diff options
Diffstat (limited to 'llvm/test/CodeGen/X86/vector-fshl-128.ll')
| -rw-r--r-- | llvm/test/CodeGen/X86/vector-fshl-128.ll | 494 |
1 files changed, 228 insertions, 266 deletions
diff --git a/llvm/test/CodeGen/X86/vector-fshl-128.ll b/llvm/test/CodeGen/X86/vector-fshl-128.ll index 4a51e3341b2..e7e187629e5 100644 --- a/llvm/test/CodeGen/X86/vector-fshl-128.ll +++ b/llvm/test/CodeGen/X86/vector-fshl-128.ll @@ -108,16 +108,14 @@ define <2 x i64> @var_funnnel_v2i64(<2 x i64> %x, <2 x i64> %y, <2 x i64> %amt) ; ; AVX512F-LABEL: var_funnnel_v2i64: ; AVX512F: # %bb.0: -; AVX512F-NEXT: # kill: def $xmm2 killed $xmm2 def $zmm2 ; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0 -; AVX512F-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63] -; AVX512F-NEXT: vpand %xmm3, %xmm2, %xmm4 -; AVX512F-NEXT: vpsllvq %xmm4, %xmm0, %xmm5 -; AVX512F-NEXT: vmovdqa {{.*#+}} xmm6 = [64,64] -; AVX512F-NEXT: vpsubq %xmm4, %xmm6, %xmm4 +; AVX512F-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2 +; AVX512F-NEXT: vpsllvq %xmm2, %xmm0, %xmm3 +; AVX512F-NEXT: vmovdqa {{.*#+}} xmm4 = [64,64] +; AVX512F-NEXT: vpsubq %xmm2, %xmm4, %xmm4 ; AVX512F-NEXT: vpsrlvq %xmm4, %xmm1, %xmm1 -; AVX512F-NEXT: vpor %xmm1, %xmm5, %xmm1 -; AVX512F-NEXT: vptestnmq %zmm3, %zmm2, %k1 +; AVX512F-NEXT: vpor %xmm1, %xmm3, %xmm1 +; AVX512F-NEXT: vptestnmq %zmm2, %zmm2, %k1 ; AVX512F-NEXT: vmovdqa64 %zmm0, %zmm1 {%k1} ; AVX512F-NEXT: vmovdqa %xmm1, %xmm0 ; AVX512F-NEXT: vzeroupper @@ -125,30 +123,27 @@ define <2 x i64> @var_funnnel_v2i64(<2 x i64> %x, <2 x i64> %y, <2 x i64> %amt) ; ; AVX512VL-LABEL: var_funnnel_v2i64: ; AVX512VL: # %bb.0: -; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63] -; AVX512VL-NEXT: vpand %xmm3, %xmm2, %xmm4 -; AVX512VL-NEXT: vpsllvq %xmm4, %xmm0, %xmm5 -; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm6 = [64,64] -; AVX512VL-NEXT: vpsubq %xmm4, %xmm6, %xmm4 +; AVX512VL-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2 +; AVX512VL-NEXT: vpsllvq %xmm2, %xmm0, %xmm3 +; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm4 = [64,64] +; AVX512VL-NEXT: vpsubq %xmm2, %xmm4, %xmm4 ; AVX512VL-NEXT: vpsrlvq %xmm4, %xmm1, %xmm1 -; AVX512VL-NEXT: vpor %xmm1, %xmm5, %xmm1 -; AVX512VL-NEXT: vptestnmq %xmm3, %xmm2, %k1 +; AVX512VL-NEXT: vpor %xmm1, %xmm3, %xmm1 +; AVX512VL-NEXT: vptestnmq %xmm2, %xmm2, %k1 ; AVX512VL-NEXT: vmovdqa64 %xmm0, %xmm1 {%k1} ; AVX512VL-NEXT: vmovdqa %xmm1, %xmm0 ; AVX512VL-NEXT: retq ; ; AVX512BW-LABEL: var_funnnel_v2i64: ; AVX512BW: # %bb.0: -; AVX512BW-NEXT: # kill: def $xmm2 killed $xmm2 def $zmm2 ; AVX512BW-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0 -; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63] -; AVX512BW-NEXT: vpand %xmm3, %xmm2, %xmm4 -; AVX512BW-NEXT: vpsllvq %xmm4, %xmm0, %xmm5 -; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm6 = [64,64] -; AVX512BW-NEXT: vpsubq %xmm4, %xmm6, %xmm4 +; AVX512BW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2 +; AVX512BW-NEXT: vpsllvq %xmm2, %xmm0, %xmm3 +; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm4 = [64,64] +; AVX512BW-NEXT: vpsubq %xmm2, %xmm4, %xmm4 ; AVX512BW-NEXT: vpsrlvq %xmm4, %xmm1, %xmm1 -; AVX512BW-NEXT: vpor %xmm1, %xmm5, %xmm1 -; AVX512BW-NEXT: vptestnmq %zmm3, %zmm2, %k1 +; AVX512BW-NEXT: vpor %xmm1, %xmm3, %xmm1 +; AVX512BW-NEXT: vptestnmq %zmm2, %zmm2, %k1 ; AVX512BW-NEXT: vmovdqa64 %zmm0, %zmm1 {%k1} ; AVX512BW-NEXT: vmovdqa %xmm1, %xmm0 ; AVX512BW-NEXT: vzeroupper @@ -156,16 +151,14 @@ define <2 x i64> @var_funnnel_v2i64(<2 x i64> %x, <2 x i64> %y, <2 x i64> %amt) ; ; AVX512VBMI2-LABEL: var_funnnel_v2i64: ; AVX512VBMI2: # %bb.0: -; AVX512VBMI2-NEXT: # kill: def $xmm2 killed $xmm2 def $zmm2 ; AVX512VBMI2-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0 -; AVX512VBMI2-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63] -; AVX512VBMI2-NEXT: vpand %xmm3, %xmm2, %xmm4 -; AVX512VBMI2-NEXT: vpsllvq %xmm4, %xmm0, %xmm5 -; AVX512VBMI2-NEXT: vmovdqa {{.*#+}} xmm6 = [64,64] -; AVX512VBMI2-NEXT: vpsubq %xmm4, %xmm6, %xmm4 +; AVX512VBMI2-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2 +; AVX512VBMI2-NEXT: vpsllvq %xmm2, %xmm0, %xmm3 +; AVX512VBMI2-NEXT: vmovdqa {{.*#+}} xmm4 = [64,64] +; AVX512VBMI2-NEXT: vpsubq %xmm2, %xmm4, %xmm4 ; AVX512VBMI2-NEXT: vpsrlvq %xmm4, %xmm1, %xmm1 -; AVX512VBMI2-NEXT: vpor %xmm1, %xmm5, %xmm1 -; AVX512VBMI2-NEXT: vptestnmq %zmm3, %zmm2, %k1 +; AVX512VBMI2-NEXT: vpor %xmm1, %xmm3, %xmm1 +; AVX512VBMI2-NEXT: vptestnmq %zmm2, %zmm2, %k1 ; AVX512VBMI2-NEXT: vmovdqa64 %zmm0, %zmm1 {%k1} ; AVX512VBMI2-NEXT: vmovdqa %xmm1, %xmm0 ; AVX512VBMI2-NEXT: vzeroupper @@ -173,14 +166,13 @@ define <2 x i64> @var_funnnel_v2i64(<2 x i64> %x, <2 x i64> %y, <2 x i64> %amt) ; ; AVX512VLBW-LABEL: var_funnnel_v2i64: ; AVX512VLBW: # %bb.0: -; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63] -; AVX512VLBW-NEXT: vpand %xmm3, %xmm2, %xmm4 -; AVX512VLBW-NEXT: vpsllvq %xmm4, %xmm0, %xmm5 -; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm6 = [64,64] -; AVX512VLBW-NEXT: vpsubq %xmm4, %xmm6, %xmm4 +; AVX512VLBW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2 +; AVX512VLBW-NEXT: vpsllvq %xmm2, %xmm0, %xmm3 +; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm4 = [64,64] +; AVX512VLBW-NEXT: vpsubq %xmm2, %xmm4, %xmm4 ; AVX512VLBW-NEXT: vpsrlvq %xmm4, %xmm1, %xmm1 -; AVX512VLBW-NEXT: vpor %xmm1, %xmm5, %xmm1 -; AVX512VLBW-NEXT: vptestnmq %xmm3, %xmm2, %k1 +; AVX512VLBW-NEXT: vpor %xmm1, %xmm3, %xmm1 +; AVX512VLBW-NEXT: vptestnmq %xmm2, %xmm2, %k1 ; AVX512VLBW-NEXT: vmovdqa64 %xmm0, %xmm1 {%k1} ; AVX512VLBW-NEXT: vmovdqa %xmm1, %xmm0 ; AVX512VLBW-NEXT: retq @@ -358,16 +350,15 @@ define <4 x i32> @var_funnnel_v4i32(<4 x i32> %x, <4 x i32> %y, <4 x i32> %amt) ; ; AVX512F-LABEL: var_funnnel_v4i32: ; AVX512F: # %bb.0: -; AVX512F-NEXT: # kill: def $xmm2 killed $xmm2 def $zmm2 ; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0 ; AVX512F-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31] -; AVX512F-NEXT: vpand %xmm3, %xmm2, %xmm4 -; AVX512F-NEXT: vpsllvd %xmm4, %xmm0, %xmm5 -; AVX512F-NEXT: vpbroadcastd {{.