adds functionality for all Zvk Instructions

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

gen_input/templates/asmjit/CORENAME.cpp.gtl

@@ -40,7 +40,11 @@
 #include <iss/instruction_decoder.h>
 <%def fcsr = registers.find {it.name=='FCSR'}
 if(fcsr != null) {%>
-#include <vm/fp_functions.h><%}%>
+#include <vm/fp_functions.h><%}
+def aes = functions.find { it.contains('aes') }
+if(aes != null) {%>
+#include <vm/aes_sbox.h>
+<%}%>
 #ifndef FMT_HEADER_ONLY
 #define FMT_HEADER_ONLY
 #endif
@@ -49,6 +53,22 @@ if(fcsr != null) {%>
 #include <array>
 #include <iss/debugger/riscv_target_adapter.h>
 
+#ifndef _MSC_VER
+using int128_t = __int128;
+using uint128_t = unsigned __int128;
+namespace std {
+template <> struct make_unsigned<__int128> { typedef unsigned __int128 type; };
+template <> class __make_unsigned_selector<__int128 unsigned, false, false> {
+public:
+    typedef unsigned __int128 __type;
+};
+template <> struct is_signed<int128_t> { static constexpr bool value = true; };
+template <> struct is_signed<uint128_t> { static constexpr bool value = false; };
+template <> struct is_unsigned<int128_t> { static constexpr bool value = false; };
+template <> struct is_unsigned<uint128_t> { static constexpr bool value = true; };
+} // namespace std
+#endif
+
 namespace iss {
 namespace asmjit {
 
@@ -116,38 +136,14 @@ protected:
         auto sign_mask = 1ULL<<(W-1);
         return (from & mask) | ((from & sign_mask) ? ~mask : 0);
     }
+    inline void raise(uint16_t trap_id, uint16_t cause){
+        auto trap_val =  0x80ULL << 24 | (cause << 16) | trap_id;
+        this->core.reg.trap_state = trap_val;
+    }
+
 <%functions.each{ it.eachLine { %>
     ${it}<%}
-}
-if(fcsr != null) {%>
-    x86_reg_t NaNBox16(jit_holder& jh , x86_reg_t NaNBox16_val){
-        x86::Compiler& cc = jh.cc;
-        if(static_cast<uint32_t>(traits::FLEN) == 16)
-            return gen_ext(cc, NaNBox16_val, traits::FLEN, false);
-        else {         
-            auto box = gen_operation(cc, bnot, (gen_ext(cc, 0, traits::FLEN, false)));
-            return gen_ext(cc, (gen_operation(jh.cc, bor, gen_ext(jh.cc, gen_operation(jh.cc, shl, box, 16), traits::FLEN, false), gen_ext(jh.cc, NaNBox16_val, traits::FLEN, false))), traits::FLEN, false);
-        } 
-    }
-    x86_reg_t NaNBox32(jit_holder& jh , x86_reg_t NaNBox32_val){
-        x86::Compiler& cc = jh.cc;
-        if(static_cast<uint32_t>(traits::FLEN) == 32)
-            return gen_ext(cc, NaNBox32_val, traits::FLEN, false);
-        else {         
-            auto box = gen_operation(cc, bnot, (gen_ext(cc, 0, /* we need to shift for 32 so need larger dt*/ 64, false)));
-            return gen_ext(cc, (gen_operation(jh.cc, bor, gen_ext(jh.cc, gen_operation(jh.cc, shl, box, 32), traits::FLEN, false), gen_ext(jh.cc, NaNBox32_val, traits::FLEN, false))), traits::FLEN, false);
-        } 
-    }
-    x86_reg_t NaNBox64(jit_holder& jh , x86_reg_t NaNBox64_val){
-        x86::Compiler& cc = jh.cc;
-        if(static_cast<uint32_t>(traits::FLEN) == 64)
-            return gen_ext(cc, NaNBox64_val, traits::FLEN, false);
-        else {         
-            auto box = gen_operation(cc, bnot, (gen_ext(cc, 0, /* we need to shift for 32 so need larger dt*/ 128, false)));
-            return gen_ext(cc, (gen_operation(jh.cc, bor, gen_ext(jh.cc, gen_operation(jh.cc, shl, box, 64), traits::FLEN, false), gen_ext(jh.cc, NaNBox64_val, traits::FLEN, false))), traits::FLEN, false);
-        } 
-    }
-<%}%>
+}%>
 private:
     /****************************************************************************
      * start opcode definitions
@@ -195,7 +191,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         <%instr.behavior.eachLine{%>${it}
         <%}%>
@@ -226,7 +221,6 @@ private:
         pc = pc + ((instr & 3) == 3 ? 4 : 2);
         mov(cc, jh.next_pc, pc.val);
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         gen_raise(jh, 0, 2);
         gen_sync(jh, POST_SYNC, instr_descr.size());
         gen_instr_epilogue(jh);
@@ -272,24 +266,25 @@ template <typename ARCH>
 void vm_impl<ARCH>::gen_instr_prologue(jit_holder& jh) {
     auto& cc = jh.cc;
 
-    cc.comment("//gen_instr_prologue");
-
     x86_reg_t current_trap_state = get_reg_for(cc, traits::TRAP_STATE);
     mov(cc, current_trap_state, get_ptr_for(jh, traits::TRAP_STATE));
     mov(cc, get_ptr_for(jh, traits::PENDING_TRAP), current_trap_state);
+    cc.comment("//Instruction prologue end");
 
 }
 template <typename ARCH>
 void vm_impl<ARCH>::gen_instr_epilogue(jit_holder& jh) {
     auto& cc = jh.cc;
 
-    cc.comment("//gen_instr_epilogue");
+    cc.comment("//Instruction epilogue begin");
     x86_reg_t current_trap_state = get_reg_for(cc, traits::TRAP_STATE);
     mov(cc, current_trap_state, get_ptr_for(jh, traits::TRAP_STATE));
     cmp(cc, current_trap_state, 0);
     cc.jne(jh.trap_entry);
     cc.inc(get_ptr_for(jh, traits::ICOUNT));
     cc.inc(get_ptr_for(jh, traits::CYCLE));
+    cc.comment("//Instruction epilogue end");
+
 }
 template <typename ARCH>
 void vm_impl<ARCH>::gen_block_prologue(jit_holder& jh){
@@ -301,7 +296,7 @@ void vm_impl<ARCH>::gen_block_prologue(jit_holder& jh){
 template <typename ARCH>
 void vm_impl<ARCH>::gen_block_epilogue(jit_holder& jh){
     x86::Compiler& cc = jh.cc;
-    cc.comment("//gen_block_epilogue");
+    cc.comment("//block epilogue begin");
     cc.ret(jh.next_pc);
 
     cc.bind(jh.trap_entry);
@@ -313,7 +308,6 @@ void vm_impl<ARCH>::gen_block_epilogue(jit_holder& jh){
     x86::Gp current_pc = get_reg_for_Gp(cc, traits::PC);
     mov(cc, current_pc, get_ptr_for(jh, traits::PC));
 
-    cc.comment("//enter trap call;");
     InvokeNode* call_enter_trap;
     cc.invoke(&call_enter_trap, &enter_trap, FuncSignature::build<uint64_t, void*, uint64_t, uint64_t, uint64_t>());
     call_enter_trap->setArg(0, jh.arch_if_ptr);
@@ -331,7 +325,6 @@ void vm_impl<ARCH>::gen_block_epilogue(jit_holder& jh){
 template <typename ARCH>
 inline void vm_impl<ARCH>::gen_raise(jit_holder& jh, uint16_t trap_id, uint16_t cause) {
     auto& cc = jh.cc;
-    cc.comment("//gen_raise");
     auto tmp1 = get_reg_for(cc, traits::TRAP_STATE);
     mov(cc, tmp1, 0x80ULL << 24 | (cause << 16) | trap_id);
     mov(cc, get_ptr_for(jh, traits::TRAP_STATE), tmp1);

gen_input/templates/interp/CORENAME.cpp.gtl

@@ -962,6 +962,20 @@ if(vector != null) {%>
                 throw new std::runtime_error("Unsupported sew bit value");
         }
     }
+    void vector_unary_op(uint8_t* V, uint8_t unary_op, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t vd, uint8_t vs2, uint8_t sew_val){
+        switch(sew_val){
+            case 0b000:
+                return softvector::vector_unary_op<${vlen}, uint8_t>(V, unary_op, vl, vstart, vtype, vm, vd, vs2);
+            case 0b001:
+                return softvector::vector_unary_op<${vlen}, uint16_t>(V, unary_op, vl, vstart, vtype, vm, vd, vs2);
+            case 0b010:
+                return softvector::vector_unary_op<${vlen}, uint32_t>(V, unary_op, vl, vstart, vtype, vm, vd, vs2);
+            case 0b011:
+                return softvector::vector_unary_op<${vlen}, uint64_t>(V, unary_op, vl, vstart, vtype, vm, vd, vs2);
+            default: 
+                throw new std::runtime_error("Unsupported sew_val");
+        }
+    }
 <%}%>
     uint64_t fetch_count{0};
     uint64_t tval{0};

gen_input/templates/llvm/CORENAME.cpp.gtl

@@ -31,12 +31,14 @@
  *******************************************************************************/
 // clang-format off
 #include <iss/arch/${coreDef.name.toLowerCase()}.h>
+// vm_base needs to be included before gdb_session as termios.h (via boost and gdb_server) has a define which clashes with a variable
+// name in ConstantRange.h
+#include <iss/llvm/vm_base.h>
+#include <iss/iss.h>
 #include <iss/debugger/gdb_session.h>
 #include <iss/debugger/server.h>
-#include <iss/iss.h>
-#include <iss/llvm/vm_base.h>
-#include <util/logging.h>
 #include <iss/instruction_decoder.h>
+#include <util/logging.h>
 <%def fcsr = registers.find {it.name=='FCSR'}
 if(fcsr != null) {%>
 #include <vm/fp_functions.h><%}%>

