diff options
Diffstat (limited to 'llvm/test/CodeGen/X86/psubus.ll')
| -rw-r--r-- | llvm/test/CodeGen/X86/psubus.ll | 1159 |
1 files changed, 480 insertions, 679 deletions
diff --git a/llvm/test/CodeGen/X86/psubus.ll b/llvm/test/CodeGen/X86/psubus.ll index 42b37312442..08d35abdee5 100644 --- a/llvm/test/CodeGen/X86/psubus.ll +++ b/llvm/test/CodeGen/X86/psubus.ll @@ -1092,40 +1092,10 @@ vector.ph: } define <8 x i16> @psubus_8i16_max(<8 x i16> %x, <8 x i16> %y) nounwind { -; SSE2-LABEL: psubus_8i16_max: -; SSE2: # %bb.0: # %vector.ph -; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [32768,32768,32768,32768,32768,32768,32768,32768] -; SSE2-NEXT: movdqa %xmm0, %xmm3 -; SSE2-NEXT: pxor %xmm2, %xmm3 -; SSE2-NEXT: pxor %xmm1, %xmm2 -; SSE2-NEXT: pcmpgtw %xmm3, %xmm2 -; SSE2-NEXT: movdqa %xmm1, %xmm3 -; SSE2-NEXT: pand %xmm2, %xmm3 -; SSE2-NEXT: pandn %xmm0, %xmm2 -; SSE2-NEXT: por %xmm3, %xmm2 -; SSE2-NEXT: psubw %xmm1, %xmm2 -; SSE2-NEXT: movdqa %xmm2, %xmm0 -; SSE2-NEXT: retq -; -; SSSE3-LABEL: psubus_8i16_max: -; SSSE3: # %bb.0: # %vector.ph -; SSSE3-NEXT: movdqa {{.*#+}} xmm2 = [32768,32768,32768,32768,32768,32768,32768,32768] -; SSSE3-NEXT: movdqa %xmm0, %xmm3 -; SSSE3-NEXT: pxor %xmm2, %xmm3 -; SSSE3-NEXT: pxor %xmm1, %xmm2 -; SSSE3-NEXT: pcmpgtw %xmm3, %xmm2 -; SSSE3-NEXT: movdqa %xmm1, %xmm3 -; SSSE3-NEXT: pand %xmm2, %xmm3 -; SSSE3-NEXT: pandn %xmm0, %xmm2 -; SSSE3-NEXT: por %xmm3, %xmm2 -; SSSE3-NEXT: psubw %xmm1, %xmm2 -; SSSE3-NEXT: movdqa %xmm2, %xmm0 -; SSSE3-NEXT: retq -; -; SSE41-LABEL: psubus_8i16_max: -; SSE41: # %bb.0: # %vector.ph -; SSE41-NEXT: psubusw %xmm1, %xmm0 -; SSE41-NEXT: retq +; SSE-LABEL: psubus_8i16_max: +; SSE: # %bb.0: # %vector.ph +; SSE-NEXT: psubusw %xmm1, %xmm0 +; SSE-NEXT: retq ; ; AVX-LABEL: psubus_8i16_max: ; AVX: # %bb.0: # %vector.ph @@ -1156,63 +1126,11 @@ vector.ph: } define <16 x i16> @psubus_16i16_max(<16 x i16> %x, <16 x i16> %y) nounwind { -; SSE2-LABEL: psubus_16i16_max: -; SSE2: # %bb.0: # %vector.ph -; SSE2-NEXT: movdqa {{.*#+}} xmm4 = [32768,32768,32768,32768,32768,32768,32768,32768] -; SSE2-NEXT: movdqa %xmm0, %xmm6 -; SSE2-NEXT: pxor %xmm4, %xmm6 -; SSE2-NEXT: movdqa %xmm2, %xmm5 -; SSE2-NEXT: pxor %xmm4, %xmm5 -; SSE2-NEXT: pcmpgtw %xmm6, %xmm5 -; SSE2-NEXT: movdqa %xmm1, %xmm6 -; SSE2-NEXT: pxor %xmm4, %xmm6 -; SSE2-NEXT: pxor %xmm3, %xmm4 -; SSE2-NEXT: pcmpgtw %xmm6, %xmm4 -; SSE2-NEXT: movdqa %xmm3, %xmm6 -; SSE2-NEXT: pand %xmm4, %xmm6 -; SSE2-NEXT: pandn %xmm1, %xmm4 -; SSE2-NEXT: por %xmm6, %xmm4 -; SSE2-NEXT: movdqa %xmm2, %xmm1 -; SSE2-NEXT: pand %xmm5, %xmm1 -; SSE2-NEXT: pandn %xmm0, %xmm5 -; SSE2-NEXT: por %xmm1, %xmm5 -; SSE2-NEXT: psubw %xmm2, %xmm5 -; SSE2-NEXT: psubw %xmm3, %xmm4 -; SSE2-NEXT: movdqa %xmm5, %xmm0 -; SSE2-NEXT: movdqa %xmm4, %xmm1 -; SSE2-NEXT: retq -; -; SSSE3-LABEL: psubus_16i16_max: -; SSSE3: # %bb.0: # %vector.ph -; SSSE3-NEXT: movdqa {{.*#+}} xmm4 = [32768,32768,32768,32768,32768,32768,32768,32768] -; SSSE3-NEXT: movdqa %xmm0, %xmm6 -; SSSE3-NEXT: pxor %xmm4, %xmm6 -; SSSE3-NEXT: movdqa %xmm2, %xmm5 -; SSSE3-NEXT: pxor %xmm4, %xmm5 -; SSSE3-NEXT: pcmpgtw %xmm6, %xmm5 -; SSSE3-NEXT: movdqa %xmm1, %xmm6 -; SSSE3-NEXT: pxor %xmm4, %xmm6 -; SSSE3-NEXT: pxor %xmm3, %xmm4 -; SSSE3-NEXT: pcmpgtw %xmm6, %xmm4 -; SSSE3-NEXT: movdqa %xmm3, %xmm6 -; SSSE3-NEXT: pand %xmm4, %xmm6 -; SSSE3-NEXT: pandn %xmm1, %xmm4 -; SSSE3-NEXT: por %xmm6, %xmm4 -; SSSE3-NEXT: movdqa %xmm2, %xmm1 -; SSSE3-NEXT: pand %xmm5, %xmm1 -; SSSE3-NEXT: pandn %xmm0, %xmm5 -; SSSE3-NEXT: por %xmm1, %xmm5 -; SSSE3-NEXT: psubw %xmm2, %xmm5 -; SSSE3-NEXT: psubw %xmm3, %xmm4 -; SSSE3-NEXT: movdqa %xmm5, %xmm0 -; SSSE3-NEXT: movdqa %xmm4, %xmm1 -; SSSE3-NEXT: retq -; -; SSE41-LABEL: psubus_16i16_max: -; SSE41: # %bb.0: # %vector.ph -; SSE41-NEXT: psubusw %xmm2, %xmm0 -; SSE41-NEXT: psubusw %xmm3, %xmm1 -; SSE41-NEXT: retq +; SSE-LABEL: psubus_16i16_max: +; SSE: # %bb.0: # %vector.ph +; SSE-NEXT: psubusw %xmm2, %xmm0 +; SSE-NEXT: psubusw %xmm3, %xmm1 +; SSE-NEXT: retq ; ; AVX1-LABEL: psubus_16i16_max: ; AVX1: # %bb.0: # %vector.ph @@ -1240,109 +1158,13 @@ vector.ph: } define <32 x i16> @psubus_32i16_max(<32 x i16> %x, <32 x i16> %y) nounwind { -; SSE2-LABEL: psubus_32i16_max: -; SSE2: # %bb.0: # %vector.ph -; SSE2-NEXT: movdqa %xmm3, %xmm11 -; SSE2-NEXT: movdqa %xmm2, %xmm10 -; SSE2-NEXT: movdqa %xmm1, %xmm9 -; SSE2-NEXT: movdqa %xmm0, %xmm8 -; SSE2-NEXT: movdqa {{.*#+}} xmm3 = [32768,32768,32768,32768,32768,32768,32768,32768] -; SSE2-NEXT: movdqa %xmm0, %xmm1 -; SSE2-NEXT: pxor %xmm3, %xmm1 -; SSE2-NEXT: movdqa %xmm4, %xmm0 -; SSE2-NEXT: pxor %xmm3, %xmm0 -; SSE2-NEXT: pcmpgtw %xmm1, %xmm0 -; SSE2-NEXT: movdqa %xmm9, %xmm2 -; SSE2-NEXT: pxor %xmm3, %xmm2 -; SSE2-NEXT: movdqa %xmm5, %xmm1 -; SSE2-NEXT: pxor %xmm3, %xmm1 -; SSE2-NEXT: pcmpgtw %xmm2, %xmm1 -; SSE2-NEXT: movdqa %xmm10, %xmm12 -; SSE2-NEXT: pxor %xmm3, %xmm12 -; SSE2-NEXT: movdqa %xmm6, %xmm2 -; SSE2-NEXT: pxor %xmm3, %xmm2 -; SSE2-NEXT: pcmpgtw %xmm12, %xmm2 -; SSE2-NEXT: movdqa %xmm11, %xmm12 -; SSE2-NEXT: pxor %xmm3, %xmm12 -; SSE2-NEXT: pxor %xmm7, %xmm3 -; SSE2-NEXT: pcmpgtw %xmm12, %xmm3 -; SSE2-NEXT: movdqa %xmm7, %xmm12 -; SSE2-NEXT: pand %xmm3, %xmm12 -; SSE2-NEXT: pandn %xmm11, %xmm3 -; SSE2-NEXT: por %xmm12, %xmm3 -; SSE2-NEXT: movdqa %xmm6, %xmm11 -; SSE2-NEXT: pand %xmm2, %xmm11 -; SSE2-NEXT: pandn %xmm10, %xmm2 -; SSE2-NEXT: por %xmm11, %xmm2 -; SSE2-NEXT: movdqa %xmm5, %xmm10 -; SSE2-NEXT: pand %xmm1, %xmm10 -; SSE2-NEXT: pandn %xmm9, %xmm1 -; SSE2-NEXT: por %xmm10, %xmm1 -; SSE2-NEXT: movdqa %xmm4, %xmm9 -; SSE2-NEXT: pand %xmm0, %xmm9 -; SSE2-NEXT: pandn %xmm8, %xmm0 -; SSE2-NEXT: por %xmm9, %xmm0 -; SSE2-NEXT: psubw %xmm4, %xmm0 -; SSE2-NEXT: psubw %xmm5, %xmm1 -; SSE2-NEXT: psubw %xmm6, %xmm2 -; SSE2-NEXT: psubw %xmm7, %xmm3 -; SSE2-NEXT: retq -; -; SSSE3-LABEL: psubus_32i16_max: -; SSSE3: # %bb.0: # %vector.ph -; SSSE3-NEXT: movdqa %xmm3, %xmm11 -; SSSE3-NEXT: movdqa %xmm2, %xmm10 -; SSSE3-NEXT: movdqa %xmm1, %xmm9 -; SSSE3-NEXT: movdqa %xmm0, %xmm8 -; SSSE3-NEXT: movdqa {{.