*#+}} xmm6 = [32,32,32,32] -; AVX512F-NEXT: vpsubd %xmm4, %xmm6, %xmm4 +; AVX512F-NEXT: vpand %xmm3, %xmm2, %xmm2 +; AVX512F-NEXT: vpsllvd %xmm2, %xmm0, %xmm3 +; AVX512F-NEXT: vpbroadcastd {{.*#+}} xmm4 = [32,32,32,32] +; AVX512F-NEXT: vpsubd %xmm2, %xmm4, %xmm4 ; AVX512F-NEXT: vpsrlvd %xmm4, %xmm1, %xmm1 -; AVX512F-NEXT: vpor %xmm1, %xmm5, %xmm1 -; AVX512F-NEXT: vptestnmd %zmm3, %zmm2, %k1 +; AVX512F-NEXT: vpor %xmm1, %xmm3, %xmm1 +; AVX512F-NEXT: vptestnmd %zmm2, %zmm2, %k1 ; AVX512F-NEXT: vmovdqa32 %zmm0, %zmm1 {%k1} ; AVX512F-NEXT: vmovdqa %xmm1, %xmm0 ; AVX512F-NEXT: vzeroupper @@ -375,30 +366,28 @@ define <4 x i32> @var_funnnel_v4i32(<4 x i32> %x, <4 x i32> %y, <4 x i32> %amt) ; ; AVX512VL-LABEL: var_funnnel_v4i32: ; AVX512VL: # %bb.0: -; AVX512VL-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31] -; AVX512VL-NEXT: vpand %xmm3, %xmm2, %xmm4 -; AVX512VL-NEXT: vpsllvd %xmm4, %xmm0, %xmm5 -; AVX512VL-NEXT: vpbroadcastd {{.*#+}} xmm6 = [32,32,32,32] -; AVX512VL-NEXT: vpsubd %xmm4, %xmm6, %xmm4 +; AVX512VL-NEXT: vpandd {{.*}}(%rip){1to4}, %xmm2, %xmm2 +; AVX512VL-NEXT: vpsllvd %xmm2, %xmm0, %xmm3 +; AVX512VL-NEXT: vpbroadcastd {{.*#+}} xmm4 = [32,32,32,32] +; AVX512VL-NEXT: vpsubd %xmm2, %xmm4, %xmm4 ; AVX512VL-NEXT: vpsrlvd %xmm4, %xmm1, %xmm1 -; AVX512VL-NEXT: vpor %xmm1, %xmm5, %xmm1 -; AVX512VL-NEXT: vptestnmd %xmm3, %xmm2, %k1 +; AVX512VL-NEXT: vpor %xmm1, %xmm3, %xmm1 +; AVX512VL-NEXT: vptestnmd %xmm2, %xmm2, %k1 ; AVX512VL-NEXT: vmovdqa32 %xmm0, %xmm1 {%k1} ; AVX512VL-NEXT: vmovdqa %xmm1, %xmm0 ; AVX512VL-NEXT: retq ; ; AVX512BW-LABEL: var_funnnel_v4i32: ; AVX512BW: # %bb.0: -; AVX512BW-NEXT: # kill: def $xmm2 killed $xmm2 def $zmm2 ; AVX512BW-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0 ; AVX512BW-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31] -; AVX512BW-NEXT: vpand %xmm3, %xmm2, %xmm4 -; AVX512BW-NEXT: vpsllvd %xmm4, %xmm0, %xmm5 -; AVX512BW-NEXT: vpbroadcastd {{.*#+}} xmm6 = [32,32,32,32] -; AVX512BW-NEXT: vpsubd %xmm4, %xmm6, %xmm4 +; AVX512BW-NEXT: vpand %xmm3, %xmm2, %xmm2 +; AVX512BW-NEXT: vpsllvd %xmm2, %xmm0, %xmm3 +; AVX512BW-NEXT: vpbroadcastd {{.*#+}} xmm4 = [32,32,32,32] +; AVX512BW-NEXT: vpsubd %xmm2, %xmm4, %xmm4 ; AVX512BW-NEXT: vpsrlvd %xmm4, %xmm1, %xmm1 -; AVX512BW-NEXT: vpor %xmm1, %xmm5, %xmm1 -; AVX512BW-NEXT: vptestnmd %zmm3, %zmm2, %k1 +; AVX512BW-NEXT: vpor %xmm1, %xmm3, %xmm1 +; AVX512BW-NEXT: vptestnmd %zmm2, %zmm2, %k1 ; AVX512BW-NEXT: vmovdqa32 %zmm0, %zmm1 {%k1} ; AVX512BW-NEXT: vmovdqa %xmm1, %xmm0 ; AVX512BW-NEXT: vzeroupper @@ -406,16 +395,15 @@ define <4 x i32> @var_funnnel_v4i32(<4 x i32> %x, <4 x i32> %y, <4 x i32> %amt) ; ; AVX512VBMI2-LABEL: var_funnnel_v4i32: ; AVX512VBMI2: # %bb.0: -; AVX512VBMI2-NEXT: # kill: def $xmm2 killed $xmm2 def $zmm2 ; AVX512VBMI2-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0 ; AVX512VBMI2-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31] -; AVX512VBMI2-NEXT: vpand %xmm3, %xmm2, %xmm4 -; AVX512VBMI2-NEXT: vpsllvd %xmm4, %xmm0, %xmm5 -; AVX512VBMI2-NEXT: vpbroadcastd {{.*#+}} xmm6 = [32,32,32,32] -; AVX512VBMI2-NEXT: vpsubd %xmm4, %xmm6, %xmm4 +; AVX512VBMI2-NEXT: vpand %xmm3, %xmm2, %xmm2 +; AVX512VBMI2-NEXT: vpsllvd %xmm2, %xmm0, %xmm3 +; AVX512VBMI2-NEXT: vpbroadcastd {{.*#+}} xmm4 = [32,32,32,32] +; AVX512VBMI2-NEXT: vpsubd %xmm2, %xmm4, %xmm4 ; AVX512VBMI2-NEXT: vpsrlvd %xmm4, %xmm1, %xmm1 -; AVX512VBMI2-NEXT: vpor %xmm1, %xmm5, %xmm1 -; AVX512VBMI2-NEXT: vptestnmd %zmm3, %zmm2, %k1 +; AVX512VBMI2-NEXT: vpor %xmm1, %xmm3, %xmm1 +; AVX512VBMI2-NEXT: vptestnmd %zmm2, %zmm2, %k1 ; AVX512VBMI2-NEXT: vmovdqa32 %zmm0, %zmm1 {%k1} ; AVX512VBMI2-NEXT: vmovdqa %xmm1, %xmm0 ; AVX512VBMI2-NEXT: vzeroupper @@ -423,14 +411,13 @@ define <4 x i32> @var_funnnel_v4i32(<4 x i32> %x, <4 x i32> %y, <4 x i32> %amt) ; ; AVX512VLBW-LABEL: var_funnnel_v4i32: ; AVX512VLBW: # %bb.0: -; AVX512VLBW-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31] -; AVX512VLBW-NEXT: vpand %xmm3, %xmm2, %xmm4 -; AVX512VLBW-NEXT: vpsllvd %xmm4, %xmm0, %xmm5 -; AVX512VLBW-NEXT: vpbroadcastd {{.*#+}} xmm6 = [32,32,32,32] -; AVX512VLBW-NEXT: vpsubd %xmm4, %xmm6, %xmm4 +; AVX512VLBW-NEXT: vpandd {{.*}}(%rip){1to4}, %xmm2, %xmm2 +; AVX512VLBW-NEXT: vpsllvd %xmm2, %xmm0, %xmm3 +; AVX512VLBW-NEXT: vpbroadcastd {{.*#+}} xmm4 = [32,32,32,32] +; AVX512VLBW-NEXT: vpsubd %xmm2, %xmm4, %xmm4 ; AVX512VLBW-NEXT: vpsrlvd %xmm4, %xmm1, %xmm1 -; AVX512VLBW-NEXT: vpor %xmm1, %xmm5, %xmm1 -; AVX512VLBW-NEXT: vptestnmd %xmm3, %xmm2, %k1 +; AVX512VLBW-NEXT: vpor %xmm1, %xmm3, %xmm1 +; AVX512VLBW-NEXT: vptestnmd %xmm2, %xmm2, %k1 ; AVX512VLBW-NEXT: vmovdqa32 %xmm0, %xmm1 {%k1} ; AVX512VLBW-NEXT: vmovdqa %xmm1, %xmm0 ; AVX512VLBW-NEXT: retq @@ -719,17 +706,15 @@ define <8 x i16> @var_funnnel_v8i16(<8 x i16> %x, <8 x i16> %y, <8 x i16> %amt) ; ; AVX512BW-LABEL: var_funnnel_v8i16: ; AVX512BW: # %bb.0: -; AVX512BW-NEXT: # kill: def $xmm2 killed $xmm2 def $zmm2 ; AVX512BW-NEXT: # kill: def $xmm1 killed $xmm1 def $zmm1 ; AVX512BW-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0 -; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15] -; AVX512BW-NEXT: vpand %xmm3, %xmm2, %xmm4 -; AVX512BW-NEXT: vpsllvw %zmm4, %zmm0, %zmm5 -; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm6 = [16,16,16,16,16,16,16,16] -; AVX512BW-NEXT: vpsubw %xmm4, %xmm6, %xmm4 +; AVX512BW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2 +; AVX512BW-NEXT: vpsllvw %zmm2, %zmm0, %zmm3 +; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm4 = [16,16,16,16,16,16,16,16] +; AVX512BW-NEXT: vpsubw %xmm2, %xmm4, %xmm4 ; AVX512BW-NEXT: vpsrlvw %zmm4, %zmm1, %zmm1 -; AVX512BW-NEXT: vpor %xmm1, %xmm5, %xmm1 -; AVX512BW-NEXT: vptestnmw %zmm3, %zmm2, %k1 +; AVX512BW-NEXT: vpor %xmm1, %xmm3, %xmm1 +; AVX512BW-NEXT: vptestnmw %zmm2, %zmm2, %k1 ; AVX512BW-NEXT: vmovdqu16 %zmm0, %zmm1 {%k1} ; AVX512BW-NEXT: vmovdqa %xmm1, %xmm0 ; AVX512BW-NEXT: vzeroupper @@ -737,17 +722,15 @@ define <8 x i16> @var_funnnel_v8i16(<8 x i16> %x, <8 x i16> %y, <8 x i16> %amt) ; ; AVX512VBMI2-LABEL: var_funnnel_v8i16: ; AVX512VBMI2: # %bb.0: -; AVX512VBMI2-NEXT: # kill: def $xmm2 killed $xmm2 def $zmm2 ; AVX512VBMI2-NEXT: # kill: def $xmm1 killed $xmm1 def $zmm1 ; AVX512VBMI2-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0 -; AVX512VBMI2-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15] -; AVX512VBMI2-NEXT: vpand %xmm3, %xmm2, %xmm4 -; AVX512VBMI2-NEXT: vpsllvw %zmm4, %zmm0, %zmm5 -; AVX512VBMI2-NEXT: vmovdqa {{.