gen_input/templates/tcc/CORENAME.cpp.gtl

@@ -159,13 +159,13 @@ if(fcsr != null) {%>
         } 
     }
     value NaNBox64(tu_builder& tu, value NaNBox64_val){
-        throw new std::runtime_error("tcc does not support Registers wider than 64 bits");
-        /*if(static_cast<uint32_t>(traits::FLEN) == 64)
+        if(static_cast<uint32_t>(traits::FLEN) == 64)
             return tu.ext(NaNBox64_val, traits::FLEN, false);
-        else {         
+        else {
+            throw new std::runtime_error("tcc does not support Registers wider than 64 bits");     
             auto box = tu.assignment(tu.logical_neg((tu.ext(0, 128, false))), traits::FLEN) ;
             return tu.ext((tu.bitwise_or(tu.shl(box, 64), NaNBox64_val)), traits::FLEN, false);
-        }*/
+        }
     }
 <%}%>
 private:
@@ -360,6 +360,7 @@ template <typename ARCH> void vm_impl<ARCH>::add_prologue(tu_builder& tu){
     os << "uint64_t (*fmadd_d)(uint64_t v1, uint64_t v2, uint64_t v3, uint32_t op, uint8_t mode)=" << (uintptr_t)&fmadd_d << ";\\n";
     os << "uint64_t (*fsel_d)(uint64_t v1, uint64_t v2, uint32_t op)=" << (uintptr_t)&fsel_d << ";\\n";
     os << "uint64_t (*fclass_d)(uint64_t v1  )=" << (uintptr_t)&fclass_d << ";\\n";
+    os << "uint64_t (*unbox_d)(uint8_t FLEN, uint64_t v)=" << (uintptr_t)&unbox_d << ";\\n";
 
     os << "uint32_t (*f64tof32)(uint64_t v1, uint8_t mode)=" << (uintptr_t)&f64tof32 << ";\\n";
     os << "uint64_t (*f32tof64)(uint32_t v1, uint8_t mode)=" << (uintptr_t)&f32tof64 << ";\\n";

src/sysc/core_complex.cpp

@@ -267,7 +267,10 @@ template <unsigned int BUSWIDTH> void core_complex<BUSWIDTH>::before_end_of_elab
     cpu = new core_wrapper(this);
     cpu->create_cpu(GET_PROP_VALUE(core_type), GET_PROP_VALUE(backend), GET_PROP_VALUE(gdb_server_port), GET_PROP_VALUE(mhartid));
     sc_assert(cpu->vm != nullptr);
-    cpu->vm->setDisassEnabled(GET_PROP_VALUE(enable_disass) || trc->m_db != nullptr);
+    auto disass = GET_PROP_VALUE(enable_disass);
+    if(disass && trc->m_db)
+        SCCINFO(SCMOD)<<"Disasssembly will only be in transaction trace database!";
+    cpu->vm->setDisassEnabled(disass || trc->m_db != nullptr);
     if(GET_PROP_VALUE(plugins).length()) {
         auto p = util::split(GET_PROP_VALUE(plugins), ';');
         for(std::string const& opt_val : p) {

src/vm/aes_sbox.h

@@ -35,7 +35,7 @@
 #ifndef _VM_AES_SBOX_H_
 #define _VM_AES_SBOX_H_
 #include <cstdint>
-
+extern "C" {
 const uint8_t AES_ENC_SBOX[] = {
     0x63, 0x7C, 0x77, 0x7B, 0xF2, 0x6B, 0x6F, 0xC5, 0x30, 0x01, 0x67, 0x2B, 0xFE, 0xD7, 0xAB, 0x76, 0xCA, 0x82, 0xC9, 0x7D, 0xFA, 0x59,
     0x47, 0xF0, 0xAD, 0xD4, 0xA2, 0xAF, 0x9C, 0xA4, 0x72, 0xC0, 0xB7, 0xFD, 0x93, 0x26, 0x36, 0x3F, 0xF7, 0xCC, 0x34, 0xA5, 0xE5, 0xF1,
@@ -65,5 +65,6 @@ const uint8_t AES_DEC_SBOX[] = {
     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/asmjit/vm_tgc5c.cpp

@@ -47,6 +47,22 @@
 #include <array>
 #include <iss/debugger/riscv_target_adapter.h>
 
+#ifndef _MSC_VER
+using int128_t = __int128;
+using uint128_t = unsigned __int128;
+namespace std {
+template <> struct make_unsigned<__int128> { typedef unsigned __int128 type; };
+template <> class __make_unsigned_selector<__int128 unsigned, false, false> {
+public:
+    typedef unsigned __int128 __type;
+};
+template <> struct is_signed<int128_t> { static constexpr bool value = true; };
+template <> struct is_signed<uint128_t> { static constexpr bool value = false; };
+template <> struct is_unsigned<int128_t> { static constexpr bool value = false; };
+template <> struct is_unsigned<uint128_t> { static constexpr bool value = true; };
+} // namespace std
+#endif
+
 namespace iss {
 namespace asmjit {
 
@@ -112,6 +128,11 @@ protected:
         auto sign_mask = 1ULL<<(W-1);
         return (from & mask) | ((from & sign_mask) ? ~mask : 0);
     }
+    inline void raise(uint16_t trap_id, uint16_t cause){
+        auto trap_val =  0x80ULL << 24 | (cause << 16) | trap_id;
+        this->core.reg.trap_state = trap_val;
+    }
+
 
 private:
     /****************************************************************************
@@ -335,7 +356,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -382,7 +402,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -429,7 +448,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         mov(jh.cc, get_ptr_for(jh, traits::LAST_BRANCH), static_cast<int>(NO_JUMP));
         if(rd>=static_cast<uint32_t>(traits::RFS)){
@@ -488,7 +506,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         mov(jh.cc, get_ptr_for(jh, traits::LAST_BRANCH), static_cast<int>(NO_JUMP));
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)){
@@ -561,7 +578,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         mov(jh.cc, get_ptr_for(jh, traits::LAST_BRANCH), static_cast<int>(NO_JUMP));
         if(rs2>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)){
@@ -625,7 +641,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         mov(jh.cc, get_ptr_for(jh, traits::LAST_BRANCH), static_cast<int>(NO_JUMP));
         if(rs2>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)){
@@ -689,7 +704,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         mov(jh.cc, get_ptr_for(jh, traits::LAST_BRANCH), static_cast<int>(NO_JUMP));
         if(rs2>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)){
@@ -755,7 +769,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         mov(jh.cc, get_ptr_for(jh, traits::LAST_BRANCH), static_cast<int>(NO_JUMP));
         if(rs2>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)){
@@ -821,7 +834,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         mov(jh.cc, get_ptr_for(jh, traits::LAST_BRANCH), static_cast<int>(NO_JUMP));
         if(rs2>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)){
@@ -885,7 +897,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         mov(jh.cc, get_ptr_for(jh, traits::LAST_BRANCH), static_cast<int>(NO_JUMP));
         if(rs2>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)){
@@ -949,7 +960,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -1003,7 +1013,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -1057,7 +1066,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -1111,7 +1119,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -1164,7 +1171,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -1217,7 +1223,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rs2>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -1266,7 +1271,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rs2>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -1315,7 +1319,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rs2>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -1364,7 +1367,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -1414,7 +1416,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -1475,7 +1476,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -1535,7 +1535,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -1584,7 +1583,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -1633,7 +1631,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -1682,7 +1679,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -1731,7 +1727,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -1780,7 +1775,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -1831,7 +1825,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -1881,7 +1874,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -1931,7 +1923,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -1981,7 +1972,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -2043,7 +2033,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -2103,7 +2092,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -2152,7 +2140,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -2202,7 +2189,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -2254,7 +2240,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -2303,7 +2288,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -2354,7 +2338,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         gen_write_mem(jh, traits::FENCE, static_cast<uint32_t>(traits::fence), (uint8_t)pred<<4|succ, 4);
         auto returnValue = CONT;
@@ -2390,7 +2373,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         mov(jh.cc, get_ptr_for(jh, traits::LAST_BRANCH), static_cast<int>(NO_JUMP));
         gen_raise(jh, 0, 11);
@@ -2427,7 +2409,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         mov(jh.cc, get_ptr_for(jh, traits::LAST_BRANCH), static_cast<int>(NO_JUMP));
         gen_raise(jh, 0, 3);
@@ -2464,7 +2445,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         mov(jh.cc, get_ptr_for(jh, traits::LAST_BRANCH), static_cast<int>(NO_JUMP));
         gen_leave(jh, 3);
@@ -2501,11 +2481,9 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         InvokeNode* call_wait_5;
-        jh.cc.comment("//call_wait");
-        jh.cc.invoke(&call_wait_5, &wait, FuncSignature::build<void, uint32_t>());
+        jh.cc.invoke(&call_wait_5,  &wait, FuncSignature::build<void, uint32_t>());
         setArg(call_wait_5, 0, 1);
         auto returnValue = CONT;
         
@@ -2544,7 +2522,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -2598,7 +2575,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -2652,7 +2628,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -2706,7 +2681,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -2756,7 +2730,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -2809,7 +2782,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -2862,7 +2834,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         gen_write_mem(jh, traits::FENCE, static_cast<uint32_t>(traits::fencei), imm, 4);
         auto returnValue = FLUSH;
@@ -2902,7 +2873,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -2955,7 +2925,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -3009,7 +2978,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -3062,7 +3030,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -3114,7 +3081,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -3203,7 +3169,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -3270,7 +3235,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -3363,7 +3327,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rs1>=static_cast<uint32_t>(traits::RFS)||rs2>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -3429,7 +3392,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(imm){
             mov(cc, get_ptr_for(jh, traits::X0+ rd+8),
@@ -3477,7 +3439,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         auto offs = gen_ext(cc, 
             (gen_operation(cc, add, load_reg_from_mem(jh, traits::X0 + rs1+8), uimm)
@@ -3523,7 +3484,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         auto offs = gen_ext(cc, 
             (gen_operation(cc, add, load_reg_from_mem(jh, traits::X0 + rs1+8), uimm)
@@ -3566,7 +3526,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rs1>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -3613,7 +3572,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         auto returnValue = CONT;
         
