[llvm] 86dabf4 - [DAG] SelectionDAG::isSplatValue - handle OR/XOR cases
Simon Pilgrim via llvm-commits
llvm-commits at lists.llvm.org
Sun Feb 7 05:28:24 PST 2021
Author: Simon Pilgrim
Date: 2021-02-07T13:27:57Z
New Revision: 86dabf4226af2dac9c62468ff5000ea93cbec764
URL: https://github.com/llvm/llvm-project/commit/86dabf4226af2dac9c62468ff5000ea93cbec764
DIFF: https://github.com/llvm/llvm-project/commit/86dabf4226af2dac9c62468ff5000ea93cbec764.diff
LOG: [DAG] SelectionDAG::isSplatValue - handle OR/XOR cases
Add OR/XOR to the basic binops that we support when checking for a splat vector value
Added:
Modified:
llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
llvm/test/CodeGen/X86/vector-fshl-128.ll
llvm/test/CodeGen/X86/vector-fshl-256.ll
llvm/test/CodeGen/X86/vector-fshl-512.ll
llvm/test/CodeGen/X86/vector-fshr-128.ll
llvm/test/CodeGen/X86/vector-fshr-256.ll
llvm/test/CodeGen/X86/vector-fshr-512.ll
Removed:
################################################################################
diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
index c9bb50cb5c8b..f01b5b3092c6 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
@@ -2430,7 +2430,9 @@ bool SelectionDAG::isSplatValue(SDValue V, const APInt &DemandedElts,
return true;
case ISD::ADD:
case ISD::SUB:
- case ISD::AND: {
+ case ISD::AND:
+ case ISD::XOR:
+ case ISD::OR: {
APInt UndefLHS, UndefRHS;
SDValue LHS = V.getOperand(0);
SDValue RHS = V.getOperand(1);
diff --git a/llvm/test/CodeGen/X86/vector-fshl-128.ll b/llvm/test/CodeGen/X86/vector-fshl-128.ll
index 9b34604eb5a4..d74db60f7865 100644
--- a/llvm/test/CodeGen/X86/vector-fshl-128.ll
+++ b/llvm/test/CodeGen/X86/vector-fshl-128.ll
@@ -1075,98 +1075,59 @@ define <16 x i8> @var_funnnel_v16i8(<16 x i8> %x, <16 x i8> %y, <16 x i8> %amt)
;
define <2 x i64> @splatvar_funnnel_v2i64(<2 x i64> %x, <2 x i64> %y, <2 x i64> %amt) nounwind {
-; SSE2-LABEL: splatvar_funnnel_v2i64:
-; SSE2: # %bb.0:
-; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm2[0,1,0,1]
-; SSE2-NEXT: movdqa {{.*#+}} xmm4 = [63,63]
-; SSE2-NEXT: pandn %xmm4, %xmm3
-; SSE2-NEXT: psrlq $1, %xmm1
-; SSE2-NEXT: movdqa %xmm1, %xmm5
-; SSE2-NEXT: psrlq %xmm3, %xmm5
-; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm3[2,3,2,3]
-; SSE2-NEXT: psrlq %xmm3, %xmm1
-; SSE2-NEXT: movsd {{.*#+}} xmm1 = xmm5[0],xmm1[1]
-; SSE2-NEXT: pand %xmm4, %xmm2
-; SSE2-NEXT: psllq %xmm2, %xmm0
-; SSE2-NEXT: por %xmm1, %xmm0
-; SSE2-NEXT: retq
-;
-; SSE41-LABEL: splatvar_funnnel_v2i64:
-; SSE41: # %bb.0:
-; SSE41-NEXT: pshufd {{.*#+}} xmm3 = xmm2[0,1,0,1]
-; SSE41-NEXT: movdqa {{.*#+}} xmm4 = [63,63]
-; SSE41-NEXT: pandn %xmm4, %xmm3
-; SSE41-NEXT: psrlq $1, %xmm1
-; SSE41-NEXT: movdqa %xmm1, %xmm5
-; SSE41-NEXT: psrlq %xmm3, %xmm5
-; SSE41-NEXT: pshufd {{.*#+}} xmm3 = xmm3[2,3,2,3]
-; SSE41-NEXT: psrlq %xmm3, %xmm1
-; SSE41-NEXT: pblendw {{.*#+}} xmm1 = xmm5[0,1,2,3],xmm1[4,5,6,7]
-; SSE41-NEXT: pand %xmm4, %xmm2
-; SSE41-NEXT: psllq %xmm2, %xmm0
-; SSE41-NEXT: por %xmm1, %xmm0
-; SSE41-NEXT: retq
-;
-; AVX1-LABEL: splatvar_funnnel_v2i64:
-; AVX1: # %bb.0:
-; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm2[0,1,0,1]
-; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [63,63]
-; AVX1-NEXT: vpandn %xmm4, %xmm3, %xmm3
-; AVX1-NEXT: vpsrlq $1, %xmm1, %xmm1
-; AVX1-NEXT: vpsrlq %xmm3, %xmm1, %xmm5
-; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[2,3,2,3]
-; AVX1-NEXT: vpsrlq %xmm3, %xmm1, %xmm1
-; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm5[0,1,2,3],xmm1[4,5,6,7]
-; AVX1-NEXT: vpand %xmm4, %xmm2, %xmm2
-; AVX1-NEXT: vpsllq %xmm2, %xmm0, %xmm0
-; AVX1-NEXT: vpor %xmm1, %xmm0, %xmm0
-; AVX1-NEXT: retq
+; SSE-LABEL: splatvar_funnnel_v2i64:
+; SSE: # %bb.0:
+; SSE-NEXT: movdqa {{.*#+}} xmm3 = [63,63]
+; SSE-NEXT: movdqa %xmm2, %xmm4
+; SSE-NEXT: pandn %xmm3, %xmm4
+; SSE-NEXT: psrlq $1, %xmm1
+; SSE-NEXT: psrlq %xmm4, %xmm1
+; SSE-NEXT: pand %xmm3, %xmm2
+; SSE-NEXT: psllq %xmm2, %xmm0
+; SSE-NEXT: por %xmm1, %xmm0
+; SSE-NEXT: retq
;
-; AVX2-LABEL: splatvar_funnnel_v2i64:
-; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastq %xmm2, %xmm3
-; AVX2-NEXT: vmovdqa {{.*#+}} xmm4 = [63,63]
-; AVX2-NEXT: vpand %xmm4, %xmm2, %xmm2
-; AVX2-NEXT: vpsllq %xmm2, %xmm0, %xmm0
-; AVX2-NEXT: vpandn %xmm4, %xmm3, %xmm2
-; AVX2-NEXT: vpsrlq $1, %xmm1, %xmm1
-; AVX2-NEXT: vpsrlvq %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpor %xmm1, %xmm0, %xmm0
-; AVX2-NEXT: retq
+; AVX-LABEL: splatvar_funnnel_v2i64:
+; AVX: # %bb.0:
+; AVX-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; AVX-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX-NEXT: vpsrlq $1, %xmm1, %xmm1
+; AVX-NEXT: vpsrlq %xmm4, %xmm1, %xmm1
+; AVX-NEXT: vpand %xmm3, %xmm2, %xmm2
+; AVX-NEXT: vpsllq %xmm2, %xmm0, %xmm0
+; AVX-NEXT: vpor %xmm1, %xmm0, %xmm0
+; AVX-NEXT: retq
;
; AVX512F-LABEL: splatvar_funnnel_v2i64:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpbroadcastq %xmm2, %xmm3
-; AVX512F-NEXT: vmovdqa {{.*#+}} xmm4 = [63,63]
-; AVX512F-NEXT: vpand %xmm4, %xmm2, %xmm2
-; AVX512F-NEXT: vpsllq %xmm2, %xmm0, %xmm0
-; AVX512F-NEXT: vpandn %xmm4, %xmm3, %xmm2
+; AVX512F-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; AVX512F-NEXT: vpandn %xmm3, %xmm2, %xmm4
; AVX512F-NEXT: vpsrlq $1, %xmm1, %xmm1
-; AVX512F-NEXT: vpsrlvq %xmm2, %xmm1, %xmm1
+; AVX512F-NEXT: vpsrlq %xmm4, %xmm1, %xmm1
+; AVX512F-NEXT: vpand %xmm3, %xmm2, %xmm2
+; AVX512F-NEXT: vpsllq %xmm2, %xmm0, %xmm0
; AVX512F-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: splatvar_funnnel_v2i64:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastq %xmm2, %xmm3
-; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm4 = [63,63]
-; AVX512VL-NEXT: vpand %xmm4, %xmm2, %xmm2
-; AVX512VL-NEXT: vpsllq %xmm2, %xmm0, %xmm0
-; AVX512VL-NEXT: vpandn %xmm4, %xmm3, %xmm2
+; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; AVX512VL-NEXT: vpandn %xmm3, %xmm2, %xmm4
; AVX512VL-NEXT: vpsrlq $1, %xmm1, %xmm1
-; AVX512VL-NEXT: vpsrlvq %xmm2, %xmm1, %xmm1
+; AVX512VL-NEXT: vpsrlq %xmm4, %xmm1, %xmm1
+; AVX512VL-NEXT: vpand %xmm3, %xmm2, %xmm2
+; AVX512VL-NEXT: vpsllq %xmm2, %xmm0, %xmm0
; AVX512VL-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: splatvar_funnnel_v2i64:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpbroadcastq %xmm2, %xmm3
-; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm4 = [63,63]
-; AVX512BW-NEXT: vpand %xmm4, %xmm2, %xmm2
-; AVX512BW-NEXT: vpsllq %xmm2, %xmm0, %xmm0
-; AVX512BW-NEXT: vpandn %xmm4, %xmm3, %xmm2
+; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; AVX512BW-NEXT: vpandn %xmm3, %xmm2, %xmm4
; AVX512BW-NEXT: vpsrlq $1, %xmm1, %xmm1
-; AVX512BW-NEXT: vpsrlvq %xmm2, %xmm1, %xmm1
+; AVX512BW-NEXT: vpsrlq %xmm4, %xmm1, %xmm1
+; AVX512BW-NEXT: vpand %xmm3, %xmm2, %xmm2
+; AVX512BW-NEXT: vpsllq %xmm2, %xmm0, %xmm0
; AVX512BW-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512BW-NEXT: retq
;
@@ -1182,13 +1143,12 @@ 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, %xmm3
-; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm4 = [63,63]
-; AVX512VLBW-NEXT: vpand %xmm4, %xmm2, %xmm2
-; AVX512VLBW-NEXT: vpsllq %xmm2, %xmm0, %xmm0
-; AVX512VLBW-NEXT: vpandn %xmm4, %xmm3, %xmm2
+; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; AVX512VLBW-NEXT: vpandn %xmm3, %xmm2, %xmm4
; AVX512VLBW-NEXT: vpsrlq $1, %xmm1, %xmm1
-; AVX512VLBW-NEXT: vpsrlvq %xmm2, %xmm1, %xmm1
+; AVX512VLBW-NEXT: vpsrlq %xmm4, %xmm1, %xmm1
+; AVX512VLBW-NEXT: vpand %xmm3, %xmm2, %xmm2
+; AVX512VLBW-NEXT: vpsllq %xmm2, %xmm0, %xmm0
; AVX512VLBW-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512VLBW-NEXT: retq
;
@@ -1198,31 +1158,16 @@ define <2 x i64> @splatvar_funnnel_v2i64(<2 x i64> %x, <2 x i64> %y, <2 x i64> %
; AVX512VLVBMI2-NEXT: vpshldvq %xmm2, %xmm1, %xmm0
; AVX512VLVBMI2-NEXT: retq
;
-; XOPAVX1-LABEL: splatvar_funnnel_v2i64:
-; XOPAVX1: # %bb.0:
-; XOPAVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm2[0,1,0,1]
-; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [63,63]
-; XOPAVX1-NEXT: vpand %xmm4, %xmm2, %xmm2
-; XOPAVX1-NEXT: vpsllq %xmm2, %xmm0, %xmm0
-; XOPAVX1-NEXT: vpandn %xmm4, %xmm3, %xmm2
-; XOPAVX1-NEXT: vpxor %xmm3, %xmm3, %xmm3
-; XOPAVX1-NEXT: vpsubq %xmm2, %xmm3, %xmm2
-; XOPAVX1-NEXT: vpsrlq $1, %xmm1, %xmm1
-; XOPAVX1-NEXT: vpshlq %xmm2, %xmm1, %xmm1
-; XOPAVX1-NEXT: vpor %xmm1, %xmm0, %xmm0
-; XOPAVX1-NEXT: retq
-;
-; XOPAVX2-LABEL: splatvar_funnnel_v2i64:
-; XOPAVX2: # %bb.0:
-; XOPAVX2-NEXT: vpbroadcastq %xmm2, %xmm3
-; XOPAVX2-NEXT: vmovdqa {{.*#+}} xmm4 = [63,63]
-; XOPAVX2-NEXT: vpand %xmm4, %xmm2, %xmm2
-; XOPAVX2-NEXT: vpsllq %xmm2, %xmm0, %xmm0
-; XOPAVX2-NEXT: vpandn %xmm4, %xmm3, %xmm2
-; XOPAVX2-NEXT: vpsrlq $1, %xmm1, %xmm1
-; XOPAVX2-NEXT: vpsrlvq %xmm2, %xmm1, %xmm1
-; XOPAVX2-NEXT: vpor %xmm1, %xmm0, %xmm0
-; XOPAVX2-NEXT: retq
+; XOP-LABEL: splatvar_funnnel_v2i64:
+; XOP: # %bb.0:
+; XOP-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; XOP-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; XOP-NEXT: vpsrlq $1, %xmm1, %xmm1
+; XOP-NEXT: vpsrlq %xmm4, %xmm1, %xmm1
+; XOP-NEXT: vpand %xmm3, %xmm2, %xmm2
+; XOP-NEXT: vpsllq %xmm2, %xmm0, %xmm0
+; XOP-NEXT: vpor %xmm1, %xmm0, %xmm0
+; XOP-NEXT: retq
;
; X86-SSE2-LABEL: splatvar_funnnel_v2i64:
; X86-SSE2: # %bb.0:
@@ -1252,53 +1197,28 @@ define <2 x i64> @splatvar_funnnel_v2i64(<2 x i64> %x, <2 x i64> %y, <2 x i64> %
define <4 x i32> @splatvar_funnnel_v4i32(<4 x i32> %x, <4 x i32> %y, <4 x i32> %amt) nounwind {
; SSE2-LABEL: splatvar_funnnel_v4i32:
; SSE2: # %bb.0:
-; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm2[0,0,0,0]
-; SSE2-NEXT: psrld $1, %xmm1
-; SSE2-NEXT: pandn {{.*}}(%rip), %xmm3
-; SSE2-NEXT: pshuflw {{.*#+}} xmm4 = xmm3[2,3,3,3,4,5,6,7]
-; SSE2-NEXT: movdqa %xmm1, %xmm5
-; SSE2-NEXT: psrld %xmm4, %xmm5
-; SSE2-NEXT: pshuflw {{.*#+}} xmm4 = xmm3[0,1,1,1,4,5,6,7]
-; SSE2-NEXT: movdqa %xmm1, %xmm6
-; SSE2-NEXT: psrld %xmm4, %xmm6
-; SSE2-NEXT: punpcklqdq {{.*#+}} xmm6 = xmm6[0],xmm5[0]
-; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm3[2,3,2,3]
-; SSE2-NEXT: pshuflw {{.*#+}} xmm4 = xmm3[2,3,3,3,4,5,6,7]
-; SSE2-NEXT: movdqa %xmm1, %xmm5
-; SSE2-NEXT: psrld %xmm4, %xmm5
-; SSE2-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,1,1,1,4,5,6,7]
-; SSE2-NEXT: psrld %xmm3, %xmm1
-; SSE2-NEXT: punpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm5[1]
-; SSE2-NEXT: shufps {{.*#+}} xmm6 = xmm6[0,3],xmm1[0,3]
; SSE2-NEXT: movd %xmm2, %eax
+; SSE2-NEXT: movl %eax, %ecx
+; SSE2-NEXT: andl $31, %ecx
+; SSE2-NEXT: movd %ecx, %xmm2
+; SSE2-NEXT: pslld %xmm2, %xmm0
+; SSE2-NEXT: psrld $1, %xmm1
+; SSE2-NEXT: notl %eax
; SSE2-NEXT: andl $31, %eax
-; SSE2-NEXT: movd %eax, %xmm1
-; SSE2-NEXT: pslld %xmm1, %xmm0
-; SSE2-NEXT: por %xmm6, %xmm0
+; SSE2-NEXT: movd %eax, %xmm2
+; SSE2-NEXT: psrld %xmm2, %xmm1
+; SSE2-NEXT: por %xmm1, %xmm0
; SSE2-NEXT: retq
;
; SSE41-LABEL: splatvar_funnnel_v4i32:
; SSE41: # %bb.0:
-; SSE41-NEXT: pshufd {{.*#+}} xmm3 = xmm2[0,0,0,0]
-; SSE41-NEXT: movdqa {{.*#+}} xmm8 = [31,31,31,31]
-; SSE41-NEXT: pandn %xmm8, %xmm3
-; SSE41-NEXT: pshuflw {{.*#+}} xmm5 = xmm3[2,3,3,3,4,5,6,7]
+; SSE41-NEXT: movdqa {{.*#+}} xmm3 = [31,31,31,31]
+; SSE41-NEXT: movdqa %xmm2, %xmm4
+; SSE41-NEXT: pandn %xmm3, %xmm4
+; SSE41-NEXT: pmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
; SSE41-NEXT: psrld $1, %xmm1
-; SSE41-NEXT: movdqa %xmm1, %xmm6
-; SSE41-NEXT: psrld %xmm5, %xmm6
-; SSE41-NEXT: pshufd {{.*#+}} xmm5 = xmm3[2,3,2,3]
-; SSE41-NEXT: pshuflw {{.*#+}} xmm7 = xmm5[2,3,3,3,4,5,6,7]
-; SSE41-NEXT: movdqa %xmm1, %xmm4
-; SSE41-NEXT: psrld %xmm7, %xmm4
-; SSE41-NEXT: pblendw {{.*#+}} xmm4 = xmm6[0,1,2,3],xmm4[4,5,6,7]
-; SSE41-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,1,1,1,4,5,6,7]
-; SSE41-NEXT: movdqa %xmm1, %xmm6
-; SSE41-NEXT: psrld %xmm3, %xmm6
-; SSE41-NEXT: pshuflw {{.*#+}} xmm3 = xmm5[0,1,1,1,4,5,6,7]
-; SSE41-NEXT: psrld %xmm3, %xmm1
-; SSE41-NEXT: pblendw {{.*#+}} xmm1 = xmm6[0,1,2,3],xmm1[4,5,6,7]
-; SSE41-NEXT: pblendw {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,3],xmm1[4,5],xmm4[6,7]
-; SSE41-NEXT: pand %xmm8, %xmm2
+; SSE41-NEXT: psrld %xmm4, %xmm1
+; SSE41-NEXT: pand %xmm3, %xmm2
; SSE41-NEXT: pmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
; SSE41-NEXT: pslld %xmm2, %xmm0
; SSE41-NEXT: por %xmm1, %xmm0