*#+}} xmm6 = [16,16,16,16,16,16,16,16] -; AVX512VBMI2-NEXT: vpsubw %xmm4, %xmm6, %xmm4 +; AVX512VBMI2-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2 +; AVX512VBMI2-NEXT: vpsllvw %zmm2, %zmm0, %zmm3 +; AVX512VBMI2-NEXT: vmovdqa {{.*#+}} xmm4 = [16,16,16,16,16,16,16,16] +; AVX512VBMI2-NEXT: vpsubw %xmm2, %xmm4, %xmm4 ; AVX512VBMI2-NEXT: vpsrlvw %zmm4, %zmm1, %zmm1 -; AVX512VBMI2-NEXT: vpor %xmm1, %xmm5, %xmm1 -; AVX512VBMI2-NEXT: vptestnmw %zmm3, %zmm2, %k1 +; AVX512VBMI2-NEXT: vpor %xmm1, %xmm3, %xmm1 +; AVX512VBMI2-NEXT: vptestnmw %zmm2, %zmm2, %k1 ; AVX512VBMI2-NEXT: vmovdqu16 %zmm0, %zmm1 {%k1} ; AVX512VBMI2-NEXT: vmovdqa %xmm1, %xmm0 ; AVX512VBMI2-NEXT: vzeroupper @@ -755,14 +738,13 @@ define <8 x i16> @var_funnnel_v8i16(<8 x i16> %x, <8 x i16> %y, <8 x i16> %amt) ; ; AVX512VLBW-LABEL: var_funnnel_v8i16: ; AVX512VLBW: # %bb.0: -; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15] -; AVX512VLBW-NEXT: vpand %xmm3, %xmm2, %xmm4 -; AVX512VLBW-NEXT: vpsllvw %xmm4, %xmm0, %xmm5 -; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm6 = [16,16,16,16,16,16,16,16] -; AVX512VLBW-NEXT: vpsubw %xmm4, %xmm6, %xmm4 +; AVX512VLBW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2 +; AVX512VLBW-NEXT: vpsllvw %xmm2, %xmm0, %xmm3 +; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm4 = [16,16,16,16,16,16,16,16] +; AVX512VLBW-NEXT: vpsubw %xmm2, %xmm4, %xmm4 ; AVX512VLBW-NEXT: vpsrlvw %xmm4, %xmm1, %xmm1 -; AVX512VLBW-NEXT: vpor %xmm1, %xmm5, %xmm1 -; AVX512VLBW-NEXT: vptestnmw %xmm3, %xmm2, %k1 +; AVX512VLBW-NEXT: vpor %xmm1, %xmm3, %xmm1 +; AVX512VLBW-NEXT: vptestnmw %xmm2, %xmm2, %k1 ; AVX512VLBW-NEXT: vmovdqu16 %xmm0, %xmm1 {%k1} ; AVX512VLBW-NEXT: vmovdqa %xmm1, %xmm0 ; AVX512VLBW-NEXT: retq @@ -1040,21 +1022,19 @@ define <16 x i8> @var_funnnel_v16i8(<16 x i8> %x, <16 x i8> %y, <16 x i8> %amt) ; ; AVX512BW-LABEL: var_funnnel_v16i8: ; AVX512BW: # %bb.0: -; AVX512BW-NEXT: # kill: def $xmm2 killed $xmm2 def $zmm2 ; AVX512BW-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0 ; AVX512BW-NEXT: vpmovzxbw {{.*#+}} ymm3 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero -; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm4 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7] -; AVX512BW-NEXT: vpand %xmm4, %xmm2, %xmm5 -; AVX512BW-NEXT: vpmovzxbw {{.*#+}} ymm6 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero,xmm5[8],zero,xmm5[9],zero,xmm5[10],zero,xmm5[11],zero,xmm5[12],zero,xmm5[13],zero,xmm5[14],zero,xmm5[15],zero -; AVX512BW-NEXT: vpsllvw %zmm6, %zmm3, %zmm3 +; AVX512BW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2 +; AVX512BW-NEXT: vpmovzxbw {{.*#+}} ymm4 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero,xmm2[8],zero,xmm2[9],zero,xmm2[10],zero,xmm2[11],zero,xmm2[12],zero,xmm2[13],zero,xmm2[14],zero,xmm2[15],zero +; AVX512BW-NEXT: vpsllvw %zmm4, %zmm3, %zmm3 ; AVX512BW-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero -; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm6 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8] -; AVX512BW-NEXT: vpsubb %xmm5, %xmm6, %xmm5 -; AVX512BW-NEXT: vpmovzxbw {{.*#+}} ymm5 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero,xmm5[8],zero,xmm5[9],zero,xmm5[10],zero,xmm5[11],zero,xmm5[12],zero,xmm5[13],zero,xmm5[14],zero,xmm5[15],zero -; AVX512BW-NEXT: vpsrlvw %zmm5, %zmm1, %zmm1 +; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm4 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8] +; AVX512BW-NEXT: vpsubb %xmm2, %xmm4, %xmm4 +; AVX512BW-NEXT: vpmovzxbw {{.*#+}} ymm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero,xmm4[8],zero,xmm4[9],zero,xmm4[10],zero,xmm4[11],zero,xmm4[12],zero,xmm4[13],zero,xmm4[14],zero,xmm4[15],zero +; AVX512BW-NEXT: vpsrlvw %zmm4, %zmm1, %zmm1 ; AVX512BW-NEXT: vpor %ymm1, %ymm3, %ymm1 ; AVX512BW-NEXT: vpmovwb %zmm1, %ymm1 -; AVX512BW-NEXT: vptestnmb %zmm4, %zmm2, %k1 +; AVX512BW-NEXT: vptestnmb %zmm2, %zmm2, %k1 ; AVX512BW-NEXT: vmovdqu8 %zmm0, %zmm1 {%k1} ; AVX512BW-NEXT: vmovdqa %xmm1, %xmm0 ; AVX512BW-NEXT: vzeroupper @@ -1062,21 +1042,19 @@ define <16 x i8> @var_funnnel_v16i8(<16 x i8> %x, <16 x i8> %y, <16 x i8> %amt) ; ; AVX512VBMI2-LABEL: var_funnnel_v16i8: ; AVX512VBMI2: # %bb.0: -; AVX512VBMI2-NEXT: # kill: def $xmm2 killed $xmm2 def $zmm2 ; AVX512VBMI2-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0 ; AVX512VBMI2-NEXT: vpmovzxbw {{.*#+}} ymm3 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero -; AVX512VBMI2-NEXT: vmovdqa {{.*#+}} xmm4 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7] -; AVX512VBMI2-NEXT: vpand %xmm4, %xmm2, %xmm5 -; AVX512VBMI2-NEXT: vpmovzxbw {{.*#+}} ymm6 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero,xmm5[8],zero,xmm5[9],zero,xmm5[10],zero,xmm5[11],zero,xmm5[12],zero,xmm5[13],zero,xmm5[14],zero,xmm5[15],zero -; AVX512VBMI2-NEXT: vpsllvw %zmm6, %zmm3, %zmm3 +; AVX512VBMI2-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2 +; AVX512VBMI2-NEXT: vpmovzxbw {{.*#+}} ymm4 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero,xmm2[8],zero,xmm2[9],zero,xmm2[10],zero,xmm2[11],zero,xmm2[12],zero,xmm2[13],zero,xmm2[14],zero,xmm2[15],zero +; AVX512VBMI2-NEXT: vpsllvw %zmm4, %zmm3, %zmm3 ; AVX512VBMI2-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero -; AVX512VBMI2-NEXT: vmovdqa {{.