Improved test17 ([DYNAREC] Added 66 0F 38 16 opcode, fixed 66 0F 38 24/34 opcodes)

4 files changed, 88 insertions, 16 deletions

diff --git a/src/dynarec/arm64/dynarec_arm64_660f.c b/src/dynarec/arm64/dynarec_arm64_660f.c
index aa8a724a..c739ee15 100755
--- a/src/dynarec/arm64/dynarec_arm64_660f.c
+++ b/src/dynarec/arm64/dynarec_arm64_660f.c
@@ -385,7 +385,7 @@ uintptr_t dynarec64_660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int n
                     GETEX(q1, 0);

                     GETGX_empty(q0);

                     SXTL_16(q0, q1);     // 16bits->32bits

-                    SXTL_32(q0, q1);     // 32bits->64bits

+                    SXTL_32(q0, q0);     // 32bits->64bits

                     break;

                 case 0x25:

                     INST_NAME("PMOVSXDQ Gx, Ex");  // SSE4 opcode!

@@ -432,7 +432,7 @@ uintptr_t dynarec64_660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int n
                     GETEX(q1, 0);

                     GETGX_empty(q0);

                     UXTL_16(q0, q1);     // 16bits->32bits

-                    UXTL_32(q0, q1);     // 32bits->64bits

+                    UXTL_32(q0, q0);     // 32bits->64bits

                     break;

                 case 0x35:

                     INST_NAME("PMOVZXDQ Gx, Ex");  // SSE4 opcode!

@@ -451,7 +451,7 @@ uintptr_t dynarec64_660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int n
                     break;

 

                 case 0x3D:

-                    INST_NAME("PMINSD Gx, Ex");  // SSE4 opcode!

+                    INST_NAME("PMAXSD Gx, Ex");  // SSE4 opcode!

                     nextop = F8;

                     GETEX(q1, 0);

                     GETGX(q0);

@@ -603,6 +603,19 @@ uintptr_t dynarec64_660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int n
                     }

                     break;

 

+                case 0x16:

+                    if(rex.w) {INST_NAME("PEXTRQ Ed, Gx, Ib");} else {INST_NAME("PEXTRD Ed, Gx, Ib");}

+                    nextop = F8;

+                    GETGX(q0);

+                    GETED(1);

+                    u8 = F8;

+                    if(rex.w) {

+                        VMOVQDto(ed, q0, (u8&1));

+                    } else {

+                        VMOVSto(ed, q0, (u8&3));

+                    }

+                    break;

+

                 case 0x22:

                     INST_NAME("PINSRD Gx, ED, Ib");

                     nextop = F8;