*#+}} xmm3 = [32768,32768,32768,32768,32768,32768,32768,32768] -; SSSE3-NEXT: movdqa %xmm0, %xmm1 -; SSSE3-NEXT: pxor %xmm3, %xmm1 -; SSSE3-NEXT: movdqa %xmm4, %xmm0 -; SSSE3-NEXT: pxor %xmm3, %xmm0 -; SSSE3-NEXT: pcmpgtw %xmm1, %xmm0 -; SSSE3-NEXT: movdqa %xmm9, %xmm2 -; SSSE3-NEXT: pxor %xmm3, %xmm2 -; SSSE3-NEXT: movdqa %xmm5, %xmm1 -; SSSE3-NEXT: pxor %xmm3, %xmm1 -; SSSE3-NEXT: pcmpgtw %xmm2, %xmm1 -; SSSE3-NEXT: movdqa %xmm10, %xmm12 -; SSSE3-NEXT: pxor %xmm3, %xmm12 -; SSSE3-NEXT: movdqa %xmm6, %xmm2 -; SSSE3-NEXT: pxor %xmm3, %xmm2 -; SSSE3-NEXT: pcmpgtw %xmm12, %xmm2 -; SSSE3-NEXT: movdqa %xmm11, %xmm12 -; SSSE3-NEXT: pxor %xmm3, %xmm12 -; SSSE3-NEXT: pxor %xmm7, %xmm3 -; SSSE3-NEXT: pcmpgtw %xmm12, %xmm3 -; SSSE3-NEXT: movdqa %xmm7, %xmm12 -; SSSE3-NEXT: pand %xmm3, %xmm12 -; SSSE3-NEXT: pandn %xmm11, %xmm3 -; SSSE3-NEXT: por %xmm12, %xmm3 -; SSSE3-NEXT: movdqa %xmm6, %xmm11 -; SSSE3-NEXT: pand %xmm2, %xmm11 -; SSSE3-NEXT: pandn %xmm10, %xmm2 -; SSSE3-NEXT: por %xmm11, %xmm2 -; SSSE3-NEXT: movdqa %xmm5, %xmm10 -; SSSE3-NEXT: pand %xmm1, %xmm10 -; SSSE3-NEXT: pandn %xmm9, %xmm1 -; SSSE3-NEXT: por %xmm10, %xmm1 -; SSSE3-NEXT: movdqa %xmm4, %xmm9 -; SSSE3-NEXT: pand %xmm0, %xmm9 -; SSSE3-NEXT: pandn %xmm8, %xmm0 -; SSSE3-NEXT: por %xmm9, %xmm0 -; SSSE3-NEXT: psubw %xmm4, %xmm0 -; SSSE3-NEXT: psubw %xmm5, %xmm1 -; SSSE3-NEXT: psubw %xmm6, %xmm2 -; SSSE3-NEXT: psubw %xmm7, %xmm3 -; SSSE3-NEXT: retq -; -; SSE41-LABEL: psubus_32i16_max: -; SSE41: # %bb.0: # %vector.ph -; SSE41-NEXT: psubusw %xmm4, %xmm0 -; SSE41-NEXT: psubusw %xmm5, %xmm1 -; SSE41-NEXT: psubusw %xmm6, %xmm2 -; SSE41-NEXT: psubusw %xmm7, %xmm3 -; SSE41-NEXT: retq +; SSE-LABEL: psubus_32i16_max: +; SSE: # %bb.0: # %vector.ph +; SSE-NEXT: psubusw %xmm4, %xmm0 +; SSE-NEXT: psubusw %xmm5, %xmm1 +; SSE-NEXT: psubusw %xmm6, %xmm2 +; SSE-NEXT: psubusw %xmm7, %xmm3 +; SSE-NEXT: retq ; ; AVX1-LABEL: psubus_32i16_max: ; AVX1: # %bb.0: # %vector.ph @@ -1451,61 +1273,55 @@ define <8 x i16> @psubus_8i32_max(<8 x i16> %x, <8 x i32> %y) nounwind { ; SSE2-LABEL: psubus_8i32_max: ; SSE2: # %bb.0: # %vector.ph ; SSE2-NEXT: movdqa %xmm0, %xmm3 -; SSE2-NEXT: pxor %xmm0, %xmm0 -; SSE2-NEXT: movdqa %xmm3, %xmm4 -; SSE2-NEXT: punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm0[4],xmm4[5],xmm0[5],xmm4[6],xmm0[6],xmm4[7],xmm0[7] -; SSE2-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3] +; SSE2-NEXT: pxor %xmm4, %xmm4 +; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3] +; SSE2-NEXT: punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7] ; SSE2-NEXT: movdqa {{.*#+}} xmm5 = [2147483648,2147483648,2147483648,2147483648] -; SSE2-NEXT: movdqa %xmm1, %xmm0 -; SSE2-NEXT: pxor %xmm5, %xmm0 -; SSE2-NEXT: movdqa %xmm3, %xmm6 -; SSE2-NEXT: por %xmm5, %xmm6 -; SSE2-NEXT: pcmpgtd %xmm6, %xmm0 ; SSE2-NEXT: movdqa %xmm2, %xmm6 ; SSE2-NEXT: pxor %xmm5, %xmm6 -; SSE2-NEXT: por %xmm4, %xmm5 -; SSE2-NEXT: pcmpgtd %xmm5, %xmm6 -; SSE2-NEXT: movdqa %xmm2, %xmm5 -; SSE2-NEXT: pand %xmm6, %xmm5 -; SSE2-NEXT: pandn %xmm4, %xmm6 -; SSE2-NEXT: por %xmm5, %xmm6 -; SSE2-NEXT: movdqa %xmm1, %xmm4 -; SSE2-NEXT: pand %xmm0, %xmm4 -; SSE2-NEXT: pandn %xmm3, %xmm0 -; SSE2-NEXT: por %xmm4, %xmm0 +; SSE2-NEXT: movdqa %xmm3, %xmm4 +; SSE2-NEXT: por %xmm5, %xmm4 +; SSE2-NEXT: pcmpgtd %xmm6, %xmm4 +; SSE2-NEXT: pand %xmm4, %xmm3 +; SSE2-NEXT: pandn %xmm2, %xmm4 +; SSE2-NEXT: por %xmm3, %xmm4 +; SSE2-NEXT: movdqa %xmm1, %xmm3 +; SSE2-NEXT: pxor %xmm5, %xmm3 +; SSE2-NEXT: por %xmm0, %xmm5 +; SSE2-NEXT: pcmpgtd %xmm3, %xmm5 +; SSE2-NEXT: pand %xmm5, %xmm0 +; SSE2-NEXT: pandn %xmm1, %xmm5 +; SSE2-NEXT: por %xmm5, %xmm0 ; SSE2-NEXT: psubd %xmm1, %xmm0 -; SSE2-NEXT: psubd %xmm2, %xmm6 -; SSE2-NEXT: pslld $16, %xmm6 -; SSE2-NEXT: psrad $16, %xmm6 +; SSE2-NEXT: psubd %xmm2, %xmm4 +; SSE2-NEXT: pslld $16, %xmm4 +; SSE2-NEXT: psrad $16, %xmm4 ; SSE2-NEXT: pslld $16, %xmm0 ; SSE2-NEXT: psrad $16, %xmm0 -; SSE2-NEXT: packssdw %xmm6, %xmm0 +; SSE2-NEXT: packssdw %xmm4, %xmm0 ; SSE2-NEXT: retq ; ; SSSE3-LABEL: psubus_8i32_max: ; SSSE3: # %bb.0: # %vector.ph ; SSSE3-NEXT: movdqa %xmm0, %xmm3 -; SSSE3-NEXT: pxor %xmm0, %xmm0 -; SSSE3-NEXT: movdqa %xmm3, %xmm4 -; SSSE3-NEXT: punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm0[4],xmm4[5],xmm0[5],xmm4[6],xmm0[6],xmm4[7],xmm0[7] -; SSSE3-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3] -; SSSE3-NEXT: movdqa {{.*#+}} xmm5 = [2147483648,2147483648,2147483648,2147483648] -; SSSE3-NEXT: movdqa %xmm1, %xmm0 -; SSSE3-NEXT: pxor %xmm5, %xmm0 +; SSSE3-NEXT: pxor %xmm4, %xmm4 +; SSSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3] +; SSSE3-NEXT: punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7] +; SSSE3-NEXT: movdqa {{.*#+}} xmm4 = [2147483648,2147483648,2147483648,2147483648] +; SSSE3-NEXT: movdqa %xmm2, %xmm5 +; SSSE3-NEXT: pxor %xmm4, %xmm5 ; SSSE3-NEXT: movdqa %xmm3, %xmm6 -; SSSE3-NEXT: por %xmm5, %xmm6 -; SSSE3-NEXT: pcmpgtd %xmm6, %xmm0 -; SSSE3-NEXT: movdqa %xmm2, %xmm6 -; SSSE3-NEXT: pxor %xmm5, %xmm6 -; SSSE3-NEXT: por %xmm4, %xmm5 +; SSSE3-NEXT: por %xmm4, %xmm6 ; SSSE3-NEXT: pcmpgtd %xmm5, %xmm6 -; SSSE3-NEXT: movdqa %xmm2, %xmm5 -; SSSE3-NEXT: pand %xmm6, %xmm5 -; SSSE3-NEXT: pandn %xmm4, %xmm6 -; SSSE3-NEXT: por %xmm5, %xmm6 -; SSSE3-NEXT: movdqa %xmm1, %xmm4 -; SSSE3-NEXT: pand %xmm0, %xmm4 -; SSSE3-NEXT: pandn %xmm3, %xmm0 +; SSSE3-NEXT: pand %xmm6, %xmm3 +; SSSE3-NEXT: pandn %xmm2, %xmm6 +; SSSE3-NEXT: por %xmm3, %xmm6 +; SSSE3-NEXT: movdqa %xmm1, %xmm3 +; SSSE3-NEXT: pxor %xmm4, %xmm3 +; SSSE3-NEXT: por %xmm0, %xmm4 +; SSSE3-NEXT: pcmpgtd %xmm3, %xmm4 +; SSSE3-NEXT: pand %xmm4, %xmm0 +; SSSE3-NEXT: pandn %xmm1, %xmm4 ; SSSE3-NEXT: por %xmm4, %xmm0 ; SSSE3-NEXT: psubd %xmm1, %xmm0 ; SSSE3-NEXT: psubd %xmm2, %xmm6 @@ -1565,265 +1381,258 @@ define <8 x i16> @psubus_8i64_max(<8 x i16> %x, <8 x i64> %y) nounwind { ; SSE2: # %bb.0: # %vector.ph ; SSE2-NEXT: pxor %xmm5, %xmm5 ; SSE2-NEXT: movdqa %xmm0, %xmm10 -; SSE2-NEXT: punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm5[0],xmm10[1],xmm5[1],xmm10[2],xmm5[2],xmm10[3],xmm5[3] -; SSE2-NEXT: movdqa %xmm10, %xmm9 -; SSE2-NEXT: punpckhdq {{.*#+}} xmm9 = xmm9[2],xmm5[2],xmm9[3],xmm5[3] -; SSE2-NEXT: punpckldq {{.*#+}} xmm10 = xmm10[0],xmm5[0],xmm10[1],xmm5[1] -; SSE2-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm5[4],xmm0[5],xmm5[5],xmm0[6],xmm5[6],xmm0[7],xmm5[7] -; SSE2-NEXT: movdqa %xmm0, %xmm8 -; SSE2-NEXT: punpckhdq {{.*#+}} xmm8 = xmm8[2],xmm5[2],xmm8[3],xmm5[3] -; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1] -; SSE2-NEXT: movdqa {{.*#+}} xmm6 = [2147483648,2147483648,2147483648,2147483648] -; SSE2-NEXT: movdqa %xmm3, %xmm5 -; SSE2-NEXT: pxor %xmm6, %xmm5 +; SSE2-NEXT: punpckhwd {{.*#+}} xmm10 = xmm10[4],xmm5[4],xmm10[5],xmm5[5],xmm10[6],xmm5[6],xmm10[7],xmm5[7] +; SSE2-NEXT: movdqa %xmm10, %xmm8 +; SSE2-NEXT: punpckldq {{.*#+}} xmm8 = xmm8[0],xmm5[0],xmm8[1],xmm5[1] +; SSE2-NEXT: punpckhdq {{.*#+}} xmm10 = xmm10[2],xmm5[2],xmm10[3],xmm5[3] +; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3] +; SSE2-NEXT: movdqa %xmm0, %xmm9 +; SSE2-NEXT: punpckldq {{.*#+}} xmm9 = xmm9[0],xmm5[0],xmm9[1],xmm5[1] +; SSE2-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm5[2],xmm0[3],xmm5[3] +; SSE2-NEXT: movdqa {{.*#+}} xmm11 = [2147483648,2147483648,2147483648,2147483648] +; SSE2-NEXT: movdqa %xmm2, %xmm6 +; SSE2-NEXT: pxor %xmm11, %xmm6 ; SSE2-NEXT: movdqa %xmm0, %xmm7 -; SSE2-NEXT: por %xmm6, %xmm7 -; SSE2-NEXT: movdqa %xmm5, %xmm11 -; SSE2-NEXT: pcmpgtd %xmm7, %xmm11 -; SSE2-NEXT: pshufd {{.*#+}} xmm12 = xmm11[0,0,2,2] -; SSE2-NEXT: pcmpeqd %xmm5, %xmm7 -; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm7[1,1,3,3] +; SSE2-NEXT: por %xmm11, %xmm7 +; SSE2-NEXT: movdqa %xmm7, %xmm5 +; SSE2-NEXT: pcmpgtd %xmm6, %xmm5 +; SSE2-NEXT: pshufd {{.*#+}} xmm12 = xmm5[0,0,2,2] +; SSE2-NEXT: pcmpeqd %xmm6, %xmm7 +; SSE2-NEXT: pshufd {{.*#+}} xmm7 = xmm7[1,1,3,3] +; SSE2-NEXT: pand %xmm12, %xmm7 +; SSE2-NEXT: pshufd {{.*#+}} xmm13 = xmm5[1,1,3,3] +; SSE2-NEXT: por %xmm7, %xmm13 +; SSE2-NEXT: pand %xmm13, %xmm0 +; SSE2-NEXT: pandn %xmm2, %xmm13 +; SSE2-NEXT: por %xmm0, %xmm13 +; SSE2-NEXT: movdqa %xmm1, %xmm0 +; SSE2-NEXT: pxor %xmm11, %xmm0 +; SSE2-NEXT: movdqa %xmm9, %xmm5 +; SSE2-NEXT: por %xmm11, %xmm5 +; SSE2-NEXT: movdqa %xmm5, %xmm7 +; SSE2-NEXT: pcmpgtd %xmm0, %xmm7 +; SSE2-NEXT: pshufd {{.