@@ -1306,23 +1226,12 @@ define <4 x i32> @splatvar_funnnel_v4i32(<4 x i32> %x, <4 x i32> %y, <4 x i32> %
;
; AVX1-LABEL: splatvar_funnnel_v4i32:
; AVX1: # %bb.0:
-; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm2[0,0,0,0]
-; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [31,31,31,31]
-; AVX1-NEXT: vpandn %xmm4, %xmm3, %xmm3
-; AVX1-NEXT: vpsrldq {{.*#+}} xmm5 = xmm3[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [31,31,31,31]
+; AVX1-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX1-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
; AVX1-NEXT: vpsrld $1, %xmm1, %xmm1
-; AVX1-NEXT: vpsrld %xmm5, %xmm1, %xmm5
-; AVX1-NEXT: vpsrlq $32, %xmm3, %xmm6
-; AVX1-NEXT: vpsrld %xmm6, %xmm1, %xmm6
-; AVX1-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0,1,2,3],xmm5[4,5,6,7]
-; AVX1-NEXT: vpxor %xmm6, %xmm6, %xmm6
-; AVX1-NEXT: vpunpckhdq {{.*#+}} xmm6 = xmm3[2],xmm6[2],xmm3[3],xmm6[3]
-; AVX1-NEXT: vpsrld %xmm6, %xmm1, %xmm6
-; AVX1-NEXT: vpmovzxdq {{.*#+}} xmm3 = xmm3[0],zero,xmm3[1],zero
-; AVX1-NEXT: vpsrld %xmm3, %xmm1, %xmm1
-; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm6[4,5,6,7]
-; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm5[2,3],xmm1[4,5],xmm5[6,7]
-; AVX1-NEXT: vpand %xmm4, %xmm2, %xmm2
+; AVX1-NEXT: vpsrld %xmm4, %xmm1, %xmm1
+; AVX1-NEXT: vpand %xmm3, %xmm2, %xmm2
; AVX1-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
; AVX1-NEXT: vpslld %xmm2, %xmm0, %xmm0
; AVX1-NEXT: vpor %xmm1, %xmm0, %xmm0
@@ -1330,12 +1239,12 @@ define <4 x i32> @splatvar_funnnel_v4i32(<4 x i32> %x, <4 x i32> %y, <4 x i32> %
;
; AVX2-LABEL: splatvar_funnnel_v4i32:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastd %xmm2, %xmm3
-; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm4 = [31,31,31,31]
-; AVX2-NEXT: vpandn %xmm4, %xmm3, %xmm3
+; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31]
+; AVX2-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX2-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
; AVX2-NEXT: vpsrld $1, %xmm1, %xmm1
-; AVX2-NEXT: vpsrlvd %xmm3, %xmm1, %xmm1
-; AVX2-NEXT: vpand %xmm4, %xmm2, %xmm2
+; AVX2-NEXT: vpsrld %xmm4, %xmm1, %xmm1
+; AVX2-NEXT: vpand %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
; AVX2-NEXT: vpslld %xmm2, %xmm0, %xmm0
; AVX2-NEXT: vpor %xmm1, %xmm0, %xmm0
@@ -1343,12 +1252,12 @@ define <4 x i32> @splatvar_funnnel_v4i32(<4 x i32> %x, <4 x i32> %y, <4 x i32> %
;
; AVX512F-LABEL: splatvar_funnnel_v4i32:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpbroadcastd %xmm2, %xmm3
-; AVX512F-NEXT: vpbroadcastd {{.*#+}} xmm4 = [31,31,31,31]
-; AVX512F-NEXT: vpandn %xmm4, %xmm3, %xmm3
+; AVX512F-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31]
+; AVX512F-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX512F-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
; AVX512F-NEXT: vpsrld $1, %xmm1, %xmm1
-; AVX512F-NEXT: vpsrlvd %xmm3, %xmm1, %xmm1
-; AVX512F-NEXT: vpand %xmm4, %xmm2, %xmm2
+; AVX512F-NEXT: vpsrld %xmm4, %xmm1, %xmm1
+; AVX512F-NEXT: vpand %xmm3, %xmm2, %xmm2
; AVX512F-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
; AVX512F-NEXT: vpslld %xmm2, %xmm0, %xmm0
; AVX512F-NEXT: vpor %xmm1, %xmm0, %xmm0
@@ -1356,12 +1265,12 @@ 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, %xmm3
-; AVX512VL-NEXT: vpbroadcastd {{.*#+}} xmm4 = [31,31,31,31]
-; AVX512VL-NEXT: vpandn %xmm4, %xmm3, %xmm3
+; AVX512VL-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31]
+; AVX512VL-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX512VL-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
; AVX512VL-NEXT: vpsrld $1, %xmm1, %xmm1
-; AVX512VL-NEXT: vpsrlvd %xmm3, %xmm1, %xmm1
-; AVX512VL-NEXT: vpand %xmm4, %xmm2, %xmm2
+; AVX512VL-NEXT: vpsrld %xmm4, %xmm1, %xmm1
+; AVX512VL-NEXT: vpand %xmm3, %xmm2, %xmm2
; AVX512VL-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
; AVX512VL-NEXT: vpslld %xmm2, %xmm0, %xmm0
; AVX512VL-NEXT: vpor %xmm1, %xmm0, %xmm0
@@ -1369,12 +1278,12 @@ define <4 x i32> @splatvar_funnnel_v4i32(<4 x i32> %x, <4 x i32> %y, <4 x i32> %
;
; AVX512BW-LABEL: splatvar_funnnel_v4i32:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpbroadcastd %xmm2, %xmm3
-; AVX512BW-NEXT: vpbroadcastd {{.*#+}} xmm4 = [31,31,31,31]
-; AVX512BW-NEXT: vpandn %xmm4, %xmm3, %xmm3
+; AVX512BW-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31]
+; AVX512BW-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX512BW-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
; AVX512BW-NEXT: vpsrld $1, %xmm1, %xmm1
-; AVX512BW-NEXT: vpsrlvd %xmm3, %xmm1, %xmm1
-; AVX512BW-NEXT: vpand %xmm4, %xmm2, %xmm2
+; AVX512BW-NEXT: vpsrld %xmm4, %xmm1, %xmm1
+; AVX512BW-NEXT: vpand %xmm3, %xmm2, %xmm2
; AVX512BW-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
; AVX512BW-NEXT: vpslld %xmm2, %xmm0, %xmm0
; AVX512BW-NEXT: vpor %xmm1, %xmm0, %xmm0
@@ -1392,12 +1301,12 @@ 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, %xmm3
-; AVX512VLBW-NEXT: vpbroadcastd {{.*#+}} xmm4 = [31,31,31,31]
-; AVX512VLBW-NEXT: vpandn %xmm4, %xmm3, %xmm3
+; AVX512VLBW-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31]
+; AVX512VLBW-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX512VLBW-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
; AVX512VLBW-NEXT: vpsrld $1, %xmm1, %xmm1
-; AVX512VLBW-NEXT: vpsrlvd %xmm3, %xmm1, %xmm1
-; AVX512VLBW-NEXT: vpand %xmm4, %xmm2, %xmm2
+; AVX512VLBW-NEXT: vpsrld %xmm4, %xmm1, %xmm1
+; AVX512VLBW-NEXT: vpand %xmm3, %xmm2, %xmm2
; AVX512VLBW-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
; AVX512VLBW-NEXT: vpslld %xmm2, %xmm0, %xmm0
; AVX512VLBW-NEXT: vpor %xmm1, %xmm0, %xmm0
@@ -1411,27 +1320,25 @@ define <4 x i32> @splatvar_funnnel_v4i32(<4 x i32> %x, <4 x i32> %y, <4 x i32> %
;
; XOPAVX1-LABEL: splatvar_funnnel_v4i32:
; XOPAVX1: # %bb.0:
-; XOPAVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm2[0,0,0,0]
-; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [31,31,31,31]
-; XOPAVX1-NEXT: vpand %xmm4, %xmm2, %xmm2
+; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [31,31,31,31]
+; XOPAVX1-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; XOPAVX1-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
+; XOPAVX1-NEXT: vpsrld $1, %xmm1, %xmm1
+; XOPAVX1-NEXT: vpsrld %xmm4, %xmm1, %xmm1
+; XOPAVX1-NEXT: vpand %xmm3, %xmm2, %xmm2
; XOPAVX1-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
; XOPAVX1-NEXT: vpslld %xmm2, %xmm0, %xmm0
-; XOPAVX1-NEXT: vpandn %xmm4, %xmm3, %xmm2
-; XOPAVX1-NEXT: vpxor %xmm3, %xmm3, %xmm3
-; XOPAVX1-NEXT: vpsubd %xmm2, %xmm3, %xmm2
-; XOPAVX1-NEXT: vpsrld $1, %xmm1, %xmm1
-; XOPAVX1-NEXT: vpshld %xmm2, %xmm1, %xmm1
; XOPAVX1-NEXT: vpor %xmm1, %xmm0, %xmm0
; XOPAVX1-NEXT: retq
;
; XOPAVX2-LABEL: splatvar_funnnel_v4i32:
; XOPAVX2: # %bb.0:
-; XOPAVX2-NEXT: vpbroadcastd %xmm2, %xmm3
-; XOPAVX2-NEXT: vpbroadcastd {{.*#+}} xmm4 = [31,31,31,31]
-; XOPAVX2-NEXT: vpandn %xmm4, %xmm3, %xmm3
+; XOPAVX2-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31]
+; XOPAVX2-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; XOPAVX2-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
; XOPAVX2-NEXT: vpsrld $1, %xmm1, %xmm1
-; XOPAVX2-NEXT: vpsrlvd %xmm3, %xmm1, %xmm1
-; XOPAVX2-NEXT: vpand %xmm4, %xmm2, %xmm2
+; XOPAVX2-NEXT: vpsrld %xmm4, %xmm1, %xmm1
+; XOPAVX2-NEXT: vpand %xmm3, %xmm2, %xmm2
; XOPAVX2-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
; XOPAVX2-NEXT: vpslld %xmm2, %xmm0, %xmm0
; XOPAVX2-NEXT: vpor %xmm1, %xmm0, %xmm0
@@ -1439,29 +1346,17 @@ define <4 x i32> @splatvar_funnnel_v4i32(<4 x i32> %x, <4 x i32> %y, <4 x i32> %
;
; X86-SSE2-LABEL: splatvar_funnnel_v4i32:
; X86-SSE2: # %bb.0:
-; X86-SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm2[0,0,0,0]
-; X86-SSE2-NEXT: psrld $1, %xmm1
-; X86-SSE2-NEXT: pandn {{\.LCPI.*}}, %xmm3
-; X86-SSE2-NEXT: pshuflw {{.*#+}} xmm4 = xmm3[2,3,3,3,4,5,6,7]
-; X86-SSE2-NEXT: movdqa %xmm1, %xmm5
-; X86-SSE2-NEXT: psrld %xmm4, %xmm5
-; X86-SSE2-NEXT: pshuflw {{.*#+}} xmm4 = xmm3[0,1,1,1,4,5,6,7]
-; X86-SSE2-NEXT: movdqa %xmm1, %xmm6
-; X86-SSE2-NEXT: psrld %xmm4, %xmm6
-; X86-SSE2-NEXT: punpcklqdq {{.*#+}} xmm6 = xmm6[0],xmm5[0]
-; X86-SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm3[2,3,2,3]
-; X86-SSE2-NEXT: pshuflw {{.*#+}} xmm4 = xmm3[2,3,3,3,4,5,6,7]
-; X86-SSE2-NEXT: movdqa %xmm1, %xmm5
-; X86-SSE2-NEXT: psrld %xmm4, %xmm5
-; X86-SSE2-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,1,1,1,4,5,6,7]
-; X86-SSE2-NEXT: psrld %xmm3, %xmm1
-; X86-SSE2-NEXT: punpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm5[1]
-; X86-SSE2-NEXT: shufps {{.*#+}} xmm6 = xmm6[0,3],xmm1[0,3]
; X86-SSE2-NEXT: movd %xmm2, %eax
+; X86-SSE2-NEXT: movl %eax, %ecx
+; X86-SSE2-NEXT: andl $31, %ecx
+; X86-SSE2-NEXT: movd %ecx, %xmm2
+; X86-SSE2-NEXT: pslld %xmm2, %xmm0
+; X86-SSE2-NEXT: psrld $1, %xmm1
+; X86-SSE2-NEXT: notl %eax
; X86-SSE2-NEXT: andl $31, %eax
-; X86-SSE2-NEXT: movd %eax, %xmm1
-; X86-SSE2-NEXT: pslld %xmm1, %xmm0
-; X86-SSE2-NEXT: por %xmm6, %xmm0
+; X86-SSE2-NEXT: movd %eax, %xmm2
+; X86-SSE2-NEXT: psrld %xmm2, %xmm1
+; X86-SSE2-NEXT: por %xmm1, %xmm0
; X86-SSE2-NEXT: retl
%splat = shufflevector <4 x i32> %amt, <4 x i32> undef, <4 x i32> zeroinitializer
%res = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %x, <4 x i32> %y, <4 x i32> %splat)
@@ -1471,178 +1366,84 @@ define <4 x i32> @splatvar_funnnel_v4i32(<4 x i32> %x, <4 x i32> %y, <4 x i32> %
define <8 x i16> @splatvar_funnnel_v8i16(<8 x i16> %x, <8 x i16> %y, <8 x i16> %amt) nounwind {
; SSE2-LABEL: splatvar_funnnel_v8i16:
; SSE2: # %bb.0:
-; SSE2-NEXT: pcmpeqd %xmm3, %xmm3
-; SSE2-NEXT: pxor %xmm2, %xmm3
-; SSE2-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,0,0,0,4,5,6,7]
-; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm3[0,0,0,0]
-; SSE2-NEXT: psllw $12, %xmm4
-; SSE2-NEXT: movdqa %xmm4, %xmm3
-; SSE2-NEXT: psraw $15, %xmm3
-; SSE2-NEXT: movdqa %xmm1, %xmm5
-; SSE2-NEXT: psrlw $1, %xmm5
-; SSE2-NEXT: psrlw $9, %xmm1
-; SSE2-NEXT: pand %xmm3, %xmm1
-; SSE2-NEXT: pandn %xmm5, %xmm3
-; SSE2-NEXT: por %xmm1, %xmm3
-; SSE2-NEXT: paddw %xmm4, %xmm4
-; SSE2-NEXT: movdqa %xmm4, %xmm1
-; SSE2-NEXT: psraw $15, %xmm1
-; SSE2-NEXT: movdqa %xmm1, %xmm5
-; SSE2-NEXT: pandn %xmm3, %xmm5
-; SSE2-NEXT: psrlw $4, %xmm3
-; SSE2-NEXT: pand %xmm1, %xmm3
-; SSE2-NEXT: por %xmm5, %xmm3
-; SSE2-NEXT: paddw %xmm4, %xmm4
-; SSE2-NEXT: movdqa %xmm4, %xmm1
-; SSE2-NEXT: psraw $15, %xmm1
-; SSE2-NEXT: movdqa %xmm1, %xmm5
-; SSE2-NEXT: pandn %xmm3, %xmm5
-; SSE2-NEXT: psrlw $2, %xmm3
-; SSE2-NEXT: pand %xmm1, %xmm3
-; SSE2-NEXT: por %xmm5, %xmm3
-; SSE2-NEXT: paddw %xmm4, %xmm4
-; SSE2-NEXT: psraw $15, %xmm4
-; SSE2-NEXT: movdqa %xmm4, %xmm1
-; SSE2-NEXT: pandn %xmm3, %xmm1
-; SSE2-NEXT: psrlw $1, %xmm3
-; SSE2-NEXT: pand %xmm4, %xmm3
-; SSE2-NEXT: pand {{.*}}(%rip), %xmm2
+; SSE2-NEXT: movdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; SSE2-NEXT: movdqa %xmm2, %xmm4
+; SSE2-NEXT: pandn %xmm3, %xmm4
+; SSE2-NEXT: pslldq {{.*#+}} xmm4 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm4[0,1]
+; SSE2-NEXT: psrldq {{.*#+}} xmm4 = xmm4[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE2-NEXT: psrlw $1, %xmm1
+; SSE2-NEXT: psrlw %xmm4, %xmm1
+; SSE2-NEXT: pand %xmm3, %xmm2
; SSE2-NEXT: pslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1]
; SSE2-NEXT: psrldq {{.*#+}} xmm2 = xmm2[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; SSE2-NEXT: psllw %xmm2, %xmm0
; SSE2-NEXT: por %xmm1, %xmm0
-; SSE2-NEXT: por %xmm3, %xmm0
; SSE2-NEXT: retq
;
; SSE41-LABEL: splatvar_funnnel_v8i16:
; SSE41: # %bb.0:
-; SSE41-NEXT: movdqa %xmm0, %xmm3
-; SSE41-NEXT: pshuflw {{.*#+}} xmm0 = xmm2[0,0,0,0,4,5,6,7]
-; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
-; SSE41-NEXT: movdqa {{.*#+}} xmm5 = [15,15,15,15,15,15,15,15]
-; SSE41-NEXT: pandn %xmm5, %xmm0
-; SSE41-NEXT: movdqa %xmm0, %xmm4
-; SSE41-NEXT: psllw $12, %xmm4
-; SSE41-NEXT: psllw $4, %xmm0
-; SSE41-NEXT: por %xmm4, %xmm0
-; SSE41-NEXT: movdqa %xmm0, %xmm4
-; SSE41-NEXT: paddw %xmm0, %xmm4
-; SSE41-NEXT: movdqa %xmm1, %xmm6
-; SSE41-NEXT: psrlw $9, %xmm6
+; SSE41-NEXT: movdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; SSE41-NEXT: movdqa %xmm2, %xmm4
+; SSE41-NEXT: pandn %xmm3, %xmm4
+; SSE41-NEXT: pmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
; SSE41-NEXT: psrlw $1, %xmm1
