_sbox_inv(bit_sub<40, 47 - 40 + 1>(x)) << 40) | ((uint64_t)aes_sbox_inv(bit_sub<32, 39 - 32 + 1>(x)) << 32) |
+           ((uint64_t)aes_sbox_inv(bit_sub<24, 31 - 24 + 1>(x)) << 24) | ((uint64_t)aes_sbox_inv(bit_sub<16, 23 - 16 + 1>(x)) << 16) |
+           ((uint64_t)aes_sbox_inv(bit_sub<8, 15 - 8 + 1>(x)) << 8) | aes_sbox_inv(bit_sub<0, 7 - 0 + 1>(x));
+}
+
+uint64_t aes_rv64_shiftrows_fwd(uint64_t rs2, uint64_t rs1) {
+    return ((uint64_t)bit_sub<24, 31 - 24 + 1>(rs1) << 56) | ((uint64_t)bit_sub<48, 55 - 48 + 1>(rs2) << 48) |
+           ((uint64_t)bit_sub<8, 15 - 8 + 1>(rs2) << 40) | ((uint64_t)bit_sub<32, 39 - 32 + 1>(rs1) << 32) |
+           ((uint64_t)bit_sub<56, 63 - 56 + 1>(rs2) << 24) | ((uint64_t)bit_sub<16, 23 - 16 + 1>(rs2) << 16) |
+           ((uint64_t)bit_sub<40, 47 - 40 + 1>(rs1) << 8) | bit_sub<0, 7 - 0 + 1>(rs1);
+}
+
+uint64_t aes_rv64_shiftrows_inv(uint64_t rs2, uint64_t rs1) {
+    return ((uint64_t)bit_sub<24, 31 - 24 + 1>(rs2) << 56) | ((uint64_t)bit_sub<48, 55 - 48 + 1>(rs2) << 48) |
+           ((uint64_t)bit_sub<8, 15 - 8 + 1>(rs1) << 40) | ((uint64_t)bit_sub<32, 39 - 32 + 1>(rs1) << 32) |
+           ((uint64_t)bit_sub<56, 63 - 56 + 1>(rs1) << 24) | ((uint64_t)bit_sub<16, 23 - 16 + 1>(rs2) << 16) |
+           ((uint64_t)bit_sub<40, 47 - 40 + 1>(rs2) << 8) | bit_sub<0, 7 - 0 + 1>(rs1);
+}
+
+uint128_t aes_shift_rows_fwd(uint128_t x) {
+    uint32_t ic3 = aes_get_column(x, 3);
+    uint32_t ic2 = aes_get_column(x, 2);
+    uint32_t ic1 = aes_get_column(x, 1);
+    uint32_t ic0 = aes_get_column(x, 0);
+    uint32_t oc0 = ((uint32_t)bit_sub<24, 31 - 24 + 1>(ic3) << 24) | ((uint32_t)bit_sub<16, 23 - 16 + 1>(ic2) << 16) |
+                   ((uint32_t)bit_sub<8, 15 - 8 + 1>(ic1) << 8) | bit_sub<0, 7 - 0 + 1>(ic0);
+    uint32_t oc1 = ((uint32_t)bit_sub<24, 31 - 24 + 1>(ic0) << 24) | ((uint32_t)bit_sub<16, 23 - 16 + 1>(ic3) << 16) |
+                   ((uint32_t)bit_sub<8, 15 - 8 + 1>(ic2) << 8) | bit_sub<0, 7 - 0 + 1>(ic1);
+    uint32_t oc2 = ((uint32_t)bit_sub<24, 31 - 24 + 1>(ic1) << 24) | ((uint32_t)bit_sub<16, 23 - 16 + 1>(ic0) << 16) |
+                   ((uint32_t)bit_sub<8, 15 - 8 + 1>(ic3) << 8) | bit_sub<0, 7 - 0 + 1>(ic2);
+    uint32_t oc3 = ((uint32_t)bit_sub<24, 31 - 24 + 1>(ic2) << 24) | ((uint32_t)bit_sub<16, 23 - 16 + 1>(ic1) << 16) |
+                   ((uint32_t)bit_sub<8, 15 - 8 + 1>(ic0) << 8) | bit_sub<0, 7 - 0 + 1>(ic3);
+    return ((uint128_t)oc3 << 96) | ((uint128_t)oc2 << 64) | ((uint128_t)oc1 << 32) | oc0;
+}
+
+uint128_t aes_shift_rows_inv(uint128_t x) {
+    uint32_t ic3 = aes_get_column(x, 3);
+    uint32_t ic2 = aes_get_column(x, 2);
+    uint32_t ic1 = aes_get_column(x, 1);
+    uint32_t ic0 = aes_get_column(x, 0);
+    uint32_t oc0 = ((uint32_t)bit_sub<24, 31 - 24 + 1>(ic1) << 24) | ((uint32_t)bit_sub<16, 23 - 16 + 1>(ic2) << 16) |
+                   ((uint32_t)bit_sub<8, 15 - 8 + 1>(ic3) << 8) | bit_sub<0, 7 - 0 + 1>(ic0);
+    uint32_t oc1 = ((uint32_t)bit_sub<24, 31 - 24 + 1>(ic2) << 24) | ((uint32_t)bit_sub<16, 23 - 16 + 1>(ic3) << 16) |
+                   ((uint32_t)bit_sub<8, 15 - 8 + 1>(ic0) << 8) | bit_sub<0, 7 - 0 + 1>(ic1);
+    uint32_t oc2 = ((uint32_t)bit_sub<24, 31 - 24 + 1>(ic3) << 24) | ((uint32_t)bit_sub<16, 23 - 16 + 1>(ic0) << 16) |