*#+}} xmm12 = xmm7[0,0,2,2] +; SSE2-NEXT: pcmpeqd %xmm0, %xmm5 +; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm5[1,1,3,3] ; SSE2-NEXT: pand %xmm12, %xmm5 -; SSE2-NEXT: pshufd {{.*#+}} xmm11 = xmm11[1,1,3,3] -; SSE2-NEXT: por %xmm5, %xmm11 +; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm7[1,1,3,3] +; SSE2-NEXT: por %xmm5, %xmm0 +; SSE2-NEXT: pand %xmm0, %xmm9 +; SSE2-NEXT: pandn %xmm1, %xmm0 +; SSE2-NEXT: por %xmm9, %xmm0 ; SSE2-NEXT: movdqa %xmm4, %xmm5 -; SSE2-NEXT: pxor %xmm6, %xmm5 -; SSE2-NEXT: movdqa %xmm8, %xmm7 -; SSE2-NEXT: por %xmm6, %xmm7 -; SSE2-NEXT: movdqa %xmm5, %xmm12 -; SSE2-NEXT: pcmpgtd %xmm7, %xmm12 -; SSE2-NEXT: pshufd {{.*#+}} xmm13 = xmm12[0,0,2,2] -; SSE2-NEXT: pcmpeqd %xmm5, %xmm7 -; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm7[1,1,3,3] -; SSE2-NEXT: pand %xmm13, %xmm5 -; SSE2-NEXT: pshufd {{.*#+}} xmm12 = xmm12[1,1,3,3] -; SSE2-NEXT: por %xmm5, %xmm12 -; SSE2-NEXT: movdqa %xmm1, %xmm5 -; SSE2-NEXT: pxor %xmm6, %xmm5 +; SSE2-NEXT: pxor %xmm11, %xmm5 ; SSE2-NEXT: movdqa %xmm10, %xmm7 -; SSE2-NEXT: por %xmm6, %xmm7 -; SSE2-NEXT: movdqa %xmm5, %xmm13 -; SSE2-NEXT: pcmpgtd %xmm7, %xmm13 -; SSE2-NEXT: pshufd {{.*#+}} xmm14 = xmm13[0,0,2,2] +; SSE2-NEXT: por %xmm11, %xmm7 +; SSE2-NEXT: movdqa %xmm7, %xmm6 +; SSE2-NEXT: pcmpgtd %xmm5, %xmm6 +; SSE2-NEXT: pshufd {{.*#+}} xmm9 = xmm6[0,0,2,2] ; SSE2-NEXT: pcmpeqd %xmm5, %xmm7 -; SSE2-NEXT: pshufd {{.*#+}} xmm7 = xmm7[1,1,3,3] -; SSE2-NEXT: pand %xmm14, %xmm7 -; SSE2-NEXT: pshufd {{.*#+}} xmm13 = xmm13[1,1,3,3] -; SSE2-NEXT: por %xmm7, %xmm13 -; SSE2-NEXT: movdqa %xmm2, %xmm7 -; SSE2-NEXT: pxor %xmm6, %xmm7 -; SSE2-NEXT: por %xmm9, %xmm6 -; SSE2-NEXT: movdqa %xmm7, %xmm5 -; SSE2-NEXT: pcmpgtd %xmm6, %xmm5 -; SSE2-NEXT: pshufd {{.*#+}} xmm14 = xmm5[0,0,2,2] -; SSE2-NEXT: pcmpeqd %xmm7, %xmm6 +; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm7[1,1,3,3] +; SSE2-NEXT: pand %xmm9, %xmm5 ; SSE2-NEXT: pshufd {{.*#+}} xmm7 = xmm6[1,1,3,3] -; SSE2-NEXT: pand %xmm14, %xmm7 -; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm5[1,1,3,3] -; SSE2-NEXT: por %xmm7, %xmm6 -; SSE2-NEXT: movdqa %xmm2, %xmm5 -; SSE2-NEXT: pand %xmm6, %xmm5 -; SSE2-NEXT: pandn %xmm9, %xmm6 -; SSE2-NEXT: por %xmm5, %xmm6 -; SSE2-NEXT: movdqa %xmm1, %xmm5 -; SSE2-NEXT: pand %xmm13, %xmm5 -; SSE2-NEXT: pandn %xmm10, %xmm13 -; SSE2-NEXT: por %xmm5, %xmm13 -; SSE2-NEXT: movdqa %xmm4, %xmm5 -; SSE2-NEXT: pand %xmm12, %xmm5 -; SSE2-NEXT: pandn %xmm8, %xmm12 -; SSE2-NEXT: por %xmm5, %xmm12 +; SSE2-NEXT: por %xmm5, %xmm7 +; SSE2-NEXT: pand %xmm7, %xmm10 +; SSE2-NEXT: pandn %xmm4, %xmm7 +; SSE2-NEXT: por %xmm10, %xmm7 ; SSE2-NEXT: movdqa %xmm3, %xmm5 -; SSE2-NEXT: pand %xmm11, %xmm5 -; SSE2-NEXT: pandn %xmm0, %xmm11 -; SSE2-NEXT: por %xmm5, %xmm11 -; SSE2-NEXT: psubq %xmm3, %xmm11 -; SSE2-NEXT: psubq %xmm4, %xmm12 -; SSE2-NEXT: psubq %xmm1, %xmm13 -; SSE2-NEXT: psubq %xmm2, %xmm6 -; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm6[0,2,2,3] -; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7] +; SSE2-NEXT: pxor %xmm11, %xmm5 +; SSE2-NEXT: por %xmm8, %xmm11 +; SSE2-NEXT: movdqa %xmm11, %xmm6 +; SSE2-NEXT: pcmpgtd %xmm5, %xmm6 +; SSE2-NEXT: pshufd {{.*#+}} xmm9 = xmm6[0,0,2,2] +; SSE2-NEXT: pcmpeqd %xmm5, %xmm11 +; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm11[1,1,3,3] +; SSE2-NEXT: pand %xmm9, %xmm5 +; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm6[1,1,3,3] +; SSE2-NEXT: por %xmm5, %xmm6 +; SSE2-NEXT: pand %xmm6, %xmm8 +; SSE2-NEXT: pandn %xmm3, %xmm6 +; SSE2-NEXT: por %xmm8, %xmm6 +; SSE2-NEXT: psubq %xmm3, %xmm6 +; SSE2-NEXT: psubq %xmm4, %xmm7 +; SSE2-NEXT: psubq %xmm1, %xmm0 +; SSE2-NEXT: psubq %xmm2, %xmm13 ; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm13[0,2,2,3] ; SSE2-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,2,2,3,4,5,6,7] -; SSE2-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] -; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm12[0,2,2,3] -; SSE2-NEXT: pshuflw {{.*#+}} xmm2 = xmm0[0,1,0,2,4,5,6,7] -; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm11[0,2,2,3] +; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3] +; SSE2-NEXT: pshuflw {{.*#+}} xmm2 = xmm0[0,2,2,3,4,5,6,7] +; SSE2-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1] +; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm7[0,2,2,3] +; SSE2-NEXT: pshuflw {{.*#+}} xmm1 = xmm0[0,1,0,2,4,5,6,7] +; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm6[0,2,2,3] ; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,2,4,5,6,7] -; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1] -; SSE2-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] +; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] +; SSE2-NEXT: movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1] ; SSE2-NEXT: retq ; ; SSSE3-LABEL: psubus_8i64_max: ; SSSE3: # %bb.0: # %vector.ph ; SSSE3-NEXT: pxor %xmm5, %xmm5 ; SSSE3-NEXT: movdqa %xmm0, %xmm10 -; SSSE3-NEXT: punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm5[0],xmm10[1],xmm5[1],xmm10[2],xmm5[2],xmm10[3],xmm5[3] -; SSSE3-NEXT: movdqa %xmm10, %xmm9 -; SSSE3-NEXT: punpckhdq {{.*#+}} xmm9 = xmm9[2],xmm5[2],xmm9[3],xmm5[3] -; SSSE3-NEXT: punpckldq {{.*#+}} xmm10 = xmm10[0],xmm5[0],xmm10[1],xmm5[1] -; SSSE3-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm5[4],xmm0[5],xmm5[5],xmm0[6],xmm5[6],xmm0[7],xmm5[7] -; SSSE3-NEXT: movdqa %xmm0, %xmm8 -; SSSE3-NEXT: punpckhdq {{.*#+}} xmm8 = xmm8[2],xmm5[2],xmm8[3],xmm5[3] -; SSSE3-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1] -; SSSE3-NEXT: movdqa {{.*#+}} xmm6 = [2147483648,2147483648,2147483648,2147483648] -; SSSE3-NEXT: movdqa %xmm3, %xmm5 -; SSSE3-NEXT: pxor %xmm6, %xmm5 +; SSSE3-NEXT: punpckhwd {{.*#+}} xmm10 = xmm10[4],xmm5[4],xmm10[5],xmm5[5],xmm10[6],xmm5[6],xmm10[7],xmm5[7] +; SSSE3-NEXT: movdqa %xmm10, %xmm8 +; SSSE3-NEXT: punpckldq {{.*#+}} xmm8 = xmm8[0],xmm5[0],xmm8[1],xmm5[1] +; SSSE3-NEXT: punpckhdq {{.*#+}} xmm10 = xmm10[2],xmm5[2],xmm10[3],xmm5[3] +; SSSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3] +; SSSE3-NEXT: movdqa %xmm0, %xmm9 +; SSSE3-NEXT: punpckldq {{.*#+}} xmm9 = xmm9[0],xmm5[0],xmm9[1],xmm5[1] +; SSSE3-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm5[2],xmm0[3],xmm5[3] +; SSSE3-NEXT: movdqa {{.*#+}} xmm11 = [2147483648,2147483648,2147483648,2147483648] +; SSSE3-NEXT: movdqa %xmm2, %xmm6 +; SSSE3-NEXT: pxor %xmm11, %xmm6 ; SSSE3-NEXT: movdqa %xmm0, %xmm7 -; SSSE3-NEXT: por %xmm6, %xmm7 -; SSSE3-NEXT: movdqa %xmm5, %xmm11 -; SSSE3-NEXT: pcmpgtd %xmm7, %xmm11 -; SSSE3-NEXT: pshufd {{.*#+}} xmm12 = xmm11[0,0,2,2] -; SSSE3-NEXT: pcmpeqd %xmm5, %xmm7 -; SSSE3-NEXT: pshufd {{.*#+}} xmm5 = xmm7[1,1,3,3] +; SSSE3-NEXT: por %xmm11, %xmm7 +; SSSE3-NEXT: movdqa %xmm7, %xmm5 +; SSSE3-NEXT: pcmpgtd %xmm6, %xmm5 +; SSSE3-NEXT: pshufd {{.*#+}} xmm12 = xmm5[0,0,2,2] +; SSSE3-NEXT: pcmpeqd %xmm6, %xmm7 +; SSSE3-NEXT: pshufd {{.