Merge remote-tracking branch 'refs/remotes/origin/master'

rvvector/benchmark_temp/Makefile

@@ -1,11 +1,11 @@
 LIB_PATH = ../../runtime
 
-COMP     = /nethome/ekim79/riscv-gnu-toolchain/drops/bin/riscv32-unknown-elf-gcc
+COMP     = /home/priya/dev/riscv_vec/riscv-gnu/bin/riscv32-unknown-elf-gcc
 
 CC_FLAGS = -ffreestanding -O0 -Wl,--gc-sections -nostartfiles -nostdlib -nostartfiles -nodefaultlibs -Wl,-Bstatic,-T,$(LIB_PATH)/mains/vortex_link.ld -march=rv32imv -mabi=ilp32
 
-DMP  = /nethome/ekim79/riscv-gnu-toolchain/drops/bin/riscv32-unknown-elf-objdump
+DMP  = /home/priya/dev/riscv_vec/riscv-gnu/bin/riscv32-unknown-elf-objdump
-CPY  = /nethome/ekim79/riscv-gnu-toolchain/drops/bin/riscv32-unknown-elf-objcopy
+CPY  = /home/priya/dev/riscv_vec/riscv-gnu/bin/riscv32-unknown-elf-objcopy
 
 # VX_STR  = ../../startup/vx_start.s
 
@@ -21,7 +21,7 @@ VX_VEC2  = vx_vec_saxpy.s       #float --> int
 VX_VEC3  = vx_vec_sgemm_float.s #float --> int
 VX_VEC4  = vx_vec_vsadd.s 
 VX_VEC5  = vx_vec_memcpy.s 
-LIBS    = /nethome/ekim79/riscv-gnu-toolchain/drops/riscv32-unknown-elf/lib/libc.a /nethome/ekim79/riscv-gnu-toolchain/drops/riscv32-unknown-elf/lib/libstdc++.a -static-libgcc -lgcc
+LIBS    = /home/priya/dev/riscv_vec/riscv-gnu/riscv32-unknown-elf/lib/libc.a /home/priya/dev/riscv_vec/riscv-gnu/riscv32-unknown-elf/lib/libstdc++.a -static-libgcc -lgcc
 
 VX_MAIN = vx_vec_benchmark
 

rvvector/benchmark_temp/vx_vec_saxpy.s

@@ -41,15 +41,21 @@ saxpy:
 #    ret
 
 # a0 n, rs1 a, a2 x, a3 y
+
+# a0 n, a1 a, a2 x, a3 y
 vx_vec_saxpy:
-    vsetvli a4, a0, e32, m8
+    vsetvli a4, a0, e32, m1
 saxpy:
     vlw.v v0, (a2)
     sub a0, a0, a4
     slli a4, a4, 2
     add a2, a2, a4
     vlw.v v1, (a3)
-    vmacc.vx v1, rs1, v0
+    vmul.vx v0, v0, a1
+#    vmul.vv v0, v0, v1
+#    li x1, 10
+#    vmul.vx v0, v0, x1
+    vadd.vv v1, v0, v1
     vsw.v v1, (a3)
     add a3, a3, a4
     bnez a0, saxpy