-; SSE41-NEXT: pblendvb %xmm0, %xmm6, %xmm1
-; SSE41-NEXT: movdqa %xmm1, %xmm6
-; SSE41-NEXT: psrlw $4, %xmm6
-; SSE41-NEXT: movdqa %xmm4, %xmm0
-; SSE41-NEXT: pblendvb %xmm0, %xmm6, %xmm1
-; SSE41-NEXT: movdqa %xmm1, %xmm6
-; SSE41-NEXT: psrlw $2, %xmm6
-; SSE41-NEXT: paddw %xmm4, %xmm4
-; SSE41-NEXT: movdqa %xmm4, %xmm0
-; SSE41-NEXT: pblendvb %xmm0, %xmm6, %xmm1
-; SSE41-NEXT: movdqa %xmm1, %xmm6
-; SSE41-NEXT: psrlw $1, %xmm6
-; SSE41-NEXT: paddw %xmm4, %xmm4
-; SSE41-NEXT: movdqa %xmm4, %xmm0
-; SSE41-NEXT: pblendvb %xmm0, %xmm6, %xmm1
-; SSE41-NEXT: pand %xmm5, %xmm2
-; SSE41-NEXT: pmovzxwq {{.*#+}} xmm0 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
-; SSE41-NEXT: psllw %xmm0, %xmm3
-; SSE41-NEXT: por %xmm1, %xmm3
-; SSE41-NEXT: movdqa %xmm3, %xmm0
+; SSE41-NEXT: psrlw %xmm4, %xmm1
+; SSE41-NEXT: pand %xmm3, %xmm2
+; SSE41-NEXT: pmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
+; SSE41-NEXT: psllw %xmm2, %xmm0
+; SSE41-NEXT: por %xmm1, %xmm0
; SSE41-NEXT: retq
;
-; AVX1-LABEL: splatvar_funnnel_v8i16:
-; AVX1: # %bb.0:
-; AVX1-NEXT: vpshuflw {{.*#+}} xmm3 = xmm2[0,0,0,0,4,5,6,7]
-; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,0,0,0]
-; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [15,15,15,15,15,15,15,15]
-; AVX1-NEXT: vpandn %xmm4, %xmm3, %xmm3
-; AVX1-NEXT: vpsllw $12, %xmm3, %xmm5
-; AVX1-NEXT: vpsllw $4, %xmm3, %xmm3
-; AVX1-NEXT: vpor %xmm5, %xmm3, %xmm3
-; AVX1-NEXT: vpaddw %xmm3, %xmm3, %xmm5
-; AVX1-NEXT: vpsrlw $9, %xmm1, %xmm6
-; AVX1-NEXT: vpsrlw $1, %xmm1, %xmm1
-; AVX1-NEXT: vpblendvb %xmm3, %xmm6, %xmm1, %xmm1
-; AVX1-NEXT: vpsrlw $4, %xmm1, %xmm3
-; AVX1-NEXT: vpblendvb %xmm5, %xmm3, %xmm1, %xmm1
-; AVX1-NEXT: vpsrlw $2, %xmm1, %xmm3
-; AVX1-NEXT: vpaddw %xmm5, %xmm5, %xmm5
-; AVX1-NEXT: vpblendvb %xmm5, %xmm3, %xmm1, %xmm1
-; AVX1-NEXT: vpsrlw $1, %xmm1, %xmm3
-; AVX1-NEXT: vpaddw %xmm5, %xmm5, %xmm5
-; AVX1-NEXT: vpblendvb %xmm5, %xmm3, %xmm1, %xmm1
-; AVX1-NEXT: vpand %xmm4, %xmm2, %xmm2
-; AVX1-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
-; AVX1-NEXT: vpsllw %xmm2, %xmm0, %xmm0
-; AVX1-NEXT: vpor %xmm1, %xmm0, %xmm0
-; AVX1-NEXT: retq
-;
-; AVX2-LABEL: splatvar_funnnel_v8i16:
-; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastw %xmm2, %xmm3
-; AVX2-NEXT: vmovdqa {{.*#+}} xmm4 = [15,15,15,15,15,15,15,15]
-; AVX2-NEXT: vpandn %xmm4, %xmm3, %xmm3
-; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero
-; AVX2-NEXT: vpsrlw $1, %xmm1, %xmm1
-; AVX2-NEXT: vpmovzxwd {{.*#+}} 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
-; AVX2-NEXT: vpsrlvd %ymm3, %ymm1, %ymm1
-; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm3
-; AVX2-NEXT: vpackusdw %xmm3, %xmm1, %xmm1
-; AVX2-NEXT: vpand %xmm4, %xmm2, %xmm2
-; AVX2-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
-; AVX2-NEXT: vpsllw %xmm2, %xmm0, %xmm0
-; AVX2-NEXT: vpor %xmm1, %xmm0, %xmm0
-; AVX2-NEXT: vzeroupper
-; AVX2-NEXT: retq
+; AVX-LABEL: splatvar_funnnel_v8i16:
+; AVX: # %bb.0:
+; AVX-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; AVX-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
+; AVX-NEXT: vpsrlw $1, %xmm1, %xmm1
+; AVX-NEXT: vpsrlw %xmm4, %xmm1, %xmm1
+; AVX-NEXT: vpand %xmm3, %xmm2, %xmm2
+; AVX-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
+; AVX-NEXT: vpsllw %xmm2, %xmm0, %xmm0
+; AVX-NEXT: vpor %xmm1, %xmm0, %xmm0
+; AVX-NEXT: retq
;
; AVX512F-LABEL: splatvar_funnnel_v8i16:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpbroadcastw %xmm2, %xmm3
-; AVX512F-NEXT: vmovdqa {{.*#+}} xmm4 = [15,15,15,15,15,15,15,15]
-; AVX512F-NEXT: vpand %xmm4, %xmm2, %xmm2
+; AVX512F-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; AVX512F-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX512F-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
+; AVX512F-NEXT: vpsrlw $1, %xmm1, %xmm1
+; AVX512F-NEXT: vpsrlw %xmm4, %xmm1, %xmm1
+; AVX512F-NEXT: vpand %xmm3, %xmm2, %xmm2
; AVX512F-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
; AVX512F-NEXT: vpsllw %xmm2, %xmm0, %xmm0
-; AVX512F-NEXT: vpandn %xmm4, %xmm3, %xmm2
-; AVX512F-NEXT: vpmovzxwd {{.*#+}} ymm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero
-; AVX512F-NEXT: vpsrlw $1, %xmm1, %xmm1
-; AVX512F-NEXT: vpmovzxwd {{.*#+}} 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
-; AVX512F-NEXT: vpsrlvd %ymm2, %ymm1, %ymm1
-; AVX512F-NEXT: vpmovdw %zmm1, %ymm1
; AVX512F-NEXT: vpor %xmm1, %xmm0, %xmm0
-; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: splatvar_funnnel_v8i16:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastw %xmm2, %xmm3
-; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm4 = [15,15,15,15,15,15,15,15]
-; AVX512VL-NEXT: vpand %xmm4, %xmm2, %xmm2
+; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; AVX512VL-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX512VL-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
+; AVX512VL-NEXT: vpsrlw $1, %xmm1, %xmm1
+; AVX512VL-NEXT: vpsrlw %xmm4, %xmm1, %xmm1
+; AVX512VL-NEXT: vpand %xmm3, %xmm2, %xmm2
; AVX512VL-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
; AVX512VL-NEXT: vpsllw %xmm2, %xmm0, %xmm0
-; AVX512VL-NEXT: vpandn %xmm4, %xmm3, %xmm2
-; AVX512VL-NEXT: vpmovzxwd {{.*#+}} ymm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero
-; AVX512VL-NEXT: vpsrlw $1, %xmm1, %xmm1
-; AVX512VL-NEXT: vpmovzxwd {{.*#+}} 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
-; AVX512VL-NEXT: vpsrlvd %ymm2, %ymm1, %ymm1
-; AVX512VL-NEXT: vpmovdw %ymm1, %xmm1
; AVX512VL-NEXT: vpor %xmm1, %xmm0, %xmm0
-; AVX512VL-NEXT: vzeroupper
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: splatvar_funnnel_v8i16:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpbroadcastw %xmm2, %xmm3
+; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX512BW-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
; AVX512BW-NEXT: vpsrlw $1, %xmm1, %xmm1
-; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm4 = [15,15,15,15,15,15,15,15]
-; AVX512BW-NEXT: vpandn %xmm4, %xmm3, %xmm3
-; AVX512BW-NEXT: vpsrlvw %zmm3, %zmm1, %zmm1
-; AVX512BW-NEXT: vpand %xmm4, %xmm2, %xmm2
+; AVX512BW-NEXT: vpsrlw %xmm4, %xmm1, %xmm1
+; AVX512BW-NEXT: vpand %xmm3, %xmm2, %xmm2
; AVX512BW-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
; AVX512BW-NEXT: vpsllw %xmm2, %xmm0, %xmm0
; AVX512BW-NEXT: vpor %xmm1, %xmm0, %xmm0
-; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
;
; AVX512VBMI2-LABEL: splatvar_funnnel_v8i16:
@@ -1657,12 +1458,12 @@ 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, %xmm3
-; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm4 = [15,15,15,15,15,15,15,15]
-; AVX512VLBW-NEXT: vpandn %xmm4, %xmm3, %xmm3
+; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; AVX512VLBW-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX512VLBW-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
; AVX512VLBW-NEXT: vpsrlw $1, %xmm1, %xmm1
-; AVX512VLBW-NEXT: vpsrlvw %xmm3, %xmm1, %xmm1
-; AVX512VLBW-NEXT: vpand %xmm4, %xmm2, %xmm2
+; AVX512VLBW-NEXT: vpsrlw %xmm4, %xmm1, %xmm1
+; AVX512VLBW-NEXT: vpand %xmm3, %xmm2, %xmm2
; AVX512VLBW-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
; AVX512VLBW-NEXT: vpsllw %xmm2, %xmm0, %xmm0
; AVX512VLBW-NEXT: vpor %xmm1, %xmm0, %xmm0
@@ -1674,80 +1475,33 @@ define <8 x i16> @splatvar_funnnel_v8i16(<8 x i16> %x, <8 x i16> %y, <8 x i16> %
; AVX512VLVBMI2-NEXT: vpshldvw %xmm2, %xmm1, %xmm0
; AVX512VLVBMI2-NEXT: retq
;
-; XOPAVX1-LABEL: splatvar_funnnel_v8i16:
-; XOPAVX1: # %bb.0:
-; XOPAVX1-NEXT: vpshuflw {{.*#+}} xmm3 = xmm2[0,0,0,0,4,5,6,7]
-; XOPAVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,0,0,0]
-; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [15,15,15,15,15,15,15,15]
-; XOPAVX1-NEXT: vpand %xmm4, %xmm2, %xmm2
-; XOPAVX1-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
-; XOPAVX1-NEXT: vpsllw %xmm2, %xmm0, %xmm0
-; XOPAVX1-NEXT: vpandn %xmm4, %xmm3, %xmm2
-; XOPAVX1-NEXT: vpxor %xmm3, %xmm3, %xmm3
-; XOPAVX1-NEXT: vpsubw %xmm2, %xmm3, %xmm2
-; XOPAVX1-NEXT: vpsrlw $1, %xmm1, %xmm1
-; XOPAVX1-NEXT: vpshlw %xmm2, %xmm1, %xmm1
-; XOPAVX1-NEXT: vpor %xmm1, %xmm0, %xmm0
-; XOPAVX1-NEXT: retq
-;
-; XOPAVX2-LABEL: splatvar_funnnel_v8i16:
-; XOPAVX2: # %bb.0:
-; XOPAVX2-NEXT: vpbroadcastw %xmm2, %xmm3
-; XOPAVX2-NEXT: vmovdqa {{.*#+}} xmm4 = [15,15,15,15,15,15,15,15]
-; XOPAVX2-NEXT: vpand %xmm4, %xmm2, %xmm2
-; XOPAVX2-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
-; XOPAVX2-NEXT: vpsllw %xmm2, %xmm0, %xmm0
-; XOPAVX2-NEXT: vpandn %xmm4, %xmm3, %xmm2
-; XOPAVX2-NEXT: vpxor %xmm3, %xmm3, %xmm3
-; XOPAVX2-NEXT: vpsubw %xmm2, %xmm3, %xmm2
-; XOPAVX2-NEXT: vpsrlw $1, %xmm1, %xmm1
-; XOPAVX2-NEXT: vpshlw %xmm2, %xmm1, %xmm1
-; XOPAVX2-NEXT: vpor %xmm1, %xmm0, %xmm0
-; XOPAVX2-NEXT: retq
+; XOP-LABEL: splatvar_funnnel_v8i16:
+; XOP: # %bb.0:
+; XOP-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; XOP-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; XOP-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
+; XOP-NEXT: vpsrlw $1, %xmm1, %xmm1
+; XOP-NEXT: vpsrlw %xmm4, %xmm1, %xmm1
+; XOP-NEXT: vpand %xmm3, %xmm2, %xmm2
+; XOP-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
+; XOP-NEXT: vpsllw %xmm2, %xmm0, %xmm0
+; XOP-NEXT: vpor %xmm1, %xmm0, %xmm0
+; XOP-NEXT: retq
;
; X86-SSE2-LABEL: splatvar_funnnel_v8i16:
; X86-SSE2: # %bb.0:
-; X86-SSE2-NEXT: pcmpeqd %xmm3, %xmm3
-; X86-SSE2-NEXT: pxor %xmm2, %xmm3
-; X86-SSE2-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,0,0,0,4,5,6,7]
-; X86-SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm3[0,0,0,0]
-; X86-SSE2-NEXT: psllw $12, %xmm4
-; X86-SSE2-NEXT: movdqa %xmm4, %xmm3
-; X86-SSE2-NEXT: psraw $15, %xmm3
-; X86-SSE2-NEXT: movdqa %xmm1, %xmm5
-; X86-SSE2-NEXT: psrlw $1, %xmm5
-; X86-SSE2-NEXT: psrlw $9, %xmm1
-; X86-SSE2-NEXT: pand %xmm3, %xmm1
-; X86-SSE2-NEXT: pandn %xmm5, %xmm3
-; X86-SSE2-NEXT: por %xmm1, %xmm3
-; X86-SSE2-NEXT: paddw %xmm4, %xmm4
-; X86-SSE2-NEXT: movdqa %xmm4, %xmm1
-; X86-SSE2-NEXT: psraw $15, %xmm1
-; X86-SSE2-NEXT: movdqa %xmm1, %xmm5
-; X86-SSE2-NEXT: pandn %xmm3, %xmm5
-; X86-SSE2-NEXT: psrlw $4, %xmm3
-; X86-SSE2-NEXT: pand %xmm1, %xmm3
-; X86-SSE2-NEXT: por %xmm5, %xmm3
-; X86-SSE2-NEXT: paddw %xmm4, %xmm4
-; X86-SSE2-NEXT: movdqa %xmm4, %xmm1
-; X86-SSE2-NEXT: psraw $15, %xmm1
-; X86-SSE2-NEXT: movdqa %xmm1, %xmm5
-; X86-SSE2-NEXT: pandn %xmm3, %xmm5
-; X86-SSE2-NEXT: psrlw $2, %xmm3
-; X86-SSE2-NEXT: pand %xmm1, %xmm3
-; X86-SSE2-NEXT: por %xmm5, %xmm3
-; X86-SSE2-NEXT: paddw %xmm4, %xmm4
-; X86-SSE2-NEXT: psraw $15, %xmm4
-; X86-SSE2-NEXT: movdqa %xmm4, %xmm1
-; X86-SSE2-NEXT: pandn %xmm3, %xmm1
-; X86-SSE2-NEXT: psrlw $1, %xmm3
-; X86-SSE2-NEXT: pand %xmm4, %xmm3
-; X86-SSE2-NEXT: pand {{\.LCPI.*}}, %xmm2
+; X86-SSE2-NEXT: movdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; X86-SSE2-NEXT: movdqa %xmm2, %xmm4
+; X86-SSE2-NEXT: pandn %xmm3, %xmm4
+; X86-SSE2-NEXT: pslldq {{.*#+}} xmm4 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm4[0,1]
+; X86-SSE2-NEXT: psrldq {{.*#+}} xmm4 = xmm4[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-SSE2-NEXT: psrlw $1, %xmm1
+; X86-SSE2-NEXT: psrlw %xmm4, %xmm1
+; X86-SSE2-NEXT: pand %xmm3, %xmm2
; X86-SSE2-NEXT: pslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1]
; X86-SSE2-NEXT: psrldq {{.*#+}} xmm2 = xmm2[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; X86-SSE2-NEXT: psllw %xmm2, %xmm0
; X86-SSE2-NEXT: por %xmm1, %xmm0
-; X86-SSE2-NEXT: por %xmm3, %xmm0
; X86-SSE2-NEXT: retl
%splat = shufflevector <8 x i16> %amt, <8 x i16> undef, <8 x i32> zeroinitializer
%res = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %x, <8 x i16> %y, <8 x i16> %splat)
@@ -1757,119 +1511,75 @@ define <8 x i16> @splatvar_funnnel_v8i16(<8 x i16> %x, <8 x i16> %y, <8 x i16> %
define <16 x i8> @splatvar_funnnel_v16i8(<16 x i8> %x, <16 x i8> %y, <16 x i8> %amt) nounwind {
; SSE2-LABEL: splatvar_funnnel_v16i8:
; SSE2: # %bb.0:
-; SSE2-NEXT: movdqa %xmm2, %xmm3
-; SSE2-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3],xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
-; SSE2-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,0,0,0,4,5,6,7]
-; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm3[0,0,0,0]
-; SSE2-NEXT: psrlw $1, %xmm1
-; SSE2-NEXT: movdqa {{.*#+}} xmm5 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127]