*#+}} xmm7 = xmm7[1,1,3,3] +; SSSE3-NEXT: pand %xmm12, %xmm7 +; SSSE3-NEXT: pshufd {{.*#+}} xmm13 = xmm5[1,1,3,3] +; SSSE3-NEXT: por %xmm7, %xmm13 +; SSSE3-NEXT: pand %xmm13, %xmm0 +; SSSE3-NEXT: pandn %xmm2, %xmm13 +; SSSE3-NEXT: por %xmm0, %xmm13 +; SSSE3-NEXT: movdqa %xmm1, %xmm0 +; SSSE3-NEXT: pxor %xmm11, %xmm0 +; SSSE3-NEXT: movdqa %xmm9, %xmm5 +; SSSE3-NEXT: por %xmm11, %xmm5 +; SSSE3-NEXT: movdqa %xmm5, %xmm7 +; SSSE3-NEXT: pcmpgtd %xmm0, %xmm7 +; SSSE3-NEXT: pshufd {{.*#+}} xmm12 = xmm7[0,0,2,2] +; SSSE3-NEXT: pcmpeqd %xmm0, %xmm5 +; SSSE3-NEXT: pshufd {{.*#+}} xmm5 = xmm5[1,1,3,3] ; SSSE3-NEXT: pand %xmm12, %xmm5 -; SSSE3-NEXT: pshufd {{.*#+}} xmm11 = xmm11[1,1,3,3] -; SSSE3-NEXT: por %xmm5, %xmm11 +; SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm7[1,1,3,3] +; SSSE3-NEXT: por %xmm5, %xmm0 +; SSSE3-NEXT: pand %xmm0, %xmm9 +; SSSE3-NEXT: pandn %xmm1, %xmm0 +; SSSE3-NEXT: por %xmm9, %xmm0 ; SSSE3-NEXT: movdqa %xmm4, %xmm5 -; SSSE3-NEXT: pxor %xmm6, %xmm5 -; SSSE3-NEXT: movdqa %xmm8, %xmm7 -; SSSE3-NEXT: por %xmm6, %xmm7 -; SSSE3-NEXT: movdqa %xmm5, %xmm12 -; SSSE3-NEXT: pcmpgtd %xmm7, %xmm12 -; SSSE3-NEXT: pshufd {{.*#+}} xmm13 = xmm12[0,0,2,2] -; SSSE3-NEXT: pcmpeqd %xmm5, %xmm7 -; SSSE3-NEXT: pshufd {{.*#+}} xmm5 = xmm7[1,1,3,3] -; SSSE3-NEXT: pand %xmm13, %xmm5 -; SSSE3-NEXT: pshufd {{.*#+}} xmm12 = xmm12[1,1,3,3] -; SSSE3-NEXT: por %xmm5, %xmm12 -; SSSE3-NEXT: movdqa %xmm1, %xmm5 -; SSSE3-NEXT: pxor %xmm6, %xmm5 +; SSSE3-NEXT: pxor %xmm11, %xmm5 ; SSSE3-NEXT: movdqa %xmm10, %xmm7 -; SSSE3-NEXT: por %xmm6, %xmm7 -; SSSE3-NEXT: movdqa %xmm5, %xmm13 -; SSSE3-NEXT: pcmpgtd %xmm7, %xmm13 -; SSSE3-NEXT: pshufd {{.*#+}} xmm14 = xmm13[0,0,2,2] +; SSSE3-NEXT: por %xmm11, %xmm7 +; SSSE3-NEXT: movdqa %xmm7, %xmm6 +; SSSE3-NEXT: pcmpgtd %xmm5, %xmm6 +; SSSE3-NEXT: pshufd {{.*#+}} xmm9 = xmm6[0,0,2,2] ; SSSE3-NEXT: pcmpeqd %xmm5, %xmm7 -; SSSE3-NEXT: pshufd {{.*#+}} xmm7 = xmm7[1,1,3,3] -; SSSE3-NEXT: pand %xmm14, %xmm7 -; SSSE3-NEXT: pshufd {{.*#+}} xmm13 = xmm13[1,1,3,3] -; SSSE3-NEXT: por %xmm7, %xmm13 -; SSSE3-NEXT: movdqa %xmm2, %xmm7 -; SSSE3-NEXT: pxor %xmm6, %xmm7 -; SSSE3-NEXT: por %xmm9, %xmm6 -; SSSE3-NEXT: movdqa %xmm7, %xmm5 -; SSSE3-NEXT: pcmpgtd %xmm6, %xmm5 -; SSSE3-NEXT: pshufd {{.*#+}} xmm14 = xmm5[0,0,2,2] -; SSSE3-NEXT: pcmpeqd %xmm7, %xmm6 +; SSSE3-NEXT: pshufd {{.*#+}} xmm5 = xmm7[1,1,3,3] +; SSSE3-NEXT: pand %xmm9, %xmm5 ; SSSE3-NEXT: pshufd {{.*#+}} xmm7 = xmm6[1,1,3,3] -; SSSE3-NEXT: pand %xmm14, %xmm7 -; SSSE3-NEXT: pshufd {{.*#+}} xmm6 = xmm5[1,1,3,3] -; SSSE3-NEXT: por %xmm7, %xmm6 -; SSSE3-NEXT: movdqa %xmm2, %xmm5 -; SSSE3-NEXT: pand %xmm6, %xmm5 -; SSSE3-NEXT: pandn %xmm9, %xmm6 -; SSSE3-NEXT: por %xmm5, %xmm6 -; SSSE3-NEXT: movdqa %xmm1, %xmm5 -; SSSE3-NEXT: pand %xmm13, %xmm5 -; SSSE3-NEXT: pandn %xmm10, %xmm13 -; SSSE3-NEXT: por %xmm5, %xmm13 -; SSSE3-NEXT: movdqa %xmm4, %xmm5 -; SSSE3-NEXT: pand %xmm12, %xmm5 -; SSSE3-NEXT: pandn %xmm8, %xmm12 -; SSSE3-NEXT: por %xmm5, %xmm12 +; SSSE3-NEXT: por %xmm5, %xmm7 +; SSSE3-NEXT: pand %xmm7, %xmm10 +; SSSE3-NEXT: pandn %xmm4, %xmm7 +; SSSE3-NEXT: por %xmm10, %xmm7 ; SSSE3-NEXT: movdqa %xmm3, %xmm5 -; SSSE3-NEXT: pand %xmm11, %xmm5 -; SSSE3-NEXT: pandn %xmm0, %xmm11 -; SSSE3-NEXT: por %xmm5, %xmm11 -; SSSE3-NEXT: psubq %xmm3, %xmm11 -; SSSE3-NEXT: psubq %xmm4, %xmm12 -; SSSE3-NEXT: psubq %xmm1, %xmm13 -; SSSE3-NEXT: psubq %xmm2, %xmm6 -; SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm6[0,2,2,3] -; SSSE3-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7] +; SSSE3-NEXT: pxor %xmm11, %xmm5 +; SSSE3-NEXT: por %xmm8, %xmm11 +; SSSE3-NEXT: movdqa %xmm11, %xmm6 +; SSSE3-NEXT: pcmpgtd %xmm5, %xmm6 +; SSSE3-NEXT: pshufd {{.*#+}} xmm9 = xmm6[0,0,2,2] +; SSSE3-NEXT: pcmpeqd %xmm5, %xmm11 +; SSSE3-NEXT: pshufd {{.*#+}} xmm5 = xmm11[1,1,3,3] +; SSSE3-NEXT: pand %xmm9, %xmm5 +; SSSE3-NEXT: pshufd {{.*#+}} xmm6 = xmm6[1,1,3,3] +; SSSE3-NEXT: por %xmm5, %xmm6 +; SSSE3-NEXT: pand %xmm6, %xmm8 +; SSSE3-NEXT: pandn %xmm3, %xmm6 +; SSSE3-NEXT: por %xmm8, %xmm6 +; SSSE3-NEXT: psubq %xmm3, %xmm6 +; SSSE3-NEXT: psubq %xmm4, %xmm7 +; SSSE3-NEXT: psubq %xmm1, %xmm0 +; SSSE3-NEXT: psubq %xmm2, %xmm13 ; SSSE3-NEXT: pshufd {{.*#+}} xmm1 = xmm13[0,2,2,3] ; SSSE3-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,2,2,3,4,5,6,7] -; SSSE3-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] -; SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm12[0,2,2,3] -; SSSE3-NEXT: pshuflw {{.*#+}} xmm2 = xmm0[0,1,0,2,4,5,6,7] -; SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm11[0,2,2,3] +; SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3] +; SSSE3-NEXT: pshuflw {{.*#+}} xmm2 = xmm0[0,2,2,3,4,5,6,7] +; SSSE3-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1] +; SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm7[0,2,2,3] +; SSSE3-NEXT: pshuflw {{.*#+}} xmm1 = xmm0[0,1,0,2,4,5,6,7] +; SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm6[0,2,2,3] ; SSSE3-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,2,4,5,6,7] -; SSSE3-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1] -; SSSE3-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] +; SSSE3-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] +; SSSE3-NEXT: movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1] ; SSSE3-NEXT: retq ; ; SSE41-LABEL: psubus_8i64_max: ; SSE41: # %bb.0: # %vector.ph -; SSE41-NEXT: pshufd {{.*#+}} xmm5 = xmm0[3,1,2,3] -; SSE41-NEXT: pmovzxwq {{.*#+}} xmm11 = xmm5[0],zero,zero,zero,xmm5[1],zero,zero,zero -; SSE41-NEXT: pshufd {{.*#+}} xmm5 = xmm0[2,3,0,1] -; SSE41-NEXT: pmovzxwq {{.*#+}} xmm12 = xmm5[0],zero,zero,zero,xmm5[1],zero,zero,zero -; SSE41-NEXT: pmovzxwq {{.*#+}} xmm13 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero -; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,2,3] +; SSE41-NEXT: pshufd {{.*#+}} xmm5 = xmm0[1,1,2,3] +; SSE41-NEXT: pmovzxwq {{.*#+}} xmm8 = xmm5[0],zero,zero,zero,xmm5[1],zero,zero,zero ; SSE41-NEXT: pmovzxwq {{.*#+}} xmm10 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero -; SSE41-NEXT: movdqa {{.*#+}} xmm0 = [2147483648,2147483648,2147483648,2147483648] -; SSE41-NEXT: movdqa %xmm2, %xmm6 -; SSE41-NEXT: pxor %xmm0, %xmm6 -; SSE41-NEXT: movdqa %xmm10, %xmm7 -; SSE41-NEXT: por %xmm0, %xmm7 -; SSE41-NEXT: movdqa %xmm6, %xmm5 -; SSE41-NEXT: pcmpgtd %xmm7, %xmm5 -; SSE41-NEXT: pshufd {{.*#+}} xmm8 = xmm5[0,0,2,2] -; SSE41-NEXT: pcmpeqd %xmm6, %xmm7 -; SSE41-NEXT: pshufd {{.*#+}} xmm6 = xmm7[1,1,3,3] -; SSE41-NEXT: pand %xmm8, %xmm6 -; SSE41-NEXT: pshufd {{.*#+}} xmm8 = xmm5[1,1,3,3] -; SSE41-NEXT: por %xmm6, %xmm8 -; SSE41-NEXT: movdqa %xmm1, %xmm5 -; SSE41-NEXT: pxor %xmm0, %xmm5 -; SSE41-NEXT: movdqa %xmm13, %xmm6 -; SSE41-NEXT: por %xmm0, %xmm6 -; SSE41-NEXT: movdqa %xmm5, %xmm7 -; SSE41-NEXT: pcmpgtd %xmm6, %xmm7 -; SSE41-NEXT: pshufd {{.*#+}} xmm9 = xmm7[0,0,2,2] -; SSE41-NEXT: pcmpeqd %xmm5, %xmm6 -; SSE41-NEXT: pshufd {{.*#+}} xmm5 = xmm6[1,1,3,3] -; SSE41-NEXT: pand %xmm9, %xmm5 -; SSE41-NEXT: pshufd {{.*#+}} xmm9 = xmm7[1,1,3,3] -; SSE41-NEXT: por %xmm5, %xmm9 -; SSE41-NEXT: movdqa %xmm3, %xmm5 -; SSE41-NEXT: pxor %xmm0, %xmm5 -; SSE41-NEXT: movdqa %xmm12, %xmm6 -; SSE41-NEXT: por %xmm0, %xmm6 +; SSE41-NEXT: pshufd {{.