*#+}} xmm6 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8] -; AVX512VBMI2-NEXT: vpsubb %xmm5, %xmm6, %xmm5 -; AVX512VBMI2-NEXT: vpmovzxbw {{.*#+}} ymm5 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero,xmm5[8],zero,xmm5[9],zero,xmm5[10],zero,xmm5[11],zero,xmm5[12],zero,xmm5[13],zero,xmm5[14],zero,xmm5[15],zero -; AVX512VBMI2-NEXT: vpsrlvw %zmm5, %zmm1, %zmm1 +; AVX512VBMI2-NEXT: vmovdqa {{.*#+}} xmm4 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8] +; AVX512VBMI2-NEXT: vpsubb %xmm2, %xmm4, %xmm4 +; AVX512VBMI2-NEXT: vpmovzxbw {{.*#+}} ymm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero,xmm4[8],zero,xmm4[9],zero,xmm4[10],zero,xmm4[11],zero,xmm4[12],zero,xmm4[13],zero,xmm4[14],zero,xmm4[15],zero +; AVX512VBMI2-NEXT: vpsrlvw %zmm4, %zmm1, %zmm1 ; AVX512VBMI2-NEXT: vpor %ymm1, %ymm3, %ymm1 ; AVX512VBMI2-NEXT: vpmovwb %zmm1, %ymm1 -; AVX512VBMI2-NEXT: vptestnmb %zmm4, %zmm2, %k1 +; AVX512VBMI2-NEXT: vptestnmb %zmm2, %zmm2, %k1 ; AVX512VBMI2-NEXT: vmovdqu8 %zmm0, %zmm1 {%k1} ; AVX512VBMI2-NEXT: vmovdqa %xmm1, %xmm0 ; AVX512VBMI2-NEXT: vzeroupper @@ -1084,19 +1062,18 @@ define <16 x i8> @var_funnnel_v16i8(<16 x i8> %x, <16 x i8> %y, <16 x i8> %amt) ; ; AVX512VLBW-LABEL: var_funnnel_v16i8: ; AVX512VLBW: # %bb.0: -; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7] -; AVX512VLBW-NEXT: vpand %xmm3, %xmm2, %xmm4 -; AVX512VLBW-NEXT: vpmovzxbw {{.*#+}} ymm5 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero,xmm4[8],zero,xmm4[9],zero,xmm4[10],zero,xmm4[11],zero,xmm4[12],zero,xmm4[13],zero,xmm4[14],zero,xmm4[15],zero -; AVX512VLBW-NEXT: vpmovzxbw {{.*#+}} ymm6 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero -; AVX512VLBW-NEXT: vpsllvw %ymm5, %ymm6, %ymm5 -; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm6 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8] -; AVX512VLBW-NEXT: vpsubb %xmm4, %xmm6, %xmm4 +; AVX512VLBW-NEXT: vpmovzxbw {{.*#+}} ymm3 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero +; AVX512VLBW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2 +; AVX512VLBW-NEXT: vpmovzxbw {{.*#+}} ymm4 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero,xmm2[8],zero,xmm2[9],zero,xmm2[10],zero,xmm2[11],zero,xmm2[12],zero,xmm2[13],zero,xmm2[14],zero,xmm2[15],zero +; AVX512VLBW-NEXT: vpsllvw %ymm4, %ymm3, %ymm3 +; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm4 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8] +; AVX512VLBW-NEXT: vpsubb %xmm2, %xmm4, %xmm4 ; AVX512VLBW-NEXT: vpmovzxbw {{.*#+}} ymm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero,xmm4[8],zero,xmm4[9],zero,xmm4[10],zero,xmm4[11],zero,xmm4[12],zero,xmm4[13],zero,xmm4[14],zero,xmm4[15],zero ; AVX512VLBW-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero ; AVX512VLBW-NEXT: vpsrlvw %ymm4, %ymm1, %ymm1 -; AVX512VLBW-NEXT: vpor %ymm1, %ymm5, %ymm1 +; AVX512VLBW-NEXT: vpor %ymm1, %ymm3, %ymm1 ; AVX512VLBW-NEXT: vpmovwb %ymm1, %xmm1 -; AVX512VLBW-NEXT: vptestnmb %xmm3, %xmm2, %k1 +; AVX512VLBW-NEXT: vptestnmb %xmm2, %xmm2, %k1 ; AVX512VLBW-NEXT: vmovdqu8 %xmm0, %xmm1 {%k1} ; AVX512VLBW-NEXT: vmovdqa %xmm1, %xmm0 ; AVX512VLBW-NEXT: vzeroupper @@ -1104,19 +1081,18 @@ define <16 x i8> @var_funnnel_v16i8(<16 x i8> %x, <16 x i8> %y, <16 x i8> %amt) ; ; AVX512VLVBMI2-LABEL: var_funnnel_v16i8: ; AVX512VLVBMI2: # %bb.0: -; AVX512VLVBMI2-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7] -; AVX512VLVBMI2-NEXT: vpand %xmm3, %xmm2, %xmm4 -; AVX512VLVBMI2-NEXT: vpmovzxbw {{.*#+}} ymm5 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero,xmm4[8],zero,xmm4[9],zero,xmm4[10],zero,xmm4[11],zero,xmm4[12],zero,xmm4[13],zero,xmm4[14],zero,xmm4[15],zero -; AVX512VLVBMI2-NEXT: vpmovzxbw {{.*#+}} ymm6 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero -; AVX512VLVBMI2-NEXT: vpsllvw %ymm5, %ymm6, %ymm5 -; AVX512VLVBMI2-NEXT: vmovdqa {{.*#+}} xmm6 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8] -; AVX512VLVBMI2-NEXT: vpsubb %xmm4, %xmm6, %xmm4 +; AVX512VLVBMI2-NEXT: vpmovzxbw {{.*#+}} ymm3 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero +; AVX512VLVBMI2-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2 +; AVX512VLVBMI2-NEXT: vpmovzxbw {{.*#+}} ymm4 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero,xmm2[8],zero,xmm2[9],zero,xmm2[10],zero,xmm2[11],zero,xmm2[12],zero,xmm2[13],zero,xmm2[14],zero,xmm2[15],zero +; AVX512VLVBMI2-NEXT: vpsllvw %ymm4, %ymm3, %ymm3 +; AVX512VLVBMI2-NEXT: vmovdqa {{.*#+}} xmm4 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8] +; AVX512VLVBMI2-NEXT: vpsubb %xmm2, %xmm4, %xmm4 ; AVX512VLVBMI2-NEXT: vpmovzxbw {{.*#+}} ymm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero,xmm4[8],zero,xmm4[9],zero,xmm4[10],zero,xmm4[11],zero,xmm4[12],zero,xmm4[13],zero,xmm4[14],zero,xmm4[15],zero ; AVX512VLVBMI2-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero ; AVX512VLVBMI2-NEXT: vpsrlvw %ymm4, %ymm1, %ymm1 -; AVX512VLVBMI2-NEXT: vpor %ymm1, %ymm5, %ymm1 +; AVX512VLVBMI2-NEXT: vpor %ymm1, %ymm3, %ymm1 ; AVX512VLVBMI2-NEXT: vpmovwb %ymm1, %xmm1 -; AVX512VLVBMI2-NEXT: vptestnmb %xmm3, %xmm2, %k1 +; AVX512VLVBMI2-NEXT: vptestnmb %xmm2, %xmm2, %k1 ; AVX512VLVBMI2-NEXT: vmovdqu8 %xmm0, %xmm1 {%k1} ; AVX512VLVBMI2-NEXT: vmovdqa %xmm1, %xmm0 ; AVX512VLVBMI2-NEXT: vzeroupper @@ -1276,14 +1252,13 @@ define <2 x i64> @splatvar_funnnel_v2i64(<2 x i64> %x, <2 x i64> %y, <2 x i64> % ; AVX512F: # %bb.0: ; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0 ; AVX512F-NEXT: vpbroadcastq %xmm2, %xmm2 -; AVX512F-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63] -; AVX512F-NEXT: vpand %xmm3, %xmm2, %xmm4 -; AVX512F-NEXT: vpsllq %xmm4, %xmm0, %xmm5 -; AVX512F-NEXT: vmovdqa {{.*#+}} xmm6 = [64,64] -; AVX512F-NEXT: vpsubq %xmm4, %xmm6, %xmm4 +; AVX512F-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2 +; AVX512F-NEXT: vpsllq %xmm2, %xmm0, %xmm3 +; AVX512F-NEXT: vmovdqa {{.