simX/instruction.cpp

@@ -2016,6 +2016,64 @@ void Instruction::executeOn(Warp &c, trace_inst_t * trace_inst) {
                 }
               }              
               break; 
+              case 37: //vmul
+              {
+                D(3, "vmul");
+                uint8_t *result_ptr;
+
+                vector<Reg<char *>> & vr1 = c.vreg[rsrc[0]];
+                vector<Reg<char *>> & vr2 = c.vreg[rsrc[1]];
+                vector<Reg<char *>> & vd  = c.vreg[rdest];
+                if(c.vtype.vsew == 8){
+                  for(uint8_t i = 0; i < c.vl; i++){
+                    uint8_t *first_ptr = (uint8_t *)vr1[i].val;
+                    uint8_t *second_ptr = (uint8_t *)vr2[i].val;
+                    uint8_t result =  (*first_ptr * *second_ptr);
+                    D(3,"Comparing " << *first_ptr << " + " << *second_ptr << " = " << result);
+
+                    result_ptr = (uint8_t *) vd[i].val;
+                    *result_ptr = result;
+                  }
+                  for(uint8_t i = c.vl; i < VLMAX; i++){
+                    result_ptr = (uint8_t *) vd[i].val;
+                    *result_ptr = 0;
+                  }
+                } 
+                else if(c.vtype.vsew == 16) {
+                  uint16_t *result_ptr;
+                  for(uint16_t i = 0; i < c.vl; i++){
+                    uint16_t *first_ptr = (uint16_t *)vr1[i].val;
+                    uint16_t *second_ptr = (uint16_t *)vr2[i].val;
+                    uint16_t result = (*first_ptr * *second_ptr);
+                    D(3,"Comparing " << *first_ptr << " + " << *second_ptr << " = " << result);
+
+                    result_ptr = (uint16_t *) vd[i].val;
+                    *result_ptr = result;
+                  }
+                  for(uint16_t i = c.vl; i < VLMAX; i++){
+                    result_ptr = (uint16_t *) vd[i].val;
+                    *result_ptr = 0;
+                  }
+
+                } else if(c.vtype.vsew == 32) {
+                  uint32_t *result_ptr;
+
+                  for(uint32_t i = 0; i < c.vl; i++){
+                    uint32_t *first_ptr = (uint32_t *)vr1[i].val;
+                    uint32_t *second_ptr = (uint32_t *)vr2[i].val;
+                    uint32_t result = (*first_ptr * *second_ptr);
+                    D(3,"Comparing " << *first_ptr << " + " << *second_ptr << " = " << result);
+
+                    result_ptr = (uint32_t *) vd[i].val;
+                    *result_ptr = result;
+                  }
+                  for(Word i = c.vl; i < VLMAX; i++){
+                    result_ptr = (uint32_t *) vd[i].val;
+                    *result_ptr = 0;
+                  }
+                }
+              }
+              break;
               case 45: //vmacc   
               {
                 D(3, "vmacc");
@@ -2077,6 +2135,129 @@ void Instruction::executeOn(Warp &c, trace_inst_t * trace_inst) {
             }
           }
           break;
+          case 6: 
+          {
+            switch(func6)
+            {
+              case 0:
+              {
+                D(3, "vmadd.vx");
+                uint8_t *result_ptr;
+
+                //vector<Reg<char *>> & vr1 = c.vreg[rsrc[0]];
+                vector<Reg<char *>> & vr2 = c.vreg[rsrc[1]];
+                vector<Reg<char *>> & vd  = c.vreg[rdest];
+                if(c.vtype.vsew == 8){
+                  for(uint8_t i = 0; i < c.vl; i++){
+                    //uint8_t *first_ptr = (uint8_t *)vr1[i].val;
+                    uint8_t *second_ptr = (uint8_t *)vr2[i].val;
+                    uint8_t result =  (reg[rsrc[0]] + *second_ptr);
+                    D(3,"Comparing " << reg[rsrc[0]] << " + " << *second_ptr << " = " << result);
+
+                    result_ptr = (uint8_t *) vd[i].val;
+                    *result_ptr = result;
+                  }
+                  for(uint8_t i = c.vl; i < VLMAX; i++){
+                    result_ptr = (uint8_t *) vd[i].val;
+                    *result_ptr = 0;
+                  }
+                } 
+                else if(c.vtype.vsew == 16) {
+                  uint16_t *result_ptr;
+                  for(uint16_t i = 0; i < c.vl; i++){
+                    //uint16_t *first_ptr = (uint16_t *)vr1[i].val;
+                    uint16_t *second_ptr = (uint16_t *)vr2[i].val;
+                    uint16_t result = (reg[rsrc[0]] + *second_ptr);