*#+}} xmm5 = xmm0[2,3,0,1] +; SSE41-NEXT: pmovzxwq {{.*#+}} xmm11 = xmm5[0],zero,zero,zero,xmm5[1],zero,zero,zero +; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,1,2,3] +; SSE41-NEXT: pmovzxwq {{.*#+}} xmm12 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero +; SSE41-NEXT: movdqa {{.*#+}} xmm6 = [2147483648,2147483648,2147483648,2147483648] +; SSE41-NEXT: movdqa %xmm4, %xmm0 +; SSE41-NEXT: pxor %xmm6, %xmm0 +; SSE41-NEXT: movdqa %xmm12, %xmm7 +; SSE41-NEXT: por %xmm6, %xmm7 +; SSE41-NEXT: movdqa %xmm7, %xmm5 +; SSE41-NEXT: pcmpgtd %xmm0, %xmm5 +; SSE41-NEXT: pshufd {{.*#+}} xmm9 = xmm5[0,0,2,2] +; SSE41-NEXT: pcmpeqd %xmm0, %xmm7 +; SSE41-NEXT: pshufd {{.*#+}} xmm7 = xmm7[1,1,3,3] +; SSE41-NEXT: pand %xmm9, %xmm7 +; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm5[1,1,3,3] +; SSE41-NEXT: por %xmm7, %xmm0 +; SSE41-NEXT: movdqa %xmm4, %xmm9 +; SSE41-NEXT: blendvpd %xmm0, %xmm12, %xmm9 +; SSE41-NEXT: movdqa %xmm3, %xmm0 +; SSE41-NEXT: pxor %xmm6, %xmm0 +; SSE41-NEXT: movdqa %xmm11, %xmm5 +; SSE41-NEXT: por %xmm6, %xmm5 ; SSE41-NEXT: movdqa %xmm5, %xmm7 -; SSE41-NEXT: pcmpgtd %xmm6, %xmm7 -; SSE41-NEXT: pshufd {{.*#+}} xmm14 = xmm7[0,0,2,2] -; SSE41-NEXT: pcmpeqd %xmm5, %xmm6 -; SSE41-NEXT: pshufd {{.*#+}} xmm5 = xmm6[1,1,3,3] -; SSE41-NEXT: pand %xmm14, %xmm5 -; SSE41-NEXT: pshufd {{.*#+}} xmm6 = xmm7[1,1,3,3] -; SSE41-NEXT: por %xmm5, %xmm6 -; SSE41-NEXT: movdqa %xmm4, %xmm5 -; SSE41-NEXT: pxor %xmm0, %xmm5 -; SSE41-NEXT: por %xmm11, %xmm0 +; SSE41-NEXT: pcmpgtd %xmm0, %xmm7 +; SSE41-NEXT: pshufd {{.*#+}} xmm12 = xmm7[0,0,2,2] +; SSE41-NEXT: pcmpeqd %xmm0, %xmm5 +; SSE41-NEXT: pshufd {{.*#+}} xmm5 = xmm5[1,1,3,3] +; SSE41-NEXT: pand %xmm12, %xmm5 +; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm7[1,1,3,3] +; SSE41-NEXT: por %xmm5, %xmm0 +; SSE41-NEXT: movdqa %xmm3, %xmm12 +; SSE41-NEXT: blendvpd %xmm0, %xmm11, %xmm12 +; SSE41-NEXT: movdqa %xmm1, %xmm0 +; SSE41-NEXT: pxor %xmm6, %xmm0 +; SSE41-NEXT: movdqa %xmm10, %xmm5 +; SSE41-NEXT: por %xmm6, %xmm5 ; SSE41-NEXT: movdqa %xmm5, %xmm7 ; SSE41-NEXT: pcmpgtd %xmm0, %xmm7 -; SSE41-NEXT: pshufd {{.*#+}} xmm14 = xmm7[0,0,2,2] -; SSE41-NEXT: pcmpeqd %xmm5, %xmm0 -; SSE41-NEXT: pshufd {{.*#+}} xmm5 = xmm0[1,1,3,3] -; SSE41-NEXT: pand %xmm14, %xmm5 +; SSE41-NEXT: pshufd {{.*#+}} xmm11 = xmm7[0,0,2,2] +; SSE41-NEXT: pcmpeqd %xmm0, %xmm5 +; SSE41-NEXT: pshufd {{.*#+}} xmm5 = xmm5[1,1,3,3] +; SSE41-NEXT: pand %xmm11, %xmm5 ; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm7[1,1,3,3] ; SSE41-NEXT: por %xmm5, %xmm0 -; SSE41-NEXT: blendvpd %xmm0, %xmm4, %xmm11 -; SSE41-NEXT: movdqa %xmm6, %xmm0 -; SSE41-NEXT: blendvpd %xmm0, %xmm3, %xmm12 -; SSE41-NEXT: movdqa %xmm9, %xmm0 -; SSE41-NEXT: blendvpd %xmm0, %xmm1, %xmm13 -; SSE41-NEXT: movdqa %xmm8, %xmm0 -; SSE41-NEXT: blendvpd %xmm0, %xmm2, %xmm10 -; SSE41-NEXT: psubq %xmm2, %xmm10 -; SSE41-NEXT: psubq %xmm1, %xmm13 +; SSE41-NEXT: movdqa %xmm1, %xmm5 +; SSE41-NEXT: blendvpd %xmm0, %xmm10, %xmm5 +; SSE41-NEXT: movdqa %xmm2, %xmm0 +; SSE41-NEXT: pxor %xmm6, %xmm0 +; SSE41-NEXT: por %xmm8, %xmm6 +; SSE41-NEXT: movdqa %xmm6, %xmm7 +; SSE41-NEXT: pcmpgtd %xmm0, %xmm7 +; SSE41-NEXT: pshufd {{.*#+}} xmm10 = xmm7[0,0,2,2] +; SSE41-NEXT: pcmpeqd %xmm0, %xmm6 +; SSE41-NEXT: pshufd {{.*#+}} xmm6 = xmm6[1,1,3,3] +; SSE41-NEXT: pand %xmm10, %xmm6 +; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm7[1,1,3,3] +; SSE41-NEXT: por %xmm6, %xmm0 +; SSE41-NEXT: movdqa %xmm2, %xmm6 +; SSE41-NEXT: blendvpd %xmm0, %xmm8, %xmm6 +; SSE41-NEXT: psubq %xmm2, %xmm6 +; SSE41-NEXT: psubq %xmm1, %xmm5 ; SSE41-NEXT: psubq %xmm3, %xmm12 -; SSE41-NEXT: psubq %xmm4, %xmm11 +; SSE41-NEXT: psubq %xmm4, %xmm9 ; SSE41-NEXT: pxor %xmm0, %xmm0 -; SSE41-NEXT: pblendw {{.*#+}} xmm11 = xmm11[0],xmm0[1,2,3],xmm11[4],xmm0[5,6,7] +; SSE41-NEXT: pblendw {{.*#+}} xmm9 = xmm9[0],xmm0[1,2,3],xmm9[4],xmm0[5,6,7] ; SSE41-NEXT: pblendw {{.*#+}} xmm12 = xmm12[0],xmm0[1,2,3],xmm12[4],xmm0[5,6,7] -; SSE41-NEXT: packusdw %xmm11, %xmm12 -; SSE41-NEXT: pblendw {{.*#+}} xmm13 = xmm13[0],xmm0[1,2,3],xmm13[4],xmm0[5,6,7] -; SSE41-NEXT: pblendw {{.*#+}} xmm10 = xmm10[0],xmm0[1,2,3],xmm10[4],xmm0[5,6,7] -; SSE41-NEXT: packusdw %xmm10, %xmm13 -; SSE41-NEXT: packusdw %xmm12, %xmm13 -; SSE41-NEXT: movdqa %xmm13, %xmm0 +; SSE41-NEXT: packusdw %xmm9, %xmm12 +; SSE41-NEXT: pblendw {{.*#+}} xmm5 = xmm5[0],xmm0[1,2,3],xmm5[4],xmm0[5,6,7] +; SSE41-NEXT: pblendw {{.*#+}} xmm6 = xmm6[0],xmm0[1,2,3],xmm6[4],xmm0[5,6,7] +; SSE41-NEXT: packusdw %xmm6, %xmm5 +; SSE41-NEXT: packusdw %xmm12, %xmm5 +; SSE41-NEXT: movdqa %xmm5, %xmm0 ; SSE41-NEXT: retq ; ; AVX1-LABEL: psubus_8i64_max: @@ -1837,31 +1646,31 @@ define <8 x i16> @psubus_8i64_max(<8 x i16> %x, <8 x i64> %y) nounwind { ; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[1,1,2,3] ; AVX1-NEXT: vpmovzxwq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm6, %ymm9 -; AVX1-NEXT: vextractf128 $1, %ymm2, %xmm10 +; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm10 ; AVX1-NEXT: vmovdqa {{.*#+}} xmm7 = [9223372036854775808,9223372036854775808] ; AVX1-NEXT: vpxor %xmm7, %xmm10, %xmm5 -; AVX1-NEXT: vpor %xmm7, %xmm4, %xmm4 -; AVX1-NEXT: vpcmpgtq %xmm4, %xmm5, %xmm4 -; AVX1-NEXT: vpxor %xmm7, %xmm2, %xmm5 -; AVX1-NEXT: vpor %xmm7, %xmm3, %xmm3 -; AVX1-NEXT: vpcmpgtq %xmm3, %xmm5, %xmm3 -; AVX1-NEXT: vinsertf128 $1, %xmm4, %ymm3, %ymm3 -; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm4 -; AVX1-NEXT: vpxor %xmm7, %xmm4, %xmm5 ; AVX1-NEXT: vpor %xmm7, %xmm0, %xmm0 -; AVX1-NEXT: vpcmpgtq %xmm0, %xmm5, %xmm0 -; AVX1-NEXT: vpor %xmm7, %xmm6, %xmm5 -; AVX1-NEXT: vpxor %xmm7, %xmm1, %xmm6 +; AVX1-NEXT: vpcmpgtq %xmm5, %xmm0, %xmm0 +; AVX1-NEXT: vpxor %xmm7, %xmm1, %xmm5 +; AVX1-NEXT: vpor %xmm7, %xmm6, %xmm6 ; AVX1-NEXT: vpcmpgtq %xmm5, %xmm6, %xmm5 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm5, %ymm0 -; AVX1-NEXT: vblendvpd %ymm0, %ymm1, %ymm9, %ymm0 -; AVX1-NEXT: vblendvpd %ymm3, %ymm2, %ymm8, %ymm3 +; AVX1-NEXT: vblendvpd %ymm0, %ymm9, %ymm1, %ymm0 +; AVX1-NEXT: vextractf128 $1, %ymm2, %xmm5 +; AVX1-NEXT: vpxor %xmm7, %xmm5, %xmm6 +; AVX1-NEXT: vpor %xmm7, %xmm4, %xmm4 +; AVX1-NEXT: vpcmpgtq %xmm6, %xmm4, %xmm4 +; AVX1-NEXT: vpxor %xmm7, %xmm2, %xmm6 +; AVX1-NEXT: vpor %xmm7, %xmm3, %xmm3 +; AVX1-NEXT: vpcmpgtq %xmm6, %xmm3, %xmm3 +; AVX1-NEXT: vinsertf128 $1, %xmm4, %ymm3, %ymm3 +; AVX1-NEXT: vblendvpd %ymm3, %ymm8, %ymm2, %ymm3 ; AVX1-NEXT: vpsubq %xmm2, %xmm3, %xmm2 ; AVX1-NEXT: vextractf128 $1, %ymm3, %xmm3 -; AVX1-NEXT: vpsubq %xmm10, %xmm3, %xmm3 +; AVX1-NEXT: vpsubq %xmm5, %xmm3, %xmm3 ; AVX1-NEXT: vpsubq %xmm1, %xmm0, %xmm1 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0 -; AVX1-NEXT: vpsubq %xmm4, %xmm0, %xmm0 +; AVX1-NEXT: vpsubq %xmm10, %xmm0, %xmm0 ; AVX1-NEXT: vpxor %xmm4, %xmm4, %xmm4 ; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm4[1,2,3],xmm0[4],xmm4[5,6,7] ; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0],xmm4[1,2,3],xmm1[4],xmm4[5,6,7] @@ -1875,25 +1684,25 @@ define <8 x i16> @psubus_8i64_max(<8 x i16> %x, <8 x i64> %y) nounwind { ; ; AVX2-SLOW-LABEL: psubus_8i64_max: ; AVX2-SLOW: # %bb.0: # %vector.ph -; AVX2-SLOW-NEXT: vpmovzxwq {{.