diff --git a/tests/ref17.txt b/tests/ref17.txt
index 2bae8c46..43882799 100644
--- a/tests/ref17.txt
+++ b/tests/ref17.txt
@@ -152,9 +152,32 @@ cmpss 7 inf, inf => 0xffffffff
 cmpss 7 -inf, inf => 0xffffffff
 cmpss 7 inf, -inf => 0xffffffff
 cmpss 7 nan, nan => 0x0
+pshufb(0xff 0x80 0x7f 0x0 0x1 0x2 0x3 0x81 0xfe 0x84 0x72 0x52 0xa5 0x0 0xc0 0x32 , 0x0 0x1 0x5 0x15 0x20 0x80 0xff 0x0 0x8 0x7 0x81 0x6 0xa 0xf 0x10 0x1 ) = 0xff 0x80 0x2 0x2 0xff 0x0 0x0 0xff 0xfe 0x81 0x0 0x3 0x72 0x32 0xff 0x80 
+phaddw(0xffff 0x8000 0x7fff 0x0 0x1 0x2 0x3 0x8001 , 0x8000 0x7fff 0xffff 0xffff 0x50 0x9000 0xfffe 0x8001 ) = 0x7fff 0x7fff 0x3 0x8004 0xffff 0xfffe 0x9050 0x7fff 
+phaddd(0xffffffff 0x80000000 0x7fffffff 0x0 , 0x1 0x80000000 0x5 0xfffffffe ) = 0x7fffffff 0x7fffffff 0x80000001 0x3 
+phaddsw(0xffff 0x8000 0x7fff 0x0 0x1 0x2 0x3 0x8001 , 0x8000 0x7fff 0xffff 0xffff 0x50 0x9000 0xfffe 0x8001 ) = 0x8000 0x7fff 0x3 0x8004 0xffff 0xfffe 0x9050 0x8000 
+pmaddubsw(0x80ff 0x7f 0x201 0x8103 0x84fe 0x5272 0xa5 0x32c0 , 0x100 0x1505 0x8020 0xff 0x708 0x681 0xf0a 0x110 ) = 0x80 0x27b 0xff20 0xfffd 0xb8c 0xc95e 0x672 0xc32 
+phsubw(0xffff 0x8000 0x7fff 0x0 0x1 0x2 0x3 0x8001 , 0x8000 0x7fff 0xffff 0xffff 0x50 0x9000 0xfffe 0x8001 ) = 0x7fff 0x7fff 0xffff 0x8002 0x1 0x0 0x7050 0x7ffd 
+psignb(0xff 0x80 0x7f 0x0 0x1 0x2 0x3 0x81 0xfe 0x84 0x72 0x52 0xa5 0x0 0xc0 0x32 , 0x0 0x1 0x5 0x15 0x20 0x80 0xff 0x0 0x8 0x7 0x81 0x6 0xa 0xf 0x10 0x1 ) = 0x0 0x80 0x7f 0x0 0x1 0xfe 0xfd 0x0 0xfe 0x84 0x8e 0x52 0xa5 0x0 0xc0 0x32 
+psignw(0xffff 0x8000 0x7fff 0x0 0x1 0x2 0x3 0x8001 , 0x8000 0x7fff 0xffff 0xffff 0x50 0x9000 0xfffe 0x8001 ) = 0x1 0x8000 0x8001 0x0 0x1 0xfffe 0xfffd 0x7fff 
+psignd(0xffffffff 0x80000000 0x7fffffff 0x0 , 0x1 0x80000000 0x5 0xfffffffe ) = 0xffffffff 0x80000000 0x7fffffff 0x0 
+pmulhrsw(0xffff 0x8000 0x7fff 0x0 0x1 0x2 0x3 0x8001 , 0x8000 0x7fff 0xffff 0xffff 0x50 0x9000 0xfffe 0x8001 ) = 0x1 0x8001 0xffff 0x0 0x0 0xfffe 0x0 0x7ffe 
+pblendvps(0xffffffff 0x80000000 0x7fffffff 0x0 , 0x1 0x80000000 0x5 0xfffffffe , 0x1 0x80000000 0x80000005 0xfffe ) = 0xffffffff 0x80000000 0x5 0x0 
+ptestz(0x80000000ffffffff 0x7fffffff , 0x8000000000000001 0xfffffffe00000005 ) = 0
+ptestc(0x80000000ffffffff 0x7fffffff , 0x8000000000000001 0xfffffffe00000005 ) = 0
+ptestnzc(0x80000000ffffffff 0x7fffffff , 0x8000000000000001 0xfffffffe00000005 ) = 1
 pabsb(0xff 0x80 0x7f 0x0 0x1 0x2 0x3 0x81 0xfe 0x84 0x72 0x52 0xa5 0x0 0xc0 0x32 ) = 0x1 0x80 0x7f 0x0 0x1 0x2 0x3 0x7f 0x2 0x7c 0x72 0x52 0x5b 0x0 0x40 0x32 
 pabsw(0xffff 0x8000 0x7fff 0x0 0x1 0x2 0x3 0x8001 ) = 0x1 0x8000 0x7fff 0x0 0x1 0x2 0x3 0x7fff 
 pabsd(0xffffffff 0x80000000 0x7fffffff 0x0 ) = 0x1 0x80000000 0x7fffffff 0x0 
-pshuffleb(0xff 0x80 0x7f 0x0 0x1 0x2 0x3 0x81 0xfe 0x84 0x72 0x52 0xa5 0x0 0xc0 0x32 , 0x0 0x1 0x5 0x15 0x20 0x80 0xff 0x0 0x8 0x7 0x81 0x6 0xa 0xf 0x10 0x1 ) = 0xff 0x80 0x2 0x2 0xff 0x0 0x0 0xff 0xfe 0x81 0x0 0x3 0x72 0x32 0xff 0x80 
-phaddw(0xffff 0x8000 0x7fff 0x0 0x1 0x2 0x3 0x8001 , 0x8000 0x7fff 0xffff 0xffff 0x50 0x9000 0xfffe 0x8001 ) = 0x7fff 0x7fff 0x3 0x8004 0xffff 0xfffe 0x9050 0x7fff 
-phaddd(0xffffffff 0x80000000 0x7fffffff 0x0 , 0x1 0x80000000 0x5 0xfffffffe ) = 0x7fffffff 0x7fffffff 0x80000001 0x3 
+pmovsxbw(0xffff 0x8000 0x7fff 0x0 0x1 0x2 0x3 0x8001 ) = 0xffff 0xffff 0x0 0xff80 0xffff 0x7f 0x0 0x0 
+pmovsxbd(0xffffffff 0x80000000 0x7fffffff 0x0 ) = 0xffffffff 0xffffffff 0xffffffff 0xffffffff 
+pmovsxbq(0xffffffffffffffff 0x8000000000000000 ) = 0xffffffffffffffff 0xffffffffffffffff 
+pmovsxwd(0xffffffff 0x80000000 0x7fffffff 0x0 ) = 0xffffffff 0xffffffff 0x0 0xffff8000 
+pmovsxwq(0xffffffffffffffff 0x8000000000000000 ) = 0xffffffffffffffff 0xffffffffffffffff 
+pmovsxdq(0xffffffffffffffff 0x8000000000000000 ) = 0xffffffffffffffff 0xffffffffffffffff 
+pmovzxbw(0xffff 0x8000 0x7fff 0x0 0x1 0x2 0x3 0x8001 ) = 0xff 0xff 0x0 0x80 0xff 0x7f 0x0 0x0 
+pmovzxbd(0xffffffff 0x80000000 0x7fffffff 0x0 ) = 0xff 0xff 0xff 0xff 
+pmovzxbq(0xffffffffffffffff 0x8000000000000000 ) = 0xff 0xff 
+pmovzxwd(0xffffffff 0x80000000 0x7fffffff 0x0 ) = 0xffff 0xffff 0x0 0x8000 
+pmovzxwq(0xffffffffffffffff 0x8000000000000000 ) = 0xffff 0xffff 
+pmovzxdq(0xffffffffffffffff 0x8000000000000000 ) = 0xffffffff 0xffffffff 
diff --git a/tests/test17 b/tests/test17
index 0d0b89e4..b6be07fc 100755
--- a/tests/test17
+++ b/tests/test17
Binary files differdiff --git a/tests/test17.c b/tests/test17.c
index e0c07685..d3992ea2 100644
--- a/tests/test17.c
+++ b/tests/test17.c
@@ -103,6 +103,9 @@ const v128 b128_16 = {.u16 = {
 const v128 b128_32 = {.u32 = {
     0x00000001, 0x80000000, 0x00000005, 0xfffffffe
 }};
+const v128 c128_32 = {.u32 = {
+    0x00000001, 0x80000000, 0x80000005, 0x0000fffe
+}};
 