-; SSE2-NEXT: pand %xmm5, %xmm1
-; SSE2-NEXT: movdqa {{.*#+}} xmm8 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; SSE2-NEXT: pandn %xmm8, %xmm6
-; SSE2-NEXT: psllw $5, %xmm6
-; SSE2-NEXT: pxor %xmm3, %xmm3
-; SSE2-NEXT: pxor %xmm7, %xmm7
-; SSE2-NEXT: pcmpgtb %xmm6, %xmm7
-; SSE2-NEXT: movdqa %xmm7, %xmm4
-; SSE2-NEXT: pandn %xmm1, %xmm4
-; SSE2-NEXT: psrlw $4, %xmm1
-; SSE2-NEXT: pand %xmm7, %xmm1
-; SSE2-NEXT: pand {{.*}}(%rip), %xmm1
-; SSE2-NEXT: por %xmm4, %xmm1
-; SSE2-NEXT: paddb %xmm6, %xmm6
-; SSE2-NEXT: pxor %xmm4, %xmm4
-; SSE2-NEXT: pcmpgtb %xmm6, %xmm4
-; SSE2-NEXT: movdqa %xmm4, %xmm7
-; SSE2-NEXT: pandn %xmm1, %xmm7
-; SSE2-NEXT: psrlw $2, %xmm1
-; SSE2-NEXT: pand %xmm4, %xmm1
-; SSE2-NEXT: pand {{.*}}(%rip), %xmm1
-; SSE2-NEXT: por %xmm7, %xmm1
-; SSE2-NEXT: paddb %xmm6, %xmm6
-; SSE2-NEXT: pcmpgtb %xmm6, %xmm3
-; SSE2-NEXT: movdqa %xmm3, %xmm4
-; SSE2-NEXT: pandn %xmm1, %xmm4
-; SSE2-NEXT: psrlw $1, %xmm1
-; SSE2-NEXT: pand %xmm5, %xmm3
-; SSE2-NEXT: pand %xmm1, %xmm3
-; SSE2-NEXT: pand %xmm8, %xmm2
+; SSE2-NEXT: movdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; SSE2-NEXT: movdqa %xmm2, %xmm4
+; SSE2-NEXT: pand %xmm3, %xmm4
+; SSE2-NEXT: pslldq {{.*#+}} xmm4 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm4[0]
+; SSE2-NEXT: psrldq {{.*#+}} xmm4 = xmm4[15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE2-NEXT: psllw %xmm4, %xmm0
+; SSE2-NEXT: pcmpeqd %xmm5, %xmm5
+; SSE2-NEXT: pcmpeqd %xmm6, %xmm6
+; SSE2-NEXT: psllw %xmm4, %xmm6
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm6 = xmm6[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE2-NEXT: pshuflw {{.*#+}} xmm4 = xmm6[0,0,0,0,4,5,6,7]
+; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm4[0,0,0,0]
+; SSE2-NEXT: pand %xmm4, %xmm0
+; SSE2-NEXT: pandn %xmm3, %xmm2
; SSE2-NEXT: pslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0]
; SSE2-NEXT: psrldq {{.*#+}} xmm2 = xmm2[15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE2-NEXT: psllw %xmm2, %xmm0
-; SSE2-NEXT: pcmpeqd %xmm1, %xmm1
-; SSE2-NEXT: psllw %xmm2, %xmm1
-; SSE2-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
-; SSE2-NEXT: pand %xmm0, %xmm1
-; SSE2-NEXT: por %xmm4, %xmm1
-; SSE2-NEXT: por %xmm3, %xmm1
-; SSE2-NEXT: movdqa %xmm1, %xmm0
+; SSE2-NEXT: psrlw $1, %xmm1
+; SSE2-NEXT: pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT: psrlw %xmm2, %xmm1
+; SSE2-NEXT: psrlw %xmm2, %xmm5
+; SSE2-NEXT: psrlw $8, %xmm5
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm5 = xmm5[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE2-NEXT: pshuflw {{.*#+}} xmm2 = xmm5[0,0,0,0,4,5,6,7]
+; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
+; SSE2-NEXT: pand %xmm1, %xmm2
+; SSE2-NEXT: por %xmm2, %xmm0
; SSE2-NEXT: retq
;
; SSE41-LABEL: splatvar_funnnel_v16i8:
; SSE41: # %bb.0:
-; SSE41-NEXT: movdqa %xmm0, %xmm3
-; SSE41-NEXT: pxor %xmm4, %xmm4
-; SSE41-NEXT: movdqa %xmm2, %xmm0
-; SSE41-NEXT: pshufb %xmm4, %xmm0
-; SSE41-NEXT: movdqa {{.*#+}} xmm5 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; SSE41-NEXT: pandn %xmm5, %xmm0
-; SSE41-NEXT: psllw $5, %xmm0
+; SSE41-NEXT: movdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; SSE41-NEXT: movdqa %xmm2, %xmm4
+; SSE41-NEXT: pandn %xmm3, %xmm4
+; SSE41-NEXT: pmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
; SSE41-NEXT: psrlw $1, %xmm1
-; SSE41-NEXT: movdqa {{.*#+}} xmm6 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127]
-; SSE41-NEXT: pand %xmm6, %xmm1
-; SSE41-NEXT: movdqa %xmm1, %xmm7
-; SSE41-NEXT: psrlw $4, %xmm7
-; SSE41-NEXT: pand {{.*}}(%rip), %xmm7
-; SSE41-NEXT: pblendvb %xmm0, %xmm7, %xmm1
-; SSE41-NEXT: movdqa %xmm1, %xmm7
-; SSE41-NEXT: psrlw $2, %xmm7
-; SSE41-NEXT: pand {{.*}}(%rip), %xmm7
-; SSE41-NEXT: paddb %xmm0, %xmm0
-; SSE41-NEXT: pblendvb %xmm0, %xmm7, %xmm1
-; SSE41-NEXT: movdqa %xmm1, %xmm7
-; SSE41-NEXT: psrlw $1, %xmm7
-; SSE41-NEXT: pand %xmm6, %xmm7
-; SSE41-NEXT: paddb %xmm0, %xmm0
-; SSE41-NEXT: pblendvb %xmm0, %xmm7, %xmm1
-; SSE41-NEXT: pand %xmm5, %xmm2
-; SSE41-NEXT: pmovzxbq {{.*#+}} xmm0 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
-; SSE41-NEXT: psllw %xmm0, %xmm3
-; SSE41-NEXT: pcmpeqd %xmm2, %xmm2
-; SSE41-NEXT: psllw %xmm0, %xmm2
-; SSE41-NEXT: pshufb %xmm4, %xmm2
-; SSE41-NEXT: pand %xmm2, %xmm3
-; SSE41-NEXT: por %xmm1, %xmm3
-; SSE41-NEXT: movdqa %xmm3, %xmm0
+; SSE41-NEXT: pand {{.*}}(%rip), %xmm1
+; SSE41-NEXT: psrlw %xmm4, %xmm1
+; SSE41-NEXT: pcmpeqd %xmm5, %xmm5
+; SSE41-NEXT: pcmpeqd %xmm6, %xmm6
+; SSE41-NEXT: psrlw %xmm4, %xmm6
+; SSE41-NEXT: pshufb {{.*#+}} xmm6 = xmm6[1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1]
+; SSE41-NEXT: pand %xmm1, %xmm6
+; SSE41-NEXT: pand %xmm3, %xmm2
+; SSE41-NEXT: pmovzxbq {{.*#+}} xmm1 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
+; SSE41-NEXT: psllw %xmm1, %xmm0
+; SSE41-NEXT: psllw %xmm1, %xmm5
+; SSE41-NEXT: pxor %xmm1, %xmm1
+; SSE41-NEXT: pshufb %xmm1, %xmm5
+; SSE41-NEXT: pand %xmm5, %xmm0
+; SSE41-NEXT: por %xmm6, %xmm0
; SSE41-NEXT: retq
;
; AVX1-LABEL: splatvar_funnnel_v16i8:
; AVX1: # %bb.0:
-; AVX1-NEXT: vpxor %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpshufb %xmm3, %xmm2, %xmm4
-; AVX1-NEXT: vmovdqa {{.*#+}} xmm5 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; AVX1-NEXT: vpandn %xmm5, %xmm4, %xmm4
-; AVX1-NEXT: vpsllw $5, %xmm4, %xmm4
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX1-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX1-NEXT: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
; AVX1-NEXT: vpsrlw $1, %xmm1, %xmm1
-; AVX1-NEXT: vmovdqa {{.*#+}} xmm6 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127]
-; AVX1-NEXT: vpand %xmm6, %xmm1, %xmm1
-; AVX1-NEXT: vpsrlw $4, %xmm1, %xmm7
-; AVX1-NEXT: vpand {{.*}}(%rip), %xmm7, %xmm7
-; AVX1-NEXT: vpblendvb %xmm4, %xmm7, %xmm1, %xmm1
-; AVX1-NEXT: vpsrlw $2, %xmm1, %xmm7
-; AVX1-NEXT: vpand {{.*}}(%rip), %xmm7, %xmm7
-; AVX1-NEXT: vpaddb %xmm4, %xmm4, %xmm4
-; AVX1-NEXT: vpblendvb %xmm4, %xmm7, %xmm1, %xmm1
-; AVX1-NEXT: vpsrlw $1, %xmm1, %xmm7
-; AVX1-NEXT: vpand %xmm6, %xmm7, %xmm6
-; AVX1-NEXT: vpaddb %xmm4, %xmm4, %xmm4
-; AVX1-NEXT: vpblendvb %xmm4, %xmm6, %xmm1, %xmm1
-; AVX1-NEXT: vpand %xmm5, %xmm2, %xmm2
+; AVX1-NEXT: vpand {{.*}}(%rip), %xmm1, %xmm1
+; AVX1-NEXT: vpsrlw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT: vpcmpeqd %xmm5, %xmm5, %xmm5
+; AVX1-NEXT: vpsrlw %xmm4, %xmm5, %xmm4
+; AVX1-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1]
+; AVX1-NEXT: vpand %xmm4, %xmm1, %xmm1
+; AVX1-NEXT: vpand %xmm3, %xmm2, %xmm2
; AVX1-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
; AVX1-NEXT: vpsllw %xmm2, %xmm0, %xmm0
-; AVX1-NEXT: vpcmpeqd %xmm4, %xmm4, %xmm4
-; AVX1-NEXT: vpsllw %xmm2, %xmm4, %xmm2
+; AVX1-NEXT: vpsllw %xmm2, %xmm5, %xmm2
+; AVX1-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX1-NEXT: vpshufb %xmm3, %xmm2, %xmm2
; AVX1-NEXT: vpand %xmm2, %xmm0, %xmm0
; AVX1-NEXT: vpor %xmm1, %xmm0, %xmm0
@@ -1877,48 +1587,39 @@ define <16 x i8> @splatvar_funnnel_v16i8(<16 x i8> %x, <16 x i8> %y, <16 x i8> %
;
; AVX2-LABEL: splatvar_funnnel_v16i8:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastb %xmm2, %xmm3
-; AVX2-NEXT: vmovdqa {{.*#+}} xmm4 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; AVX2-NEXT: vpandn %xmm4, %xmm3, %xmm3
-; AVX2-NEXT: vpsllw $5, %xmm3, %xmm3
-; AVX2-NEXT: vpsrlw $1, %xmm1, %xmm1
-; AVX2-NEXT: vmovdqa {{.*#+}} xmm5 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127]
-; AVX2-NEXT: vpand %xmm5, %xmm1, %xmm1
-; AVX2-NEXT: vpsrlw $4, %xmm1, %xmm6
-; AVX2-NEXT: vpand {{.*}}(%rip), %xmm6, %xmm6
-; AVX2-NEXT: vpblendvb %xmm3, %xmm6, %xmm1, %xmm1
-; AVX2-NEXT: vpsrlw $2, %xmm1, %xmm6
-; AVX2-NEXT: vpand {{.*}}(%rip), %xmm6, %xmm6
-; AVX2-NEXT: vpaddb %xmm3, %xmm3, %xmm3
-; AVX2-NEXT: vpblendvb %xmm3, %xmm6, %xmm1, %xmm1
-; AVX2-NEXT: vpsrlw $1, %xmm1, %xmm6
-; AVX2-NEXT: vpand %xmm5, %xmm6, %xmm5
-; AVX2-NEXT: vpaddb %xmm3, %xmm3, %xmm3
-; AVX2-NEXT: vpblendvb %xmm3, %xmm5, %xmm1, %xmm1
-; AVX2-NEXT: vpand %xmm4, %xmm2, %xmm2
+; AVX2-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX2-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX2-NEXT: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
+; AVX2-NEXT: vpsllw %xmm4, %xmm0, %xmm0
+; AVX2-NEXT: vpcmpeqd %xmm5, %xmm5, %xmm5
+; AVX2-NEXT: vpsllw %xmm4, %xmm5, %xmm4
+; AVX2-NEXT: vpbroadcastb %xmm4, %xmm4
+; AVX2-NEXT: vpand %xmm4, %xmm0, %xmm0
+; AVX2-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
-; AVX2-NEXT: vpsllw %xmm2, %xmm0, %xmm0
-; AVX2-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX2-NEXT: vpsllw %xmm2, %xmm3, %xmm2
+; AVX2-NEXT: vpsrlw $1, %xmm1, %xmm1
+; AVX2-NEXT: vpand {{.*}}(%rip), %xmm1, %xmm1
+; AVX2-NEXT: vpsrlw %xmm2, %xmm1, %xmm1
+; AVX2-NEXT: vpsrlw %xmm2, %xmm5, %xmm2
+; AVX2-NEXT: vpsrlw $8, %xmm2, %xmm2
; AVX2-NEXT: vpbroadcastb %xmm2, %xmm2
-; AVX2-NEXT: vpand %xmm2, %xmm0, %xmm0
+; AVX2-NEXT: vpand %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX2-NEXT: retq
;
; AVX512F-LABEL: splatvar_funnnel_v16i8:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpbroadcastb %xmm2, %xmm3
-; AVX512F-NEXT: vmovdqa {{.*#+}} xmm4 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; AVX512F-NEXT: vpand %xmm4, %xmm2, %xmm2
-; AVX512F-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512F-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512F-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512F-NEXT: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
; AVX512F-NEXT: vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
-; AVX512F-NEXT: vpslld %xmm2, %zmm0, %zmm0
-; AVX512F-NEXT: vpandn %xmm4, %xmm3, %xmm2
-; AVX512F-NEXT: vpmovzxbd {{.*#+}} zmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero,xmm2[2],zero,zero,zero,xmm2[3],zero,zero,zero,xmm2[4],zero,zero,zero,xmm2[5],zero,zero,zero,xmm2[6],zero,zero,zero,xmm2[7],zero,zero,zero,xmm2[8],zero,zero,zero,xmm2[9],zero,zero,zero,xmm2[10],zero,zero,zero,xmm2[11],zero,zero,zero,xmm2[12],zero,zero,zero,xmm2[13],zero,zero,zero,xmm2[14],zero,zero,zero,xmm2[15],zero,zero,zero
+; AVX512F-NEXT: vpslld %xmm4, %zmm0, %zmm0
+; AVX512F-NEXT: vpandn %xmm3, %xmm2, %xmm2
+; AVX512F-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
; AVX512F-NEXT: vpsrlw $1, %xmm1, %xmm1
; AVX512F-NEXT: vpand {{.*}}(%rip), %xmm1, %xmm1
; AVX512F-NEXT: vpmovzxbd {{.*#+}} zmm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero,xmm1[2],zero,zero,zero,xmm1[3],zero,zero,zero,xmm1[4],zero,zero,zero,xmm1[5],zero,zero,zero,xmm1[6],zero,zero,zero,xmm1[7],zero,zero,zero,xmm1[8],zero,zero,zero,xmm1[9],zero,zero,zero,xmm1[10],zero,zero,zero,xmm1[11],zero,zero,zero,xmm1[12],zero,zero,zero,xmm1[13],zero,zero,zero,xmm1[14],zero,zero,zero,xmm1[15],zero,zero,zero
-; AVX512F-NEXT: vpsrlvd %zmm2, %zmm1, %zmm1
+; AVX512F-NEXT: vpsrld %xmm2, %zmm1, %zmm1
; AVX512F-NEXT: vpord %zmm1, %zmm0, %zmm0
; AVX512F-NEXT: vpmovdb %zmm0, %xmm0
; AVX512F-NEXT: vzeroupper
@@ -1926,18 +1627,17 @@ define <16 x i8> @splatvar_funnnel_v16i8(<16 x i8> %x, <16 x i8> %y, <16 x i8> %
;
; AVX512VL-LABEL: splatvar_funnnel_v16i8:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastb %xmm2, %xmm3
-; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm4 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; AVX512VL-NEXT: vpand %xmm4, %xmm2, %xmm2
-; AVX512VL-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512VL-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512VL-NEXT: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
; AVX512VL-NEXT: vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
-; AVX512VL-NEXT: vpslld %xmm2, %zmm0, %zmm0
-; AVX512VL-NEXT: vpandn %xmm4, %xmm3, %xmm2
-; AVX512VL-NEXT: vpmovzxbd {{.*#+}} zmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero,xmm2[2],zero,zero,zero,xmm2[3],zero,zero,zero,xmm2[4],zero,zero,zero,xmm2[5],zero,zero,zero,xmm2[6],zero,zero,zero,xmm2[7],zero,zero,zero,xmm2[8],zero,zero,zero,xmm2[9],zero,zero,zero,xmm2[10],zero,zero,zero,xmm2[11],zero,zero,zero,xmm2[12],zero,zero,zero,xmm2[13],zero,zero,zero,xmm2[14],zero,zero,zero,xmm2[15],zero,zero,zero
+; AVX512VL-NEXT: vpslld %xmm4, %zmm0, %zmm0
+; AVX512VL-NEXT: vpandn %xmm3, %xmm2, %xmm2
+; AVX512VL-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
; AVX512VL-NEXT: vpsrlw $1, %xmm1, %xmm1
