adds integer extension and add/substract with carry vector instructions

gen_input/templates/interp/CORENAME.cpp.gtl

@@ -188,38 +188,30 @@ if(vector != null) {%>
     uint64_t vsxseg(uint8_t* V, uint8_t vs3, uint8_t vs2, uint64_t rs1_val, uint64_t vl, uint64_t vstart, softvector::vtype_t vtype, bool vm, uint8_t elem_byte_size, uint16_t elem_count, uint8_t segment_size, bool ordered){
         return softvector::vector_load_store_index(this->get_arch(), softvector::softvec_write, V, traits::VLEN, traits::XLEN, vs3, vs2, rs1_val, vl, vstart, vtype, vm, elem_byte_size, elem_count, segment_size, ordered);
     }
-    void vector_vector_op(uint8_t* V, uint8_t funct, 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){
+    void vector_vector_op(uint8_t* V, uint8_t funct, 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, int8_t carry = 0){
         switch(sew_val){
             case 0b000:
-                softvector::vector_vector_op<${vlen}, uint8_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, vs1);
+                return softvector::vector_vector_op<${vlen}, uint8_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, vs1, static_cast<softvector::carry_t>(carry));
-                break;
             case 0b001:
-                softvector::vector_vector_op<${vlen}, uint16_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, vs1);
+                return softvector::vector_vector_op<${vlen}, uint16_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, vs1, static_cast<softvector::carry_t>(carry));
-                break;
             case 0b010:
-                softvector::vector_vector_op<${vlen}, uint32_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, vs1);
+                return softvector::vector_vector_op<${vlen}, uint32_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, vs1, static_cast<softvector::carry_t>(carry));
-                break;
             case 0b011:
-                softvector::vector_vector_op<${vlen}, uint64_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, vs1);
+                return softvector::vector_vector_op<${vlen}, uint64_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, vs1, static_cast<softvector::carry_t>(carry));
-                break;
             default:
                 throw new std::runtime_error("Unsupported sew bit value");
         }
     }
-    void vector_imm_op(uint8_t* V, uint8_t funct, 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){
+    void vector_imm_op(uint8_t* V, uint8_t funct, 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, int8_t carry = 0){
         switch(sew_val){
             case 0b000:
-                softvector::vector_imm_op<${vlen}, uint8_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, imm);
+                return softvector::vector_imm_op<${vlen}, uint8_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, imm, static_cast<softvector::carry_t>(carry));
-                break;
             case 0b001:
-                softvector::vector_imm_op<${vlen}, uint16_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, imm);
+                return softvector::vector_imm_op<${vlen}, uint16_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, imm, static_cast<softvector::carry_t>(carry));
-                break;
             case 0b010:
-                softvector::vector_imm_op<${vlen}, uint32_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, imm);
+                return softvector::vector_imm_op<${vlen}, uint32_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, imm, static_cast<softvector::carry_t>(carry));
-                break;
             case 0b011:
-                softvector::vector_imm_op<${vlen}, uint64_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, imm);
+                return softvector::vector_imm_op<${vlen}, uint64_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, imm, static_cast<softvector::carry_t>(carry));
-                break;
             default:
                 throw new std::runtime_error("Unsupported sew bit value");
         }
@@ -227,14 +219,11 @@ if(vector != null) {%>
     void vector_vector_wv(uint8_t* V, uint8_t funct, 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:
-                softvector::vector_vector_op<${vlen}, uint16_t, uint8_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, vs1 );
+                return softvector::vector_vector_op<${vlen}, uint16_t, uint8_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, vs1 );