+                   ((uint32_t)bit_sub<8, 15 - 8 + 1>(ic1) << 8) | bit_sub<0, 7 - 0 + 1>(ic2);
+    uint32_t oc3 = ((uint32_t)bit_sub<24, 31 - 24 + 1>(ic0) << 24) | ((uint32_t)bit_sub<16, 23 - 16 + 1>(ic1) << 16) |
+                   ((uint32_t)bit_sub<8, 15 - 8 + 1>(ic2) << 8) | bit_sub<0, 7 - 0 + 1>(ic3);
+    return ((uint128_t)oc3 << 96) | ((uint128_t)oc2 << 64) | ((uint128_t)oc1 << 32) | oc0;
+}
+
+uint128_t aes_subbytes_fwd(uint128_t x) {
+    uint32_t oc0 = aes_subword_fwd(aes_get_column(x, 0));
+    uint32_t oc1 = aes_subword_fwd(aes_get_column(x, 1));
+    uint32_t oc2 = aes_subword_fwd(aes_get_column(x, 2));
+    uint32_t oc3 = aes_subword_fwd(aes_get_column(x, 3));
+    return ((uint128_t)oc3 << 96) | ((uint128_t)oc2 << 64) | ((uint128_t)oc1 << 32) | oc0;
+}
+
+uint128_t aes_subbytes_inv(uint128_t x) {
+    uint32_t oc0 = aes_subword_inv(aes_get_column(x, 0));
+    uint32_t oc1 = aes_subword_inv(aes_get_column(x, 1));
+    uint32_t oc2 = aes_subword_inv(aes_get_column(x, 2));
+    uint32_t oc3 = aes_subword_inv(aes_get_column(x, 3));
+    return ((uint128_t)oc3 << 96) | ((uint128_t)oc2 << 64) | ((uint128_t)oc1 << 32) | oc0;
+}
+
+uint128_t aes_mixcolumns_fwd(uint128_t x) {
+    uint32_t oc0 = aes_mixcolumn_fwd(aes_get_column(x, 0));
+    uint32_t oc1 = aes_mixcolumn_fwd(aes_get_column(x, 1));
+    uint32_t oc2 = aes_mixcolumn_fwd(aes_get_column(x, 2));
+    uint32_t oc3 = aes_mixcolumn_fwd(aes_get_column(x, 3));
+    return ((uint128_t)oc3 << 96) | ((uint128_t)oc2 << 64) | ((uint128_t)oc1 << 32) | oc0;
+}
+
+uint128_t aes_mixcolumns_inv(uint128_t x) {
+    uint32_t oc0 = aes_mixcolumn_inv(aes_get_column(x, 0));
+    uint32_t oc1 = aes_mixcolumn_inv(aes_get_column(x, 1));
+    uint32_t oc2 = aes_mixcolumn_inv(aes_get_column(x, 2));
+    uint32_t oc3 = aes_mixcolumn_inv(aes_get_column(x, 3));
+    return ((uint128_t)oc3 << 96) | ((uint128_t)oc2 << 64) | ((uint128_t)oc1 << 32) | oc0;
+}
+
+uint32_t aes_rotword(uint32_t x) {
+    uint8_t a0 = bit_sub<0, 7 - 0 + 1>(x);
+    uint8_t a1 = bit_sub<8, 15 - 8 + 1>(x);
+    uint8_t a2 = bit_sub<16, 23 - 16 + 1>(x);
+    uint8_t a3 = bit_sub<24, 31 - 24 + 1>(x);
+    return ((uint32_t)a0 << 24) | ((uint32_t)a3 << 16) | ((uint32_t)a2 << 8) | a1;
+}
+
+std::function<uint128_t(uint128_t, uint128_t, uint128_t)> get_crypto_funct(unsigned funct6, unsigned vs1) {
+    switch(funct6) {
+    case 0b101000: // VAES.VV
+    case 0b101001: // VAES.VS
+        switch(vs1) {
+        case 0b00000: // VAESDM
+            return [](uint128_t state, uint128_t rkey, uint128_t) {
+                uint128_t sr = aes_shift_rows_inv(state);
+                uint128_t sb = aes_subbytes_inv(sr);
+                uint128_t ark = sb ^ rkey;
+                uint128_t mix = aes_mixcolumns_inv(ark);
+                return mix;
+            };
+        case 0b00001: // VAESDF
+            return [](uint128_t state, uint128_t rkey, uint128_t) {
+                uint128_t sr = aes_shift_rows_inv(state);
+                uint128_t sb = aes_subbytes_inv(sr);
+                uint128_t ark = sb ^ rkey;
+                return ark;
+            };
+        case 0b00010: // VAESEM
+            return [](uint128_t state, uint128_t rkey, uint128_t) {
+                uint128_t sb = aes_subbytes_fwd(state);
