Implement PADDQ

emu/decode.h

@@ -310,6 +310,9 @@ __no_instrument DECODER_RET glue(DECODER_NAME, OP_SIZE)(DECODER_ARGS) {
                 case 0xc5: TRACEI("pextrw xmm, modrm_val, imm8");
                            READMODRM_NOMEM; READIMM8; V_OP_IMM(extract_w, xmm_modrm_val, modrm_reg,128); break;
 
+                case 0xd4: TRACEI("paddq xmm:modrm, xmm");
+                           READMODRM; V_OP(add, xmm_modrm_val, xmm_modrm_reg, 64); break;
+
                 case 0xd6: TRACEI("movq xmm, xmm:modrm");
                            READMODRM; VMOV(xmm_modrm_reg, xmm_modrm_val,64); break;
 

emu/vec.c

@@ -81,6 +81,11 @@ void vec_xor128(NO_CPU, union xmm_reg *src, union xmm_reg *dst) {
     dst->qw[1] ^= src->qw[1];
 }
 
+void vec_add64(NO_CPU, union xmm_reg *src, union xmm_reg *dst) {
+    dst->qw[0] += src->qw[0];
+    dst->qw[1] += src->qw[1];
+}
+
 void vec_fadds64(NO_CPU, const double *src, double *dst) {
     *dst += *src;
 }

emu/vec.h

@@ -22,6 +22,8 @@ void vec_imm_shiftl64(NO_CPU, const uint8_t amount, union xmm_reg *dst);
 void vec_imm_shiftr64(NO_CPU, const uint8_t amount, union xmm_reg *dst);
 void vec_xor128(NO_CPU, union xmm_reg *src, union xmm_reg *dst);
 
+void vec_add64(NO_CPU, union xmm_reg *src, union xmm_reg *dst);
+
 void vec_fadds64(NO_CPU, const double *src, double *dst);
 void vec_fmuls64(NO_CPU, const double *src, double *dst);
 void vec_fsubs64(NO_CPU, const double *src, double *dst);

tests/e2e/sse2/expected.txt

@@ -25,3 +25,7 @@ psrlq
 01234 05678
 00617 02839
 00000 00000
+paddq
+01234 05678
+11111 11111
+12345 16789

tests/e2e/sse2/paddq.c

@@ -0,0 +1,29 @@
+#include <stdint.h>
+#include <stdio.h>
+#include <emmintrin.h>
+#include <xmmintrin.h>
+
+#define printout() printf("%05lld %05lld\n", (long long) out[0], (long long) out[1])
+
+void main(void) {
+    int64_t out[2] = { 0, 0 };
+    int64_t buf1234[2] = {  1234,  5678 };
+    int64_t buf1111[2] = { 11111, 11111 };
+
+    // xmm1 Initially 1234
+    __m128i xmm1 = _mm_load_si128((__m128i*) buf1234);
+    _mm_store_si128((__m128i*) out, xmm1);
+    printout();
+
+    // xmm2 Initially 1111
+    __m128i xmm2 = _mm_load_si128((__m128i*) buf1111);
+    _mm_store_si128((__m128i*) out, xmm2);
+    printout();
+
+    // Result is just each added by 1
+    asm volatile(	"paddq %[vec2], %[vec1]\n\t"
+            : [vec1] "+x" (xmm1), [vec2] "+x" (xmm2)
+            :);
+    _mm_store_si128((__m128i*) out, xmm1);
+    printout();
+}

tests/e2e/sse2/test.sh

@@ -18,3 +18,7 @@ gcc -msse2 psllq.c -o test_psllq
 echo psrlq
 gcc -msse2 psrlq.c -o test_psrlq
 ./test_psrlq
+
+echo paddq
+gcc -msse2 paddq.c -o test_paddq
+./test_paddq