@@ -3650,7 +3608,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         mov(jh.cc, get_ptr_for(jh, traits::LAST_BRANCH), static_cast<int>(NO_JUMP));
         mov(cc, get_ptr_for(jh, traits::X0+ 1),
@@ -3694,7 +3651,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -3741,7 +3697,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(imm==0||rd>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -3785,7 +3740,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(nzimm){
             mov(cc, get_ptr_for(jh, traits::X0+ 2),
@@ -3830,7 +3784,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         auto returnValue = CONT;
@@ -3869,7 +3822,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         mov(cc, get_ptr_for(jh, traits::X0+ rs1+8),
               gen_operation(cc, shr, load_reg_from_mem(jh, traits::X0 + rs1+8), shamt)
@@ -3910,7 +3862,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(shamt){
             mov(cc, get_ptr_for(jh, traits::X0+ rs1+8),
@@ -3964,7 +3915,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         mov(cc, get_ptr_for(jh, traits::X0+ rs1+8),
               gen_ext(cc, 
@@ -4006,7 +3956,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         mov(cc, get_ptr_for(jh, traits::X0+ rd+8),
               gen_ext(cc, 
@@ -4048,7 +3997,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         mov(cc, get_ptr_for(jh, traits::X0+ rd+8),
               gen_operation(cc, bxor, load_reg_from_mem(jh, traits::X0 + rd+8), load_reg_from_mem(jh, traits::X0 + rs2+8))
@@ -4089,7 +4037,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         mov(cc, get_ptr_for(jh, traits::X0+ rd+8),
               gen_operation(cc, bor, load_reg_from_mem(jh, traits::X0 + rd+8), load_reg_from_mem(jh, traits::X0 + rs2+8))
@@ -4130,7 +4077,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         mov(cc, get_ptr_for(jh, traits::X0+ rd+8),
               gen_operation(cc, band, load_reg_from_mem(jh, traits::X0 + rd+8), load_reg_from_mem(jh, traits::X0 + rs2+8))
@@ -4170,7 +4116,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         mov(jh.cc, get_ptr_for(jh, traits::LAST_BRANCH), static_cast<int>(NO_JUMP));
         auto PC_val_v = (uint32_t)(PC+(int16_t)sext<12>(imm));
@@ -4212,7 +4157,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         mov(jh.cc, get_ptr_for(jh, traits::LAST_BRANCH), static_cast<int>(NO_JUMP));
         {
@@ -4263,7 +4207,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         mov(jh.cc, get_ptr_for(jh, traits::LAST_BRANCH), static_cast<int>(NO_JUMP));
         {
@@ -4314,7 +4257,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rs1>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -4362,7 +4304,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)||rd==0){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -4412,7 +4353,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -4458,7 +4398,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         mov(jh.cc, get_ptr_for(jh, traits::LAST_BRANCH), static_cast<int>(NO_JUMP));
         if(rs1&&rs1<static_cast<uint32_t>(traits::RFS)){
@@ -4504,7 +4443,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         gen_raise(jh, 0, 2);
         auto returnValue = CONT;
@@ -4543,7 +4481,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rd>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -4591,7 +4528,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         mov(jh.cc, get_ptr_for(jh, traits::LAST_BRANCH), static_cast<int>(NO_JUMP));
         if(rs1>=static_cast<uint32_t>(traits::RFS)){
@@ -4640,7 +4576,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         gen_raise(jh, 0, 3);
         auto returnValue = CONT;
@@ -4679,7 +4614,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         if(rs2>=static_cast<uint32_t>(traits::RFS)){
             gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
@@ -4724,7 +4658,6 @@ private:
         mov(cc, jh.next_pc, pc.val);
 
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         /*generate behavior*/
         gen_raise(jh, 0, static_cast<uint32_t>(traits::RV_CAUSE_ILLEGAL_INSTRUCTION));
         auto returnValue = CONT;
@@ -4756,7 +4689,6 @@ private:
         pc = pc + ((instr & 3) == 3 ? 4 : 2);
         mov(cc, jh.next_pc, pc.val);
         gen_instr_prologue(jh);
-        cc.comment("//behavior:");
         gen_raise(jh, 0, 2);
         gen_sync(jh, POST_SYNC, instr_descr.size());
         gen_instr_epilogue(jh);
@@ -4802,24 +4734,25 @@ template <typename ARCH>
 void vm_impl<ARCH>::gen_instr_prologue(jit_holder& jh) {
     auto& cc = jh.cc;
 
-    cc.comment("//gen_instr_prologue");
-
     x86_reg_t current_trap_state = get_reg_for(cc, traits::TRAP_STATE);
     mov(cc, current_trap_state, get_ptr_for(jh, traits::TRAP_STATE));
     mov(cc, get_ptr_for(jh, traits::PENDING_TRAP), current_trap_state);
+    cc.comment("//Instruction prologue end");
 
 }
 template <typename ARCH>
 void vm_impl<ARCH>::gen_instr_epilogue(jit_holder& jh) {
     auto& cc = jh.cc;
 
-    cc.comment("//gen_instr_epilogue");
+    cc.comment("//Instruction epilogue begin");
     x86_reg_t current_trap_state = get_reg_for(cc, traits::TRAP_STATE);
     mov(cc, current_trap_state, get_ptr_for(jh, traits::TRAP_STATE));
     cmp(cc, current_trap_state, 0);
     cc.jne(jh.trap_entry);
     cc.inc(get_ptr_for(jh, traits::ICOUNT));
     cc.inc(get_ptr_for(jh, traits::CYCLE));
+    cc.comment("//Instruction epilogue end");
+
 }
 template <typename ARCH>
 void vm_impl<ARCH>::gen_block_prologue(jit_holder& jh){
@@ -4831,7 +4764,7 @@ void vm_impl<ARCH>::gen_block_prologue(jit_holder& jh){
 template <typename ARCH>
 void vm_impl<ARCH>::gen_block_epilogue(jit_holder& jh){
     x86::Compiler& cc = jh.cc;
-    cc.comment("//gen_block_epilogue");
+    cc.comment("//block epilogue begin");
     cc.ret(jh.next_pc);
 
     cc.bind(jh.trap_entry);
@@ -4843,7 +4776,6 @@ void vm_impl<ARCH>::gen_block_epilogue(jit_holder& jh){
     x86::Gp current_pc = get_reg_for_Gp(cc, traits::PC);
     mov(cc, current_pc, get_ptr_for(jh, traits::PC));
 
-    cc.comment("//enter trap call;");
     InvokeNode* call_enter_trap;
     cc.invoke(&call_enter_trap, &enter_trap, FuncSignature::build<uint64_t, void*, uint64_t, uint64_t, uint64_t>());
     call_enter_trap->setArg(0, jh.arch_if_ptr);
@@ -4861,7 +4793,6 @@ void vm_impl<ARCH>::gen_block_epilogue(jit_holder& jh){
 template <typename ARCH>
 inline void vm_impl<ARCH>::gen_raise(jit_holder& jh, uint16_t trap_id, uint16_t cause) {
     auto& cc = jh.cc;
-    cc.comment("//gen_raise");
     auto tmp1 = get_reg_for(cc, traits::TRAP_STATE);
     mov(cc, tmp1, 0x80ULL << 24 | (cause << 16) | trap_id);
     mov(cc, get_ptr_for(jh, traits::TRAP_STATE), tmp1);

src/vm/llvm/fp_impl.cpp

@@ -102,6 +102,7 @@ void add_fp_functions_2_module(Module* mod, uint32_t flen, uint32_t xlen) {
         FDECL(fmadd_s, INT_TYPE(32), INT_TYPE(32), INT_TYPE(32), INT_TYPE(32), INT_TYPE(32), INT_TYPE(8));
         FDECL(fsel_s, INT_TYPE(32), INT_TYPE(32), INT_TYPE(32), INT_TYPE(32));
         FDECL(fclass_s, INT_TYPE(32), INT_TYPE(32));
+        FDECL(unbox_s, INT_TYPE(32), INT_TYPE(32), INT_TYPE(64)); // technically the first arg is only 8 bits
 
         FDECL(f32toi32, INT_TYPE(32), INT_TYPE(32), INT_TYPE(8));
         FDECL(f32toui32, INT_TYPE(32), INT_TYPE(32), INT_TYPE(8));
@@ -112,7 +113,6 @@ void add_fp_functions_2_module(Module* mod, uint32_t flen, uint32_t xlen) {
         FDECL(i64tof32, INT_TYPE(32), INT_TYPE(64), INT_TYPE(8));
         FDECL(ui64tof32, INT_TYPE(32), INT_TYPE(64), INT_TYPE(8));
         if(flen > 32) {
-            FDECL(unbox_s, INT_TYPE(32), INT_TYPE(32), INT_TYPE(64)); // technically the first arg is only 8 bits
 