*#+}} ymm3 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero -; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1] +; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[2,3,0,1] +; AVX2-SLOW-NEXT: vpmovzxwq {{.*#+}} ymm3 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero,xmm3[2],zero,zero,zero,xmm3[3],zero,zero,zero ; AVX2-SLOW-NEXT: vpmovzxwq {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero ; AVX2-SLOW-NEXT: vpbroadcastq {{.*#+}} ymm4 = [9223372036854775808,9223372036854775808,9223372036854775808,9223372036854775808] -; AVX2-SLOW-NEXT: vpxor %ymm4, %ymm2, %ymm5 +; AVX2-SLOW-NEXT: vpxor %ymm4, %ymm1, %ymm5 ; AVX2-SLOW-NEXT: vpor %ymm4, %ymm0, %ymm6 -; AVX2-SLOW-NEXT: vpcmpgtq %ymm6, %ymm5, %ymm5 -; AVX2-SLOW-NEXT: vpxor %ymm4, %ymm1, %ymm6 +; AVX2-SLOW-NEXT: vpcmpgtq %ymm5, %ymm6, %ymm5 +; AVX2-SLOW-NEXT: vblendvpd %ymm5, %ymm0, %ymm1, %ymm0 +; AVX2-SLOW-NEXT: vpxor %ymm4, %ymm2, %ymm5 ; AVX2-SLOW-NEXT: vpor %ymm4, %ymm3, %ymm4 -; AVX2-SLOW-NEXT: vpcmpgtq %ymm4, %ymm6, %ymm4 -; AVX2-SLOW-NEXT: vblendvpd %ymm4, %ymm1, %ymm3, %ymm3 -; AVX2-SLOW-NEXT: vblendvpd %ymm5, %ymm2, %ymm0, %ymm0 -; AVX2-SLOW-NEXT: vpsubq %ymm2, %ymm0, %ymm0 -; AVX2-SLOW-NEXT: vpsubq %ymm1, %ymm3, %ymm1 -; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7] -; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3] +; AVX2-SLOW-NEXT: vpcmpgtq %ymm5, %ymm4, %ymm4 +; AVX2-SLOW-NEXT: vblendvpd %ymm4, %ymm3, %ymm2, %ymm3 +; AVX2-SLOW-NEXT: vpsubq %ymm2, %ymm3, %ymm2 +; AVX2-SLOW-NEXT: vpsubq %ymm1, %ymm0, %ymm0 ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3] -; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 +; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm2[0,2,2,3,4,6,6,7] +; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3] +; AVX2-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 ; AVX2-SLOW-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3] ; AVX2-SLOW-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0 @@ -1902,24 +1711,24 @@ define <8 x i16> @psubus_8i64_max(<8 x i16> %x, <8 x i64> %y) nounwind { ; ; AVX2-FAST-LABEL: psubus_8i64_max: ; AVX2-FAST: # %bb.0: # %vector.ph -; AVX2-FAST-NEXT: vpmovzxwq {{.*#+}} ymm3 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero -; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1] +; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[2,3,0,1] +; AVX2-FAST-NEXT: vpmovzxwq {{.*#+}} ymm3 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero,xmm3[2],zero,zero,zero,xmm3[3],zero,zero,zero ; AVX2-FAST-NEXT: vpmovzxwq {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero ; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} ymm4 = [9223372036854775808,9223372036854775808,9223372036854775808,9223372036854775808] -; AVX2-FAST-NEXT: vpxor %ymm4, %ymm2, %ymm5 +; AVX2-FAST-NEXT: vpxor %ymm4, %ymm1, %ymm5 ; AVX2-FAST-NEXT: vpor %ymm4, %ymm0, %ymm6 -; AVX2-FAST-NEXT: vpcmpgtq %ymm6, %ymm5, %ymm5 -; AVX2-FAST-NEXT: vpxor %ymm4, %ymm1, %ymm6 +; AVX2-FAST-NEXT: vpcmpgtq %ymm5, %ymm6, %ymm5 +; AVX2-FAST-NEXT: vblendvpd %ymm5, %ymm0, %ymm1, %ymm0 +; AVX2-FAST-NEXT: vpxor %ymm4, %ymm2, %ymm5 ; AVX2-FAST-NEXT: vpor %ymm4, %ymm3, %ymm4 -; AVX2-FAST-NEXT: vpcmpgtq %ymm4, %ymm6, %ymm4 -; AVX2-FAST-NEXT: vblendvpd %ymm4, %ymm1, %ymm3, %ymm3 -; AVX2-FAST-NEXT: vblendvpd %ymm5, %ymm2, %ymm0, %ymm0 -; AVX2-FAST-NEXT: vpsubq %ymm2, %ymm0, %ymm0 -; AVX2-FAST-NEXT: vpsubq %ymm1, %ymm3, %ymm1 -; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = [0,2,4,6,4,6,6,7] -; AVX2-FAST-NEXT: vpermd %ymm1, %ymm2, %ymm1 -; AVX2-FAST-NEXT: vpermd %ymm0, %ymm2, %ymm0 -; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 +; AVX2-FAST-NEXT: vpcmpgtq %ymm5, %ymm4, %ymm4 +; AVX2-FAST-NEXT: vblendvpd %ymm4, %ymm3, %ymm2, %ymm3 +; AVX2-FAST-NEXT: vpsubq %ymm2, %ymm3, %ymm2 +; AVX2-FAST-NEXT: vpsubq %ymm1, %ymm0, %ymm0 +; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7] +; AVX2-FAST-NEXT: vpermd %ymm0, %ymm1, %ymm0 +; AVX2-FAST-NEXT: vpermd %ymm2, %ymm1, %ymm1 +; AVX2-FAST-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 ; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31] ; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3] ; AVX2-FAST-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0 @@ -1945,127 +1754,115 @@ define <16 x i16> @psubus_16i32_max(<16 x i16> %x, <16 x i32> %y) nounwind { ; SSE2-LABEL: psubus_16i32_max: ; SSE2: # %bb.0: # %vector.ph ; SSE2-NEXT: movdqa %xmm1, %xmm8 -; SSE2-NEXT: movdqa %xmm0, %xmm9 -; SSE2-NEXT: pxor %xmm0, %xmm0 -; SSE2-NEXT: movdqa %xmm9, %xmm11 -; SSE2-NEXT: punpckhwd {{.*#+}} xmm11 = xmm11[4],xmm0[4],xmm11[5],xmm0[5],xmm11[6],xmm0[6],xmm11[7],xmm0[7] -; SSE2-NEXT: punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1],xmm9[2],xmm0[2],xmm9[3],xmm0[3] -; SSE2-NEXT: movdqa %xmm1, %xmm10 -; SSE2-NEXT: punpckhwd {{.*#+}} xmm10 = xmm10[4],xmm0[4],xmm10[5],xmm0[5],xmm10[6],xmm0[6],xmm10[7],xmm0[7] -; SSE2-NEXT: punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1],xmm8[2],xmm0[2],xmm8[3],xmm0[3] -; SSE2-NEXT: movdqa {{.*#+}} xmm6 = [2147483648,2147483648,2147483648,2147483648] -; SSE2-NEXT: movdqa %xmm4, %xmm1 -; SSE2-NEXT: pxor %xmm6, %xmm1 -; SSE2-NEXT: movdqa %xmm8, %xmm0 -; SSE2-NEXT: por %xmm6, %xmm0 -; SSE2-NEXT: pcmpgtd %xmm0, %xmm1 -; SSE2-NEXT: movdqa %xmm5, %xmm12 -; SSE2-NEXT: pxor %xmm6, %xmm12 -; SSE2-NEXT: movdqa %xmm10, %xmm0 -; SSE2-NEXT: por %xmm6, %xmm0 -; SSE2-NEXT: pcmpgtd %xmm0, %xmm12 -; SSE2-NEXT: movdqa %xmm2, %xmm0 -; SSE2-NEXT: pxor %xmm6, %xmm0 -; SSE2-NEXT: movdqa %xmm9, %xmm7 -; SSE2-NEXT: por %xmm6, %xmm7 -; SSE2-NEXT: pcmpgtd %xmm7, %xmm0 -; SSE2-NEXT: movdqa %xmm3, %xmm7 -; SSE2-NEXT: pxor %xmm6, %xmm7 -; SSE2-NEXT: por %xmm11, %xmm6 -; SSE2-NEXT: pcmpgtd %xmm6, %xmm7 +; SSE2-NEXT: pxor %xmm7, %xmm7 +; SSE2-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1],xmm1[2],xmm7[2],xmm1[3],xmm7[3] +; SSE2-NEXT: punpckhwd {{.*#+}} xmm8 = xmm8[4],xmm7[4],xmm8[5],xmm7[5],xmm8[6],xmm7[6],xmm8[7],xmm7[7] +; SSE2-NEXT: movdqa %xmm0, %xmm10 +; SSE2-NEXT: punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm7[0],xmm10[1],xmm7[1],xmm10[2],xmm7[2],xmm10[3],xmm7[3] +; SSE2-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm7[4],xmm0[5],xmm7[5],xmm0[6],xmm7[6],xmm0[7],xmm7[7] +; SSE2-NEXT: movdqa {{.