*#+}} xmm4 = [64,64] +; AVX512F-NEXT: vpsubq %xmm2, %xmm4, %xmm4 ; AVX512F-NEXT: vpsrlq %xmm4, %xmm1, %xmm1 -; AVX512F-NEXT: vpor %xmm1, %xmm5, %xmm1 -; AVX512F-NEXT: vptestnmq %zmm3, %zmm2, %k1 +; AVX512F-NEXT: vpor %xmm1, %xmm3, %xmm1 +; AVX512F-NEXT: vptestnmq %zmm2, %zmm2, %k1 ; AVX512F-NEXT: vmovdqa64 %zmm0, %zmm1 {%k1} ; AVX512F-NEXT: vmovdqa %xmm1, %xmm0 ; AVX512F-NEXT: vzeroupper @@ -1292,14 +1267,13 @@ define <2 x i64> @splatvar_funnnel_v2i64(<2 x i64> %x, <2 x i64> %y, <2 x i64> % ; AVX512VL-LABEL: splatvar_funnnel_v2i64: ; AVX512VL: # %bb.0: ; AVX512VL-NEXT: vpbroadcastq %xmm2, %xmm2 -; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63] -; AVX512VL-NEXT: vpand %xmm3, %xmm2, %xmm4 -; AVX512VL-NEXT: vpsllq %xmm4, %xmm0, %xmm5 -; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm6 = [64,64] -; AVX512VL-NEXT: vpsubq %xmm4, %xmm6, %xmm4 +; AVX512VL-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2 +; AVX512VL-NEXT: vpsllq %xmm2, %xmm0, %xmm3 +; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm4 = [64,64] +; AVX512VL-NEXT: vpsubq %xmm2, %xmm4, %xmm4 ; AVX512VL-NEXT: vpsrlq %xmm4, %xmm1, %xmm1 -; AVX512VL-NEXT: vpor %xmm1, %xmm5, %xmm1 -; AVX512VL-NEXT: vptestnmq %xmm3, %xmm2, %k1 +; AVX512VL-NEXT: vpor %xmm1, %xmm3, %xmm1 +; AVX512VL-NEXT: vptestnmq %xmm2, %xmm2, %k1 ; AVX512VL-NEXT: vmovdqa64 %xmm0, %xmm1 {%k1} ; AVX512VL-NEXT: vmovdqa %xmm1, %xmm0 ; AVX512VL-NEXT: retq @@ -1308,14 +1282,13 @@ define <2 x i64> @splatvar_funnnel_v2i64(<2 x i64> %x, <2 x i64> %y, <2 x i64> % ; AVX512BW: # %bb.0: ; AVX512BW-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0 ; AVX512BW-NEXT: vpbroadcastq %xmm2, %xmm2 -; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63] -; AVX512BW-NEXT: vpand %xmm3, %xmm2, %xmm4 -; AVX512BW-NEXT: vpsllq %xmm4, %xmm0, %xmm5 -; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm6 = [64,64] -; AVX512BW-NEXT: vpsubq %xmm4, %xmm6, %xmm4 +; AVX512BW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2 +; AVX512BW-NEXT: vpsllq %xmm2, %xmm0, %xmm3 +; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm4 = [64,64] +; AVX512BW-NEXT: vpsubq %xmm2, %xmm4, %xmm4 ; AVX512BW-NEXT: vpsrlq %xmm4, %xmm1, %xmm1 -; AVX512BW-NEXT: vpor %xmm1, %xmm5, %xmm1 -; AVX512BW-NEXT: vptestnmq %zmm3, %zmm2, %k1 +; AVX512BW-NEXT: vpor %xmm1, %xmm3, %xmm1 +; AVX512BW-NEXT: vptestnmq %zmm2, %zmm2, %k1 ; AVX512BW-NEXT: vmovdqa64 %zmm0, %zmm1 {%k1} ; AVX512BW-NEXT: vmovdqa %xmm1, %xmm0 ; AVX512BW-NEXT: vzeroupper @@ -1325,14 +1298,13 @@ define <2 x i64> @splatvar_funnnel_v2i64(<2 x i64> %x, <2 x i64> %y, <2 x i64> % ; AVX512VBMI2: # %bb.0: ; AVX512VBMI2-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0 ; AVX512VBMI2-NEXT: vpbroadcastq %xmm2, %xmm2 -; AVX512VBMI2-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63] -; AVX512VBMI2-NEXT: vpand %xmm3, %xmm2, %xmm4 -; AVX512VBMI2-NEXT: vpsllq %xmm4, %xmm0, %xmm5 -; AVX512VBMI2-NEXT: vmovdqa {{.*#+}} xmm6 = [64,64] -; AVX512VBMI2-NEXT: vpsubq %xmm4, %xmm6, %xmm4 +; AVX512VBMI2-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2 +; AVX512VBMI2-NEXT: vpsllq %xmm2, %xmm0, %xmm3 +; AVX512VBMI2-NEXT: vmovdqa {{.*#+}} xmm4 = [64,64] +; AVX512VBMI2-NEXT: vpsubq %xmm2, %xmm4, %xmm4 ; AVX512VBMI2-NEXT: vpsrlq %xmm4, %xmm1, %xmm1 -; AVX512VBMI2-NEXT: vpor %xmm1, %xmm5, %xmm1 -; AVX512VBMI2-NEXT: vptestnmq %zmm3, %zmm2, %k1 +; AVX512VBMI2-NEXT: vpor %xmm1, %xmm3, %xmm1 +; AVX512VBMI2-NEXT: vptestnmq %zmm2, %zmm2, %k1 ; AVX512VBMI2-NEXT: vmovdqa64 %zmm0, %zmm1 {%k1} ; AVX512VBMI2-NEXT: vmovdqa %xmm1, %xmm0 ; AVX512VBMI2-NEXT: vzeroupper @@ -1341,14 +1313,13 @@ define <2 x i64> @splatvar_funnnel_v2i64(<2 x i64> %x, <2 x i64> %y, <2 x i64> % ; AVX512VLBW-LABEL: splatvar_funnnel_v2i64: ; AVX512VLBW: # %bb.0: ; AVX512VLBW-NEXT: vpbroadcastq %xmm2, %xmm2 -; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63] -; AVX512VLBW-NEXT: vpand %xmm3, %xmm2, %xmm4 -; AVX512VLBW-NEXT: vpsllq %xmm4, %xmm0, %xmm5 -; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm6 = [64,64] -; AVX512VLBW-NEXT: vpsubq %xmm4, %xmm6, %xmm4 +; AVX512VLBW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2 +; AVX512VLBW-NEXT: vpsllq %xmm2, %xmm0, %xmm3 +; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm4 = [64,64] +; AVX512VLBW-NEXT: vpsubq %xmm2, %xmm4, %xmm4 ; AVX512VLBW-NEXT: vpsrlq %xmm4, %xmm1, %xmm1 -; AVX512VLBW-NEXT: vpor %xmm1, %xmm5, %xmm1 -; AVX512VLBW-NEXT: vptestnmq %xmm3, %xmm2, %k1 +; AVX512VLBW-NEXT: vpor %xmm1, %xmm3, %xmm1 +; AVX512VLBW-NEXT: vptestnmq %xmm2, %xmm2, %k1 ; AVX512VLBW-NEXT: vmovdqa64 %xmm0, %xmm1 {%k1} ; AVX512VLBW-NEXT: vmovdqa %xmm1, %xmm0 ; AVX512VLBW-NEXT: retq @@ -1497,15 +1468,15 @@ define <4 x i32> @splatvar_funnnel_v4i32(<4 x i32> %x, <4 x i32> %y, <4 x i32> % ; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0 ; AVX512F-NEXT: vpbroadcastd %xmm2, %xmm2 ; AVX512F-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31] -; AVX512F-NEXT: vpand %xmm3, %xmm2, %xmm4 -; AVX512F-NEXT: vpmovzxdq {{.*#+}} xmm5 = xmm4[0],zero,xmm4[1],zero -; AVX512F-NEXT: vpslld %xmm5, %xmm0, %xmm5 -; AVX512F-NEXT: vpbroadcastd {{.*#+}} xmm6 = [32,32,32,32] -; AVX512F-NEXT: vpsubd %xmm4, %xmm6, %xmm4 +; AVX512F-NEXT: vpand %xmm3, %xmm2, %xmm2 +; AVX512F-NEXT: vpmovzxdq {{.*#+}} xmm3 = xmm2[0],zero,xmm2[1],zero +; AVX512F-NEXT: vpslld %xmm3, %xmm0, %xmm3 +; AVX512F-NEXT: vpbroadcastd {{.*#+}} xmm4 = [32,32,32,32] +; AVX512F-NEXT: vpsubd %xmm2, %xmm4, %xmm4 ; AVX512F-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero ; AVX512F-NEXT: vpsrld %xmm4, %xmm1, %xmm1 -; AVX512F-NEXT: vpor %xmm1, %xmm5, %xmm1 -; AVX512F-NEXT: vptestnmd %zmm3, %zmm2, %k1 +; AVX512F-NEXT: vpor %xmm1, %xmm3, %xmm1 +; AVX512F-NEXT: vptestnmd %zmm2, %zmm2, %k1 ; AVX512F-NEXT: vmovdqa32 %zmm0, %zmm1 {%k1} ; AVX512F-NEXT: vmovdqa %xmm1, %xmm0 ; AVX512F-NEXT: vzeroupper @@ -1514,16 +1485,15 @@ define <4 x i32> @splatvar_funnnel_v4i32(<4 x i32> %x, <4 x i32> %y, <4 x i32> % ; AVX512VL-LABEL: splatvar_funnnel_v4i32: ; AVX512VL: # %bb.0: ; AVX512VL-NEXT: vpbroadcastd %xmm2, %xmm2 -; AVX512VL-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31] -; AVX512VL-NEXT: vpand %xmm3, %xmm2, %xmm4 -; AVX512VL-NEXT: vpmovzxdq {{.