             FDECL(fadd_d, INT_TYPE(64), INT_TYPE(64), INT_TYPE(64), INT_TYPE(8));
             FDECL(fsub_d, INT_TYPE(64), INT_TYPE(64), INT_TYPE(64), INT_TYPE(8));
@@ -134,6 +134,8 @@ void add_fp_functions_2_module(Module* mod, uint32_t flen, uint32_t xlen) {
             FDECL(ui32tof64, INT_TYPE(64), INT_TYPE(32), INT_TYPE(8));
             FDECL(f64toi32, INT_TYPE(32), INT_TYPE(64), INT_TYPE(8));
             FDECL(f64toui32, INT_TYPE(32), INT_TYPE(64), INT_TYPE(8));
+
+            FDECL(unbox_d, INT_TYPE(64), INT_TYPE(32), INT_TYPE(64)); // technically the first arg is only 8 bits
         }
     }
 }

src/vm/llvm/vm_tgc5c.cpp

@@ -31,10 +31,12 @@
  *******************************************************************************/
 // clang-format off
 #include <iss/arch/tgc5c.h>
+// vm_base needs to be included before gdb_session as termios.h (via boost and gdb_server) has a define which clashes with a variable
+// name in ConstantRange.h
+#include <iss/llvm/vm_base.h>
 #include <iss/debugger/gdb_session.h>
 #include <iss/debugger/server.h>
 #include <iss/iss.h>
-#include <iss/llvm/vm_base.h>
 #include <util/logging.h>
 #include <iss/instruction_decoder.h>
 