*#+}} xmm7 = [2147483648,2147483648,2147483648,2147483648] ; SSE2-NEXT: movdqa %xmm3, %xmm6 -; SSE2-NEXT: pand %xmm7, %xmm6 -; SSE2-NEXT: pandn %xmm11, %xmm7 -; SSE2-NEXT: por %xmm6, %xmm7 +; SSE2-NEXT: pxor %xmm7, %xmm6 +; SSE2-NEXT: movdqa %xmm0, %xmm9 +; SSE2-NEXT: por %xmm7, %xmm9 +; SSE2-NEXT: pcmpgtd %xmm6, %xmm9 +; SSE2-NEXT: pand %xmm9, %xmm0 +; SSE2-NEXT: pandn %xmm3, %xmm9 +; SSE2-NEXT: por %xmm0, %xmm9 ; SSE2-NEXT: movdqa %xmm2, %xmm6 -; SSE2-NEXT: pand %xmm0, %xmm6 -; SSE2-NEXT: pandn %xmm9, %xmm0 -; SSE2-NEXT: por %xmm6, %xmm0 -; SSE2-NEXT: movdqa %xmm5, %xmm6 -; SSE2-NEXT: pand %xmm12, %xmm6 -; SSE2-NEXT: pandn %xmm10, %xmm12 -; SSE2-NEXT: por %xmm6, %xmm12 -; SSE2-NEXT: movdqa %xmm4, %xmm6 -; SSE2-NEXT: pand %xmm1, %xmm6 -; SSE2-NEXT: pandn %xmm8, %xmm1 -; SSE2-NEXT: por %xmm6, %xmm1 +; SSE2-NEXT: pxor %xmm7, %xmm6 +; SSE2-NEXT: movdqa %xmm10, %xmm0 +; SSE2-NEXT: por %xmm7, %xmm0 +; SSE2-NEXT: pcmpgtd %xmm6, %xmm0 +; SSE2-NEXT: pand %xmm0, %xmm10 +; SSE2-NEXT: pandn %xmm2, %xmm0 +; SSE2-NEXT: por %xmm10, %xmm0 +; SSE2-NEXT: movdqa %xmm5, %xmm10 +; SSE2-NEXT: pxor %xmm7, %xmm10 +; SSE2-NEXT: movdqa %xmm8, %xmm6 +; SSE2-NEXT: por %xmm7, %xmm6 +; SSE2-NEXT: pcmpgtd %xmm10, %xmm6 +; SSE2-NEXT: pand %xmm6, %xmm8 +; SSE2-NEXT: pandn %xmm5, %xmm6 +; SSE2-NEXT: por %xmm8, %xmm6 +; SSE2-NEXT: movdqa %xmm4, %xmm8 +; SSE2-NEXT: pxor %xmm7, %xmm8 +; SSE2-NEXT: por %xmm1, %xmm7 +; SSE2-NEXT: pcmpgtd %xmm8, %xmm7 +; SSE2-NEXT: pand %xmm7, %xmm1 +; SSE2-NEXT: pandn %xmm4, %xmm7 +; SSE2-NEXT: por %xmm7, %xmm1 ; SSE2-NEXT: psubd %xmm4, %xmm1 -; SSE2-NEXT: psubd %xmm5, %xmm12 +; SSE2-NEXT: psubd %xmm5, %xmm6 ; SSE2-NEXT: psubd %xmm2, %xmm0 -; SSE2-NEXT: psubd %xmm3, %xmm7 -; SSE2-NEXT: pslld $16, %xmm7 -; SSE2-NEXT: psrad $16, %xmm7 +; SSE2-NEXT: psubd %xmm3, %xmm9 +; SSE2-NEXT: pslld $16, %xmm9 +; SSE2-NEXT: psrad $16, %xmm9 ; SSE2-NEXT: pslld $16, %xmm0 ; SSE2-NEXT: psrad $16, %xmm0 -; SSE2-NEXT: packssdw %xmm7, %xmm0 -; SSE2-NEXT: pslld $16, %xmm12 -; SSE2-NEXT: psrad $16, %xmm12 +; SSE2-NEXT: packssdw %xmm9, %xmm0 +; SSE2-NEXT: pslld $16, %xmm6 +; SSE2-NEXT: psrad $16, %xmm6 ; SSE2-NEXT: pslld $16, %xmm1 ; SSE2-NEXT: psrad $16, %xmm1 -; SSE2-NEXT: packssdw %xmm12, %xmm1 +; SSE2-NEXT: packssdw %xmm6, %xmm1 ; SSE2-NEXT: retq ; ; SSSE3-LABEL: psubus_16i32_max: ; SSSE3: # %bb.0: # %vector.ph ; SSSE3-NEXT: movdqa %xmm1, %xmm8 -; SSSE3-NEXT: movdqa %xmm0, %xmm9 -; SSSE3-NEXT: pxor %xmm0, %xmm0 -; SSSE3-NEXT: movdqa %xmm9, %xmm11 -; SSSE3-NEXT: punpckhwd {{.*#+}} xmm11 = xmm11[4],xmm0[4],xmm11[5],xmm0[5],xmm11[6],xmm0[6],xmm11[7],xmm0[7] -; SSSE3-NEXT: punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1],xmm9[2],xmm0[2],xmm9[3],xmm0[3] -; SSSE3-NEXT: movdqa %xmm1, %xmm10 -; SSSE3-NEXT: punpckhwd {{.*#+}} xmm10 = xmm10[4],xmm0[4],xmm10[5],xmm0[5],xmm10[6],xmm0[6],xmm10[7],xmm0[7] -; SSSE3-NEXT: punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1],xmm8[2],xmm0[2],xmm8[3],xmm0[3] -; SSSE3-NEXT: movdqa {{.*#+}} xmm6 = [2147483648,2147483648,2147483648,2147483648] -; SSSE3-NEXT: movdqa %xmm4, %xmm1 -; SSSE3-NEXT: pxor %xmm6, %xmm1 -; SSSE3-NEXT: movdqa %xmm8, %xmm0 -; SSSE3-NEXT: por %xmm6, %xmm0 -; SSSE3-NEXT: pcmpgtd %xmm0, %xmm1 -; SSSE3-NEXT: movdqa %xmm5, %xmm12 -; SSSE3-NEXT: pxor %xmm6, %xmm12 -; SSSE3-NEXT: movdqa %xmm10, %xmm0 -; SSSE3-NEXT: por %xmm6, %xmm0 -; SSSE3-NEXT: pcmpgtd %xmm0, %xmm12 -; SSSE3-NEXT: movdqa %xmm2, %xmm0 -; SSSE3-NEXT: pxor %xmm6, %xmm0 -; SSSE3-NEXT: movdqa %xmm9, %xmm7 -; SSSE3-NEXT: por %xmm6, %xmm7 -; SSSE3-NEXT: pcmpgtd %xmm7, %xmm0 -; SSSE3-NEXT: movdqa %xmm3, %xmm7 -; SSSE3-NEXT: pxor %xmm6, %xmm7 -; SSSE3-NEXT: por %xmm11, %xmm6 -; SSSE3-NEXT: pcmpgtd %xmm6, %xmm7 +; SSSE3-NEXT: pxor %xmm7, %xmm7 +; SSSE3-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1],xmm1[2],xmm7[2],xmm1[3],xmm7[3] +; SSSE3-NEXT: punpckhwd {{.*#+}} xmm8 = xmm8[4],xmm7[4],xmm8[5],xmm7[5],xmm8[6],xmm7[6],xmm8[7],xmm7[7] +; SSSE3-NEXT: movdqa %xmm0, %xmm10 +; SSSE3-NEXT: punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm7[0],xmm10[1],xmm7[1],xmm10[2],xmm7[2],xmm10[3],xmm7[3] +; SSSE3-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm7[4],xmm0[5],xmm7[5],xmm0[6],xmm7[6],xmm0[7],xmm7[7] +; SSSE3-NEXT: movdqa {{.*#+}} xmm7 = [2147483648,2147483648,2147483648,2147483648] ; SSSE3-NEXT: movdqa %xmm3, %xmm6 -; SSSE3-NEXT: pand %xmm7, %xmm6 -; SSSE3-NEXT: pandn %xmm11, %xmm7 -; SSSE3-NEXT: por %xmm6, %xmm7 +; SSSE3-NEXT: pxor %xmm7, %xmm6 +; SSSE3-NEXT: movdqa %xmm0, %xmm9 +; SSSE3-NEXT: por %xmm7, %xmm9 +; SSSE3-NEXT: pcmpgtd %xmm6, %xmm9 +; SSSE3-NEXT: pand %xmm9, %xmm0 +; SSSE3-NEXT: pandn %xmm3, %xmm9 +; SSSE3-NEXT: por %xmm0, %xmm9 ; SSSE3-NEXT: movdqa %xmm2, %xmm6 -; SSSE3-NEXT: pand %xmm0, %xmm6 -; SSSE3-NEXT: pandn %xmm9, %xmm0 -; SSSE3-NEXT: por %xmm6, %xmm0 -; SSSE3-NEXT: movdqa %xmm5, %xmm6 -; SSSE3-NEXT: pand %xmm12, %xmm6 -; SSSE3-NEXT: pandn %xmm10, %xmm12 -; SSSE3-NEXT: por %xmm6, %xmm12 -; SSSE3-NEXT: movdqa %xmm4, %xmm6 -; SSSE3-NEXT: pand %xmm1, %xmm6 -; SSSE3-NEXT: pandn %xmm8, %xmm1 -; SSSE3-NEXT: por %xmm6, %xmm1 +; SSSE3-NEXT: pxor %xmm7, %xmm6 +; SSSE3-NEXT: movdqa %xmm10, %xmm0 +; SSSE3-NEXT: por %xmm7, %xmm0 +; SSSE3-NEXT: pcmpgtd %xmm6, %xmm0 +; SSSE3-NEXT: pand %xmm0, %xmm10 +; SSSE3-NEXT: pandn %xmm2, %xmm0 +; SSSE3-NEXT: por %xmm10, %xmm0 +; SSSE3-NEXT: movdqa %xmm5, %xmm10 +; SSSE3-NEXT: pxor %xmm7, %xmm10 +; SSSE3-NEXT: movdqa %xmm8, %xmm6 +; SSSE3-NEXT: por %xmm7, %xmm6 +; SSSE3-NEXT: pcmpgtd %xmm10, %xmm6 +; SSSE3-NEXT: pand %xmm6, %xmm8 +; SSSE3-NEXT: pandn %xmm5, %xmm6 +; SSSE3-NEXT: por %xmm8, %xmm6 +; SSSE3-NEXT: movdqa %xmm4, %xmm8 +; SSSE3-NEXT: pxor %xmm7, %xmm8 +; SSSE3-NEXT: por %xmm1, %xmm7 +; SSSE3-NEXT: pcmpgtd %xmm8, %xmm7 +; SSSE3-NEXT: pand %xmm7, %xmm1 +; SSSE3-NEXT: pandn %xmm4, %xmm7 +; SSSE3-NEXT: por %xmm7, %xmm1 ; SSSE3-NEXT: psubd %xmm4, %xmm1 -; SSSE3-NEXT: psubd %xmm5, %xmm12 +; SSSE3-NEXT: psubd %xmm5, %xmm6 ; SSSE3-NEXT: psubd %xmm2, %xmm0 -; SSSE3-NEXT: psubd %xmm3, %xmm7 -; SSSE3-NEXT: pslld $16, %xmm7 -; SSSE3-NEXT: psrad $16, %xmm7 +; SSSE3-NEXT: psubd %xmm3, %xmm9 +; SSSE3-NEXT: pslld $16, %xmm9 +; SSSE3-NEXT: psrad $16, %xmm9 ; SSSE3-NEXT: pslld $16, %xmm0 ; SSSE3-NEXT: psrad $16, %xmm0 -; SSSE3-NEXT: packssdw %xmm7, %xmm0 -; SSSE3-NEXT: pslld $16, %xmm12 -; SSSE3-NEXT: psrad $16, %xmm12 +; SSSE3-NEXT: packssdw %xmm9, %xmm0 +; SSSE3-NEXT: pslld $16, %xmm6 +; SSSE3-NEXT: psrad $16, %xmm6 ; SSSE3-NEXT: pslld $16, %xmm1 ; SSSE3-NEXT: psrad $16, %xmm1 -; SSSE3-NEXT: packssdw %xmm12, %xmm1 +; SSSE3-NEXT: packssdw %xmm6, %xmm1 ; SSSE3-NEXT: retq ; ; SSE41-LABEL: psubus_16i32_max: @@ -2148,65 +1945,67 @@ vector.ph: define <8 x i16> @psubus_i16_i32_max_swapped(<8 x i16> %x, <8 x i32> %y) nounwind { ; SSE2-LABEL: psubus_i16_i32_max_swapped: ; SSE2: # %bb.0: # %vector.ph -; SSE2-NEXT: movdqa %xmm0, %xmm3 -; SSE2-NEXT: pxor %xmm0, %xmm0 -; SSE2-NEXT: movdqa %xmm3, %xmm5 -; SSE2-NEXT: punpckhwd {{.*#+}} xmm5 = xmm5[4],xmm0[4],xmm5[5],xmm0[5],xmm5[6],xmm0[6],xmm5[7],xmm0[7] -; SSE2-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3] -; SSE2-NEXT: movdqa {{.*#+}} xmm4 = [2147483648,2147483648,2147483648,2147483648] -; SSE2-NEXT: movdqa %xmm1, %xmm6 -; SSE2-NEXT: pxor %xmm4, %xmm6 -; SSE2-NEXT: movdqa %xmm3, %xmm0 -; SSE2-NEXT: por %xmm4, %xmm0 -; SSE2-NEXT: pcmpgtd %xmm6, %xmm0 +; SSE2-NEXT: pxor %xmm3, %xmm3 +; SSE2-NEXT: movdqa %xmm0, %xmm4 +; SSE2-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3] +; SSE2-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7] +; SSE2-NEXT: movdqa {{.