; AVX512VL-NEXT: vpand {{.*}}(%rip), %xmm1, %xmm1
; AVX512VL-NEXT: vpmovzxbd {{.*#+}} zmm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero,xmm1[2],zero,zero,zero,xmm1[3],zero,zero,zero,xmm1[4],zero,zero,zero,xmm1[5],zero,zero,zero,xmm1[6],zero,zero,zero,xmm1[7],zero,zero,zero,xmm1[8],zero,zero,zero,xmm1[9],zero,zero,zero,xmm1[10],zero,zero,zero,xmm1[11],zero,zero,zero,xmm1[12],zero,zero,zero,xmm1[13],zero,zero,zero,xmm1[14],zero,zero,zero,xmm1[15],zero,zero,zero
-; AVX512VL-NEXT: vpsrlvd %zmm2, %zmm1, %zmm1
+; AVX512VL-NEXT: vpsrld %xmm2, %zmm1, %zmm1
; AVX512VL-NEXT: vpord %zmm1, %zmm0, %zmm0
; AVX512VL-NEXT: vpmovdb %zmm0, %xmm0
; AVX512VL-NEXT: vzeroupper
@@ -1945,18 +1645,17 @@ define <16 x i8> @splatvar_funnnel_v16i8(<16 x i8> %x, <16 x i8> %y, <16 x i8> %
;
; AVX512BW-LABEL: splatvar_funnnel_v16i8:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpbroadcastb %xmm2, %xmm3
-; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm4 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; AVX512BW-NEXT: vpandn %xmm4, %xmm3, %xmm3
-; AVX512BW-NEXT: vpmovzxbw {{.*#+}} ymm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero,xmm3[8],zero,xmm3[9],zero,xmm3[10],zero,xmm3[11],zero,xmm3[12],zero,xmm3[13],zero,xmm3[14],zero,xmm3[15],zero
+; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512BW-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512BW-NEXT: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512BW-NEXT: vpmovzxbw {{.*#+}} ymm0 = 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: vpsllw %xmm4, %ymm0, %ymm0
+; AVX512BW-NEXT: vpandn %xmm3, %xmm2, %xmm2
+; AVX512BW-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
; AVX512BW-NEXT: vpsrlw $1, %xmm1, %xmm1
; AVX512BW-NEXT: vpand {{.*}}(%rip), %xmm1, %xmm1
; 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: vpsrlvw %zmm3, %zmm1, %zmm1
-; AVX512BW-NEXT: vpand %xmm4, %xmm2, %xmm2
-; AVX512BW-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512BW-NEXT: vpmovzxbw {{.*#+}} ymm0 = 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: vpsllw %xmm2, %ymm0, %ymm0
+; AVX512BW-NEXT: vpsrlw %xmm2, %ymm1, %ymm1
; AVX512BW-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX512BW-NEXT: vpmovwb %zmm0, %ymm0
; AVX512BW-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0
@@ -1965,18 +1664,17 @@ define <16 x i8> @splatvar_funnnel_v16i8(<16 x i8> %x, <16 x i8> %y, <16 x i8> %
;
; AVX512VBMI2-LABEL: splatvar_funnnel_v16i8:
; AVX512VBMI2: # %bb.0:
-; AVX512VBMI2-NEXT: vpbroadcastb %xmm2, %xmm3
-; AVX512VBMI2-NEXT: vmovdqa {{.*#+}} xmm4 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; AVX512VBMI2-NEXT: vpandn %xmm4, %xmm3, %xmm3
-; AVX512VBMI2-NEXT: vpmovzxbw {{.*#+}} ymm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero,xmm3[8],zero,xmm3[9],zero,xmm3[10],zero,xmm3[11],zero,xmm3[12],zero,xmm3[13],zero,xmm3[14],zero,xmm3[15],zero
+; AVX512VBMI2-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512VBMI2-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512VBMI2-NEXT: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512VBMI2-NEXT: vpmovzxbw {{.*#+}} ymm0 = 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: vpsllw %xmm4, %ymm0, %ymm0
+; AVX512VBMI2-NEXT: vpandn %xmm3, %xmm2, %xmm2
+; AVX512VBMI2-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
; AVX512VBMI2-NEXT: vpsrlw $1, %xmm1, %xmm1
; AVX512VBMI2-NEXT: vpand {{.*}}(%rip), %xmm1, %xmm1
; 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: vpsrlvw %zmm3, %zmm1, %zmm1
-; AVX512VBMI2-NEXT: vpand %xmm4, %xmm2, %xmm2
-; AVX512VBMI2-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512VBMI2-NEXT: vpmovzxbw {{.*#+}} ymm0 = 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: vpsllw %xmm2, %ymm0, %ymm0
+; AVX512VBMI2-NEXT: vpsrlw %xmm2, %ymm1, %ymm1
; AVX512VBMI2-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX512VBMI2-NEXT: vpmovwb %zmm0, %ymm0
; AVX512VBMI2-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0
@@ -1985,18 +1683,17 @@ 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, %xmm3
-; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm4 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; AVX512VLBW-NEXT: vpand %xmm4, %xmm2, %xmm2
-; AVX512VLBW-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
+; 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: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
; AVX512VLBW-NEXT: vpmovzxbw {{.*#+}} ymm0 = 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: vpsllw %xmm2, %ymm0, %ymm0
-; AVX512VLBW-NEXT: vpandn %xmm4, %xmm3, %xmm2
-; AVX512VLBW-NEXT: vpmovzxbw {{.*#+}} ymm2 = 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: vpsllw %xmm4, %ymm0, %ymm0
+; AVX512VLBW-NEXT: vpandn %xmm3, %xmm2, %xmm2
+; AVX512VLBW-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
; AVX512VLBW-NEXT: vpsrlw $1, %xmm1, %xmm1
; AVX512VLBW-NEXT: vpand {{.*}}(%rip), %xmm1, %xmm1
; 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 %ymm2, %ymm1, %ymm1
+; AVX512VLBW-NEXT: vpsrlw %xmm2, %ymm1, %ymm1
; AVX512VLBW-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX512VLBW-NEXT: vpmovwb %ymm0, %xmm0
; AVX512VLBW-NEXT: vzeroupper
@@ -2004,18 +1701,17 @@ 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, %xmm3
-; AVX512VLVBMI2-NEXT: vmovdqa {{.*#+}} xmm4 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; AVX512VLVBMI2-NEXT: vpand %xmm4, %xmm2, %xmm2
-; AVX512VLVBMI2-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
+; 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: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
; AVX512VLVBMI2-NEXT: vpmovzxbw {{.*#+}} ymm0 = 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: vpsllw %xmm2, %ymm0, %ymm0
-; AVX512VLVBMI2-NEXT: vpandn %xmm4, %xmm3, %xmm2
-; AVX512VLVBMI2-NEXT: vpmovzxbw {{.*#+}} ymm2 = 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: vpsllw %xmm4, %ymm0, %ymm0
+; AVX512VLVBMI2-NEXT: vpandn %xmm3, %xmm2, %xmm2
+; AVX512VLVBMI2-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
; AVX512VLVBMI2-NEXT: vpsrlw $1, %xmm1, %xmm1
; AVX512VLVBMI2-NEXT: vpand {{.*}}(%rip), %xmm1, %xmm1
; 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 %ymm2, %ymm1, %ymm1
+; AVX512VLVBMI2-NEXT: vpsrlw %xmm2, %ymm1, %ymm1
; AVX512VLVBMI2-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX512VLVBMI2-NEXT: vpmovwb %ymm0, %xmm0
; AVX512VLVBMI2-NEXT: vzeroupper
@@ -2053,54 +1749,32 @@ define <16 x i8> @splatvar_funnnel_v16i8(<16 x i8> %x, <16 x i8> %y, <16 x i8> %
;
; X86-SSE2-LABEL: splatvar_funnnel_v16i8:
; X86-SSE2: # %bb.0:
-; X86-SSE2-NEXT: movdqa %xmm2, %xmm3
-; X86-SSE2-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3],xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
-; X86-SSE2-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,0,0,0,4,5,6,7]
-; X86-SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm3[0,0,0,0]
-; X86-SSE2-NEXT: psrlw $1, %xmm1
-; X86-SSE2-NEXT: movdqa {{.*#+}} xmm4 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127]
-; X86-SSE2-NEXT: pand %xmm4, %xmm1
; X86-SSE2-NEXT: movdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; X86-SSE2-NEXT: pandn %xmm3, %xmm5
-; X86-SSE2-NEXT: psllw $5, %xmm5
-; X86-SSE2-NEXT: pxor %xmm6, %xmm6
-; X86-SSE2-NEXT: pcmpgtb %xmm5, %xmm6
-; X86-SSE2-NEXT: movdqa %xmm6, %xmm7
-; X86-SSE2-NEXT: pandn %xmm1, %xmm7
-; X86-SSE2-NEXT: psrlw $4, %xmm1
-; X86-SSE2-NEXT: pand %xmm6, %xmm1
-; X86-SSE2-NEXT: pand {{\.LCPI.*}}, %xmm1
-; X86-SSE2-NEXT: por %xmm7, %xmm1
-; X86-SSE2-NEXT: paddb %xmm5, %xmm5
-; X86-SSE2-NEXT: pxor %xmm6, %xmm6
-; X86-SSE2-NEXT: pcmpgtb %xmm5, %xmm6
-; X86-SSE2-NEXT: movdqa %xmm6, %xmm7
-; X86-SSE2-NEXT: pandn %xmm1, %xmm7
-; X86-SSE2-NEXT: psrlw $2, %xmm1
-; X86-SSE2-NEXT: pand %xmm6, %xmm1
-; X86-SSE2-NEXT: pand {{\.LCPI.*}}, %xmm1
-; X86-SSE2-NEXT: por %xmm7, %xmm1
-; X86-SSE2-NEXT: pxor %xmm6, %xmm6
-; X86-SSE2-NEXT: paddb %xmm5, %xmm5
-; X86-SSE2-NEXT: pcmpgtb %xmm5, %xmm6
-; X86-SSE2-NEXT: movdqa %xmm6, %xmm5
-; X86-SSE2-NEXT: pand %xmm4, %xmm6
-; X86-SSE2-NEXT: pandn %xmm1, %xmm5
-; X86-SSE2-NEXT: psrlw $1, %xmm1
-; X86-SSE2-NEXT: pand %xmm1, %xmm6
-; X86-SSE2-NEXT: pand %xmm3, %xmm2
+; X86-SSE2-NEXT: movdqa %xmm2, %xmm4
+; X86-SSE2-NEXT: pand %xmm3, %xmm4
+; X86-SSE2-NEXT: pslldq {{.*#+}} xmm4 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm4[0]
+; X86-SSE2-NEXT: psrldq {{.*#+}} xmm4 = xmm4[15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-SSE2-NEXT: psllw %xmm4, %xmm0
+; X86-SSE2-NEXT: pcmpeqd %xmm5, %xmm5
+; X86-SSE2-NEXT: pcmpeqd %xmm6, %xmm6
+; X86-SSE2-NEXT: psllw %xmm4, %xmm6
+; X86-SSE2-NEXT: punpcklbw {{.*#+}} xmm6 = xmm6[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; X86-SSE2-NEXT: pshuflw {{.*#+}} xmm4 = xmm6[0,0,0,0,4,5,6,7]
+; X86-SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm4[0,0,0,0]
+; X86-SSE2-NEXT: pand %xmm4, %xmm0
+; X86-SSE2-NEXT: pandn %xmm3, %xmm2
; X86-SSE2-NEXT: pslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0]
; X86-SSE2-NEXT: psrldq {{.*#+}} xmm2 = xmm2[15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; X86-SSE2-NEXT: psllw %xmm2, %xmm0
-; X86-SSE2-NEXT: pcmpeqd %xmm1, %xmm1
-; X86-SSE2-NEXT: psllw %xmm2, %xmm1
-; X86-SSE2-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; X86-SSE2-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
-; X86-SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
-; X86-SSE2-NEXT: pand %xmm0, %xmm1
-; X86-SSE2-NEXT: por %xmm5, %xmm1
-; X86-SSE2-NEXT: por %xmm6, %xmm1
-; X86-SSE2-NEXT: movdqa %xmm1, %xmm0
+; X86-SSE2-NEXT: psrlw $1, %xmm1
+; X86-SSE2-NEXT: pand {{\.LCPI.*}}, %xmm1
+; X86-SSE2-NEXT: psrlw %xmm2, %xmm1
+; X86-SSE2-NEXT: psrlw %xmm2, %xmm5
+; X86-SSE2-NEXT: psrlw $8, %xmm5
+; X86-SSE2-NEXT: punpcklbw {{.*#+}} xmm5 = xmm5[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; X86-SSE2-NEXT: pshuflw {{.*#+}} xmm2 = xmm5[0,0,0,0,4,5,6,7]
+; X86-SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
+; X86-SSE2-NEXT: pand %xmm1, %xmm2
+; X86-SSE2-NEXT: por %xmm2, %xmm0
; X86-SSE2-NEXT: retl
%splat = shufflevector <16 x i8> %amt, <16 x i8> undef, <16 x i32> zeroinitializer
%res = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %x, <16 x i8> %y, <16 x i8> %splat)
diff --git a/llvm/test/CodeGen/X86/vector-fshl-256.ll b/llvm/test/CodeGen/X86/vector-fshl-256.ll
index 6308a04ea079..ae2cdf3a24e1 100644
--- a/llvm/test/CodeGen/X86/vector-fshl-256.ll
+++ b/llvm/test/CodeGen/X86/vector-fshl-256.ll
@@ -821,24 +821,16 @@ define <32 x i8> @var_funnnel_v32i8(<32 x i8> %x, <32 x i8> %y, <32 x i8> %amt)
define <4 x i64> @splatvar_funnnel_v4i64(<4 x i64> %x, <4 x i64> %y, <4 x i64> %amt) nounwind {
; AVX1-LABEL: splatvar_funnnel_v4i64:
; AVX1: # %bb.0:
-; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm2[0,1,0,1]
-; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm3, %ymm3
-; AVX1-NEXT: vandnps {{.*}}(%rip), %ymm3, %ymm3
-; AVX1-NEXT: vextractf128 $1, %ymm3, %xmm4
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; AVX1-NEXT: vpandn %xmm3, %xmm2, %xmm4
; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm5
; AVX1-NEXT: vpsrlq $1, %xmm5, %xmm5
-; AVX1-NEXT: vpsrlq %xmm4, %xmm5, %xmm6
-; AVX1-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[2,3,2,3]
-; AVX1-NEXT: vpsrlq %xmm4, %xmm5, %xmm4
-; AVX1-NEXT: vpblendw {{.*#+}} xmm4 = xmm6[0,1,2,3],xmm4[4,5,6,7]
+; AVX1-NEXT: vpsrlq %xmm4, %xmm5, %xmm5
; AVX1-NEXT: vpsrlq $1, %xmm1, %xmm1
-; AVX1-NEXT: vpsrlq %xmm3, %xmm1, %xmm5
-; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[2,3,2,3]
-; AVX1-NEXT: vpsrlq %xmm3, %xmm1, %xmm1
-; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm5[0,1,2,3],xmm1[4,5,6,7]
-; AVX1-NEXT: vinsertf128 $1, %xmm4, %ymm1, %ymm1
+; AVX1-NEXT: vpsrlq %xmm4, %xmm1, %xmm1
+; AVX1-NEXT: vinsertf128 $1, %xmm5, %ymm1, %ymm1
+; AVX1-NEXT: vpand %xmm3, %xmm2, %xmm2
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
-; AVX1-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
; AVX1-NEXT: vpsllq %xmm2, %xmm3, %xmm3
; AVX1-NEXT: vpsllq %xmm2, %xmm0, %xmm0
; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
@@ -847,48 +839,45 @@ define <4 x i64> @splatvar_funnnel_v4i64(<4 x i64> %x, <4 x i64> %y, <4 x i64> %
;
; AVX2-LABEL: splatvar_funnnel_v4i64:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastq %xmm2, %ymm3
-; AVX2-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX2-NEXT: vpsllq %xmm2, %ymm0, %ymm0