+                    D(3,"Comparing " << reg[rsrc[0]] << " + " << *second_ptr << " = " << result);
+
+                    result_ptr = (uint16_t *) vd[i].val;
+                    *result_ptr = result;
+                  }
+                  for(uint16_t i = c.vl; i < VLMAX; i++){
+                    result_ptr = (uint16_t *) vd[i].val;
+                    *result_ptr = 0;
+                  }
+
+                } else if(c.vtype.vsew == 32) {
+                  uint32_t *result_ptr;
+
+                  for(uint32_t i = 0; i < c.vl; i++){
+                    //uint32_t *first_ptr = (uint32_t *)vr1[i].val;
+                    uint32_t *second_ptr = (uint32_t *)vr2[i].val;
+                    uint32_t result = (reg[rsrc[0]] + *second_ptr);
+                    D(3,"Comparing " << reg[rsrc[0]] << " + " << *second_ptr << " = " << result);
+
+                    result_ptr = (uint32_t *) vd[i].val;
+                    *result_ptr = result;
+                  }
+                  for(Word i = c.vl; i < VLMAX; i++){
+                    result_ptr = (uint32_t *) vd[i].val;
+                    *result_ptr = 0;
+                  }
+                }                
+              }
+              break;
+              case 37: //vmul.vx
+              {
+                D(3, "vmul.vx");
+                uint8_t *result_ptr;
+
+                //vector<Reg<char *>> & vr1 = c.vreg[rsrc[0]];
+                vector<Reg<char *>> & vr2 = c.vreg[rsrc[1]];
+                vector<Reg<char *>> & vd  = c.vreg[rdest];
+                if(c.vtype.vsew == 8){
+                  for(uint8_t i = 0; i < c.vl; i++){
+                    //uint8_t *first_ptr = (uint8_t *)vr1[i].val;
+                    uint8_t *second_ptr = (uint8_t *)vr2[i].val;
+                    uint8_t result =  (reg[rsrc[0]] * *second_ptr);
+                    D(3,"Comparing " << reg[rsrc[0]] << " + " << *second_ptr << " = " << result);
+
+                    result_ptr = (uint8_t *) vd[i].val;
+                    *result_ptr = result;
+                  }
+                  for(uint8_t i = c.vl; i < VLMAX; i++){
+                    result_ptr = (uint8_t *) vd[i].val;
+                    *result_ptr = 0;
+                  }
+                } 
+                else if(c.vtype.vsew == 16) {
+                  uint16_t *result_ptr;
+                  for(uint16_t i = 0; i < c.vl; i++){
+                    //uint16_t *first_ptr = (uint16_t *)vr1[i].val;
+                    uint16_t *second_ptr = (uint16_t *)vr2[i].val;
+                    uint16_t result = (reg[rsrc[0]] * *second_ptr);
+                    D(3,"Comparing " << reg[rsrc[0]] << " + " << *second_ptr << " = " << result);
+
+                    result_ptr = (uint16_t *) vd[i].val;
+                    *result_ptr = result;
+                  }
+                  for(uint16_t i = c.vl; i < VLMAX; i++){
+                    result_ptr = (uint16_t *) vd[i].val;
+                    *result_ptr = 0;
+                  }
+
+                } else if(c.vtype.vsew == 32) {
+                  uint32_t *result_ptr;
+
+                  for(uint32_t i = 0; i < c.vl; i++){
+                    //uint32_t *first_ptr = (uint32_t *)vr1[i].val;
+                    uint32_t *second_ptr = (uint32_t *)vr2[i].val;
+                    uint32_t result = (reg[rsrc[0]] * *second_ptr);
+                    D(3,"Comparing " << reg[rsrc[0]] << " + " << *second_ptr << " = " << result);
+
+                    result_ptr = (uint32_t *) vd[i].val;
+                    *result_ptr = result;
+                  }
+                  for(Word i = c.vl; i < VLMAX; i++){
+                    result_ptr = (uint32_t *) vd[i].val;
+                    *result_ptr = 0;
+                  }
+                }                
+              }
+              break;
+            }
+          }
+          break;
           case 7:
           {
             is_vec = true;

simX/test_riscv.sh

@@ -3,4 +3,5 @@ echo start > results.txt
 # echo ../kernel/vortex_test.hex
 make
 printf "Fasten your seatbelts ladies and gentelmen!!\n\n\n\n"
-cd obj_dir && ./Vcache_simX -E -a rv32i --core ../../runtime/mains/simple/vx_simple_main.hex  -s -b 1> emulator.debug
+#cd obj_dir && ./Vcache_simX -E -a rv32i --core ../../runtime/mains/simple/vx_simple_main.hex  -s -b 1> emulator.debug
+cd obj_dir && ./Vcache_simX -E -a rv32i --core /home/priya/Desktop/new_vortex/Vortex/rvvector/benchmark_temp/vx_vec_benchmark.hex -s -b 1> emulator.debug