*#+}} xmm5 = [2147483648,2147483648,2147483648,2147483648] +; SSE2-NEXT: movdqa %xmm2, %xmm3 +; SSE2-NEXT: pxor %xmm5, %xmm3 +; SSE2-NEXT: movdqa %xmm0, %xmm6 +; SSE2-NEXT: por %xmm5, %xmm6 +; SSE2-NEXT: pcmpgtd %xmm6, %xmm3 ; SSE2-NEXT: movdqa %xmm2, %xmm6 -; SSE2-NEXT: pxor %xmm4, %xmm6 -; SSE2-NEXT: por %xmm5, %xmm4 -; SSE2-NEXT: pcmpgtd %xmm6, %xmm4 -; SSE2-NEXT: pand %xmm4, %xmm5 -; SSE2-NEXT: pandn %xmm2, %xmm4 -; SSE2-NEXT: por %xmm5, %xmm4 -; SSE2-NEXT: pand %xmm0, %xmm3 -; SSE2-NEXT: pandn %xmm1, %xmm0 -; SSE2-NEXT: por %xmm3, %xmm0 +; SSE2-NEXT: pand %xmm3, %xmm6 +; SSE2-NEXT: pandn %xmm0, %xmm3 +; SSE2-NEXT: por %xmm6, %xmm3 +; SSE2-NEXT: movdqa %xmm1, %xmm0 +; SSE2-NEXT: pxor %xmm5, %xmm0 +; SSE2-NEXT: por %xmm4, %xmm5 +; SSE2-NEXT: pcmpgtd %xmm5, %xmm0 +; SSE2-NEXT: movdqa %xmm1, %xmm5 +; SSE2-NEXT: pand %xmm0, %xmm5 +; SSE2-NEXT: pandn %xmm4, %xmm0 +; SSE2-NEXT: por %xmm5, %xmm0 ; SSE2-NEXT: psubd %xmm1, %xmm0 -; SSE2-NEXT: psubd %xmm2, %xmm4 -; SSE2-NEXT: pslld $16, %xmm4 -; SSE2-NEXT: psrad $16, %xmm4 +; SSE2-NEXT: psubd %xmm2, %xmm3 +; SSE2-NEXT: pslld $16, %xmm3 +; SSE2-NEXT: psrad $16, %xmm3 ; SSE2-NEXT: pslld $16, %xmm0 ; SSE2-NEXT: psrad $16, %xmm0 -; SSE2-NEXT: packssdw %xmm4, %xmm0 +; SSE2-NEXT: packssdw %xmm3, %xmm0 ; SSE2-NEXT: retq ; ; SSSE3-LABEL: psubus_i16_i32_max_swapped: ; SSSE3: # %bb.0: # %vector.ph -; SSSE3-NEXT: movdqa %xmm0, %xmm3 -; SSSE3-NEXT: pxor %xmm0, %xmm0 -; SSSE3-NEXT: movdqa %xmm3, %xmm5 -; SSSE3-NEXT: punpckhwd {{.*#+}} xmm5 = xmm5[4],xmm0[4],xmm5[5],xmm0[5],xmm5[6],xmm0[6],xmm5[7],xmm0[7] -; SSSE3-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3] -; SSSE3-NEXT: movdqa {{.*#+}} xmm4 = [2147483648,2147483648,2147483648,2147483648] -; SSSE3-NEXT: movdqa %xmm1, %xmm6 -; SSSE3-NEXT: pxor %xmm4, %xmm6 -; SSSE3-NEXT: movdqa %xmm3, %xmm0 -; SSSE3-NEXT: por %xmm4, %xmm0 -; SSSE3-NEXT: pcmpgtd %xmm6, %xmm0 +; SSSE3-NEXT: pxor %xmm3, %xmm3 +; SSSE3-NEXT: movdqa %xmm0, %xmm4 +; SSSE3-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3] +; SSSE3-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7] +; SSSE3-NEXT: movdqa {{.*#+}} xmm5 = [2147483648,2147483648,2147483648,2147483648] +; SSSE3-NEXT: movdqa %xmm2, %xmm3 +; SSSE3-NEXT: pxor %xmm5, %xmm3 +; SSSE3-NEXT: movdqa %xmm0, %xmm6 +; SSSE3-NEXT: por %xmm5, %xmm6 +; SSSE3-NEXT: pcmpgtd %xmm6, %xmm3 ; SSSE3-NEXT: movdqa %xmm2, %xmm6 -; SSSE3-NEXT: pxor %xmm4, %xmm6 -; SSSE3-NEXT: por %xmm5, %xmm4 -; SSSE3-NEXT: pcmpgtd %xmm6, %xmm4 -; SSSE3-NEXT: pand %xmm4, %xmm5 -; SSSE3-NEXT: pandn %xmm2, %xmm4 -; SSSE3-NEXT: por %xmm5, %xmm4 -; SSSE3-NEXT: pand %xmm0, %xmm3 -; SSSE3-NEXT: pandn %xmm1, %xmm0 -; SSSE3-NEXT: por %xmm3, %xmm0 +; SSSE3-NEXT: pand %xmm3, %xmm6 +; SSSE3-NEXT: pandn %xmm0, %xmm3 +; SSSE3-NEXT: por %xmm6, %xmm3 +; SSSE3-NEXT: movdqa %xmm1, %xmm0 +; SSSE3-NEXT: pxor %xmm5, %xmm0 +; SSSE3-NEXT: por %xmm4, %xmm5 +; SSSE3-NEXT: pcmpgtd %xmm5, %xmm0 +; SSSE3-NEXT: movdqa %xmm1, %xmm5 +; SSSE3-NEXT: pand %xmm0, %xmm5 +; SSSE3-NEXT: pandn %xmm4, %xmm0 +; SSSE3-NEXT: por %xmm5, %xmm0 ; SSSE3-NEXT: psubd %xmm1, %xmm0 -; SSSE3-NEXT: psubd %xmm2, %xmm4 +; SSSE3-NEXT: psubd %xmm2, %xmm3 ; SSSE3-NEXT: movdqa {{.*#+}} xmm1 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15] -; SSSE3-NEXT: pshufb %xmm1, %xmm4 +; SSSE3-NEXT: pshufb %xmm1, %xmm3 ; SSSE3-NEXT: pshufb %xmm1, %xmm0 -; SSSE3-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm4[0] +; SSSE3-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm3[0] ; SSSE3-NEXT: retq ; ; SSE41-LABEL: psubus_i16_i32_max_swapped: @@ -2259,65 +2058,67 @@ define <8 x i16> @psubus_i16_i32_min(<8 x i16> %x, <8 x i32> %y) nounwind { ; SSE2: # %bb.0: # %vector.ph ; SSE2-NEXT: pxor %xmm4, %xmm4 ; SSE2-NEXT: movdqa %xmm0, %xmm3 -; SSE2-NEXT: punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7] -; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3] +; SSE2-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3] +; SSE2-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7] ; SSE2-NEXT: movdqa {{.*#+}} xmm4 = [2147483648,2147483648,2147483648,2147483648] -; SSE2-NEXT: movdqa %xmm1, %xmm5 +; SSE2-NEXT: movdqa %xmm2, %xmm5 ; SSE2-NEXT: pxor %xmm4, %xmm5 ; SSE2-NEXT: movdqa %xmm0, %xmm6 ; SSE2-NEXT: por %xmm4, %xmm6 ; SSE2-NEXT: pcmpgtd %xmm6, %xmm5 -; SSE2-NEXT: movdqa %xmm2, %xmm6 -; SSE2-NEXT: pxor %xmm4, %xmm6 +; SSE2-NEXT: movdqa %xmm0, %xmm6 +; SSE2-NEXT: pand %xmm5, %xmm6 +; SSE2-NEXT: pandn %xmm2, %xmm5 +; SSE2-NEXT: por %xmm6, %xmm5 +; SSE2-NEXT: movdqa %xmm1, %xmm2 +; SSE2-NEXT: pxor %xmm4, %xmm2 ; SSE2-NEXT: por %xmm3, %xmm4 -; SSE2-NEXT: pcmpgtd %xmm4, %xmm6 +; SSE2-NEXT: pcmpgtd %xmm4, %xmm2 ; SSE2-NEXT: movdqa %xmm3, %xmm4 -; SSE2-NEXT: pand %xmm6, %xmm4 -; SSE2-NEXT: pandn %xmm2, %xmm6 -; SSE2-NEXT: por %xmm4, %xmm6 -; SSE2-NEXT: movdqa %xmm0, %xmm2 -; SSE2-NEXT: pand %xmm5, %xmm2 -; SSE2-NEXT: pandn %xmm1, %xmm5 -; SSE2-NEXT: por %xmm2, %xmm5 +; SSE2-NEXT: pand %xmm2, %xmm4 +; SSE2-NEXT: pandn %xmm1, %xmm2 +; SSE2-NEXT: por %xmm4, %xmm2 +; SSE2-NEXT: psubd %xmm2, %xmm3 ; SSE2-NEXT: psubd %xmm5, %xmm0 -; SSE2-NEXT: psubd %xmm6, %xmm3 -; SSE2-NEXT: pslld $16, %xmm3 -; SSE2-NEXT: psrad $16, %xmm3 ; SSE2-NEXT: pslld $16, %xmm0 ; SSE2-NEXT: psrad $16, %xmm0 -; SSE2-NEXT: packssdw %xmm3, %xmm0 +; SSE2-NEXT: pslld $16, %xmm3 +; SSE2-NEXT: psrad $16, %xmm3 +; SSE2-NEXT: packssdw %xmm0, %xmm3 +; SSE2-NEXT: movdqa %xmm3, %xmm0 ; SSE2-NEXT: retq ; ; SSSE3-LABEL: psubus_i16_i32_min: ; SSSE3: # %bb.0: # %vector.ph ; SSSE3-NEXT: pxor %xmm4, %xmm4 ; SSSE3-NEXT: movdqa %xmm0, %xmm3 -; SSSE3-NEXT: punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7] -; SSSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3] +; SSSE3-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3] +; SSSE3-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7] ; SSSE3-NEXT: movdqa {{.*#+}} xmm4 = [2147483648,2147483648,2147483648,2147483648] -; SSSE3-NEXT: movdqa %xmm1, %xmm5 +; SSSE3-NEXT: movdqa %xmm2, %xmm5 ; SSSE3-NEXT: pxor %xmm4, %xmm5 ; SSSE3-NEXT: movdqa %xmm0, %xmm6 ; SSSE3-NEXT: por %xmm4, %xmm6 ; SSSE3-NEXT: pcmpgtd %xmm6, %xmm5 -; SSSE3-NEXT: movdqa %xmm2, %xmm6 -; SSSE3-NEXT: pxor %xmm4, %xmm6 +; SSSE3-NEXT: movdqa %xmm0, %xmm6 +; SSSE3-NEXT: pand %xmm5, %xmm6 +; SSSE3-NEXT: pandn %xmm2, %xmm5 +; SSSE3-NEXT: por %xmm6, %xmm5 +; SSSE3-NEXT: movdqa %xmm1, %xmm2 +; SSSE3-NEXT: pxor %xmm4, %xmm2 ; SSSE3-NEXT: por %xmm3, %xmm4 -; SSSE3-NEXT: pcmpgtd %xmm4, %xmm6 +; SSSE3-NEXT: pcmpgtd %xmm4, %xmm2 ; SSSE3-NEXT: movdqa %xmm3, %xmm4 -; SSSE3-NEXT: pand %xmm6, %xmm4 -; SSSE3-NEXT: pandn %xmm2, %xmm6 -; SSSE3-NEXT: por %xmm4, %xmm6 -; SSSE3-NEXT: movdqa %xmm0, %xmm2 -; SSSE3-NEXT: pand %xmm5, %xmm2 -; SSSE3-NEXT: pandn %xmm1, %xmm5 -; SSSE3-NEXT: por %xmm2, %xmm5 +; SSSE3-NEXT: pand %xmm2, %xmm4 +; SSSE3-NEXT: pandn %xmm1, %xmm2 +; SSSE3-NEXT: por %xmm4, %xmm2 +; SSSE3-NEXT: psubd %xmm2, %xmm3 ; SSSE3-NEXT: psubd %xmm5, %xmm0 -; SSSE3-NEXT: psubd %xmm6, %xmm3 ; SSSE3-NEXT: movdqa {{.*#+}} xmm1 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15] -; SSSE3-NEXT: pshufb %xmm1, %xmm3 ; SSSE3-NEXT: pshufb %xmm1, %xmm0 -; SSSE3-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm3[0] +; SSSE3-NEXT: pshufb %xmm1, %xmm3 +; SSSE3-NEXT: punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm0[0] +; SSSE3-NEXT: movdqa %xmm3, %xmm0 ; SSSE3-NEXT: retq ; ; SSE41-LABEL: psubus_i16_i32_min: |