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm2 = [63,63,63,63]
-; AVX2-NEXT: vpandn %ymm2, %ymm3, %ymm2
+; AVX2-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; AVX2-NEXT: vpandn %xmm3, %xmm2, %xmm4
; AVX2-NEXT: vpsrlq $1, %ymm1, %ymm1
-; AVX2-NEXT: vpsrlvq %ymm2, %ymm1, %ymm1
+; AVX2-NEXT: vpsrlq %xmm4, %ymm1, %ymm1
+; AVX2-NEXT: vpand %xmm3, %xmm2, %xmm2
+; AVX2-NEXT: vpsllq %xmm2, %ymm0, %ymm0
; AVX2-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX2-NEXT: retq
;
; AVX512F-LABEL: splatvar_funnnel_v4i64:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpbroadcastq %xmm2, %ymm3
-; AVX512F-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX512F-NEXT: vpsllq %xmm2, %ymm0, %ymm0
-; AVX512F-NEXT: vpbroadcastq {{.*#+}} ymm2 = [63,63,63,63]
-; AVX512F-NEXT: vpandn %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; AVX512F-NEXT: vpandn %xmm3, %xmm2, %xmm4
; AVX512F-NEXT: vpsrlq $1, %ymm1, %ymm1
-; AVX512F-NEXT: vpsrlvq %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT: vpsrlq %xmm4, %ymm1, %ymm1
+; AVX512F-NEXT: vpand %xmm3, %xmm2, %xmm2
+; AVX512F-NEXT: vpsllq %xmm2, %ymm0, %ymm0
; AVX512F-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: splatvar_funnnel_v4i64:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastq %xmm2, %ymm3
-; AVX512VL-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX512VL-NEXT: vpsllq %xmm2, %ymm0, %ymm0
+; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; AVX512VL-NEXT: vpandn %xmm3, %xmm2, %xmm4
; AVX512VL-NEXT: vpsrlq $1, %ymm1, %ymm1
-; AVX512VL-NEXT: vpandnq {{.*}}(%rip){1to4}, %ymm3, %ymm2
-; AVX512VL-NEXT: vpsrlvq %ymm2, %ymm1, %ymm1
+; AVX512VL-NEXT: vpsrlq %xmm4, %ymm1, %ymm1
+; AVX512VL-NEXT: vpand %xmm3, %xmm2, %xmm2
+; AVX512VL-NEXT: vpsllq %xmm2, %ymm0, %ymm0
; AVX512VL-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: splatvar_funnnel_v4i64:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpbroadcastq %xmm2, %ymm3
-; AVX512BW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX512BW-NEXT: vpsllq %xmm2, %ymm0, %ymm0
-; AVX512BW-NEXT: vpbroadcastq {{.*#+}} ymm2 = [63,63,63,63]
-; AVX512BW-NEXT: vpandn %ymm2, %ymm3, %ymm2
+; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; AVX512BW-NEXT: vpandn %xmm3, %xmm2, %xmm4
; AVX512BW-NEXT: vpsrlq $1, %ymm1, %ymm1
-; AVX512BW-NEXT: vpsrlvq %ymm2, %ymm1, %ymm1
+; AVX512BW-NEXT: vpsrlq %xmm4, %ymm1, %ymm1
+; AVX512BW-NEXT: vpand %xmm3, %xmm2, %xmm2
+; AVX512BW-NEXT: vpsllq %xmm2, %ymm0, %ymm0
; AVX512BW-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX512BW-NEXT: retq
;
@@ -903,12 +892,12 @@ define <4 x i64> @splatvar_funnnel_v4i64(<4 x i64> %x, <4 x i64> %y, <4 x i64> %
;
; AVX512VLBW-LABEL: splatvar_funnnel_v4i64:
; AVX512VLBW: # %bb.0:
-; AVX512VLBW-NEXT: vpbroadcastq %xmm2, %ymm3
-; AVX512VLBW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX512VLBW-NEXT: vpsllq %xmm2, %ymm0, %ymm0
+; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; AVX512VLBW-NEXT: vpandn %xmm3, %xmm2, %xmm4
; AVX512VLBW-NEXT: vpsrlq $1, %ymm1, %ymm1
-; AVX512VLBW-NEXT: vpandnq {{.*}}(%rip){1to4}, %ymm3, %ymm2
-; AVX512VLBW-NEXT: vpsrlvq %ymm2, %ymm1, %ymm1
+; AVX512VLBW-NEXT: vpsrlq %xmm4, %ymm1, %ymm1
+; AVX512VLBW-NEXT: vpand %xmm3, %xmm2, %xmm2
+; AVX512VLBW-NEXT: vpsllq %xmm2, %ymm0, %ymm0
; AVX512VLBW-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX512VLBW-NEXT: retq
;
@@ -920,36 +909,30 @@ define <4 x i64> @splatvar_funnnel_v4i64(<4 x i64> %x, <4 x i64> %y, <4 x i64> %
;
; XOPAVX1-LABEL: splatvar_funnnel_v4i64:
; XOPAVX1: # %bb.0:
-; XOPAVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm2[0,1,0,1]
-; XOPAVX1-NEXT: vinsertf128 $1, %xmm3, %ymm3, %ymm3
-; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm4
-; XOPAVX1-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
-; XOPAVX1-NEXT: vpsllq %xmm2, %xmm4, %xmm4
-; XOPAVX1-NEXT: vpsllq %xmm2, %xmm0, %xmm0
-; XOPAVX1-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm0
-; XOPAVX1-NEXT: vandnps {{.*}}(%rip), %ymm3, %ymm2
-; XOPAVX1-NEXT: vextractf128 $1, %ymm2, %xmm3
-; XOPAVX1-NEXT: vpxor %xmm4, %xmm4, %xmm4
-; XOPAVX1-NEXT: vpsubq %xmm3, %xmm4, %xmm3
+; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; XOPAVX1-NEXT: vpandn %xmm3, %xmm2, %xmm4
; XOPAVX1-NEXT: vextractf128 $1, %ymm1, %xmm5
; XOPAVX1-NEXT: vpsrlq $1, %xmm5, %xmm5
-; XOPAVX1-NEXT: vpshlq %xmm3, %xmm5, %xmm3
-; XOPAVX1-NEXT: vpsubq %xmm2, %xmm4, %xmm2
+; XOPAVX1-NEXT: vpsrlq %xmm4, %xmm5, %xmm5
; XOPAVX1-NEXT: vpsrlq $1, %xmm1, %xmm1
-; XOPAVX1-NEXT: vpshlq %xmm2, %xmm1, %xmm1
-; XOPAVX1-NEXT: vinsertf128 $1, %xmm3, %ymm1, %ymm1
+; XOPAVX1-NEXT: vpsrlq %xmm4, %xmm1, %xmm1
+; XOPAVX1-NEXT: vinsertf128 $1, %xmm5, %ymm1, %ymm1
+; XOPAVX1-NEXT: vpand %xmm3, %xmm2, %xmm2
+; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
+; XOPAVX1-NEXT: vpsllq %xmm2, %xmm3, %xmm3
+; XOPAVX1-NEXT: vpsllq %xmm2, %xmm0, %xmm0
+; XOPAVX1-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; XOPAVX1-NEXT: vorps %ymm1, %ymm0, %ymm0
; XOPAVX1-NEXT: retq
;
; XOPAVX2-LABEL: splatvar_funnnel_v4i64:
; XOPAVX2: # %bb.0:
-; XOPAVX2-NEXT: vpbroadcastq %xmm2, %ymm3
-; XOPAVX2-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
-; XOPAVX2-NEXT: vpsllq %xmm2, %ymm0, %ymm0
-; XOPAVX2-NEXT: vpbroadcastq {{.*#+}} ymm2 = [63,63,63,63]
-; XOPAVX2-NEXT: vpandn %ymm2, %ymm3, %ymm2
+; XOPAVX2-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; XOPAVX2-NEXT: vpandn %xmm3, %xmm2, %xmm4
; XOPAVX2-NEXT: vpsrlq $1, %ymm1, %ymm1
-; XOPAVX2-NEXT: vpsrlvq %ymm2, %ymm1, %ymm1
+; XOPAVX2-NEXT: vpsrlq %xmm4, %ymm1, %ymm1
+; XOPAVX2-NEXT: vpand %xmm3, %xmm2, %xmm2
+; XOPAVX2-NEXT: vpsllq %xmm2, %ymm0, %ymm0
; XOPAVX2-NEXT: vpor %ymm1, %ymm0, %ymm0
; XOPAVX2-NEXT: retq
%splat = shufflevector <4 x i64> %amt, <4 x i64> undef, <4 x i32> zeroinitializer
@@ -960,40 +943,19 @@ define <4 x i64> @splatvar_funnnel_v4i64(<4 x i64> %x, <4 x i64> %y, <4 x i64> %
define <8 x i32> @splatvar_funnnel_v8i32(<8 x i32> %x, <8 x i32> %y, <8 x i32> %amt) nounwind {
; AVX1-LABEL: splatvar_funnnel_v8i32:
; AVX1: # %bb.0:
-; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm2[0,0,0,0]
-; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm3, %ymm3
-; AVX1-NEXT: vandnps {{.*}}(%rip), %ymm3, %ymm3
-; AVX1-NEXT: vextractf128 $1, %ymm3, %xmm4
-; AVX1-NEXT: vpsrldq {{.*#+}} xmm5 = xmm4[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm6
-; AVX1-NEXT: vpsrld $1, %xmm6, %xmm6
-; AVX1-NEXT: vpsrld %xmm5, %xmm6, %xmm5
-; AVX1-NEXT: vpsrlq $32, %xmm4, %xmm7
-; AVX1-NEXT: vpsrld %xmm7, %xmm6, %xmm7
-; AVX1-NEXT: vpblendw {{.*#+}} xmm5 = xmm7[0,1,2,3],xmm5[4,5,6,7]
-; AVX1-NEXT: vpxor %xmm8, %xmm8, %xmm8
-; AVX1-NEXT: vpunpckhdq {{.*#+}} xmm7 = xmm4[2],xmm8[2],xmm4[3],xmm8[3]
-; AVX1-NEXT: vpsrld %xmm7, %xmm6, %xmm7
+; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [31,31,31,31]
+; AVX1-NEXT: vpandn %xmm3, %xmm2, %xmm4
; AVX1-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
-; AVX1-NEXT: vpsrld %xmm4, %xmm6, %xmm4
-; AVX1-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3],xmm7[4,5,6,7]
-; AVX1-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm5[2,3],xmm4[4,5],xmm5[6,7]
-; AVX1-NEXT: vpsrldq {{.*#+}} xmm5 = xmm3[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm5
+; AVX1-NEXT: vpsrld $1, %xmm5, %xmm5
+; AVX1-NEXT: vpsrld %xmm4, %xmm5, %xmm5
; AVX1-NEXT: vpsrld $1, %xmm1, %xmm1
-; AVX1-NEXT: vpsrld %xmm5, %xmm1, %xmm5
-; AVX1-NEXT: vpsrlq $32, %xmm3, %xmm6
-; AVX1-NEXT: vpsrld %xmm6, %xmm1, %xmm6
-; AVX1-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0,1,2,3],xmm5[4,5,6,7]
-; AVX1-NEXT: vpunpckhdq {{.*#+}} xmm6 = xmm3[2],xmm8[2],xmm3[3],xmm8[3]
-; AVX1-NEXT: vpsrld %xmm6, %xmm1, %xmm6
-; AVX1-NEXT: vpmovzxdq {{.*#+}} xmm3 = xmm3[0],zero,xmm3[1],zero
-; AVX1-NEXT: vpsrld %xmm3, %xmm1, %xmm1
-; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm6[4,5,6,7]
-; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm5[2,3],xmm1[4,5],xmm5[6,7]
-; AVX1-NEXT: vinsertf128 $1, %xmm4, %ymm1, %ymm1
-; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
-; AVX1-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; AVX1-NEXT: vpsrld %xmm4, %xmm1, %xmm1
+; AVX1-NEXT: vinsertf128 $1, %xmm5, %ymm1, %ymm1
+; AVX1-NEXT: vpand %xmm3, %xmm2, %xmm2
; AVX1-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX1-NEXT: vpslld %xmm2, %xmm3, %xmm3
; AVX1-NEXT: vpslld %xmm2, %xmm0, %xmm0
; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
@@ -1002,11 +964,12 @@ define <8 x i32> @splatvar_funnnel_v8i32(<8 x i32> %x, <8 x i32> %y, <8 x i32> %
;
; AVX2-LABEL: splatvar_funnnel_v8i32:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastd %xmm2, %ymm2
-; AVX2-NEXT: vpbroadcastd {{.*#+}} ymm3 = [31,31,31,31,31,31,31,31]
-; AVX2-NEXT: vpandn %ymm3, %ymm2, %ymm4
+; AVX2-NEXT: vpbroadcastd %xmm2, %xmm2
+; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31]
+; AVX2-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX2-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
; AVX2-NEXT: vpsrld $1, %ymm1, %ymm1
-; AVX2-NEXT: vpsrlvd %ymm4, %ymm1, %ymm1
+; AVX2-NEXT: vpsrld %xmm4, %ymm1, %ymm1
; AVX2-NEXT: vpand %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
; AVX2-NEXT: vpslld %xmm2, %ymm0, %ymm0
@@ -1015,11 +978,12 @@ define <8 x i32> @splatvar_funnnel_v8i32(<8 x i32> %x, <8 x i32> %y, <8 x i32> %
;
; AVX512F-LABEL: splatvar_funnnel_v8i32:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpbroadcastd %xmm2, %ymm2
-; AVX512F-NEXT: vpbroadcastd {{.*#+}} ymm3 = [31,31,31,31,31,31,31,31]
-; AVX512F-NEXT: vpandn %ymm3, %ymm2, %ymm4
+; AVX512F-NEXT: vpbroadcastd %xmm2, %xmm2
+; AVX512F-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31]
+; AVX512F-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX512F-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
; AVX512F-NEXT: vpsrld $1, %ymm1, %ymm1
-; AVX512F-NEXT: vpsrlvd %ymm4, %ymm1, %ymm1
+; AVX512F-NEXT: vpsrld %xmm4, %ymm1, %ymm1
; AVX512F-NEXT: vpand %xmm3, %xmm2, %xmm2
; AVX512F-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
; AVX512F-NEXT: vpslld %xmm2, %ymm0, %ymm0
@@ -1028,11 +992,12 @@ define <8 x i32> @splatvar_funnnel_v8i32(<8 x i32> %x, <8 x i32> %y, <8 x i32> %
;
; AVX512VL-LABEL: splatvar_funnnel_v8i32:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastd %xmm2, %ymm2
-; AVX512VL-NEXT: vpbroadcastd {{.*#+}} ymm3 = [31,31,31,31,31,31,31,31]
-; AVX512VL-NEXT: vpandn %ymm3, %ymm2, %ymm4
+; AVX512VL-NEXT: vpbroadcastd %xmm2, %xmm2
+; AVX512VL-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31]
+; AVX512VL-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX512VL-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
; AVX512VL-NEXT: vpsrld $1, %ymm1, %ymm1
-; AVX512VL-NEXT: vpsrlvd %ymm4, %ymm1, %ymm1
+; AVX512VL-NEXT: vpsrld %xmm4, %ymm1, %ymm1
; AVX512VL-NEXT: vpand %xmm3, %xmm2, %xmm2
; AVX512VL-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
; AVX512VL-NEXT: vpslld %xmm2, %ymm0, %ymm0
@@ -1041,11 +1006,12 @@ define <8 x i32> @splatvar_funnnel_v8i32(<8 x i32> %x, <8 x i32> %y, <8 x i32> %
;
; AVX512BW-LABEL: splatvar_funnnel_v8i32:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpbroadcastd %xmm2, %ymm2
-; AVX512BW-NEXT: vpbroadcastd {{.*#+}} ymm3 = [31,31,31,31,31,31,31,31]
-; AVX512BW-NEXT: vpandn %ymm3, %ymm2, %ymm4
+; AVX512BW-NEXT: vpbroadcastd %xmm2, %xmm2
+; AVX512BW-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31]
+; AVX512BW-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX512BW-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
; AVX512BW-NEXT: vpsrld $1, %ymm1, %ymm1
-; AVX512BW-NEXT: vpsrlvd %ymm4, %ymm1, %ymm1
+; AVX512BW-NEXT: vpsrld %xmm4, %ymm1, %ymm1
; AVX512BW-NEXT: vpand %xmm3, %xmm2, %xmm2
; AVX512BW-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
; AVX512BW-NEXT: vpslld %xmm2, %ymm0, %ymm0
@@ -1063,11 +1029,12 @@ define <8 x i32> @splatvar_funnnel_v8i32(<8 x i32> %x, <8 x i32> %y, <8 x i32> %
;
; AVX512VLBW-LABEL: splatvar_funnnel_v8i32:
; AVX512VLBW: # %bb.0:
-; AVX512VLBW-NEXT: vpbroadcastd %xmm2, %ymm2
-; AVX512VLBW-NEXT: vpbroadcastd {{.*#+}} ymm3 = [31,31,31,31,31,31,31,31]
-; AVX512VLBW-NEXT: vpandn %ymm3, %ymm2, %ymm4
+; AVX512VLBW-NEXT: vpbroadcastd %xmm2, %xmm2
+; AVX512VLBW-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31]