-                break;
             case 0b001:
-                softvector::vector_vector_op<${vlen}, uint32_t, uint16_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, vs1 );
+                return softvector::vector_vector_op<${vlen}, uint32_t, uint16_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, vs1 );
-                break;
             case 0b010:
-                softvector::vector_vector_op<${vlen}, uint64_t, uint32_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, vs1 );
+                return softvector::vector_vector_op<${vlen}, uint64_t, uint32_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, vs1 );
-                break;
             case 0b011: // would widen to 128 bits
             default:
                 throw new std::runtime_error("Unsupported sew bit value");
@@ -243,14 +232,11 @@ if(vector != null) {%>
     void vector_imm_wv(uint8_t* V, uint8_t funct, 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:
-                softvector::vector_imm_op<${vlen}, uint16_t, uint8_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, imm);
+                return softvector::vector_imm_op<${vlen}, uint16_t, uint8_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, imm);
-                break;
             case 0b001:
-                softvector::vector_imm_op<${vlen}, uint32_t, uint16_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, imm);
+                return softvector::vector_imm_op<${vlen}, uint32_t, uint16_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, imm);
-                break;
             case 0b010:
-                softvector::vector_imm_op<${vlen}, uint64_t, uint32_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, imm);
+                return softvector::vector_imm_op<${vlen}, uint64_t, uint32_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, imm);
-                break;
             case 0b011: // would widen to 128 bits
             default:
                 throw new std::runtime_error("Unsupported sew bit value");
@@ -259,14 +245,11 @@ if(vector != null) {%>
     void vector_vector_ww(uint8_t* V, uint8_t funct, 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:
-                softvector::vector_vector_op<${vlen}, uint16_t, uint16_t, uint8_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, vs1);
+                return softvector::vector_vector_op<${vlen}, uint16_t, uint16_t, uint8_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, vs1);
-                break;
             case 0b001:
-                softvector::vector_vector_op<${vlen}, uint32_t, uint32_t, uint16_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, vs1);
+                return softvector::vector_vector_op<${vlen}, uint32_t, uint32_t, uint16_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, vs1);
-                break;
             case 0b010:
-                softvector::vector_vector_op<${vlen}, uint64_t, uint64_t, uint32_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, vs1);
+                return softvector::vector_vector_op<${vlen}, uint64_t, uint64_t, uint32_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, vs1);
-                break;
             case 0b011: // would widen to 128 bits
             default:
                 throw new std::runtime_error("Unsupported sew bit value");
@@ -275,19 +258,80 @@ if(vector != null) {%>
     void vector_imm_ww(uint8_t* V, uint8_t funct, 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:
-                softvector::vector_imm_op<${vlen}, uint16_t, uint16_t, uint8_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, imm);
+                return softvector::vector_imm_op<${vlen}, uint16_t, uint16_t, uint8_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, imm);
-                break;
             case 0b001:
-                softvector::vector_imm_op<${vlen}, uint32_t, uint32_t, uint16_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, imm);
+                return softvector::vector_imm_op<${vlen}, uint32_t, uint32_t, uint16_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, imm);
-                break;
             case 0b010:
-                softvector::vector_imm_op<${vlen}, uint64_t, uint64_t, uint32_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, imm);
+                return softvector::vector_imm_op<${vlen}, uint64_t, uint64_t, uint32_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, imm);