src/vm/vector_functions.cpp

@@ -34,6 +34,7 @@
 
 #include "vector_functions.h"
 #include "iss/vm_types.h"
+#include "vm/aes_sbox.h"
 #include <algorithm>
 #include <cassert>
 #include <cstddef>
@@ -98,4 +99,309 @@ vmask_view read_vmask(uint8_t* V, uint16_t VLEN, uint16_t elem_count, uint8_t re
     assert(mask_start + elem_count / 8 <= V + VLEN * RFS / 8);
     return {mask_start, elem_count};
 }
+uint8_t xt2(uint8_t x) { return (x << 1) ^ (bit_sub<7, 1>(x) ? 27 : 0); }
+
+uint8_t xt3(uint8_t x) { return x ^ xt2(x); }
+
+uint8_t gfmul(uint8_t x, uint8_t y) {
+    return (bit_sub<0, 1>(y) ? x : 0) ^ (bit_sub<1, 1>(y) ? xt2(x) : 0) ^ (bit_sub<2, 1>(y) ? xt2(xt2(x)) : 0) ^
+           (bit_sub<3, 1>(y) ? xt2(xt2(xt2(x))) : 0);
+}
+
+uint32_t aes_mixcolumn_byte_fwd(uint8_t so) {
+    return ((uint32_t)gfmul(so, 3) << 24) | ((uint32_t)so << 16) | ((uint32_t)so << 8) | gfmul(so, 2);
+}
+
+uint32_t aes_mixcolumn_byte_inv(uint8_t so) {
+    return ((uint32_t)gfmul(so, 11) << 24) | ((uint32_t)gfmul(so, 13) << 16) | ((uint32_t)gfmul(so, 9) << 8) | gfmul(so, 14);
+}
+
+uint32_t aes_mixcolumn_fwd(uint32_t x) {
+    uint8_t s0 = bit_sub<0, 7 - 0 + 1>(x);
+    uint8_t s1 = bit_sub<8, 15 - 8 + 1>(x);
+    uint8_t s2 = bit_sub<16, 23 - 16 + 1>(x);
+    uint8_t s3 = bit_sub<24, 31 - 24 + 1>(x);
+    uint8_t b0 = xt2(s0) ^ xt3(s1) ^ (s2) ^ (s3);
+    uint8_t b1 = (s0) ^ xt2(s1) ^ xt3(s2) ^ (s3);
+    uint8_t b2 = (s0) ^ (s1) ^ xt2(s2) ^ xt3(s3);
+    uint8_t b3 = xt3(s0) ^ (s1) ^ (s2) ^ xt2(s3);
+    return ((uint32_t)b3 << 24) | ((uint32_t)b2 << 16) | ((uint32_t)b1 << 8) | b0;
+}
+
+uint32_t aes_mixcolumn_inv(uint32_t x) {
+    uint8_t s0 = bit_sub<0, 7 - 0 + 1>(x);
+    uint8_t s1 = bit_sub<8, 15 - 8 + 1>(x);
+    uint8_t s2 = bit_sub<16, 23 - 16 + 1>(x);
+    uint8_t s3 = bit_sub<24, 31 - 24 + 1>(x);
+    uint8_t b0 = gfmul(s0, 14) ^ gfmul(s1, 11) ^ gfmul(s2, 13) ^ gfmul(s3, 9);
+    uint8_t b1 = gfmul(s0, 9) ^ gfmul(s1, 14) ^ gfmul(s2, 11) ^ gfmul(s3, 13);
+    uint8_t b2 = gfmul(s0, 13) ^ gfmul(s1, 9) ^ gfmul(s2, 14) ^ gfmul(s3, 11);
+    uint8_t b3 = gfmul(s0, 11) ^ gfmul(s1, 13) ^ gfmul(s2, 9) ^ gfmul(s3, 14);
+    return ((uint32_t)b3 << 24) | ((uint32_t)b2 << 16) | ((uint32_t)b1 << 8) | b0;
+}
+
+uint32_t aes_decode_rcon(uint8_t r) {
+    switch(r) {
+    case 0:
+        return 1;
+    case 1:
+        return 2;
+    case 2:
+        return 4;
+    case 3:
+        return 8;
+    case 4:
+        return 16;
+    case 5:
+        return 32;
+    case 6:
+        return 64;
+    case 7:
+        return 128;
+    case 8:
+        return 27;
+    case 9:
+        return 54;
+    }
+    return 0;
+}
+
+uint32_t aes_subword_fwd(uint32_t x) {
+    return ((uint32_t)aes_sbox_fwd(bit_sub<24, 31 - 24 + 1>(x)) << 24) | ((uint32_t)aes_sbox_fwd(bit_sub<16, 23 - 16 + 1>(x)) << 16) |
+           ((uint32_t)aes_sbox_fwd(bit_sub<8, 15 - 8 + 1>(x)) << 8) | aes_sbox_fwd(bit_sub<0, 7 - 0 + 1>(x));
+}
+
+uint32_t aes_subword_inv(uint32_t x) {
+    return ((uint32_t)aes_sbox_inv(bit_sub<24, 31 - 24 + 1>(x)) << 24) | ((uint32_t)aes_sbox_inv(bit_sub<16, 23 - 16 + 1>(x)) << 16) |
+           ((uint32_t)aes_sbox_inv(bit_sub<8, 15 - 8 + 1>(x)) << 8) | aes_sbox_inv(bit_sub<0, 7 - 0 + 1>(x));
+}
+
+uint32_t aes_get_column(__uint128_t state, unsigned c) {
+    assert(c < 4);
+    return static_cast<uint32_t>(state >> (32 * c));
+};
+
+uint64_t aes_apply_fwd_sbox_to_each_byte(uint64_t x) {
+    return ((uint64_t)aes_sbox_fwd(bit_sub<56, 63 - 56 + 1>(x)) << 56) | ((uint64_t)aes_sbox_fwd(bit_sub<48, 55 - 48 + 1>(x)) << 48) |
+           ((uint64_t)aes_sbox_fwd(bit_sub<40, 47 - 40 + 1>(x)) << 40) | ((uint64_t)aes_sbox_fwd(bit_sub<32, 39 - 32 + 1>(x)) << 32) |
+           ((uint64_t)aes_sbox_fwd(bit_sub<24, 31 - 24 + 1>(x)) << 24) | ((uint64_t)aes_sbox_fwd(bit_sub<16, 23 - 16 + 1>(x)) << 16) |
+           ((uint64_t)aes_sbox_fwd(bit_sub<8, 15 - 8 + 1>(x)) << 8) | aes_sbox_fwd(bit_sub<0, 7 - 0 + 1>(x));
+}
+
+uint64_t aes_apply_inv_sbox_to_each_byte(uint64_t x) {
+    return ((uint64_t)aes_sbox_inv(bit_sub<56, 63 - 56 + 1>(x)) << 56) | ((uint64_t)aes_sbox_inv(bit_sub<48, 55 - 48 + 1>(x)) << 48) |
+           ((uint64_t)aes_sbox_inv(bit_sub<40, 47 - 40 + 1>(x)) << 40) | ((uint64_t)aes_sbox_inv(bit_sub<32, 39 - 32 + 1>(x)) << 32) |
+           ((uint64_t)aes_sbox_inv(bit_sub<24, 31 - 24 + 1>(x)) << 24) | ((uint64_t)aes_sbox_inv(bit_sub<16, 23 - 16 + 1>(x)) << 16) |
+           ((uint64_t)aes_sbox_inv(bit_sub<8, 15 - 8 + 1>(x)) << 8) | aes_sbox_inv(bit_sub<0, 7 - 0 + 1>(x));
+}
+
+uint64_t aes_rv64_shiftrows_fwd(uint64_t rs2, uint64_t rs1) {
+    return ((uint64_t)bit_sub<24, 31 - 24 + 1>(rs1) << 56) | ((uint64_t)bit_sub<48, 55 - 48 + 1>(rs2) << 48) |
+           ((uint64_t)bit_sub<8, 15 - 8 + 1>(rs2) << 40) | ((uint64_t)bit_sub<32, 39 - 32 + 1>(rs1) << 32) |
+           ((uint64_t)bit_sub<56, 63 - 56 + 1>(rs2) << 24) | ((uint64_t)bit_sub<16, 23 - 16 + 1>(rs2) << 16) |
+           ((uint64_t)bit_sub<40, 47 - 40 + 1>(rs1) << 8) | bit_sub<0, 7 - 0 + 1>(rs1);
+}
+
+uint64_t aes_rv64_shiftrows_inv(uint64_t rs2, uint64_t rs1) {
+    return ((uint64_t)bit_sub<24, 31 - 24 + 1>(rs2) << 56) | ((uint64_t)bit_sub<48, 55 - 48 + 1>(rs2) << 48) |
+           ((uint64_t)bit_sub<8, 15 - 8 + 1>(rs1) << 40) | ((uint64_t)bit_sub<32, 39 - 32 + 1>(rs1) << 32) |
+           ((uint64_t)bit_sub<56, 63 - 56 + 1>(rs1) << 24) | ((uint64_t)bit_sub<16, 23 - 16 + 1>(rs2) << 16) |
+           ((uint64_t)bit_sub<40, 47 - 40 + 1>(rs2) << 8) | bit_sub<0, 7 - 0 + 1>(rs1);
+}
+
+uint128_t aes_shift_rows_fwd(uint128_t x) {
+    uint32_t ic3 = aes_get_column(x, 3);
+    uint32_t ic2 = aes_get_column(x, 2);
+    uint32_t ic1 = aes_get_column(x, 1);
+    uint32_t ic0 = aes_get_column(x, 0);
+    uint32_t oc0 = ((uint32_t)bit_sub<24, 31 - 24 + 1>(ic3) << 24) | ((uint32_t)bit_sub<16, 23 - 16 + 1>(ic2) << 16) |
+                   ((uint32_t)bit_sub<8, 15 - 8 + 1>(ic1) << 8) | bit_sub<0, 7 - 0 + 1>(ic0);
+    uint32_t oc1 = ((uint32_t)bit_sub<24, 31 - 24 + 1>(ic0) << 24) | ((uint32_t)bit_sub<16, 23 - 16 + 1>(ic3) << 16) |
+                   ((uint32_t)bit_sub<8, 15 - 8 + 1>(ic2) << 8) | bit_sub<0, 7 - 0 + 1>(ic1);
+    uint32_t oc2 = ((uint32_t)bit_sub<24, 31 - 24 + 1>(ic1) << 24) | ((uint32_t)bit_sub<16, 23 - 16 + 1>(ic0) << 16) |
+                   ((uint32_t)bit_sub<8, 15 - 8 + 1>(ic3) << 8) | bit_sub<0, 7 - 0 + 1>(ic2);
+    uint32_t oc3 = ((uint32_t)bit_sub<24, 31 - 24 + 1>(ic2) << 24) | ((uint32_t)bit_sub<16, 23 - 16 + 1>(ic1) << 16) |
+                   ((uint32_t)bit_sub<8, 15 - 8 + 1>(ic0) << 8) | bit_sub<0, 7 - 0 + 1>(ic3);
+    return ((uint128_t)oc3 << 96) | ((uint128_t)oc2 << 64) | ((uint128_t)oc1 << 32) | oc0;
+}
+
+uint128_t aes_shift_rows_inv(uint128_t x) {
+    uint32_t ic3 = aes_get_column(x, 3);
+    uint32_t ic2 = aes_get_column(x, 2);
+    uint32_t ic1 = aes_get_column(x, 1);
+    uint32_t ic0 = aes_get_column(x, 0);
+    uint32_t oc0 = ((uint32_t)bit_sub<24, 31 - 24 + 1>(ic1) << 24) | ((uint32_t)bit_sub<16, 23 - 16 + 1>(ic2) << 16) |
+                   ((uint32_t)bit_sub<8, 15 - 8 + 1>(ic3) << 8) | bit_sub<0, 7 - 0 + 1>(ic0);
+    uint32_t oc1 = ((uint32_t)bit_sub<24, 31 - 24 + 1>(ic2) << 24) | ((uint32_t)bit_sub<16, 23 - 16 + 1>(ic3) << 16) |
+                   ((uint32_t)bit_sub<8, 15 - 8 + 1>(ic0) << 8) | bit_sub<0, 7 - 0 + 1>(ic1);
+    uint32_t oc2 = ((uint32_t)bit_sub<24, 31 - 24 + 1>(ic3) << 24) | ((uint32_t)bit_sub<16, 23 - 16 + 1>(ic0) << 16) |
+                   ((uint32_t)bit_sub<8, 15 - 8 + 1>(ic1) << 8) | bit_sub<0, 7 - 0 + 1>(ic2);
+    uint32_t oc3 = ((uint32_t)bit_sub<24, 31 - 24 + 1>(ic0) << 24) | ((uint32_t)bit_sub<16, 23 - 16 + 1>(ic1) << 16) |
+                   ((uint32_t)bit_sub<8, 15 - 8 + 1>(ic2) << 8) | bit_sub<0, 7 - 0 + 1>(ic3);
+    return ((uint128_t)oc3 << 96) | ((uint128_t)oc2 << 64) | ((uint128_t)oc1 << 32) | oc0;
+}
+
+uint128_t aes_subbytes_fwd(uint128_t x) {
+    uint32_t oc0 = aes_subword_fwd(aes_get_column(x, 0));
+    uint32_t oc1 = aes_subword_fwd(aes_get_column(x, 1));
+    uint32_t oc2 = aes_subword_fwd(aes_get_column(x, 2));
+    uint32_t oc3 = aes_subword_fwd(aes_get_column(x, 3));
+    return ((uint128_t)oc3 << 96) | ((uint128_t)oc2 << 64) | ((uint128_t)oc1 << 32) | oc0;
+}
+
+uint128_t aes_subbytes_inv(uint128_t x) {
+    uint32_t oc0 = aes_subword_inv(aes_get_column(x, 0));
+    uint32_t oc1 = aes_subword_inv(aes_get_column(x, 1));
+    uint32_t oc2 = aes_subword_inv(aes_get_column(x, 2));
+    uint32_t oc3 = aes_subword_inv(aes_get_column(x, 3));
+    return ((uint128_t)oc3 << 96) | ((uint128_t)oc2 << 64) | ((uint128_t)oc1 << 32) | oc0;
+}
+
+uint128_t aes_mixcolumns_fwd(uint128_t x) {
+    uint32_t oc0 = aes_mixcolumn_fwd(aes_get_column(x, 0));
+    uint32_t oc1 = aes_mixcolumn_fwd(aes_get_column(x, 1));
+    uint32_t oc2 = aes_mixcolumn_fwd(aes_get_column(x, 2));
+    uint32_t oc3 = aes_mixcolumn_fwd(aes_get_column(x, 3));
+    return ((uint128_t)oc3 << 96) | ((uint128_t)oc2 << 64) | ((uint128_t)oc1 << 32) | oc0;
+}
+
+uint128_t aes_mixcolumns_inv(uint128_t x) {
+    uint32_t oc0 = aes_mixcolumn_inv(aes_get_column(x, 0));
+    uint32_t oc1 = aes_mixcolumn_inv(aes_get_column(x, 1));
+    uint32_t oc2 = aes_mixcolumn_inv(aes_get_column(x, 2));
+    uint32_t oc3 = aes_mixcolumn_inv(aes_get_column(x, 3));
+    return ((uint128_t)oc3 << 96) | ((uint128_t)oc2 << 64) | ((uint128_t)oc1 << 32) | oc0;
+}
+
+uint32_t aes_rotword(uint32_t x) {
+    uint8_t a0 = bit_sub<0, 7 - 0 + 1>(x);
+    uint8_t a1 = bit_sub<8, 15 - 8 + 1>(x);
+    uint8_t a2 = bit_sub<16, 23 - 16 + 1>(x);
+    uint8_t a3 = bit_sub<24, 31 - 24 + 1>(x);
+    return ((uint32_t)a0 << 24) | ((uint32_t)a3 << 16) | ((uint32_t)a2 << 8) | a1;
+}
+
+std::function<uint128_t(uint128_t, uint128_t, uint128_t)> get_crypto_funct(unsigned funct6, unsigned vs1) {
+    switch(funct6) {
+    case 0b101000: // VAES.VV
+    case 0b101001: // VAES.VS
+        switch(vs1) {
+        case 0b00000: // VAESDM
+            return [](uint128_t state, uint128_t rkey, uint128_t) {
+                uint128_t sr = aes_shift_rows_inv(state);
+                uint128_t sb = aes_subbytes_inv(sr);
+                uint128_t ark = sb ^ rkey;
+                uint128_t mix = aes_mixcolumns_inv(ark);
+                return mix;
+            };
+        case 0b00001: // VAESDF
+            return [](uint128_t state, uint128_t rkey, uint128_t) {
+                uint128_t sr = aes_shift_rows_inv(state);
+                uint128_t sb = aes_subbytes_inv(sr);
+                uint128_t ark = sb ^ rkey;
+                return ark;
+            };
+        case 0b00010: // VAESEM
+            return [](uint128_t state, uint128_t rkey, uint128_t) {
+                uint128_t sb = aes_subbytes_fwd(state);
+                uint128_t sr = aes_shift_rows_fwd(sb);
+                uint128_t mix = aes_mixcolumns_fwd(sr);
+                uint128_t ark = mix ^ rkey;
+                return ark;
+            };
+        case 0b00011: // VAESEF
+            return [](uint128_t state, uint128_t rkey, uint128_t) {
+                uint128_t sb = aes_subbytes_fwd(state);
+                uint128_t sr = aes_shift_rows_fwd(sb);
+                uint128_t ark = sr ^ rkey;
+                return ark;
+            };
+        case 0b00111: // VAESZ
+            return [](uint128_t state, uint128_t rkey, uint128_t) {
+                uint128_t ark = state ^ rkey;
+                return ark;
+            };
+        case 0b10000: // VSM4R
+            throw new std::runtime_error("Unsupported operation in get_crypto_funct");
+        case 0b10001: // VGMUL
+            return [](uint128_t vd, uint128_t vs2, uint128_t) {
+                uint128_t Y = brev8<uint128_t>(vd);
+                uint128_t H = brev8<uint128_t>(vs2);
+                uint128_t Z = 0;
+                for(size_t bit = 0; bit < 128; bit++) {
+                    if((Y >> bit) & 1)
+                        Z ^= H;
+                    bool reduce = (H >> 127) & 1;
+                    H = H << 1;
+                    if(reduce)
+                        H ^= 0x87;
+                }
+                uint128_t result = brev8<uint128_t>(Z);
+                return result;
+            };
+        default:
+            throw new std::runtime_error("Unsupported operation in get_crypto_funct");
+        }
+    case 0b100000: // VSM3ME
+    case 0b100001: // VSM4K
+        throw new std::runtime_error("Unsupported operation in get_crypto_funct");
+    case 0b100010: // VAESKF1
+        return [](uint128_t vd, uint128_t vs2, uint128_t r) {
+            auto extract_word = [](const uint128_t& value, int index) -> uint32_t {
+                return static_cast<uint32_t>((value >> (32 * index)) & std::numeric_limits<uint32_t>::max());
+            };
+
+            uint32_t k0 = (vs2 >> 32 * 0) & std::numeric_limits<uint32_t>::max();
+            uint32_t k1 = (vs2 >> 32 * 1) & std::numeric_limits<uint32_t>::max();
+            uint32_t k2 = (vs2 >> 32 * 2) & std::numeric_limits<uint32_t>::max();
+            uint32_t k3 = (vs2 >> 32 * 3) & std::numeric_limits<uint32_t>::max();
+            uint32_t w0 = aes_subword_fwd(aes_rotword(k3)) ^ aes_decode_rcon(r) ^ k0;
+            uint32_t w1 = w0 ^ k1;
+            uint32_t w2 = w1 ^ k2;
+            uint32_t w3 = w2 ^ k3;
+            uint128_t result = (uint128_t(w3) << 96) | (uint128_t(w2) << 64) | (uint128_t(w1) << 32) | (uint128_t(w0));
+            return result;
+        };
+    case 0b101010: // VAESKF2
+        return [](uint128_t vd, uint128_t vs2, uint128_t r) {
+            uint32_t k0 = (vs2 >> 32 * 0) & std::numeric_limits<uint32_t>::max();
+            uint32_t k1 = (vs2 >> 32 * 1) & std::numeric_limits<uint32_t>::max();
+            uint32_t k2 = (vs2 >> 32 * 2) & std::numeric_limits<uint32_t>::max();
+            uint32_t k3 = (vs2 >> 32 * 3) & std::numeric_limits<uint32_t>::max();
+            uint32_t rkb0 = (vd >> 32 * 0) & std::numeric_limits<uint32_t>::max();
+            uint32_t rkb1 = (vd >> 32 * 1) & std::numeric_limits<uint32_t>::max();
+            uint32_t rkb2 = (vd >> 32 * 2) & std::numeric_limits<uint32_t>::max();
+            uint32_t rkb3 = (vd >> 32 * 3) & std::numeric_limits<uint32_t>::max();
+            uint32_t w0 = r & 1 ? aes_subword_fwd(k3) ^ rkb0 : aes_subword_fwd(aes_rotword(k3)) ^ aes_decode_rcon((r >> 1) - 1) ^ rkb0;
+            uint32_t w1 = w0 ^ rkb1;
+            uint32_t w2 = w1 ^ rkb2;
+            uint32_t w3 = w2 ^ rkb3;
+            uint128_t result = (uint128_t(w3) << 96) | (uint128_t(w2) << 64) | (uint128_t(w1) << 32) | (uint128_t(w0));
+            return result;
+        };
+    case 0b101011: // VSM3C
+        throw new std::runtime_error("Unsupported operation in get_crypto_funct");
+    case 0b101100: // VGHSH
+        return [](uint128_t Y, uint128_t vs2, uint128_t X) {
+            auto H = brev8<uint128_t>(vs2);
+            uint128_t Z = 0;
+            uint128_t S = brev8<uint128_t>(Y ^ X);
+            for(size_t bit = 0; bit < 128; bit++) {
+                if((S >> bit) & 1)
+                    Z ^= H;
+                bool reduce = (H >> 127) & 1;
+                H = H << 1;
+                if(reduce)
+                    H ^= 0x87;
+            }
+            uint128_t result = brev8<uint128_t>(Z);
+            return result;
+        };
+    case 0b101101: // VSHA2MS
+    case 0b101110: // VSHA2CH
+    case 0b101111: // VSHA2CL
+    default:
+        throw new std::runtime_error("Unknown funct6 in get_crypto_funct");
+    }
+}
 } // namespace softvector