+                uint128_t sr = aes_shift_rows_fwd(sb);
+                uint128_t mix = aes_mixcolumns_fwd(sr);
+                uint128_t ark = mix ^ rkey;
+                return ark;
+            };
+        case 0b00011: // VAESEF
+            return [](uint128_t state, uint128_t rkey, uint128_t) {
+                uint128_t sb = aes_subbytes_fwd(state);
+                uint128_t sr = aes_shift_rows_fwd(sb);
+                uint128_t ark = sr ^ rkey;
+                return ark;
+            };
+        case 0b00111: // VAESZ
+            return [](uint128_t state, uint128_t rkey, uint128_t) {
+                uint128_t ark = state ^ rkey;
+                return ark;
+            };
+        case 0b10000: // VSM4R
+            throw new std::runtime_error("Unsupported operation in get_crypto_funct");
+        case 0b10001: // VGMUL
+            return [](uint128_t vd, uint128_t vs2, uint128_t) {
+                uint128_t Y = brev8<uint128_t>(vd);
+                uint128_t H = brev8<uint128_t>(vs2);
+                uint128_t Z = 0;
+                for(size_t bit = 0; bit < 128; bit++) {
+                    if((Y >> bit) & 1)
+                        Z ^= H;
+                    bool reduce = (H >> 127) & 1;
+                    H = H << 1;
+                    if(reduce)
+                        H ^= 0x87;
+                }
+                uint128_t result = brev8<uint128_t>(Z);
+                return result;
+            };
+        default:
+            throw new std::runtime_error("Unsupported operation in get_crypto_funct");
+        }
+    case 0b100000: // VSM3ME
+    case 0b100001: // VSM4K
+        throw new std::runtime_error("Unsupported operation in get_crypto_funct");
+    case 0b100010: // VAESKF1
+        return [](uint128_t vd, uint128_t vs2, uint128_t r) {
+            auto extract_word = [](const uint128_t& value, int index) -> uint32_t {
+                return static_cast<uint32_t>((value >> (32 * index)) & std::numeric_limits<uint32_t>::max());
+            };
+
+            uint32_t k0 = (vs2 >> 32 * 0) & std::numeric_limits<uint32_t>::max();
+            uint32_t k1 = (vs2 >> 32 * 1) & std::numeric_limits<uint32_t>::max();
+            uint32_t k2 = (vs2 >> 32 * 2) & std::numeric_limits<uint32_t>::max();
+            uint32_t k3 = (vs2 >> 32 * 3) & std::numeric_limits<uint32_t>::max();
+            uint32_t w0 = aes_subword_fwd(aes_rotword(k3)) ^ aes_decode_rcon(r) ^ k0;
+            uint32_t w1 = w0 ^ k1;
+            uint32_t w2 = w1 ^ k2;
+            uint32_t w3 = w2 ^ k3;
+            uint128_t result = (uint128_t(w3) << 96) | (uint128_t(w2) << 64) | (uint128_t(w1) << 32) | (uint128_t(w0));
+            return result;
+        };
+    case 0b101010: // VAESKF2
+        return [](uint128_t vd, uint128_t vs2, uint128_t r) {
+            uint32_t k0 = (vs2 >> 32 * 0) & std::numeric_limits<uint32_t>::max();
+            uint32_t k1 = (vs2 >> 32 * 1) & std::numeric_limits<uint32_t>::max();
+            uint32_t k2 = (vs2 >> 32 * 2) & std::numeric_limits<uint32_t>::max();
+            uint32_t k3 = (vs2 >> 32 * 3) & std::numeric_limits<uint32_t>::max();
+            uint32_t rkb0 = (vd >> 32 * 0) & std::numeric_limits<uint32_t>::max();
+            uint32_t rkb1 = (vd >> 32 * 1) & std::numeric_limits<uint32_t>::max();