 void print_8(v128 v) {
     for(int i=0; i<16; ++i)
@@ -219,25 +222,58 @@ printf(N " %g, %g => %g\n", b, a, *(float*)&r);
  #undef GO1
  #undef GO2
  v128 a128;
+ int i;
 
  #define GO1(A, N, C)                               \
  a128.mm = _mm_##A##_epi##N(a128_##N.mm);           \
- printf("p%s%s(", #A, #C); print_##N(a128_##N);     \
+ printf("%s(", #C); print_##N(a128_##N);           \
  printf(") = "); print_##N(a128); printf("\n");
  #define GO2(A, N, C, A1, A2)                       \
  a128.mm = _mm_##A##_epi##N(A1.mm, A2.mm);          \
- printf("p%s%s(", #A, #C); print_##N(A1);           \
+ printf("%s(", #C); print_##N(A1);                 \
  printf(", "); print_##N(A2);                       \
  printf(") = "); print_##N(a128); printf("\n");
+ #define GO2i(A, A1, A2)                            \
+ i = _mm_##A##_si128(A1.mm, A2.mm);                 \
+ printf("p%s(", #A); print_64(A1);                  \
+ printf(", "); print_64(A2);                        \
+ printf(") = %d\n", i);
+ #define GO3PS(A, N, A1, A2, A3)                    \
+ a128.mf = _mm_##A##_ps(A1.mf, A2.mf, A3.mf);       \
+ printf("p%s%s(", #A, "ps"); print_##N(A1);         \
+ printf(", "); print_##N(A2);                       \
+ printf(", "); print_##N(A3);                       \
+ printf(") = "); print_##N(a128); printf("\n");
 
 
- GO1(abs, 8, b)
- GO1(abs, 16, w)
- GO1(abs, 32, d)
- GO2(shuffle, 8, b, a128_8, b128_8)
- GO2(hadd, 16, w, a128_16, b128_16)
- GO2(hadd, 32, d, a128_32, b128_32)
-
-
+ GO2(shuffle, 8, pshufb, a128_8, b128_8)
+ GO2(hadd, 16, phaddw, a128_16, b128_16)
+ GO2(hadd, 32, phaddd, a128_32, b128_32)
+ GO2(hadds, 16, phaddsw, a128_16, b128_16)
+ GO2(maddubs, 16, pmaddubsw, a128_8, b128_8)
+ GO2(hsub, 16, phsubw, a128_16, b128_16)
+ GO2(sign, 8, psignb, a128_8, b128_8)
+ GO2(sign, 16, psignw, a128_16, b128_16)
+ GO2(sign, 32, psignd, a128_32, b128_32)
+ GO2(mulhrs, 16, pmulhrsw, a128_16, b128_16)
+ GO3PS(blendv, 32, a128_32, b128_32, c128_32)
+ GO2i(testz, a128_32, b128_32)
+ GO2i(testc, a128_32, b128_32)
+ GO2i(testnzc, a128_32, b128_32)
+ GO1(abs, 8, pabsb)
+ GO1(abs, 16, pabsw)
+ GO1(abs, 32, pabsd)
+ GO1(cvtepi8, 16, pmovsxbw);
+ GO1(cvtepi8, 32, pmovsxbd);
+ GO1(cvtepi8, 64, pmovsxbq);
+ GO1(cvtepi16, 32, pmovsxwd);
+ GO1(cvtepi16, 64, pmovsxwq);
+ GO1(cvtepi32, 64, pmovsxdq);
+ GO1(cvtepu8, 16, pmovzxbw);
+ GO1(cvtepu8, 32, pmovzxbd);
+ GO1(cvtepu8, 64, pmovzxbq);
+ GO1(cvtepu16, 32, pmovzxwd);
+ GO1(cvtepu16, 64, pmovzxwq);
+ GO1(cvtepu32, 64, pmovzxdq);
  return 0;
 }