src/vm/vector_functions.h

@@ -41,6 +41,10 @@
 #include <functional>
 #include <stdint.h>
 namespace softvector {
+#ifndef _MSC_VER
+using int128_t = __int128;
+using uint128_t = unsigned __int128;
+#endif
 const unsigned RFS = 32;
 
 struct vtype_t {
@@ -70,9 +74,14 @@ struct vmask_view {
 };
 vmask_view read_vmask(uint8_t* V, uint16_t VLEN, uint16_t elem_count, uint8_t reg_idx = 0);
 template <unsigned VLEN> vmask_view read_vmask(uint8_t* V, uint16_t elem_count, uint8_t reg_idx = 0);
+std::function<uint128_t(uint128_t, uint128_t, uint128_t)> get_crypto_funct(unsigned funct6, unsigned vs1);
+
+template <typename dest_elem_t, typename src_elem_t = dest_elem_t> dest_elem_t brev(src_elem_t vs2);
+template <typename dest_elem_t, typename src_elem_t = dest_elem_t> dest_elem_t brev8(src_elem_t vs2);
 
 bool softvec_read(void* core, uint64_t addr, uint64_t length, uint8_t* data);
 bool softvec_write(void* core, uint64_t addr, uint64_t length, uint8_t* data);
+
 template <unsigned VLEN, typename eew_t>
 uint64_t vector_load_store(void* core, std::function<bool(void*, uint64_t, uint64_t, uint8_t*)> load_store_fn, uint8_t* V, uint64_t vl,
                            uint64_t vstart, vtype_t vtype, bool vm, uint8_t vd, uint64_t rs1, uint8_t segment_size, int64_t stride = 0,
@@ -167,6 +176,17 @@ void mask_fp_vector_vector_op(uint8_t* V, unsigned funct6, uint64_t vl, uint64_t
 template <unsigned VLEN, typename elem_t>
 void mask_fp_vector_imm_op(uint8_t* V, unsigned funct6, uint64_t vl, uint64_t vstart, vtype_t vtype, bool vm, unsigned vd, unsigned vs2,
                            elem_t imm, uint8_t rm);
+template <unsigned VLEN, unsigned EGS>
+void vector_vector_crypto(uint8_t* V, unsigned funct6, uint64_t eg_len, uint64_t eg_start, vtype_t vtype, unsigned vd, unsigned vs2,
+                          unsigned vs1);
+template <unsigned VLEN, unsigned EGS>
+void vector_scalar_crypto(uint8_t* V, unsigned funct6, uint64_t eg_len, uint64_t eg_start, vtype_t vtype, unsigned vd, unsigned vs2,
+                          unsigned vs1);
+template <unsigned VLEN, unsigned EGS>
+void vector_imm_crypto(uint8_t* V, unsigned funct6, uint64_t eg_len, uint64_t eg_start, vtype_t vtype, unsigned vd, unsigned vs2,
+                       uint8_t imm);
+template <unsigned VLEN, unsigned EGS, typename elem_type_t>
+void vector_crypto(uint8_t* V, unsigned funct6, uint64_t eg_len, uint64_t eg_start, vtype_t vtype, unsigned vd, unsigned vs2, unsigned vs1);
 } // namespace softvector
 #include "vm/vector_functions.hpp"
 #endif /* _VM_VECTOR_FUNCTIONS_H_ */

src/vm/vector_functions.hpp

@@ -32,6 +32,7 @@
 //       alex@minres.com - initial API and implementation
 ////////////////////////////////////////////////////////////////////////////////
 #pragma once
+#include "iss/interp/vm_base.h"
 extern "C" {
 #include <softfloat.h>
 }
@@ -55,6 +56,14 @@ extern "C" {
 #ifdef __SIZEOF_INT128__
 template <> struct std::make_signed<__uint128_t> { using type = __int128_t; };
 template <> struct std::make_signed<__int128_t> { using type = __int128_t; };
+// helper struct to make calling twice<T> on 128-bit datatypes legal at compile time
+struct poison128_t {
+    poison128_t() { throw std::runtime_error("Attempt to use twice<__uint128_t>::type at runtime"); }
+    poison128_t(const poison128_t&) { throw std::runtime_error("Copy of poison128_t is not allowed"); }
+    template <typename U> poison128_t(U) { throw std::runtime_error("Conversion to poison128_t is not allowed"); }
+    operator __uint128_t() const { throw std::runtime_error("Use of poison128_t as __uint128_t is not allowed"); }
+};
+template <> struct std::make_signed<poison128_t> { using type = poison128_t; };
 #endif
 