+            uint32_t rkb2 = (vd >> 32 * 2) & std::numeric_limits<uint32_t>::max();
+            uint32_t rkb3 = (vd >> 32 * 3) & std::numeric_limits<uint32_t>::max();
+            uint32_t w0 = r & 1 ? aes_subword_fwd(k3) ^ rkb0 : aes_subword_fwd(aes_rotword(k3)) ^ aes_decode_rcon((r >> 1) - 1) ^ rkb0;
+            uint32_t w1 = w0 ^ rkb1;
+            uint32_t w2 = w1 ^ rkb2;
+            uint32_t w3 = w2 ^ rkb3;
+            uint128_t result = (uint128_t(w3) << 96) | (uint128_t(w2) << 64) | (uint128_t(w1) << 32) | (uint128_t(w0));
+            return result;
+        };
+    case 0b101011: // VSM3C
+        throw new std::runtime_error("Unsupported operation in get_crypto_funct");
+    case 0b101100: // VGHSH
+        return [](uint128_t Y, uint128_t vs2, uint128_t X) {
+            auto H = brev8<uint128_t>(vs2);
+            uint128_t Z = 0;
+            uint128_t S = brev8<uint128_t>(Y ^ X);
+            for(size_t bit = 0; bit < 128; bit++) {
+                if((S >> bit) & 1)
+                    Z ^= H;
+                bool reduce = (H >> 127) & 1;
+                H = H << 1;
+                if(reduce)
+                    H ^= 0x87;
+            }
+            uint128_t result = brev8<uint128_t>(Z);
+            return result;
+        };
+    case 0b101101: // VSHA2MS
+    case 0b101110: // VSHA2CH
+    case 0b101111: // VSHA2CL
+    default:
+        throw new std::runtime_error("Unknown funct6 in get_crypto_funct");
+    }
+}
+} // namespace softvector

src/vm/vector_functions.h

@@ -0,0 +1,192 @@
+////////////////////////////////////////////////////////////////////////////////
+// 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:
+//       alex@minres.com - initial API and implementation
+////////////////////////////////////////////////////////////////////////////////
+
+#ifndef _VM_VECTOR_FUNCTIONS_H_
+#define _VM_VECTOR_FUNCTIONS_H_
+
+#include "iss/arch_if.h"
+#include "iss/vm_types.h"
+#include <cstdint>
+#include <functional>
+#include <stdint.h>
+namespace softvector {
+#ifndef _MSC_VER
+using int128_t = __int128;
+using uint128_t = unsigned __int128;
+#endif
+const unsigned RFS = 32;
+
+struct vtype_t {
+    uint64_t underlying;
+    vtype_t(uint32_t vtype_val);
+    vtype_t(uint64_t vtype_val);
+    unsigned sew();
+    double lmul();
+    bool vill();
+    bool vma();
+    bool vta();
+};
+class mask_bit_reference {
+    uint8_t* start;
+    uint8_t pos;
+
+public:
+    mask_bit_reference& operator=(const bool new_value);
+    mask_bit_reference(uint8_t* start, uint8_t pos);
+    operator bool() const;
+};
+
+struct vmask_view {
+    uint8_t* start;
+    size_t elem_count;
+    mask_bit_reference operator[](size_t) const;
+};
+vmask_view read_vmask(uint8_t* V, uint16_t VLEN, uint16_t elem_count, uint8_t reg_idx = 0);
+template <unsigned VLEN> vmask_view read_vmask(uint8_t* V, uint16_t elem_count, uint8_t reg_idx = 0);
+std::function<uint128_t(uint128_t, uint128_t, uint128_t)> get_crypto_funct(unsigned funct6, unsigned vs1);
+
+template <typename dest_elem_t, typename src_elem_t = dest_elem_t> dest_elem_t brev(src_elem_t vs2);
+template <typename dest_elem_t, typename src_elem_t = dest_elem_t> dest_elem_t brev8(src_elem_t vs2);
+
+bool softvec_read(void* core, uint64_t addr, uint64_t length, uint8_t* data);