*#+}} xmm5 = xmm4[0],zero,xmm4[1],zero -; AVX512VL-NEXT: vpslld %xmm5, %xmm0, %xmm5 -; AVX512VL-NEXT: vpbroadcastd {{.*#+}} xmm6 = [32,32,32,32] -; AVX512VL-NEXT: vpsubd %xmm4, %xmm6, %xmm4 +; AVX512VL-NEXT: vpandd {{.*}}(%rip){1to4}, %xmm2, %xmm2 +; AVX512VL-NEXT: vpmovzxdq {{.*#+}} xmm3 = xmm2[0],zero,xmm2[1],zero +; AVX512VL-NEXT: vpslld %xmm3, %xmm0, %xmm3 +; AVX512VL-NEXT: vpbroadcastd {{.*#+}} xmm4 = [32,32,32,32] +; AVX512VL-NEXT: vpsubd %xmm2, %xmm4, %xmm4 ; AVX512VL-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero ; AVX512VL-NEXT: vpsrld %xmm4, %xmm1, %xmm1 -; AVX512VL-NEXT: vpor %xmm1, %xmm5, %xmm1 -; AVX512VL-NEXT: vptestnmd %xmm3, %xmm2, %k1 +; AVX512VL-NEXT: vpor %xmm1, %xmm3, %xmm1 +; AVX512VL-NEXT: vptestnmd %xmm2, %xmm2, %k1 ; AVX512VL-NEXT: vmovdqa32 %xmm0, %xmm1 {%k1} ; AVX512VL-NEXT: vmovdqa %xmm1, %xmm0 ; AVX512VL-NEXT: retq @@ -1533,15 +1503,15 @@ define <4 x i32> @splatvar_funnnel_v4i32(<4 x i32> %x, <4 x i32> %y, <4 x i32> % ; AVX512BW-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0 ; AVX512BW-NEXT: vpbroadcastd %xmm2, %xmm2 ; AVX512BW-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31] -; AVX512BW-NEXT: vpand %xmm3, %xmm2, %xmm4 -; AVX512BW-NEXT: vpmovzxdq {{.*#+}} xmm5 = xmm4[0],zero,xmm4[1],zero -; AVX512BW-NEXT: vpslld %xmm5, %xmm0, %xmm5 -; AVX512BW-NEXT: vpbroadcastd {{.*#+}} xmm6 = [32,32,32,32] -; AVX512BW-NEXT: vpsubd %xmm4, %xmm6, %xmm4 +; AVX512BW-NEXT: vpand %xmm3, %xmm2, %xmm2 +; AVX512BW-NEXT: vpmovzxdq {{.*#+}} xmm3 = xmm2[0],zero,xmm2[1],zero +; AVX512BW-NEXT: vpslld %xmm3, %xmm0, %xmm3 +; AVX512BW-NEXT: vpbroadcastd {{.*#+}} xmm4 = [32,32,32,32] +; AVX512BW-NEXT: vpsubd %xmm2, %xmm4, %xmm4 ; AVX512BW-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero ; AVX512BW-NEXT: vpsrld %xmm4, %xmm1, %xmm1 -; AVX512BW-NEXT: vpor %xmm1, %xmm5, %xmm1 -; AVX512BW-NEXT: vptestnmd %zmm3, %zmm2, %k1 +; AVX512BW-NEXT: vpor %xmm1, %xmm3, %xmm1 +; AVX512BW-NEXT: vptestnmd %zmm2, %zmm2, %k1 ; AVX512BW-NEXT: vmovdqa32 %zmm0, %zmm1 {%k1} ; AVX512BW-NEXT: vmovdqa %xmm1, %xmm0 ; AVX512BW-NEXT: vzeroupper @@ -1552,15 +1522,15 @@ define <4 x i32> @splatvar_funnnel_v4i32(<4 x i32> %x, <4 x i32> %y, <4 x i32> % ; AVX512VBMI2-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0 ; AVX512VBMI2-NEXT: vpbroadcastd %xmm2, %xmm2 ; AVX512VBMI2-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31] -; AVX512VBMI2-NEXT: vpand %xmm3, %xmm2, %xmm4 -; AVX512VBMI2-NEXT: vpmovzxdq {{.*#+}} xmm5 = xmm4[0],zero,xmm4[1],zero -; AVX512VBMI2-NEXT: vpslld %xmm5, %xmm0, %xmm5 -; AVX512VBMI2-NEXT: vpbroadcastd {{.*#+}} xmm6 = [32,32,32,32] -; AVX512VBMI2-NEXT: vpsubd %xmm4, %xmm6, %xmm4 +; AVX512VBMI2-NEXT: vpand %xmm3, %xmm2, %xmm2 +; AVX512VBMI2-NEXT: vpmovzxdq {{.*#+}} xmm3 = xmm2[0],zero,xmm2[1],zero +; AVX512VBMI2-NEXT: vpslld %xmm3, %xmm0, %xmm3 +; AVX512VBMI2-NEXT: vpbroadcastd {{.*#+}} xmm4 = [32,32,32,32] +; AVX512VBMI2-NEXT: vpsubd %xmm2, %xmm4, %xmm4 ; AVX512VBMI2-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero ; AVX512VBMI2-NEXT: vpsrld %xmm4, %xmm1, %xmm1 -; AVX512VBMI2-NEXT: vpor %xmm1, %xmm5, %xmm1 -; AVX512VBMI2-NEXT: vptestnmd %zmm3, %zmm2, %k1 +; AVX512VBMI2-NEXT: vpor %xmm1, %xmm3, %xmm1 +; AVX512VBMI2-NEXT: vptestnmd %zmm2, %zmm2, %k1 ; AVX512VBMI2-NEXT: vmovdqa32 %zmm0, %zmm1 {%k1} ; AVX512VBMI2-NEXT: vmovdqa %xmm1, %xmm0 ; AVX512VBMI2-NEXT: vzeroupper @@ -1569,16 +1539,15 @@ define <4 x i32> @splatvar_funnnel_v4i32(<4 x i32> %x, <4 x i32> %y, <4 x i32> % ; AVX512VLBW-LABEL: splatvar_funnnel_v4i32: ; AVX512VLBW: # %bb.0: ; AVX512VLBW-NEXT: vpbroadcastd %xmm2, %xmm2 -; AVX512VLBW-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31] -; AVX512VLBW-NEXT: vpand %xmm3, %xmm2, %xmm4 -; AVX512VLBW-NEXT: vpmovzxdq {{.*#+}} xmm5 = xmm4[0],zero,xmm4[1],zero -; AVX512VLBW-NEXT: vpslld %xmm5, %xmm0, %xmm5 -; AVX512VLBW-NEXT: vpbroadcastd {{.*#+}} xmm6 = [32,32,32,32] -; AVX512VLBW-NEXT: vpsubd %xmm4, %xmm6, %xmm4 +; AVX512VLBW-NEXT: vpandd {{.*}}(%rip){1to4}, %xmm2, %xmm2 +; AVX512VLBW-NEXT: vpmovzxdq {{.*#+}} xmm3 = xmm2[0],zero,xmm2[1],zero +; AVX512VLBW-NEXT: vpslld %xmm3, %xmm0, %xmm3 +; AVX512VLBW-NEXT: vpbroadcastd {{.*#+}} xmm4 = [32,32,32,32] +; AVX512VLBW-NEXT: vpsubd %xmm2, %xmm4, %xmm4 ; AVX512VLBW-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero ; AVX512VLBW-NEXT: vpsrld %xmm4, %xmm1, %xmm1 -; AVX512VLBW-NEXT: vpor %xmm1, %xmm5, %xmm1 -; AVX512VLBW-NEXT: vptestnmd %xmm3, %xmm2, %k1 +; AVX512VLBW-NEXT: vpor %xmm1, %xmm3, %xmm1 +; AVX512VLBW-NEXT: vptestnmd %xmm2, %xmm2, %k1 ; AVX512VLBW-NEXT: vmovdqa32 %xmm0, %xmm1 {%k1} ; AVX512VLBW-NEXT: vmovdqa %xmm1, %xmm0 ; AVX512VLBW-NEXT: retq @@ -1759,16 +1728,15 @@ define <8 x i16> @splatvar_funnnel_v8i16(<8 x i16> %x, <8 x i16> %y, <8 x i16> % ; AVX512BW: # %bb.0: ; AVX512BW-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0 ; AVX512BW-NEXT: vpbroadcastw %xmm2, %xmm2 -; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15] -; AVX512BW-NEXT: vpand %xmm3, %xmm2, %xmm4 -; AVX512BW-NEXT: vpmovzxwq {{.*#+}} xmm5 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero -; AVX512BW-NEXT: vpsllw %xmm5, %xmm0, %xmm5 -; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm6 = [16,16,16,16,16,16,16,16] -; AVX512BW-NEXT: vpsubw %xmm4, %xmm6, %xmm4 +; AVX512BW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2 +; AVX512BW-NEXT: vpmovzxwq {{.