-                break;
             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_m(uint8_t* V, uint8_t funct, 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){
+        vector_vector_op(V, funct, vl, vstart, vtype, 0, vd, vs2, vs1, sew_val, carry);
+    }
+    void vector_imm_m(uint8_t* V, uint8_t funct, 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){
+        vector_imm_op(V, funct, vl, vstart, vtype, 0, vd, vs2, imm, sew_val, carry);
+    }
+    void mask_vector_vector_op(uint8_t* V, unsigned funct, 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<${vlen}, uint8_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b001:
+                return softvector::mask_vector_vector_op<${vlen}, uint16_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b010:
+                return softvector::mask_vector_vector_op<${vlen}, uint32_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, vs1);
+            case 0b011:
+                return softvector::mask_vector_vector_op<${vlen}, uint64_t>(V, funct, 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 funct, 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<${vlen}, uint8_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b001:
+                return softvector::mask_vector_imm_op<${vlen}, uint16_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b010:
+                return softvector::mask_vector_imm_op<${vlen}, uint32_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, imm);
+            case 0b011:
+                return softvector::mask_vector_imm_op<${vlen}, uint64_t>(V, funct, vl, vstart, vtype, vm, vd, vs2, imm);
+            default:
+                throw new std::runtime_error("Unsupported sew bit value");
+        }
+    }
+
 <%}%>
     uint64_t fetch_count{0};
     uint64_t tval{0};

src/vm/vector_functions.h

@@ -56,8 +56,9 @@ struct vtype_t {
 struct vmask_view {
     uint8_t* start;
     size_t elem_count;
-    bool operator[](size_t);
+    bool operator[](size_t) const;
 };
+enum class carry_t { NO_CARRY = 0, ADD_CARRY = 1, SUB_CARRY = 2 };
 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);
 
@@ -73,10 +74,18 @@ uint64_t vector_load_store_index(void* core, std::function<bool(void*, uint64_t,
                                  uint8_t segment_size, bool ordered);
 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, uint64_t vl, uint64_t vstart, vtype_t vtype, bool vm, unsigned vd, unsigned vs2,
-                      unsigned vs1);
+                      unsigned vs1, carry_t carry = carry_t::NO_CARRY);
 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, uint64_t vl, uint64_t vstart, vtype_t vtype, bool vm, unsigned vd, unsigned vs2,
-                   typename std::make_signed<src1_elem_t>::type imm);
+                   typename std::make_signed<src1_elem_t>::type imm, carry_t carry = carry_t::NO_CARRY);
+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, 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, uint64_t vl, uint64_t vstart, vtype_t vtype, bool vm, unsigned vd, unsigned vs2,
+                        typename std::make_signed<elem_t>::type imm);
 } // namespace softvector
 #include "vm/vector_functions.hpp"
 #endif /* _VM_VECTOR_FUNCTIONS_H_ */

src/vm/vector_functions.hpp

@@ -34,6 +34,7 @@
 #pragma once
 #include "vm/vector_functions.h"
 #include <functional>
+#include <limits>
 #include <stdexcept>
 #include <type_traits>
 #ifndef _VM_VECTOR_FUNCTIONS_H_
@@ -68,6 +69,7 @@ template <typename dest_elem_t, typename src2_elem_t = dest_elem_t, typename src
 std::function<dest_elem_t(src2_elem_t, src1_elem_t)> get_funct(unsigned funct) {
     switch(funct) {
     case 0b000000: // VADD
+    case 0b010000: // VADC
     case 0b110000: // VWADDU
     case 0b110100: // VWADDU.W
         return [](src2_elem_t vs2, src1_elem_t vs1) { return vs2 + vs1; };
@@ -89,6 +91,7 @@ std::function<dest_elem_t(src2_elem_t, src1_elem_t)> get_funct(unsigned funct) {
             return static_cast<typename std::make_signed_t<dest_elem_t>>(static_cast<typename std::make_signed_t<src2_elem_t>>(vs2) +
                                                                          static_cast<typename std::make_signed_t<src1_elem_t>>(vs1));
         };
+    case 0b010010: // VSBC
     case 0b110011: // VWSUB
     case 0b110111: // VWSUB.W
         return [](src2_elem_t vs2, src1_elem_t vs1) {
@@ -101,7 +104,7 @@ std::function<dest_elem_t(src2_elem_t, src1_elem_t)> get_funct(unsigned funct) {