 namespace softvector {
@@ -108,6 +117,10 @@ template <> struct twice<uint32_t> { using type = uint64_t; };
 #ifdef __SIZEOF_INT128__
 template <> struct twice<int64_t> { using type = __int128_t; };
 template <> struct twice<uint64_t> { using type = __uint128_t; };
+
+template <> struct twice<__uint128_t> { using type = poison128_t; };
+template <> struct twice<__int128_t> { using type = poison128_t; };
+
 #endif
 template <class T> using twice_t = typename twice<T>::type; // for convenience
 template <typename TO, typename FROM> constexpr TO sext(FROM val) {
@@ -180,6 +193,8 @@ std::function<dest_elem_t(dest_elem_t, src2_elem_t, src1_elem_t)> get_funct(unsi
         switch(funct6) {
         case 0b000000: // VADD
             return [](dest_elem_t vd, src2_elem_t vs2, src1_elem_t vs1) { return vs2 + vs1; };
+        case 0b000001: // VANDN
+            return [](dest_elem_t vd, src2_elem_t vs2, src1_elem_t vs1) { return vs2 & ~vs1; };
         case 0b000010: // VSUB
             return [](dest_elem_t vd, src2_elem_t vs2, src1_elem_t vs1) { return vs2 - vs1; };
         case 0b000011: // VRSUB
@@ -205,6 +220,22 @@ std::function<dest_elem_t(dest_elem_t, src2_elem_t, src1_elem_t)> get_funct(unsi
                 return static_cast<std::make_signed_t<dest_elem_t>>(static_cast<std::make_signed_t<src2_elem_t>>(vs2) -
                                                                     static_cast<std::make_signed_t<src1_elem_t>>(vs1));
             };
+        case 0b010100: // VROR
+            return [](dest_elem_t vd, src2_elem_t vs2, src1_elem_t vs1) {
+                constexpr dest_elem_t bits = sizeof(src2_elem_t) * 8;
+                auto shamt = vs1 & shift_mask<src1_elem_t>();
+                return (vs2 >> shamt) | (vs2 << (bits - shamt));
+            };
+        case 0b010101: { // VROL
+            if(funct3 == OPIVI)
+                return get_funct<dest_elem_t>(0b010100, funct3);
+            else
+                return [](dest_elem_t vd, src2_elem_t vs2, src1_elem_t vs1) {
+                    constexpr dest_elem_t bits = sizeof(src2_elem_t) * 8;
+                    auto shamt = vs1 & shift_mask<src1_elem_t>();
+                    return (vs2 << shamt) | (vs2 >> (bits - shamt));
+                };
+        }
         case 0b100101: // VSLL
             return [](dest_elem_t vd, src2_elem_t vs2, src1_elem_t vs1) { return vs2 << (vs1 & shift_mask<src2_elem_t>()); };
         case 0b101000: // VSRL
@@ -219,6 +250,10 @@ std::function<dest_elem_t(dest_elem_t, src2_elem_t, src1_elem_t)> get_funct(unsi
             return [](dest_elem_t vd, src2_elem_t vs2, src1_elem_t vs1) {
                 return static_cast<std::make_signed_t<src2_elem_t>>(vs2) >> (vs1 & shift_mask<src2_elem_t>());
             };
+        case 0b110101: // VWSLL
+            return [](dest_elem_t vd, src2_elem_t vs2, src1_elem_t vs1) {
+                return static_cast<dest_elem_t>(vs2) << (vs1 & (shift_mask<dest_elem_t>()));
+            };
         default:
             throw new std::runtime_error("Unknown funct6 in get_funct");
         }
@@ -328,6 +363,24 @@ std::function<dest_elem_t(dest_elem_t, src2_elem_t, src1_elem_t)> get_funct(unsi
             return [](dest_elem_t vd, src2_elem_t vs2, src1_elem_t vs1) {
                 return sext<dest_elem_t>(vs1) * static_cast<dest_elem_t>(vs2) + vd;
             };
+        case 0b001100: // VCLMUL
+            return [](dest_elem_t vd, src2_elem_t vs2, src1_elem_t vs1) {
+                dest_elem_t output = 0;
+                for(size_t i = 0; i <= sizeof(dest_elem_t) * 8 - 1; i++) {
+                    if((vs2 >> i) & 1)
+                        output = output ^ (vs1 << i);
+                }
+                return output;
+            };
+        case 0b001101: // VCLMULH
+            return [](dest_elem_t vd, src2_elem_t vs2, src1_elem_t vs1) {
+                dest_elem_t output = 0;
+                for(size_t i = 1; i < sizeof(dest_elem_t) * 8; i++) {
+                    if((vs2 >> i) & 1)
+                        output = output ^ (vs1 >> (sizeof(dest_elem_t) * 8 - i));
+                }
+                return output;
+            };
         default:
             throw new std::runtime_error("Unknown funct6 in get_funct");
         }
@@ -364,11 +417,10 @@ void vector_imm_op(uint8_t* V, unsigned funct6, unsigned funct3, uint64_t vl, ui
     auto fn = get_funct<dest_elem_t, src2_elem_t, src1_elem_t>(funct6, funct3);
     for(size_t idx = vstart; idx < vl; idx++) {
         bool mask_active = vm ? 1 : mask_reg[idx];
-        if(mask_active) {
+        if(mask_active)
             vd_view[idx] = fn(vd_view[idx], vs2_view[idx], imm);
-        } else {
-            vd_view[idx] = vtype.vma() ? vd_view[idx] : vd_view[idx];
-        }
+        else if(vtype.vma())
+            vd_view[idx] = agnostic_behavior(vd_view[idx]);
     }
     if(vtype.vta())
         for(size_t idx = vl; idx < vlmax; idx++)
@@ -535,6 +587,79 @@ std::function<dest_elem_t(src2_elem_t)> get_unary_fn(unsigned unary_op) {
     case 0b00100: // VZEXT.VF4
     case 0b00010: // VZEXT.VF8
         return [](src2_elem_t vs2) { return vs2; };
+    case 0b01000: // VBREV8
+        return [](src2_elem_t vs2) { return brev8<dest_elem_t>(vs2); };
+    case 0b01001: // VREV8
+        return [](src2_elem_t vs2) {
+            constexpr unsigned byte_count = sizeof(src2_elem_t);
+            dest_elem_t result = 0;
+            for(size_t i = 0; i < byte_count; ++i) {
+                result <<= 8;
+                result |= (vs2 & 0xFF);
+                vs2 >>= 8;
+            }
+            return result;
+        };
+    case 0b01010: // VBREV
+        return [](src2_elem_t vs2) { return brev<dest_elem_t>(vs2); };
+    case 0b01100: // VCLZ
+        return [](src2_elem_t vs2) {
+            if(std::is_same_v<src2_elem_t, unsigned int>)
+                return static_cast<dest_elem_t>(__builtin_clz(vs2));
+            else if(std::is_same_v<src2_elem_t, unsigned long>)
+                return static_cast<dest_elem_t>(__builtin_clzl(vs2));
+            else if(std::is_same_v<src2_elem_t, unsigned long long>)
+                return static_cast<dest_elem_t>(__builtin_clzll(vs2));
+            else {
+                constexpr dest_elem_t bits = sizeof(src2_elem_t) * 8;
+                if(vs2 == 0)
+                    return bits;
+                dest_elem_t count = 0;
+                for(size_t i = bits - 1; i >= 0; --i) {
+                    if((vs2 >> i) & 1)
+                        break;
+                    ++count;
+                }
+                return count;
+            }
+        };
+    case 0b01101: // VCTZ
+        return [](src2_elem_t vs2) {
+            if(std::is_same_v<src2_elem_t, unsigned int>)
+                return static_cast<dest_elem_t>(__builtin_ctz(vs2));
+            else if(std::is_same_v<src2_elem_t, unsigned long>)
+                return static_cast<dest_elem_t>(__builtin_ctzl(vs2));
+            else if(std::is_same_v<src2_elem_t, unsigned long long>)
+                return static_cast<dest_elem_t>(__builtin_ctzll(vs2));
+            else {
+                constexpr dest_elem_t bits = sizeof(src2_elem_t) * 8;
+                if(vs2 == 0)
+                    return bits;
+                dest_elem_t count = 0;
+                while((vs2 & 1) == 0) {
+                    ++count;
+                    vs2 >>= 1;
+                }
+                return count;
+            }
+        };
+    case 0b01110: // VCPOP
+        return [](src2_elem_t vs2) {
+            if(std::is_same_v<src2_elem_t, unsigned int>)
+                return static_cast<dest_elem_t>(__builtin_popcount(vs2));
+            else if(std::is_same_v<src2_elem_t, unsigned long>)
+                return static_cast<dest_elem_t>(__builtin_popcountl(vs2));
+            else if(std::is_same_v<src2_elem_t, unsigned long long>)
+                return static_cast<dest_elem_t>(__builtin_popcountll(vs2));
+            else {
+                dest_elem_t count = 0;
+                while(vs2) {
+                    count += vs2 & 1;
+                    vs2 >>= 1;
+                }
+                return count;
+            }
+        };
     default:
         throw new std::runtime_error("Unknown funct in get_unary_fn");
     }
@@ -550,8 +675,8 @@ void vector_unary_op(uint8_t* V, unsigned unary_op, uint64_t vl, uint64_t vstart
         bool mask_active = vm ? 1 : mask_reg[idx];
         if(mask_active)
             vd_view[idx] = fn(vs2_view[idx]);
-        else
-            vd_view[idx] = vtype.vma() ? vd_view[idx] : vd_view[idx];
+        else if(vtype.vma())
+            vd_view[idx] = agnostic_behavior(vd_view[idx]);
     }
     if(vtype.vta())
         for(size_t idx = vl; idx < vlmax; idx++)
@@ -774,11 +899,10 @@ bool sat_vector_vector_op(uint8_t* V, unsigned funct6, unsigned funct3, uint64_t
     auto fn = get_sat_funct<dest_elem_t, src2_elem_t, src1_elem_t>(funct6, funct3);
     for(size_t idx = vstart; idx < vl; idx++) {
         bool mask_active = vm ? 1 : mask_reg[idx];
-        if(mask_active) {
+        if(mask_active)
             saturated |= fn(vxrm, vtype, vd_view[idx], vs2_view[idx], vs1_view[idx]);
-        } else {
-            vd_view[idx] = vtype.vma() ? vd_view[idx] : vd_view[idx];
-        }
+        else if(vtype.vma())
+            vd_view[idx] = agnostic_behavior(vd_view[idx]);
     }
     if(vtype.vta())
         for(size_t idx = vl; idx < vlmax; idx++) {
@@ -797,11 +921,10 @@ bool sat_vector_imm_op(uint8_t* V, unsigned funct6, unsigned funct3, uint64_t vl
     auto fn = get_sat_funct<dest_elem_t, src2_elem_t, src1_elem_t>(funct6, funct3);
     for(size_t idx = vstart; idx < vl; idx++) {
         bool mask_active = vm ? 1 : mask_reg[idx];
-        if(mask_active) {
+        if(mask_active)
             saturated |= fn(vxrm, vtype, vd_view[idx], vs2_view[idx], imm);
-        } else {
-            vd_view[idx] = vtype.vma() ? vd_view[idx] : vd_view[idx];
-        }
+        else if(vtype.vma())
+            vd_view[idx] = agnostic_behavior(vd_view[idx]);
     }
     if(vtype.vta())
         for(size_t idx = vl; idx < vlmax; idx++) {
@@ -1939,4 +2062,221 @@ template <unsigned VLEN> void vector_whole_move(uint8_t* V, unsigned vd, unsigne
     memcpy(vd_view.start, vs2_view.start, VLEN / 8 * count);
 }
 