*#+}} xmm3 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero +; AVX512BW-NEXT: vpsllw %xmm3, %xmm0, %xmm3 +; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm4 = [16,16,16,16,16,16,16,16] +; AVX512BW-NEXT: vpsubw %xmm2, %xmm4, %xmm4 ; AVX512BW-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero ; AVX512BW-NEXT: vpsrlw %xmm4, %xmm1, %xmm1 -; AVX512BW-NEXT: vpor %xmm1, %xmm5, %xmm1 -; AVX512BW-NEXT: vptestnmw %zmm3, %zmm2, %k1 +; AVX512BW-NEXT: vpor %xmm1, %xmm3, %xmm1 +; AVX512BW-NEXT: vptestnmw %zmm2, %zmm2, %k1 ; AVX512BW-NEXT: vmovdqu16 %zmm0, %zmm1 {%k1} ; AVX512BW-NEXT: vmovdqa %xmm1, %xmm0 ; AVX512BW-NEXT: vzeroupper @@ -1778,16 +1746,15 @@ define <8 x i16> @splatvar_funnnel_v8i16(<8 x i16> %x, <8 x i16> %y, <8 x i16> % ; AVX512VBMI2: # %bb.0: ; AVX512VBMI2-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0 ; AVX512VBMI2-NEXT: vpbroadcastw %xmm2, %xmm2 -; AVX512VBMI2-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15] -; AVX512VBMI2-NEXT: vpand %xmm3, %xmm2, %xmm4 -; AVX512VBMI2-NEXT: vpmovzxwq {{.*#+}} xmm5 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero -; AVX512VBMI2-NEXT: vpsllw %xmm5, %xmm0, %xmm5 -; AVX512VBMI2-NEXT: vmovdqa {{.*#+}} xmm6 = [16,16,16,16,16,16,16,16] -; AVX512VBMI2-NEXT: vpsubw %xmm4, %xmm6, %xmm4 +; AVX512VBMI2-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2 +; AVX512VBMI2-NEXT: vpmovzxwq {{.*#+}} xmm3 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero +; AVX512VBMI2-NEXT: vpsllw %xmm3, %xmm0, %xmm3 +; AVX512VBMI2-NEXT: vmovdqa {{.*#+}} xmm4 = [16,16,16,16,16,16,16,16] +; AVX512VBMI2-NEXT: vpsubw %xmm2, %xmm4, %xmm4 ; AVX512VBMI2-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero ; AVX512VBMI2-NEXT: vpsrlw %xmm4, %xmm1, %xmm1 -; AVX512VBMI2-NEXT: vpor %xmm1, %xmm5, %xmm1 -; AVX512VBMI2-NEXT: vptestnmw %zmm3, %zmm2, %k1 +; AVX512VBMI2-NEXT: vpor %xmm1, %xmm3, %xmm1 +; AVX512VBMI2-NEXT: vptestnmw %zmm2, %zmm2, %k1 ; AVX512VBMI2-NEXT: vmovdqu16 %zmm0, %zmm1 {%k1} ; AVX512VBMI2-NEXT: vmovdqa %xmm1, %xmm0 ; AVX512VBMI2-NEXT: vzeroupper @@ -1796,16 +1763,15 @@ define <8 x i16> @splatvar_funnnel_v8i16(<8 x i16> %x, <8 x i16> %y, <8 x i16> % ; AVX512VLBW-LABEL: splatvar_funnnel_v8i16: ; AVX512VLBW: # %bb.0: ; AVX512VLBW-NEXT: vpbroadcastw %xmm2, %xmm2 -; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15] -; AVX512VLBW-NEXT: vpand %xmm3, %xmm2, %xmm4 -; AVX512VLBW-NEXT: vpmovzxwq {{.*#+}} xmm5 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero -; AVX512VLBW-NEXT: vpsllw %xmm5, %xmm0, %xmm5 -; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm6 = [16,16,16,16,16,16,16,16] -; AVX512VLBW-NEXT: vpsubw %xmm4, %xmm6, %xmm4 +; AVX512VLBW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2 +; AVX512VLBW-NEXT: vpmovzxwq {{.*#+}} xmm3 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero +; AVX512VLBW-NEXT: vpsllw %xmm3, %xmm0, %xmm3 +; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm4 = [16,16,16,16,16,16,16,16] +; AVX512VLBW-NEXT: vpsubw %xmm2, %xmm4, %xmm4 ; AVX512VLBW-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero ; AVX512VLBW-NEXT: vpsrlw %xmm4, %xmm1, %xmm1 -; AVX512VLBW-NEXT: vpor %xmm1, %xmm5, %xmm1 -; AVX512VLBW-NEXT: vptestnmw %xmm3, %xmm2, %k1 +; AVX512VLBW-NEXT: vpor %xmm1, %xmm3, %xmm1 +; AVX512VLBW-NEXT: vptestnmw %xmm2, %xmm2, %k1 ; AVX512VLBW-NEXT: vmovdqu16 %xmm0, %xmm1 {%k1} ; AVX512VLBW-NEXT: vmovdqa %xmm1, %xmm0 ; AVX512VLBW-NEXT: retq @@ -2031,18 +1997,17 @@ define <16 x i8> @splatvar_funnnel_v16i8(<16 x i8> %x, <16 x i8> %y, <16 x i8> % ; AVX512BW-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0 ; AVX512BW-NEXT: vpbroadcastb %xmm2, %xmm2 ; AVX512BW-NEXT: vpmovzxbw {{.*#+}} ymm3 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero -; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm4 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7] -; AVX512BW-NEXT: vpand %xmm4, %xmm2, %xmm5 -; AVX512BW-NEXT: vpmovzxbw {{.*#+}} ymm6 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero,xmm5[8],zero,xmm5[9],zero,xmm5[10],zero,xmm5[11],zero,xmm5[12],zero,xmm5[13],zero,xmm5[14],zero,xmm5[15],zero -; AVX512BW-NEXT: vpsllvw %zmm6, %zmm3, %zmm3 +; AVX512BW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2 +; AVX512BW-NEXT: vpmovzxbw {{.*#+}} ymm4 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero,xmm2[8],zero,xmm2[9],zero,xmm2[10],zero,xmm2[11],zero,xmm2[12],zero,xmm2[13],zero,xmm2[14],zero,xmm2[15],zero +; AVX512BW-NEXT: vpsllvw %zmm4, %zmm3, %zmm3 ; AVX512BW-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero -; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm6 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8] -; AVX512BW-NEXT: vpsubb %xmm5, %xmm6, %xmm5 -; AVX512BW-NEXT: vpmovzxbw {{.*#+}} ymm5 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero,xmm5[8],zero,xmm5[9],zero,xmm5[10],zero,xmm5[11],zero,xmm5[12],zero,xmm5[13],zero,xmm5[14],zero,xmm5[15],zero -; AVX512BW-NEXT: vpsrlvw %zmm5, %zmm1, %zmm1 +; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm4 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8] +; AVX512BW-NEXT: vpsubb %xmm2, %xmm4, %xmm4 +; AVX512BW-NEXT: vpmovzxbw {{.*#+}} ymm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero,xmm4[8],zero,xmm4[9],zero,xmm4[10],zero,xmm4[11],zero,xmm4[12],zero,xmm4[13],zero,xmm4[14],zero,xmm4[15],zero +; AVX512BW-NEXT: vpsrlvw %zmm4, %zmm1, %zmm1 ; AVX512BW-NEXT: vpor %ymm1, %ymm3, %ymm1 ; AVX512BW-NEXT: vpmovwb %zmm1, %ymm1 -; AVX512BW-NEXT: vptestnmb %zmm4, %zmm2, %k1 +; AVX512BW-NEXT: vptestnmb %zmm2, %zmm2, %k1 ; AVX512BW-NEXT: vmovdqu8 %zmm0, %zmm1 {%k1} ; AVX512BW-NEXT: vmovdqa %xmm1, %xmm0 ; AVX512BW-NEXT: vzeroupper @@ -2053,18 +2018,17 @@ define <16 x i8> @splatvar_funnnel_v16i8(<16 x i8> %x, <16 x i8> %y, <16 x i8> % ; AVX512VBMI2-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0 ; AVX512VBMI2-NEXT: vpbroadcastb %xmm2, %xmm2 ; AVX512VBMI2-NEXT: vpmovzxbw {{.*#+}} ymm3 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero -; AVX512VBMI2-NEXT: vmovdqa {{.*#+}} xmm4 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7] -; AVX512VBMI2-NEXT: vpand %xmm4, %xmm2, %xmm5 -; AVX512VBMI2-NEXT: vpmovzxbw {{.