 }
 template <unsigned VLEN, typename dest_elem_t, typename src2_elem_t, typename src1_elem_t>
 void 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) {
+                      unsigned vs1, carry_t carry) {
     uint64_t elem_count = VLEN * vtype.lmul() / vtype.sew();
     vmask_view mask_reg = read_vmask<VLEN>(V, elem_count);
     auto vs1_view = get_vreg<VLEN, src1_elem_t>(V, vs1, elem_count);
@@ -110,12 +113,22 @@ void vector_vector_op(uint8_t* V, unsigned funct6, uint64_t vl, uint64_t vstart,
     auto fn = get_funct<dest_elem_t, src2_elem_t, src1_elem_t>(funct6);
     // elements w/ index smaller than vstart are in the prestart and get skipped
     // body is from vstart to min(elem_count, vl)
-    for(unsigned idx = vstart; idx < std::min(elem_count, vl); idx++) {
+    if(carry == carry_t::NO_CARRY) {
-        bool mask_active = vm ? 1 : mask_reg[idx];
+        for(unsigned idx = vstart; idx < std::min(elem_count, vl); idx++) {
-        if(mask_active) {
+            bool mask_active = vm ? 1 : mask_reg[idx];
-            vd_view[idx] = fn(vs2_view[idx], vs1_view[idx]);
+            if(mask_active) {
-        } else {
+                vd_view[idx] = fn(vs2_view[idx], vs1_view[idx]);
-            vd_view[idx] = vtype.vma() ? vd_view[idx] : vd_view[idx];
+            } else {
+                vd_view[idx] = vtype.vma() ? vd_view[idx] : vd_view[idx];
+            }
+        }
+    } else if(carry == carry_t::SUB_CARRY) {
+        for(unsigned idx = vstart; idx < std::min(elem_count, vl); idx++) {
+            vd_view[idx] = fn(vs2_view[idx], vs1_view[idx]) - mask_reg[idx];
+        }
+    } else {
+        for(unsigned idx = vstart; idx < std::min(elem_count, vl); idx++) {
+            vd_view[idx] = fn(vs2_view[idx], vs1_view[idx]) + mask_reg[idx];
         }
     }
     // elements w/ index larger than elem_count are in the tail (fractional LMUL)
@@ -128,7 +141,7 @@ void vector_vector_op(uint8_t* V, unsigned funct6, uint64_t vl, uint64_t vstart,
 }
 template <unsigned VLEN, typename dest_elem_t, typename src2_elem_t, typename src1_elem_t>
 void vector_imm_op(uint8_t* V, unsigned funct6, uint64_t vl, uint64_t vstart, vtype_t vtype, bool vm, unsigned vd, unsigned vs2,
-                   typename std::make_signed<src1_elem_t>::type imm) {
+                   typename std::make_signed<src1_elem_t>::type imm, carry_t carry) {
     uint64_t elem_count = VLEN * vtype.lmul() / vtype.sew();
     vmask_view mask_reg = read_vmask<VLEN>(V, elem_count);
     auto vs2_view = get_vreg<VLEN, src2_elem_t>(V, vs2, elem_count);
@@ -136,10 +149,127 @@ void vector_imm_op(uint8_t* V, unsigned funct6, uint64_t vl, uint64_t vstart, vt
     auto fn = get_funct<dest_elem_t, src2_elem_t, src1_elem_t>(funct6);
     // elements w/ index smaller than vstart are in the prestart and get skipped
     // body is from vstart to min(elem_count, vl)
+    if(carry == carry_t::NO_CARRY) {
+        for(unsigned idx = vstart; idx < std::min(elem_count, vl); idx++) {
+            bool mask_active = vm ? 1 : mask_reg[idx];
+            if(mask_active) {
+                vd_view[idx] = fn(vs2_view[idx], imm);
+            } else {
+                vd_view[idx] = vtype.vma() ? vd_view[idx] : vd_view[idx];
+            }
+        }
+    } else if(carry == carry_t::SUB_CARRY) {
+        for(unsigned idx = vstart; idx < std::min(elem_count, vl); idx++) {
+            auto val1 = fn(vs2_view[idx], imm);
+            auto val2 = static_cast<std::make_signed_t<dest_elem_t>>(mask_reg[idx]);
+            auto diff = val1 - val2;
+            vd_view[idx] = fn(vs2_view[idx], imm) - mask_reg[idx];
+        }
+    } else {
+        for(unsigned idx = vstart; idx < std::min(elem_count, vl); idx++) {
+            vd_view[idx] = fn(vs2_view[idx], imm) + mask_reg[idx];
+        }
+    }
+    // elements w/ index larger than elem_count are in the tail (fractional LMUL)
+    // elements w/ index larger than vl are in the tail
+    unsigned maximum_elems = VLEN * vtype.lmul() / (sizeof(dest_elem_t) * 8);
+    for(unsigned idx = std::min(elem_count, vl); idx < maximum_elems; idx++) {
+        vd_view[idx] = vtype.vta() ? vd_view[idx] : vd_view[idx];
+    }
+    return;
+}
+template <typename elem_t> std::function<bool(elem_t, elem_t, elem_t)> get_mask_funct(unsigned funct) {
+    switch(funct) {
+    case 0b010001: // VMADC
+        return [](elem_t vs2, elem_t vs1, elem_t carry) {