+template <unsigned VLEN, unsigned EGS>
+void vector_vector_crypto(uint8_t* V, unsigned funct6, uint64_t eg_len, uint64_t eg_start, vtype_t vtype, unsigned vd, unsigned vs2,
+                          unsigned vs1) {
+    uint64_t vlmax = VLEN * vtype.lmul() / (vtype.sew() * EGS);
+    auto vs1_view = get_vreg<VLEN, uint128_t>(V, vs1, vlmax);
+    auto vs2_view = get_vreg<VLEN, uint128_t>(V, vs2, vlmax);
+    auto vd_view = get_vreg<VLEN, uint128_t>(V, vd, vlmax);
+    auto fn = get_crypto_funct(funct6, vs1);
+    for(size_t idx = eg_start; idx < eg_len; idx++) {
+        vd_view[idx] = fn(vd_view[idx], vs2_view[idx], vs1_view[idx]);
+    }
+    if(vtype.vta())
+        for(size_t idx = eg_len; idx < vlmax; idx++)
+            vd_view[idx] = agnostic_behavior(vd_view[idx]);
+}
+template <unsigned VLEN, unsigned EGS>
+void vector_scalar_crypto(uint8_t* V, unsigned funct6, uint64_t eg_len, uint64_t eg_start, vtype_t vtype, unsigned vd, unsigned vs2,
+                          unsigned vs1) {
+    uint64_t vlmax = VLEN * vtype.lmul() / (vtype.sew() * EGS);
+    auto vs2_val = get_vreg<VLEN, uint128_t>(V, vs2, vlmax)[0];
+    auto vd_view = get_vreg<VLEN, uint128_t>(V, vd, vlmax);
+    auto fn = get_crypto_funct(funct6, vs1);
+    for(size_t idx = eg_start; idx < eg_len; idx++) {
+        vd_view[idx] = fn(vd_view[idx], vs2_val, -1);
+    }
+    if(vtype.vta())
+        for(size_t idx = eg_len; idx < vlmax; idx++)
+            vd_view[idx] = agnostic_behavior(vd_view[idx]);
+}
+
+template <unsigned VLEN, unsigned EGS>
+void vector_imm_crypto(uint8_t* V, unsigned funct6, uint64_t eg_len, uint64_t eg_start, vtype_t vtype, unsigned vd, unsigned vs2,
+                       uint8_t imm) {
+    uint64_t vlmax = VLEN * vtype.lmul() / (vtype.sew() * EGS);
+    auto vs2_view = get_vreg<VLEN, uint128_t>(V, vs2, vlmax);
+    auto vd_view = get_vreg<VLEN, uint128_t>(V, vd, vlmax);
+    auto fn = get_crypto_funct(funct6, -1);
+    for(size_t idx = eg_start; idx < eg_len; idx++) {
+        vd_view[idx] = fn(vd_view[idx], vs2_view[idx], imm);
+    }
+    if(vtype.vta())
+        for(size_t idx = eg_len; idx < vlmax; idx++)
+            vd_view[idx] = agnostic_behavior(vd_view[idx]);
+}
+template <typename T> T rotr(T x, unsigned n) {
+    assert(n < sizeof(T) * 8);
+    return (x >> n) | (x << (sizeof(T) * 8 - n));
+}
+template <typename T> T shr(T x, unsigned n) {
+    assert(n < sizeof(T) * 8);
+    return (x >> n);
+}
+template <typename T> T sum0(T);
+template <> inline uint32_t sum0(uint32_t x) { return rotr(x, 2) ^ rotr(x, 13) ^ rotr(x, 22); }
+template <> inline uint64_t sum0(uint64_t x) { return rotr(x, 28) ^ rotr(x, 34) ^ rotr(x, 39); }
+template <typename T> T sum1(T);
+template <> inline uint32_t sum1(uint32_t x) { return rotr(x, 6) ^ rotr(x, 11) ^ rotr(x, 25); }
+template <> inline uint64_t sum1(uint64_t x) { return rotr(x, 14) ^ rotr(x, 18) ^ rotr(x, 41); }
+template <typename T> T ch(T x, T y, T z) { return ((x & y) ^ ((~x) & z)); }
+template <typename T> T maj(T x, T y, T z) { return ((x & y) ^ (x & z) ^ (y & z)); }
+template <typename T> T sig0(T);
+template <> inline uint32_t sig0(uint32_t x) { return rotr(x, 7) ^ rotr(x, 18) ^ shr(x, 3); }
+template <> inline uint64_t sig0(uint64_t x) { return rotr(x, 1) ^ rotr(x, 8) ^ shr(x, 7); }
+template <typename T> T sig1(T);
+template <> inline uint32_t sig1(uint32_t x) { return rotr(x, 17) ^ rotr(x, 19) ^ shr(x, 10); }
+template <> inline uint64_t sig1(uint64_t x) { return rotr(x, 19) ^ rotr(x, 61) ^ shr(x, 6); }
+template <typename T> std::function<void(vreg_view<T>&, vreg_view<T>&, vreg_view<T>&)> get_crypto_funct(unsigned int funct6) {
+    switch(funct6) {
+    case 0b101110: // VSHA2CH
+        return [](vreg_view<T>& vd_view, vreg_view<T>& vs2_view, vreg_view<T>& vs1_view) {
+            T a = vs2_view[3];
+            T b = vs2_view[2];
+            T c = vd_view[3];
+            T d = vd_view[2];
+            T e = vs2_view[1];
+            T f = vs2_view[0];
+            T g = vd_view[1];
+            T h = vd_view[0];
+            T W0 = vs1_view[2];
+            T W1 = vs1_view[3];
+            T T1 = h + sum1(e) + ch(e, f, g) + W0;
+            T T2 = sum0(a) + maj(a, b, c);
+            h = g;
+            g = f;
+            f = e;
+            e = d + T1;
+            d = c;
+            c = b;
+            b = a;
+            a = T1 + T2;
+            T1 = h + sum1(e) + ch(e, f, g) + W1;
+            T2 = sum0(a) + maj(a, b, c);
+            h = g;
+            g = f;
+            f = e;
+            e = d + T1;
+            d = c;
+            c = b;
+            b = a;
+            a = T1 + T2;
+            vd_view[0] = f;
+            vd_view[1] = e;
+            vd_view[2] = b;
+            vd_view[3] = a;
+        };
+    case 0b101111: // VSHA2CL
+        return [](vreg_view<T>& vd_view, vreg_view<T>& vs2_view, vreg_view<T>& vs1_view) {
+            T a = vs2_view[3];
+            T b = vs2_view[2];
+            T c = vd_view[3];
+            T d = vd_view[2];
+            T e = vs2_view[1];
+            T f = vs2_view[0];
+            T g = vd_view[1];
+            T h = vd_view[0];
+            T W0 = vs1_view[0];
+            T W1 = vs1_view[1];
+            T T1 = h + sum1(e) + ch(e, f, g) + W0;
+            T T2 = sum0(a) + maj(a, b, c);
+            h = g;
+            g = f;
+            f = e;
+            e = d + T1;
+            d = c;
+            c = b;
+            b = a;
+            a = T1 + T2;
+            T1 = h + sum1(e) + ch(e, f, g) + W1;
+            T2 = sum0(a) + maj(a, b, c);
+            h = g;
+            g = f;
+            f = e;
+            e = d + T1;
+            d = c;
+            c = b;
+            b = a;
+            a = T1 + T2;
+            vd_view[0] = f;
+            vd_view[1] = e;
+            vd_view[2] = b;
+            vd_view[3] = a;
+        };
+    case 0b101101: // VSHA2MS
+        return [](vreg_view<T>& vd_view, vreg_view<T>& vs2_view, vreg_view<T>& vs1_view) {
+            T W0 = vd_view[0];
+            T W1 = vd_view[1];
+            T W2 = vd_view[2];
+            T W3 = vd_view[3];
+
+            T W4 = vs2_view[0];
+            T W9 = vs2_view[1];
+            T W10 = vs2_view[2];
+            T W11 = vs2_view[3];
+
+            T W12 = vs1_view[0];
+            T W13 = vs1_view[1];
+            T W14 = vs1_view[2];
+            T W15 = vs1_view[3];
+
+            T W16 = sig1(W14) + W9 + sig0(W1) + W0;
+            T W17 = sig1(W15) + W10 + sig0(W2) + W1;
+            T W18 = sig1(W16) + W11 + sig0(W3) + W2;
+            T W19 = sig1(W17) + W12 + sig0(W4) + W3;
+
+            vd_view[0] = W16;
+            vd_view[1] = W17;
+            vd_view[2] = W18;
+            vd_view[3] = W19;
+        };
+    default:
+        throw new std::runtime_error("Unsupported operation in get_crypto_funct");
+    }
+}
+template <unsigned VLEN, unsigned EGS, typename elem_type_t>
+void vector_crypto(uint8_t* V, unsigned funct6, uint64_t eg_len, uint64_t eg_start, vtype_t vtype, unsigned vd, unsigned vs2,
+                   unsigned vs1) {
+    auto fn = get_crypto_funct<elem_type_t>(funct6);
+    auto vd_view = get_vreg<VLEN, elem_type_t>(V, vd, EGS);
+    auto vs2_view = get_vreg<VLEN, elem_type_t>(V, vs2, EGS);
+    auto vs1_view = get_vreg<VLEN, elem_type_t>(V, vs1, EGS);
+    for(size_t idx = eg_start; idx < eg_len; idx++) {
+        fn(vd_view, vs2_view, vs1_view);
+        // We cannot use views in case EGW < VLEN, as views can only address the start of a register
+        vd_view.start += EGS * sizeof(elem_type_t);
+        vs2_view.start += EGS * sizeof(elem_type_t);
+        vs1_view.start += EGS * sizeof(elem_type_t);
+    }
+    if(vtype.vta()) {
+        uint64_t vlmax = VLEN * vtype.lmul() / (vtype.sew());
+        auto vd_view = get_vreg<VLEN, elem_type_t>(V, vd, vlmax);
+        for(size_t idx = eg_len * EGS; idx < vlmax; idx++)
+            vd_view[idx] = agnostic_behavior(vd_view[idx]);
+    }
+}
+
+template <typename dest_elem_t, typename src_elem_t> dest_elem_t brev(src_elem_t vs2) {
+    constexpr dest_elem_t bits = sizeof(src_elem_t) * 8;
+    dest_elem_t result = 0;
+    for(size_t i = 0; i < bits; ++i) {
+        result <<= 1;
+        result |= (vs2 & 1);
+        vs2 >>= 1;
+    }
+    return result;
+};
+template <typename dest_elem_t, typename src_elem_t> dest_elem_t brev8(src_elem_t vs2) {
+    constexpr unsigned byte_count = sizeof(src_elem_t);
+    dest_elem_t result = 0;
+    for(size_t i = 0; i < byte_count; ++i) {
+        dest_elem_t byte = (vs2 >> (i * 8)) & 0xFF;
+        byte = ((byte & 0xF0) >> 4) | ((byte & 0x0F) << 4);
+        byte = ((byte & 0xCC) >> 2) | ((byte & 0x33) << 2);
+        byte = ((byte & 0xAA) >> 1) | ((byte & 0x55) << 1);
+        result |= byte << (i * 8);
+    }
+    return result;
+};
 } // namespace softvector