+bool softvec_write(void* core, uint64_t addr, uint64_t length, uint8_t* data);
+
+template <unsigned VLEN, typename eew_t>
+uint64_t vector_load_store(void* core, std::function<bool(void*, uint64_t, uint64_t, uint8_t*)> load_store_fn, uint8_t* V, uint64_t vl,
+                           uint64_t vstart, vtype_t vtype, bool vm, uint8_t vd, uint64_t rs1, uint8_t segment_size, int64_t stride = 0,
+                           bool use_stride = false);
+template <unsigned XLEN, unsigned VLEN, typename eew_t, typename sew_t>
+uint64_t vector_load_store_index(void* core, std::function<bool(void*, uint64_t, uint64_t, uint8_t*)> load_store_fn, uint8_t* V,
+                                 uint64_t vl, uint64_t vstart, vtype_t vtype, bool vm, uint8_t vd, uint64_t rs1, uint8_t vs2,
+                                 uint8_t segment_size);
+template <unsigned VLEN, typename dest_elem_t, typename src2_elem_t = dest_elem_t, typename src1_elem_t = src2_elem_t>
+void vector_vector_op(uint8_t* V, unsigned funct6, unsigned funct3, uint64_t vl, uint64_t vstart, vtype_t vtype, bool vm, unsigned vd,
+                      unsigned vs2, unsigned vs1);
+template <unsigned VLEN, typename dest_elem_t, typename src2_elem_t = dest_elem_t, typename src1_elem_t = src2_elem_t>
+void vector_imm_op(uint8_t* V, unsigned funct6, unsigned funct3, uint64_t vl, uint64_t vstart, vtype_t vtype, bool vm, unsigned vd,
+                   unsigned vs2, typename std::make_signed<src1_elem_t>::type imm);
+template <unsigned VLEN, typename elem_t>
+void vector_vector_carry(uint8_t* V, unsigned funct6, unsigned funct3, uint64_t vl, uint64_t vstart, vtype_t vtype, unsigned vd,
+                         unsigned vs2, unsigned vs1, signed carry);
+template <unsigned VLEN, typename elem_t>
+void vector_imm_carry(uint8_t* V, unsigned funct6, unsigned funct3, uint64_t vl, uint64_t vstart, vtype_t vtype, unsigned vd, unsigned vs2,
+                      typename std::make_signed<elem_t>::type imm, signed carry);
+template <unsigned VLEN, typename scr_elem_t>
+void vector_vector_merge(uint8_t* V, uint64_t vl, uint64_t vstart, vtype_t vtype, bool vm, unsigned vd, unsigned vs2, unsigned vs1);
+template <unsigned VLEN, typename scr_elem_t>
+void vector_imm_merge(uint8_t* V, uint64_t vl, uint64_t vstart, vtype_t vtype, bool vm, unsigned vd, unsigned vs2, uint64_t imm);
+template <unsigned VLEN, typename dest_elem_t, typename src2_elem_t = dest_elem_t>
+void vector_unary_op(uint8_t* V, unsigned unary_op, uint64_t vl, uint64_t vstart, vtype_t vtype, bool vm, unsigned vd, unsigned vs2);
+template <unsigned VLEN, typename elem_t>
+void mask_vector_vector_op(uint8_t* V, unsigned funct, unsigned funct3, uint64_t vl, uint64_t vstart, vtype_t vtype, bool vm, unsigned vd,
+                           unsigned vs2, unsigned vs1);
+template <unsigned VLEN, typename elem_t>
+void mask_vector_imm_op(uint8_t* V, unsigned funct, unsigned funct3, uint64_t vl, uint64_t vstart, vtype_t vtype, bool vm, unsigned vd,
+                        unsigned vs2, typename std::make_signed<elem_t>::type imm);
+void carry_vector_vector_op(uint8_t* V, unsigned funct, uint64_t vl, uint64_t vstart, vtype_t vtype, bool vm, unsigned vd, unsigned vs2,
+                            unsigned vs1);
+template <unsigned VLEN, typename elem_t>