+            return static_cast<elem_t>(vs2 + vs1 + carry) < std::max(vs1, vs2) || static_cast<elem_t>(vs2 + vs1) < std::max(vs1, vs2);
+        };
+    case 0b010011: // VMSBC
+        return [](elem_t vs2, elem_t vs1, elem_t carry) {
+            return vs2 < static_cast<elem_t>(vs1 + carry) || (vs1 == std::numeric_limits<elem_t>::max() && carry);
+        };
+    default:
+        throw new std::runtime_error("Uknown funct in get_mask_funct");
+    }
+}
+template <unsigned VLEN, typename elem_t>
+void mask_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) {
+    uint64_t elem_count = VLEN * vtype.lmul() / vtype.sew();
+    vmask_view mask_reg = read_vmask<VLEN>(V, elem_count);
+    auto vs1_view = get_vreg<VLEN, elem_t>(V, vs1, elem_count);
+    auto vs2_view = get_vreg<VLEN, elem_t>(V, vs2, elem_count);
+    vmask_view vd_mask_view = read_vmask<VLEN>(V, elem_count, vd);
+    auto fn = get_mask_funct<elem_t>(funct);
+    // elements w/ index smaller than vstart are in the prestart and get skipped
+    // body is from vstart to min(elem_count, vl)
+    for(unsigned idx = vstart; idx < std::min(elem_count, vl); idx++) {
+        elem_t carry = vm ? 0 : mask_reg[idx];
+        bool new_bit_value = fn(vs2_view[idx], vs1_view[idx], carry);
+        uint8_t* cur_mask_byte_addr = vd_mask_view.start + idx / 8;
+        unsigned cur_bit = idx % 8;
+        *cur_mask_byte_addr = *cur_mask_byte_addr & ~(1U << cur_bit) | static_cast<unsigned>(new_bit_value) << cur_bit;
+    }
+    // elements w/ index larger than elem_count are in the tail (fractional LMUL)
+    // elements w/ index larger than vl are in the tail
+    for(unsigned idx = std::min(elem_count, vl); idx < VLEN; idx++) {
+        // always tail agnostic
+    }
+    return;
+}
+template <unsigned VLEN, typename elem_t>
+void mask_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) {
+    uint64_t elem_count = VLEN * vtype.lmul() / vtype.sew();
+    vmask_view mask_reg = read_vmask<VLEN>(V, elem_count);
+    auto vs2_view = get_vreg<VLEN, elem_t>(V, vs2, elem_count);
+    vmask_view vd_mask_view = read_vmask<VLEN>(V, elem_count, vd);
+    auto fn = get_mask_funct<elem_t>(funct);
+    // elements w/ index smaller than vstart are in the prestart and get skipped
+    // body is from vstart to min(elem_count, vl)
+    for(unsigned idx = vstart; idx < std::min(elem_count, vl); idx++) {
+        elem_t carry = vm ? 0 : mask_reg[idx];
+        bool new_bit_value = fn(vs2_view[idx], imm, carry);
+        uint8_t* cur_mask_byte_addr = vd_mask_view.start + idx / 8;
+        unsigned cur_bit = idx % 8;
+        *cur_mask_byte_addr = *cur_mask_byte_addr & ~(1U << cur_bit) | static_cast<unsigned>(new_bit_value) << cur_bit;
+    }
+    // elements w/ index larger than elem_count are in the tail (fractional LMUL)
+    // elements w/ index larger than vl are in the tail
+    for(unsigned idx = std::min(elem_count, vl); idx < VLEN; idx++) {
+        // always tail agnostic
+    }
+    return;
+}
+template <typename dest_elem_t, typename src2_elem_t = dest_elem_t>
+std::function<dest_elem_t(src2_elem_t)> get_unary_fn(unsigned unary_op) {
+    switch(unary_op) {
+    case 0b00111: // vsext.vf2
+    case 0b00101: // vsext.vf4
+    case 0b00011: // vsext.vf8
+        return [](src2_elem_t vs2) { return static_cast<typename std::make_signed_t<src2_elem_t>>(vs2); };
+    case 0b00110: // vzext.vf2
+    case 0b00100: // vzext.vf4
+    case 0b00010: // vzext.vf8
+        return [](src2_elem_t vs2) { return vs2; };
+    default:
+        throw new std::runtime_error("Uknown funct in get_unary_fn");
+    }
+}
+
+template <unsigned VLEN, typename dest_elem_t, typename src2_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) {
+    uint64_t elem_count = VLEN * vtype.lmul() / vtype.sew();
+    vmask_view mask_reg = read_vmask<VLEN>(V, elem_count);
+    auto vs2_view = get_vreg<VLEN, src2_elem_t>(V, vs2, elem_count);
+    auto vd_view = get_vreg<VLEN, dest_elem_t>(V, vd, elem_count);
+    auto fn = get_unary_fn<dest_elem_t, src2_elem_t>(unary_op);
+    // elements w/ index smaller than vstart are in the prestart and get skipped
+    // body is from vstart to min(elem_count, vl)
     for(unsigned idx = vstart; idx < std::min(elem_count, vl); idx++) {
         bool mask_active = vm ? 1 : mask_reg[idx];
         if(mask_active) {
-            vd_view[idx] = fn(vs2_view[idx], imm);
+            vd_view[idx] = fn(vs2_view[idx]);
         } else {
             vd_view[idx] = vtype.vma() ? vd_view[idx] : vd_view[idx];
         }