*#+}} ymm6 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero,xmm5[8],zero,xmm5[9],zero,xmm5[10],zero,xmm5[11],zero,xmm5[12],zero,xmm5[13],zero,xmm5[14],zero,xmm5[15],zero -; AVX512VBMI2-NEXT: vpsllvw %zmm6, %zmm3, %zmm3 +; AVX512VBMI2-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2 +; AVX512VBMI2-NEXT: vpmovzxbw {{.*#+}} ymm4 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero,xmm2[8],zero,xmm2[9],zero,xmm2[10],zero,xmm2[11],zero,xmm2[12],zero,xmm2[13],zero,xmm2[14],zero,xmm2[15],zero +; AVX512VBMI2-NEXT: vpsllvw %zmm4, %zmm3, %zmm3 ; AVX512VBMI2-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero -; AVX512VBMI2-NEXT: vmovdqa {{.*#+}} xmm6 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8] -; AVX512VBMI2-NEXT: vpsubb %xmm5, %xmm6, %xmm5 -; AVX512VBMI2-NEXT: vpmovzxbw {{.*#+}} ymm5 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero,xmm5[8],zero,xmm5[9],zero,xmm5[10],zero,xmm5[11],zero,xmm5[12],zero,xmm5[13],zero,xmm5[14],zero,xmm5[15],zero -; AVX512VBMI2-NEXT: vpsrlvw %zmm5, %zmm1, %zmm1 +; AVX512VBMI2-NEXT: vmovdqa {{.*#+}} xmm4 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8] +; AVX512VBMI2-NEXT: vpsubb %xmm2, %xmm4, %xmm4 +; AVX512VBMI2-NEXT: vpmovzxbw {{.*#+}} ymm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero,xmm4[8],zero,xmm4[9],zero,xmm4[10],zero,xmm4[11],zero,xmm4[12],zero,xmm4[13],zero,xmm4[14],zero,xmm4[15],zero +; AVX512VBMI2-NEXT: vpsrlvw %zmm4, %zmm1, %zmm1 ; AVX512VBMI2-NEXT: vpor %ymm1, %ymm3, %ymm1 ; AVX512VBMI2-NEXT: vpmovwb %zmm1, %ymm1 -; AVX512VBMI2-NEXT: vptestnmb %zmm4, %zmm2, %k1 +; AVX512VBMI2-NEXT: vptestnmb %zmm2, %zmm2, %k1 ; AVX512VBMI2-NEXT: vmovdqu8 %zmm0, %zmm1 {%k1} ; AVX512VBMI2-NEXT: vmovdqa %xmm1, %xmm0 ; AVX512VBMI2-NEXT: vzeroupper @@ -2073,19 +2037,18 @@ define <16 x i8> @splatvar_funnnel_v16i8(<16 x i8> %x, <16 x i8> %y, <16 x i8> % ; AVX512VLBW-LABEL: splatvar_funnnel_v16i8: ; AVX512VLBW: # %bb.0: ; AVX512VLBW-NEXT: vpbroadcastb %xmm2, %xmm2 -; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7] -; AVX512VLBW-NEXT: vpand %xmm3, %xmm2, %xmm4 -; AVX512VLBW-NEXT: vpmovzxbw {{.*#+}} ymm5 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero,xmm4[8],zero,xmm4[9],zero,xmm4[10],zero,xmm4[11],zero,xmm4[12],zero,xmm4[13],zero,xmm4[14],zero,xmm4[15],zero -; AVX512VLBW-NEXT: vpmovzxbw {{.*#+}} ymm6 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero -; AVX512VLBW-NEXT: vpsllvw %ymm5, %ymm6, %ymm5 -; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm6 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8] -; AVX512VLBW-NEXT: vpsubb %xmm4, %xmm6, %xmm4 +; AVX512VLBW-NEXT: vpmovzxbw {{.*#+}} ymm3 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero +; AVX512VLBW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2 +; AVX512VLBW-NEXT: vpmovzxbw {{.*#+}} ymm4 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero,xmm2[8],zero,xmm2[9],zero,xmm2[10],zero,xmm2[11],zero,xmm2[12],zero,xmm2[13],zero,xmm2[14],zero,xmm2[15],zero +; AVX512VLBW-NEXT: vpsllvw %ymm4, %ymm3, %ymm3 +; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm4 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8] +; AVX512VLBW-NEXT: vpsubb %xmm2, %xmm4, %xmm4 ; AVX512VLBW-NEXT: vpmovzxbw {{.*#+}} ymm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero,xmm4[8],zero,xmm4[9],zero,xmm4[10],zero,xmm4[11],zero,xmm4[12],zero,xmm4[13],zero,xmm4[14],zero,xmm4[15],zero ; AVX512VLBW-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero ; AVX512VLBW-NEXT: vpsrlvw %ymm4, %ymm1, %ymm1 -; AVX512VLBW-NEXT: vpor %ymm1, %ymm5, %ymm1 +; AVX512VLBW-NEXT: vpor %ymm1, %ymm3, %ymm1 ; AVX512VLBW-NEXT: vpmovwb %ymm1, %xmm1 -; AVX512VLBW-NEXT: vptestnmb %xmm3, %xmm2, %k1 +; AVX512VLBW-NEXT: vptestnmb %xmm2, %xmm2, %k1 ; AVX512VLBW-NEXT: vmovdqu8 %xmm0, %xmm1 {%k1} ; AVX512VLBW-NEXT: vmovdqa %xmm1, %xmm0 ; AVX512VLBW-NEXT: vzeroupper @@ -2094,19 +2057,18 @@ define <16 x i8> @splatvar_funnnel_v16i8(<16 x i8> %x, <16 x i8> %y, <16 x i8> % ; AVX512VLVBMI2-LABEL: splatvar_funnnel_v16i8: ; AVX512VLVBMI2: # %bb.0: ; AVX512VLVBMI2-NEXT: vpbroadcastb %xmm2, %xmm2 -; AVX512VLVBMI2-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7] -; AVX512VLVBMI2-NEXT: vpand %xmm3, %xmm2, %xmm4 -; AVX512VLVBMI2-NEXT: vpmovzxbw {{.*#+}} ymm5 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero,xmm4[8],zero,xmm4[9],zero,xmm4[10],zero,xmm4[11],zero,xmm4[12],zero,xmm4[13],zero,xmm4[14],zero,xmm4[15],zero -; AVX512VLVBMI2-NEXT: vpmovzxbw {{.*#+}} ymm6 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero -; AVX512VLVBMI2-NEXT: vpsllvw %ymm5, %ymm6, %ymm5 -; AVX512VLVBMI2-NEXT: vmovdqa {{.*#+}} xmm6 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8] -; AVX512VLVBMI2-NEXT: vpsubb %xmm4, %xmm6, %xmm4 +; AVX512VLVBMI2-NEXT: vpmovzxbw {{.*#+}} ymm3 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero +; AVX512VLVBMI2-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2 +; AVX512VLVBMI2-NEXT: vpmovzxbw {{.*#+}} ymm4 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero,xmm2[8],zero,xmm2[9],zero,xmm2[10],zero,xmm2[11],zero,xmm2[12],zero,xmm2[13],zero,xmm2[14],zero,xmm2[15],zero +; AVX512VLVBMI2-NEXT: vpsllvw %ymm4, %ymm3, %ymm3 +; AVX512VLVBMI2-NEXT: vmovdqa {{.*#+}} xmm4 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8] +; AVX512VLVBMI2-NEXT: vpsubb %xmm2, %xmm4, %xmm4 ; AVX512VLVBMI2-NEXT: vpmovzxbw {{.*#+}} ymm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero,xmm4[8],zero,xmm4[9],zero,xmm4[10],zero,xmm4[11],zero,xmm4[12],zero,xmm4[13],zero,xmm4[14],zero,xmm4[15],zero ; AVX512VLVBMI2-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero ; AVX512VLVBMI2-NEXT: vpsrlvw %ymm4, %ymm1, %ymm1 -; AVX512VLVBMI2-NEXT: vpor %ymm1, %ymm5, %ymm1 +; AVX512VLVBMI2-NEXT: vpor %ymm1, %ymm3, %ymm1 ; AVX512VLVBMI2-NEXT: vpmovwb %ymm1, %xmm1 -; AVX512VLVBMI2-NEXT: vptestnmb %xmm3, %xmm2, %k1 +; AVX512VLVBMI2-NEXT: vptestnmb %xmm2, %xmm2, %k1 ; AVX512VLVBMI2-NEXT: vmovdqu8 %xmm0, %xmm1 {%k1} ; AVX512VLVBMI2-NEXT: vmovdqa %xmm1, %xmm0 ; AVX512VLVBMI2-NEXT: vzeroupper |