+void carry_vector_imm_op(uint8_t* V, unsigned funct, uint64_t vl, uint64_t vstart, vtype_t vtype, bool vm, unsigned vd, unsigned vs2,
+                         typename std::make_signed<elem_t>::type imm);
+template <unsigned VLEN, typename dest_elem_t, typename src2_elem_t = dest_elem_t, typename src1_elem_t = dest_elem_t>
+bool sat_vector_vector_op(uint8_t* V, unsigned funct6, unsigned funct3, uint64_t vl, uint64_t vstart, vtype_t vtype, int64_t vxrm, bool vm,
+                          unsigned vd, unsigned vs2, unsigned vs1);
+template <unsigned VLEN, typename dest_elem_t, typename src2_elem_t = dest_elem_t, typename src1_elem_t = dest_elem_t>
+bool sat_vector_imm_op(uint8_t* V, unsigned funct6, unsigned funct3, uint64_t vl, uint64_t vstart, vtype_t vtype, int64_t vxrm, bool vm,
+                       unsigned vd, unsigned vs2, typename std::make_signed<src1_elem_t>::type imm);
+template <unsigned VLEN, typename dest_elem_t, typename src_elem_t = dest_elem_t>
+void vector_red_op(uint8_t* V, unsigned funct6, unsigned funct3, uint64_t vl, uint64_t vstart, vtype_t vtype, bool vm, unsigned vd,
+                   unsigned vs2, unsigned vs1);
+template <unsigned VLEN>
+void mask_mask_op(uint8_t* V, unsigned funct6, unsigned funct3, uint64_t vl, uint64_t vstart, unsigned vd, unsigned vs2, unsigned vs1);
+template <unsigned VLEN> uint64_t vcpop(uint8_t* V, uint64_t vl, uint64_t vstart, bool vm, unsigned vs2);
+template <unsigned VLEN> uint64_t vfirst(uint8_t* V, uint64_t vl, uint64_t vstart, bool vm, unsigned vs2);
+template <unsigned VLEN> void mask_set_op(uint8_t* V, unsigned enc, uint64_t vl, uint64_t vstart, bool vm, unsigned vd, unsigned vs2);
+template <unsigned VLEN, typename src_elem_t>
+void viota(uint8_t* V, uint64_t vl, uint64_t vstart, vtype_t vtype, bool vm, unsigned vd, unsigned vs2);
+template <unsigned VLEN, typename src_elem_t> void vid(uint8_t* V, uint64_t vl, uint64_t vstart, vtype_t vtype, bool vm, unsigned vd);
+template <unsigned VLEN, typename src_elem_t> uint64_t scalar_move(uint8_t* V, vtype_t vtype, unsigned vd, uint64_t val, bool to_vector);
+template <unsigned VLEN, typename src_elem_t>
+void vector_slideup(uint8_t* V, uint64_t vl, uint64_t vstart, vtype_t vtype, bool vm, unsigned vd, unsigned vs2, uint64_t imm);
+template <unsigned VLEN, typename src_elem_t>
+void vector_slidedown(uint8_t* V, uint64_t vl, uint64_t vstart, vtype_t vtype, bool vm, unsigned vd, unsigned vs2, uint64_t imm);
+template <unsigned VLEN, typename src_elem_t>
+void vector_slide1up(uint8_t* V, uint64_t vl, uint64_t vstart, vtype_t vtype, bool vm, unsigned vd, unsigned vs2, uint64_t imm);
+template <unsigned VLEN, typename src_elem_t>
+void vector_slide1down(uint8_t* V, uint64_t vl, uint64_t vstart, vtype_t vtype, bool vm, unsigned vd, unsigned vs2, uint64_t imm);
+template <unsigned VLEN, typename dest_elem_t, typename scr_elem_t = dest_elem_t>
+void vector_vector_gather(uint8_t* V, uint64_t vl, uint64_t vstart, vtype_t vtype, bool vm, unsigned vd, unsigned vs2, unsigned vs1);
+template <unsigned VLEN, typename scr_elem_t>
+void vector_imm_gather(uint8_t* V, uint64_t vl, uint64_t vstart, vtype_t vtype, bool vm, unsigned vd, unsigned vs2, uint64_t imm);
+template <unsigned VLEN, typename scr_elem_t>
