corrects mistake from rebasing, adds newly generated templates

CMakeLists.txt

@@ -21,6 +21,7 @@ set(LIB_SOURCES
     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
@@ -142,8 +148,10 @@ 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 ${reg.name} = 0;<%
         }}%>
         uint32_t trap_state = 0, pending_trap = 0;
         uint64_t icount = 0;      // counts number of instructions undisturbed
@@ -164,6 +172,31 @@ if(fcsr != null) {%>
 <%} else { %>
     uint32_t get_fcsr(){return 0;}
     void set_fcsr(uint32_t val){}
+<%}
+def vstart = registers.find {it.name=='vstart'}
+def vl = registers.find {it.name=='vl'}
+def vtype = registers.find {it.name=='vtype'}
+def vxsat = registers.find {it.name=='vxsat'}
+def vxrm = registers.find {it.name=='vxrm'}
+
+if(vtype != null) {%>
+    uint${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/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>
@@ -102,9 +114,47 @@ protected:
 
     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 +177,792 @@ 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");
+        }
+    }
+<%}%>
     uint64_t fetch_count{0};
     uint64_t tval{0};
 
@@ -271,8 +1106,8 @@ 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}<%}%>

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/mem/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>
@@ -195,7 +194,17 @@ 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 { PRIV_U = 0, PRIV_S = 1, PRIV_M = 3, PRIV_D = 4 };
@@ -314,6 +323,19 @@ 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);
@@ -390,8 +412,10 @@ 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)
@@ -684,6 +708,63 @@ 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); },

src/iss/arch/riscv_hart_m_p.h

@@ -45,10 +45,10 @@
 #ifndef FMT_HEADER_ONLY
 #define FMT_HEADER_ONLY
 #endif
-#include <iss/mem/memory_with_htif.h>
 #include <array>
 #include <elfio/elfio.hpp>
 #include <fmt/format.h>
+#include <iss/mem/memory_with_htif.h>
 #include <unordered_map>
 
 namespace iss {

src/iss/arch/riscv_hart_mu_p.h

@@ -47,8 +47,8 @@
 #ifndef FMT_HEADER_ONLY
 #define FMT_HEADER_ONLY
 #endif
-#include <iss/mem/memory_with_htif.h>
 #include <fmt/format.h>
+#include <iss/mem/memory_with_htif.h>
 #include <unordered_map>
 
 namespace iss {

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
@@ -258,6 +259,7 @@ 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/vm/aes_sbox.h

@@ -0,0 +1,69 @@
+////////////////////////////////////////////////////////////////////////////////
+// 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>
+
+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 +39,88 @@
 
 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 fsqrt_h(uint16_t v1, uint8_t mode);
+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/vector_functions.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:
+//       alex@minres.com - initial API and implementation
+////////////////////////////////////////////////////////////////////////////////
+
+#include "vector_functions.h"
+#include "iss/vm_types.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};
+}
+} // namespace softvector

src/vm/vector_functions.h

@@ -0,0 +1,172 @@
+////////////////////////////////////////////////////////////////////////////////
+// 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 {
+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);
+
+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);
+} // namespace softvector
+#include "vm/vector_functions.hpp"
+#endif /* _VM_VECTOR_FUNCTIONS_H_ */