+void vector_compress(uint8_t* V, uint64_t vl, uint64_t vstart, vtype_t vtype, unsigned vd, unsigned vs2, unsigned vs1);
+template <unsigned VLEN> void vector_whole_move(uint8_t* V, unsigned vd, unsigned vs2, unsigned count);
+template <unsigned VLEN, typename dest_elem_t, typename src_elem_t = dest_elem_t>
+void fp_vector_red_op(uint8_t* V, unsigned funct6, unsigned funct3, uint64_t vl, uint64_t vstart, vtype_t vtype, bool vm, unsigned vd,
+                      unsigned vs2, unsigned vs1, uint8_t rm);
+template <unsigned VLEN, typename dest_elem_t, typename src2_elem_t = dest_elem_t, typename src1_elem_t = src2_elem_t>
+void fp_vector_vector_op(uint8_t* V, unsigned funct6, unsigned funct3, uint64_t vl, uint64_t vstart, vtype_t vtype, bool vm, unsigned vd,
+                         unsigned vs2, unsigned vs1, uint8_t rm);
+template <unsigned VLEN, typename dest_elem_t, typename src2_elem_t = dest_elem_t, typename src1_elem_t = src2_elem_t>
+void fp_vector_imm_op(uint8_t* V, unsigned funct6, unsigned funct3, uint64_t vl, uint64_t vstart, vtype_t vtype, bool vm, unsigned vd,
+                      unsigned vs2, src1_elem_t imm, uint8_t rm);
+template <unsigned VLEN, typename elem_t>
+void fp_vector_unary_op(uint8_t* V, unsigned encoding_space, unsigned unary_op, uint64_t vl, uint64_t vstart, vtype_t vtype, bool vm,
+                        unsigned vd, unsigned vs2, uint8_t rm);
+template <unsigned VLEN, typename dest_elem_t, typename src_elem_t>
+void fp_vector_unary_w(uint8_t* V, unsigned unary_op, uint64_t vl, uint64_t vstart, vtype_t vtype, bool vm, unsigned vd, unsigned vs2,
+                       uint8_t rm);
+template <unsigned VLEN, typename dest_elem_t, typename src_elem_t>
+void fp_vector_unary_n(uint8_t* V, unsigned unary_op, uint64_t vl, uint64_t vstart, vtype_t vtype, bool vm, unsigned vd, unsigned vs2,
+                       uint8_t rm);
+template <unsigned VLEN, typename elem_t>
+void mask_fp_vector_vector_op(uint8_t* V, unsigned funct6, uint64_t vl, uint64_t vstart, vtype_t vtype, bool vm, unsigned vd, unsigned vs2,
+                              unsigned vs1, uint8_t rm);
+template <unsigned VLEN, typename elem_t>
+void mask_fp_vector_imm_op(uint8_t* V, unsigned funct6, uint64_t vl, uint64_t vstart, vtype_t vtype, bool vm, unsigned vd, unsigned vs2,
+                           elem_t imm, uint8_t rm);
+template <unsigned VLEN, unsigned EGS>
+void vector_vector_crypto(uint8_t* V, unsigned funct6, uint64_t eg_len, uint64_t eg_start, vtype_t vtype, unsigned vd, unsigned vs2,
+                          unsigned vs1);
+template <unsigned VLEN, unsigned EGS>
+void vector_scalar_crypto(uint8_t* V, unsigned funct6, uint64_t eg_len, uint64_t eg_start, vtype_t vtype, unsigned vd, unsigned vs2,
+                          unsigned vs1);
+template <unsigned VLEN, unsigned EGS>
+void vector_imm_crypto(uint8_t* V, unsigned funct6, uint64_t eg_len, uint64_t eg_start, vtype_t vtype, unsigned vd, unsigned vs2,
+                       uint8_t imm);
+template <unsigned VLEN, unsigned EGS, typename elem_type_t>
+void vector_crypto(uint8_t* V, unsigned funct6, uint64_t eg_len, uint64_t eg_start, vtype_t vtype, unsigned vd, unsigned vs2, unsigned vs1);
+} // namespace softvector
+#include "vm/vector_functions.hpp"
+#endif /* _VM_VECTOR_FUNCTIONS_H_ */