+; AVX512VLBW-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX512VLBW-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
; AVX512VLBW-NEXT: vpsrld $1, %ymm1, %ymm1
-; AVX512VLBW-NEXT: vpsrlvd %ymm4, %ymm1, %ymm1
+; AVX512VLBW-NEXT: vpsrld %xmm4, %ymm1, %ymm1
; AVX512VLBW-NEXT: vpand %xmm3, %xmm2, %xmm2
; AVX512VLBW-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
; AVX512VLBW-NEXT: vpslld %xmm2, %ymm0, %ymm0
@@ -1082,35 +1049,33 @@ define <8 x i32> @splatvar_funnnel_v8i32(<8 x i32> %x, <8 x i32> %y, <8 x i32> %
;
; XOPAVX1-LABEL: splatvar_funnnel_v8i32:
; XOPAVX1: # %bb.0:
-; XOPAVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm2[0,0,0,0]
-; XOPAVX1-NEXT: vinsertf128 $1, %xmm3, %ymm3, %ymm3
-; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm4
-; XOPAVX1-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
-; XOPAVX1-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
-; XOPAVX1-NEXT: vpslld %xmm2, %xmm4, %xmm4
-; XOPAVX1-NEXT: vpslld %xmm2, %xmm0, %xmm0
-; XOPAVX1-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm0
-; XOPAVX1-NEXT: vandnps {{.*}}(%rip), %ymm3, %ymm2
-; XOPAVX1-NEXT: vextractf128 $1, %ymm2, %xmm3
-; XOPAVX1-NEXT: vpxor %xmm4, %xmm4, %xmm4
-; XOPAVX1-NEXT: vpsubd %xmm3, %xmm4, %xmm3
+; XOPAVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
+; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [31,31,31,31]
+; XOPAVX1-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; XOPAVX1-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
; XOPAVX1-NEXT: vextractf128 $1, %ymm1, %xmm5
; XOPAVX1-NEXT: vpsrld $1, %xmm5, %xmm5
-; XOPAVX1-NEXT: vpshld %xmm3, %xmm5, %xmm3
-; XOPAVX1-NEXT: vpsubd %xmm2, %xmm4, %xmm2
+; XOPAVX1-NEXT: vpsrld %xmm4, %xmm5, %xmm5
; XOPAVX1-NEXT: vpsrld $1, %xmm1, %xmm1
-; XOPAVX1-NEXT: vpshld %xmm2, %xmm1, %xmm1
-; XOPAVX1-NEXT: vinsertf128 $1, %xmm3, %ymm1, %ymm1
+; XOPAVX1-NEXT: vpsrld %xmm4, %xmm1, %xmm1
+; XOPAVX1-NEXT: vinsertf128 $1, %xmm5, %ymm1, %ymm1
+; XOPAVX1-NEXT: vpand %xmm3, %xmm2, %xmm2
+; XOPAVX1-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
+; XOPAVX1-NEXT: vpslld %xmm2, %xmm3, %xmm3
+; XOPAVX1-NEXT: vpslld %xmm2, %xmm0, %xmm0
+; XOPAVX1-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; XOPAVX1-NEXT: vorps %ymm1, %ymm0, %ymm0
; XOPAVX1-NEXT: retq
;
; XOPAVX2-LABEL: splatvar_funnnel_v8i32:
; XOPAVX2: # %bb.0:
-; XOPAVX2-NEXT: vpbroadcastd %xmm2, %ymm2
-; XOPAVX2-NEXT: vpbroadcastd {{.*#+}} ymm3 = [31,31,31,31,31,31,31,31]
-; XOPAVX2-NEXT: vpandn %ymm3, %ymm2, %ymm4
+; XOPAVX2-NEXT: vpbroadcastd %xmm2, %xmm2
+; XOPAVX2-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31]
+; XOPAVX2-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; XOPAVX2-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
; XOPAVX2-NEXT: vpsrld $1, %ymm1, %ymm1
-; XOPAVX2-NEXT: vpsrlvd %ymm4, %ymm1, %ymm1
+; XOPAVX2-NEXT: vpsrld %xmm4, %ymm1, %ymm1
; XOPAVX2-NEXT: vpand %xmm3, %xmm2, %xmm2
; XOPAVX2-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
; XOPAVX2-NEXT: vpslld %xmm2, %ymm0, %ymm0
@@ -1125,45 +1090,19 @@ define <16 x i16> @splatvar_funnnel_v16i16(<16 x i16> %x, <16 x i16> %y, <16 x i
; AVX1-LABEL: splatvar_funnnel_v16i16:
; AVX1: # %bb.0:
; AVX1-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,0,0,0,4,5,6,7]
-; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm2[0,0,0,0]
-; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm3, %ymm3
-; AVX1-NEXT: vandnps {{.*}}(%rip), %ymm3, %ymm3
-; AVX1-NEXT: vextractf128 $1, %ymm3, %xmm4
-; AVX1-NEXT: vpsllw $12, %xmm4, %xmm5
-; AVX1-NEXT: vpsllw $4, %xmm4, %xmm4
-; AVX1-NEXT: vpor %xmm5, %xmm4, %xmm4
-; AVX1-NEXT: vpaddw %xmm4, %xmm4, %xmm5
-; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm6
-; AVX1-NEXT: vpsrlw $9, %xmm6, %xmm7
-; AVX1-NEXT: vpsrlw $1, %xmm6, %xmm6
-; AVX1-NEXT: vpblendvb %xmm4, %xmm7, %xmm6, %xmm4
-; AVX1-NEXT: vpsrlw $4, %xmm4, %xmm6
-; AVX1-NEXT: vpblendvb %xmm5, %xmm6, %xmm4, %xmm4
-; AVX1-NEXT: vpsrlw $2, %xmm4, %xmm6
-; AVX1-NEXT: vpaddw %xmm5, %xmm5, %xmm5
-; AVX1-NEXT: vpblendvb %xmm5, %xmm6, %xmm4, %xmm4
-; AVX1-NEXT: vpsrlw $1, %xmm4, %xmm6
-; AVX1-NEXT: vpaddw %xmm5, %xmm5, %xmm5
-; AVX1-NEXT: vpblendvb %xmm5, %xmm6, %xmm4, %xmm4
-; AVX1-NEXT: vpsllw $12, %xmm3, %xmm5
-; AVX1-NEXT: vpsllw $4, %xmm3, %xmm3
-; AVX1-NEXT: vpor %xmm5, %xmm3, %xmm3
-; AVX1-NEXT: vpaddw %xmm3, %xmm3, %xmm5
-; AVX1-NEXT: vpsrlw $9, %xmm1, %xmm6
+; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; AVX1-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX1-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
+; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm5
+; AVX1-NEXT: vpsrlw $1, %xmm5, %xmm5
+; AVX1-NEXT: vpsrlw %xmm4, %xmm5, %xmm5
; AVX1-NEXT: vpsrlw $1, %xmm1, %xmm1
-; AVX1-NEXT: vpblendvb %xmm3, %xmm6, %xmm1, %xmm1
-; AVX1-NEXT: vpsrlw $4, %xmm1, %xmm3
-; AVX1-NEXT: vpblendvb %xmm5, %xmm3, %xmm1, %xmm1
-; AVX1-NEXT: vpsrlw $2, %xmm1, %xmm3
-; AVX1-NEXT: vpaddw %xmm5, %xmm5, %xmm5
-; AVX1-NEXT: vpblendvb %xmm5, %xmm3, %xmm1, %xmm1
-; AVX1-NEXT: vpsrlw $1, %xmm1, %xmm3
-; AVX1-NEXT: vpaddw %xmm5, %xmm5, %xmm5
-; AVX1-NEXT: vpblendvb %xmm5, %xmm3, %xmm1, %xmm1
-; AVX1-NEXT: vinsertf128 $1, %xmm4, %ymm1, %ymm1
-; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
-; AVX1-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; AVX1-NEXT: vpsrlw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT: vinsertf128 $1, %xmm5, %ymm1, %ymm1
+; AVX1-NEXT: vpand %xmm3, %xmm2, %xmm2
; AVX1-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX1-NEXT: vpsllw %xmm2, %xmm3, %xmm3
; AVX1-NEXT: vpsllw %xmm2, %xmm0, %xmm0
; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
@@ -1172,20 +1111,13 @@ define <16 x i16> @splatvar_funnnel_v16i16(<16 x i16> %x, <16 x i16> %y, <16 x i
;
; AVX2-LABEL: splatvar_funnnel_v16i16:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastw %xmm2, %ymm2
+; AVX2-NEXT: vpbroadcastw %xmm2, %xmm2
+; AVX2-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; AVX2-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX2-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
; AVX2-NEXT: vpsrlw $1, %ymm1, %ymm1
-; AVX2-NEXT: vpxor %xmm3, %xmm3, %xmm3
-; AVX2-NEXT: vpunpckhwd {{.*#+}} ymm4 = ymm3[4],ymm1[4],ymm3[5],ymm1[5],ymm3[6],ymm1[6],ymm3[7],ymm1[7],ymm3[12],ymm1[12],ymm3[13],ymm1[13],ymm3[14],ymm1[14],ymm3[15],ymm1[15]
-; AVX2-NEXT: vpandn {{.*}}(%rip), %ymm2, %ymm5
-; AVX2-NEXT: vpunpckhwd {{.*#+}} ymm6 = ymm5[4],ymm3[4],ymm5[5],ymm3[5],ymm5[6],ymm3[6],ymm5[7],ymm3[7],ymm5[12],ymm3[12],ymm5[13],ymm3[13],ymm5[14],ymm3[14],ymm5[15],ymm3[15]
-; AVX2-NEXT: vpsrlvd %ymm6, %ymm4, %ymm4
-; AVX2-NEXT: vpsrld $16, %ymm4, %ymm4
-; AVX2-NEXT: vpunpcklwd {{.*#+}} ymm1 = ymm3[0],ymm1[0],ymm3[1],ymm1[1],ymm3[2],ymm1[2],ymm3[3],ymm1[3],ymm3[8],ymm1[8],ymm3[9],ymm1[9],ymm3[10],ymm1[10],ymm3[11],ymm1[11]
-; AVX2-NEXT: vpunpcklwd {{.*#+}} ymm3 = ymm5[0],ymm3[0],ymm5[1],ymm3[1],ymm5[2],ymm3[2],ymm5[3],ymm3[3],ymm5[8],ymm3[8],ymm5[9],ymm3[9],ymm5[10],ymm3[10],ymm5[11],ymm3[11]
-; AVX2-NEXT: vpsrlvd %ymm3, %ymm1, %ymm1
-; AVX2-NEXT: vpsrld $16, %ymm1, %ymm1
-; AVX2-NEXT: vpackusdw %ymm4, %ymm1, %ymm1
-; AVX2-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; AVX2-NEXT: vpsrlw %xmm4, %ymm1, %ymm1
+; AVX2-NEXT: vpand %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
; AVX2-NEXT: vpsllw %xmm2, %ymm0, %ymm0
; AVX2-NEXT: vpor %ymm1, %ymm0, %ymm0
@@ -1193,41 +1125,41 @@ define <16 x i16> @splatvar_funnnel_v16i16(<16 x i16> %x, <16 x i16> %y, <16 x i
;
; AVX512F-LABEL: splatvar_funnnel_v16i16:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpbroadcastw %xmm2, %ymm2
-; AVX512F-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm3
-; AVX512F-NEXT: vpmovzxwq {{.*#+}} xmm3 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero
-; AVX512F-NEXT: vpsllw %xmm3, %ymm0, %ymm0
+; AVX512F-NEXT: vpbroadcastw %xmm2, %xmm2
+; AVX512F-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; AVX512F-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX512F-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
; AVX512F-NEXT: vpsrlw $1, %ymm1, %ymm1
-; AVX512F-NEXT: vpmovzxwd {{.*#+}} zmm1 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero
-; AVX512F-NEXT: vpandn {{.*}}(%rip), %ymm2, %ymm2
-; AVX512F-NEXT: vpmovzxwd {{.*#+}} zmm2 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero
-; AVX512F-NEXT: vpsrlvd %zmm2, %zmm1, %zmm1
-; AVX512F-NEXT: vpmovdw %zmm1, %ymm1
+; AVX512F-NEXT: vpsrlw %xmm4, %ymm1, %ymm1
+; AVX512F-NEXT: vpand %xmm3, %xmm2, %xmm2
+; AVX512F-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
+; AVX512F-NEXT: vpsllw %xmm2, %ymm0, %ymm0
; AVX512F-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: splatvar_funnnel_v16i16:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastw %xmm2, %ymm2
-; AVX512VL-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm3
-; AVX512VL-NEXT: vpmovzxwq {{.*#+}} xmm3 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero
-; AVX512VL-NEXT: vpsllw %xmm3, %ymm0, %ymm0
+; AVX512VL-NEXT: vpbroadcastw %xmm2, %xmm2
+; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; AVX512VL-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX512VL-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
; AVX512VL-NEXT: vpsrlw $1, %ymm1, %ymm1
-; AVX512VL-NEXT: vpmovzxwd {{.*#+}} zmm1 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero
-; AVX512VL-NEXT: vpandn {{.*}}(%rip), %ymm2, %ymm2
-; AVX512VL-NEXT: vpmovzxwd {{.*#+}} zmm2 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero
-; AVX512VL-NEXT: vpsrlvd %zmm2, %zmm1, %zmm1
-; AVX512VL-NEXT: vpmovdw %zmm1, %ymm1
+; AVX512VL-NEXT: vpsrlw %xmm4, %ymm1, %ymm1
+; AVX512VL-NEXT: vpand %xmm3, %xmm2, %xmm2
+; AVX512VL-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
+; AVX512VL-NEXT: vpsllw %xmm2, %ymm0, %ymm0
; AVX512VL-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: splatvar_funnnel_v16i16:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpbroadcastw %xmm2, %ymm2
+; AVX512BW-NEXT: vpbroadcastw %xmm2, %xmm2
+; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX512BW-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
; AVX512BW-NEXT: vpsrlw $1, %ymm1, %ymm1
-; AVX512BW-NEXT: vpandn {{.*}}(%rip), %ymm2, %ymm3
-; AVX512BW-NEXT: vpsrlvw %zmm3, %zmm1, %zmm1
-; AVX512BW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; AVX512BW-NEXT: vpsrlw %xmm4, %ymm1, %ymm1
+; AVX512BW-NEXT: vpand %xmm3, %xmm2, %xmm2
; AVX512BW-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
; AVX512BW-NEXT: vpsllw %xmm2, %ymm0, %ymm0
; AVX512BW-NEXT: vpor %ymm1, %ymm0, %ymm0
@@ -1244,11 +1176,13 @@ define <16 x i16> @splatvar_funnnel_v16i16(<16 x i16> %x, <16 x i16> %y, <16 x i
;
; AVX512VLBW-LABEL: splatvar_funnnel_v16i16:
; AVX512VLBW: # %bb.0:
-; AVX512VLBW-NEXT: vpbroadcastw %xmm2, %ymm2
+; AVX512VLBW-NEXT: vpbroadcastw %xmm2, %xmm2
+; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; AVX512VLBW-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX512VLBW-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
; AVX512VLBW-NEXT: vpsrlw $1, %ymm1, %ymm1
-; AVX512VLBW-NEXT: vpandn {{.*}}(%rip), %ymm2, %ymm3
-; AVX512VLBW-NEXT: vpsrlvw %ymm3, %ymm1, %ymm1
-; AVX512VLBW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; AVX512VLBW-NEXT: vpsrlw %xmm4, %ymm1, %ymm1
+; AVX512VLBW-NEXT: vpand %xmm3, %xmm2, %xmm2
; AVX512VLBW-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
; AVX512VLBW-NEXT: vpsllw %xmm2, %ymm0, %ymm0
; AVX512VLBW-NEXT: vpor %ymm1, %ymm0, %ymm0
@@ -1263,42 +1197,34 @@ define <16 x i16> @splatvar_funnnel_v16i16(<16 x i16> %x, <16 x i16> %y, <16 x i
; XOPAVX1-LABEL: splatvar_funnnel_v16i16:
; XOPAVX1: # %bb.0:
; XOPAVX1-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,0,0,0,4,5,6,7]
-; XOPAVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm2[0,0,0,0]
-; XOPAVX1-NEXT: vinsertf128 $1, %xmm3, %ymm3, %ymm3
-; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm4
-; XOPAVX1-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
-; XOPAVX1-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
-; XOPAVX1-NEXT: vpsllw %xmm2, %xmm4, %xmm4
-; XOPAVX1-NEXT: vpsllw %xmm2, %xmm0, %xmm0
-; XOPAVX1-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm0
-; XOPAVX1-NEXT: vandnps {{.*}}(%rip), %ymm3, %ymm2
-; XOPAVX1-NEXT: vextractf128 $1, %ymm2, %xmm3
-; XOPAVX1-NEXT: vpxor %xmm4, %xmm4, %xmm4
-; XOPAVX1-NEXT: vpsubw %xmm3, %xmm4, %xmm3
+; XOPAVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
+; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; XOPAVX1-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; XOPAVX1-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
; XOPAVX1-NEXT: vextractf128 $1, %ymm1, %xmm5
; XOPAVX1-NEXT: vpsrlw $1, %xmm5, %xmm5
-; XOPAVX1-NEXT: vpshlw %xmm3, %xmm5, %xmm3
-; XOPAVX1-NEXT: vpsubw %xmm2, %xmm4, %xmm2
+; XOPAVX1-NEXT: vpsrlw %xmm4, %xmm5, %xmm5
; XOPAVX1-NEXT: vpsrlw $1, %xmm1, %xmm1
-; XOPAVX1-NEXT: vpshlw %xmm2, %xmm1, %xmm1
-; XOPAVX1-NEXT: vinsertf128 $1, %xmm3, %ymm1, %ymm1
+; XOPAVX1-NEXT: vpsrlw %xmm4, %xmm1, %xmm1
+; XOPAVX1-NEXT: vinsertf128 $1, %xmm5, %ymm1, %ymm1
+; XOPAVX1-NEXT: vpand %xmm3, %xmm2, %xmm2
+; XOPAVX1-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
+; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
+; XOPAVX1-NEXT: vpsllw %xmm2, %xmm3, %xmm3
+; XOPAVX1-NEXT: vpsllw %xmm2, %xmm0, %xmm0
+; XOPAVX1-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; XOPAVX1-NEXT: vorps %ymm1, %ymm0, %ymm0
; XOPAVX1-NEXT: retq
;
; XOPAVX2-LABEL: splatvar_funnnel_v16i16:
; XOPAVX2: # %bb.0:
-; XOPAVX2-NEXT: vpbroadcastw %xmm2, %ymm2
-; XOPAVX2-NEXT: vpandn {{.*}}(%rip), %ymm2, %ymm3
-; XOPAVX2-NEXT: vextracti128 $1, %ymm3, %xmm4
-; XOPAVX2-NEXT: vpxor %xmm5, %xmm5, %xmm5
-; XOPAVX2-NEXT: vpsubw %xmm4, %xmm5, %xmm4
+; XOPAVX2-NEXT: vpbroadcastw %xmm2, %xmm2
+; XOPAVX2-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; XOPAVX2-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; XOPAVX2-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
; XOPAVX2-NEXT: vpsrlw $1, %ymm1, %ymm1
-; XOPAVX2-NEXT: vextracti128 $1, %ymm1, %xmm6
-; XOPAVX2-NEXT: vpshlw %xmm4, %xmm6, %xmm4
-; XOPAVX2-NEXT: vpsubw %xmm3, %xmm5, %xmm3
-; XOPAVX2-NEXT: vpshlw %xmm3, %xmm1, %xmm1
-; XOPAVX2-NEXT: vinserti128 $1, %xmm4, %ymm1, %ymm1
-; XOPAVX2-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; XOPAVX2-NEXT: vpsrlw %xmm4, %ymm1, %ymm1
+; XOPAVX2-NEXT: vpand %xmm3, %xmm2, %xmm2
; XOPAVX2-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
; XOPAVX2-NEXT: vpsllw %xmm2, %ymm0, %ymm0
; XOPAVX2-NEXT: vpor %ymm1, %ymm0, %ymm0
@@ -1312,49 +1238,29 @@ define <32 x i8> @splatvar_funnnel_v32i8(<32 x i8> %x, <32 x i8> %y, <32 x i8> %
; AVX1-LABEL: splatvar_funnnel_v32i8:
; AVX1: # %bb.0:
; AVX1-NEXT: vpxor %xmm8, %xmm8, %xmm8
-; AVX1-NEXT: vpshufb %xmm8, %xmm2, %xmm4
-; AVX1-NEXT: vinsertf128 $1, %xmm4, %ymm4, %ymm4
-; AVX1-NEXT: vandnps {{.*}}(%rip), %ymm4, %ymm4
-; AVX1-NEXT: vextractf128 $1, %ymm4, %xmm5
-; AVX1-NEXT: vpsllw $5, %xmm5, %xmm5
-; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm6
-; AVX1-NEXT: vpsrlw $1, %xmm6, %xmm6
-; AVX1-NEXT: vmovdqa {{.*#+}} xmm7 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127]
-; AVX1-NEXT: vpand %xmm7, %xmm6, %xmm6
-; AVX1-NEXT: vpsrlw $4, %xmm6, %xmm3
-; AVX1-NEXT: vmovdqa {{.*#+}} xmm9 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
-; AVX1-NEXT: vpand %xmm3, %xmm9, %xmm3
-; AVX1-NEXT: vpblendvb %xmm5, %xmm3, %xmm6, %xmm3
-; AVX1-NEXT: vpsrlw $2, %xmm3, %xmm6
-; AVX1-NEXT: vmovdqa {{.*#+}} xmm10 = [63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63]
-; AVX1-NEXT: vpand %xmm6, %xmm10, %xmm6
-; AVX1-NEXT: vpaddb %xmm5, %xmm5, %xmm5
-; AVX1-NEXT: vpblendvb %xmm5, %xmm6, %xmm3, %xmm3
-; AVX1-NEXT: vpsrlw $1, %xmm3, %xmm6
-; AVX1-NEXT: vpand %xmm7, %xmm6, %xmm6
-; AVX1-NEXT: vpaddb %xmm5, %xmm5, %xmm5
-; AVX1-NEXT: vpblendvb %xmm5, %xmm6, %xmm3, %xmm3
-; AVX1-NEXT: vpsllw $5, %xmm4, %xmm4
+; AVX1-NEXT: vpshufb %xmm8, %xmm2, %xmm2
+; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm4
+; AVX1-NEXT: vpsrlw $1, %xmm4, %xmm4
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm5 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127]
+; AVX1-NEXT: vpand %xmm5, %xmm4, %xmm4
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm6 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX1-NEXT: vpandn %xmm6, %xmm2, %xmm7
+; AVX1-NEXT: vpmovzxbq {{.*#+}} xmm7 = xmm7[0],zero,zero,zero,zero,zero,zero,zero,xmm7[1],zero,zero,zero,zero,zero,zero,zero
+; AVX1-NEXT: vpsrlw %xmm7, %xmm4, %xmm4
+; AVX1-NEXT: vpcmpeqd %xmm9, %xmm9, %xmm9
+; AVX1-NEXT: vpsrlw %xmm7, %xmm9, %xmm3
+; AVX1-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1]
+; AVX1-NEXT: vpand %xmm3, %xmm4, %xmm4
; AVX1-NEXT: vpsrlw $1, %xmm1, %xmm1
-; AVX1-NEXT: vpand %xmm7, %xmm1, %xmm1
-; AVX1-NEXT: vpsrlw $4, %xmm1, %xmm5
-; AVX1-NEXT: vpand %xmm5, %xmm9, %xmm5
-; AVX1-NEXT: vpblendvb %xmm4, %xmm5, %xmm1, %xmm1
-; AVX1-NEXT: vpsrlw $2, %xmm1, %xmm5
-; AVX1-NEXT: vpand %xmm5, %xmm10, %xmm5
-; AVX1-NEXT: vpaddb %xmm4, %xmm4, %xmm4
-; AVX1-NEXT: vpblendvb %xmm4, %xmm5, %xmm1, %xmm1
-; AVX1-NEXT: vpsrlw $1, %xmm1, %xmm5
-; AVX1-NEXT: vpand %xmm7, %xmm5, %xmm5
-; AVX1-NEXT: vpaddb %xmm4, %xmm4, %xmm4
-; AVX1-NEXT: vpblendvb %xmm4, %xmm5, %xmm1, %xmm1
-; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm1, %ymm1
-; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
-; AVX1-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; AVX1-NEXT: vpand %xmm5, %xmm1, %xmm1
+; AVX1-NEXT: vpsrlw %xmm7, %xmm1, %xmm1
+; AVX1-NEXT: vpand %xmm3, %xmm1, %xmm1
+; AVX1-NEXT: vinsertf128 $1, %xmm4, %ymm1, %ymm1
+; AVX1-NEXT: vpand %xmm6, %xmm2, %xmm2
; AVX1-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX1-NEXT: vpsllw %xmm2, %xmm3, %xmm3
-; AVX1-NEXT: vpcmpeqd %xmm4, %xmm4, %xmm4
-; AVX1-NEXT: vpsllw %xmm2, %xmm4, %xmm4
+; AVX1-NEXT: vpsllw %xmm2, %xmm9, %xmm4
; AVX1-NEXT: vpshufb %xmm8, %xmm4, %xmm4
; AVX1-NEXT: vpand %xmm4, %xmm3, %xmm3
; AVX1-NEXT: vpsllw %xmm2, %xmm0, %xmm0
@@ -1365,154 +1271,140 @@ define <32 x i8> @splatvar_funnnel_v32i8(<32 x i8> %x, <32 x i8> %y, <32 x i8> %
;
; AVX2-LABEL: splatvar_funnnel_v32i8:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastb %xmm2, %ymm2
-; AVX2-NEXT: vpsrlw $1, %ymm1, %ymm1
-; AVX2-NEXT: vmovdqa {{.*#+}} ymm3 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127]
-; AVX2-NEXT: vpand %ymm3, %ymm1, %ymm1
-; AVX2-NEXT: vpsrlw $4, %ymm1, %ymm4
-; AVX2-NEXT: vpand {{.*}}(%rip), %ymm4, %ymm4
-; AVX2-NEXT: vpandn {{.*}}(%rip), %ymm2, %ymm5
-; AVX2-NEXT: vpsllw $5, %ymm5, %ymm5
-; AVX2-NEXT: vpblendvb %ymm5, %ymm4, %ymm1, %ymm1
-; AVX2-NEXT: vpsrlw $2, %ymm1, %ymm4
-; AVX2-NEXT: vpand {{.*}}(%rip), %ymm4, %ymm4
-; AVX2-NEXT: vpaddb %ymm5, %ymm5, %ymm5
-; AVX2-NEXT: vpblendvb %ymm5, %ymm4, %ymm1, %ymm1
-; AVX2-NEXT: vpsrlw $1, %ymm1, %ymm4
-; AVX2-NEXT: vpand %ymm3, %ymm4, %ymm3
-; AVX2-NEXT: vpaddb %ymm5, %ymm5, %ymm4
-; AVX2-NEXT: vpblendvb %ymm4, %ymm3, %ymm1, %ymm1
-; AVX2-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; AVX2-NEXT: vpbroadcastb %xmm2, %xmm2
+; AVX2-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX2-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX2-NEXT: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
+; AVX2-NEXT: vpsllw %xmm4, %ymm0, %ymm0
+; AVX2-NEXT: vpcmpeqd %xmm5, %xmm5, %xmm5
+; AVX2-NEXT: vpsllw %xmm4, %xmm5, %xmm4
+; AVX2-NEXT: vpbroadcastb %xmm4, %ymm4
+; AVX2-NEXT: vpand %ymm4, %ymm0, %ymm0
+; AVX2-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
-; AVX2-NEXT: vpsllw %xmm2, %ymm0, %ymm0
-; AVX2-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX2-NEXT: vpsllw %xmm2, %xmm3, %xmm2
+; AVX2-NEXT: vpsrlw $1, %ymm1, %ymm1
+; AVX2-NEXT: vpand {{.*}}(%rip), %ymm1, %ymm1
+; AVX2-NEXT: vpsrlw %xmm2, %ymm1, %ymm1
+; AVX2-NEXT: vpsrlw %xmm2, %xmm5, %xmm2
+; AVX2-NEXT: vpsrlw $8, %xmm2, %xmm2
; AVX2-NEXT: vpbroadcastb %xmm2, %ymm2
-; AVX2-NEXT: vpand %ymm2, %ymm0, %ymm0
+; AVX2-NEXT: vpand %ymm2, %ymm1, %ymm1
; AVX2-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX2-NEXT: retq
;
; AVX512F-LABEL: splatvar_funnnel_v32i8:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpbroadcastb %xmm2, %ymm2
-; AVX512F-NEXT: vpsrlw $1, %ymm1, %ymm1
-; AVX512F-NEXT: vmovdqa {{.*#+}} ymm3 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127]
-; AVX512F-NEXT: vpand %ymm3, %ymm1, %ymm1
-; AVX512F-NEXT: vpsrlw $4, %ymm1, %ymm4
-; AVX512F-NEXT: vpand {{.*}}(%rip), %ymm4, %ymm4
-; AVX512F-NEXT: vpandn {{.*}}(%rip), %ymm2, %ymm5
-; AVX512F-NEXT: vpsllw $5, %ymm5, %ymm5
-; AVX512F-NEXT: vpblendvb %ymm5, %ymm4, %ymm1, %ymm1
-; AVX512F-NEXT: vpsrlw $2, %ymm1, %ymm4
-; AVX512F-NEXT: vpand {{.*}}(%rip), %ymm4, %ymm4
-; AVX512F-NEXT: vpaddb %ymm5, %ymm5, %ymm5
-; AVX512F-NEXT: vpblendvb %ymm5, %ymm4, %ymm1, %ymm1
-; AVX512F-NEXT: vpsrlw $1, %ymm1, %ymm4
-; AVX512F-NEXT: vpand %ymm3, %ymm4, %ymm3
-; AVX512F-NEXT: vpaddb %ymm5, %ymm5, %ymm4
-; AVX512F-NEXT: vpblendvb %ymm4, %ymm3, %ymm1, %ymm1
-; AVX512F-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; AVX512F-NEXT: vpbroadcastb %xmm2, %xmm2
+; AVX512F-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512F-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512F-NEXT: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512F-NEXT: vpsllw %xmm4, %ymm0, %ymm0
+; AVX512F-NEXT: vpcmpeqd %xmm5, %xmm5, %xmm5
+; AVX512F-NEXT: vpsllw %xmm4, %xmm5, %xmm4
+; AVX512F-NEXT: vpbroadcastb %xmm4, %ymm4
+; AVX512F-NEXT: vpand %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512F-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512F-NEXT: vpsllw %xmm2, %ymm0, %ymm0
-; AVX512F-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX512F-NEXT: vpsllw %xmm2, %xmm3, %xmm2
+; AVX512F-NEXT: vpsrlw $1, %ymm1, %ymm1
+; AVX512F-NEXT: vpand {{.*}}(%rip), %ymm1, %ymm1
+; AVX512F-NEXT: vpsrlw %xmm2, %ymm1, %ymm1
+; AVX512F-NEXT: vpsrlw %xmm2, %xmm5, %xmm2
+; AVX512F-NEXT: vpsrlw $8, %xmm2, %xmm2
; AVX512F-NEXT: vpbroadcastb %xmm2, %ymm2
-; AVX512F-NEXT: vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT: vpand %ymm2, %ymm1, %ymm1
; AVX512F-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: splatvar_funnnel_v32i8:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastb %xmm2, %ymm2
+; AVX512VL-NEXT: vpbroadcastb %xmm2, %xmm2
+; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512VL-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512VL-NEXT: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512VL-NEXT: vpsllw %xmm4, %ymm0, %ymm0
+; AVX512VL-NEXT: vpcmpeqd %xmm5, %xmm5, %xmm5
+; AVX512VL-NEXT: vpsllw %xmm4, %xmm5, %xmm4
+; AVX512VL-NEXT: vpbroadcastb %xmm4, %ymm4
+; AVX512VL-NEXT: vpand %ymm4, %ymm0, %ymm4
+; AVX512VL-NEXT: vpandn %xmm3, %xmm2, %xmm0
+; AVX512VL-NEXT: vpmovzxbq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero
; AVX512VL-NEXT: vpsrlw $1, %ymm1, %ymm1
-; AVX512VL-NEXT: vmovdqa {{.*#+}} ymm3 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127]
-; AVX512VL-NEXT: vpand %ymm3, %ymm1, %ymm1
-; AVX512VL-NEXT: vpsrlw $4, %ymm1, %ymm4
-; AVX512VL-NEXT: vpand {{.*}}(%rip), %ymm4, %ymm4
-; AVX512VL-NEXT: vpandn {{.*}}(%rip), %ymm2, %ymm5
-; AVX512VL-NEXT: vpsllw $5, %ymm5, %ymm5
-; AVX512VL-NEXT: vpblendvb %ymm5, %ymm4, %ymm1, %ymm1
-; AVX512VL-NEXT: vpsrlw $2, %ymm1, %ymm4
-; AVX512VL-NEXT: vpand {{.*}}(%rip), %ymm4, %ymm4
-; AVX512VL-NEXT: vpaddb %ymm5, %ymm5, %ymm5
-; AVX512VL-NEXT: vpblendvb %ymm5, %ymm4, %ymm1, %ymm1
-; AVX512VL-NEXT: vpsrlw $1, %ymm1, %ymm4
-; AVX512VL-NEXT: vpand %ymm3, %ymm4, %ymm3
-; AVX512VL-NEXT: vpaddb %ymm5, %ymm5, %ymm4
-; AVX512VL-NEXT: vpblendvb %ymm4, %ymm3, %ymm1, %ymm1
-; AVX512VL-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX512VL-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512VL-NEXT: vpsllw %xmm2, %ymm0, %ymm3
-; AVX512VL-NEXT: vpcmpeqd %xmm0, %xmm0, %xmm0
-; AVX512VL-NEXT: vpsllw %xmm2, %xmm0, %xmm0
+; AVX512VL-NEXT: vpand {{.*}}(%rip), %ymm1, %ymm1
+; AVX512VL-NEXT: vpsrlw %xmm0, %ymm1, %ymm1
+; AVX512VL-NEXT: vpsrlw %xmm0, %xmm5, %xmm0
+; AVX512VL-NEXT: vpsrlw $8, %xmm0, %xmm0
; AVX512VL-NEXT: vpbroadcastb %xmm0, %ymm0
-; AVX512VL-NEXT: vpternlogq $236, %ymm3, %ymm1, %ymm0
+; AVX512VL-NEXT: vpternlogq $236, %ymm1, %ymm4, %ymm0
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: splatvar_funnnel_v32i8:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpbroadcastb %xmm2, %ymm3
-; AVX512BW-NEXT: vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512BW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX512BW-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512BW-NEXT: vpsllw %xmm2, %zmm0, %zmm0
+; AVX512BW-NEXT: vpbroadcastb %xmm2, %xmm2
+; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512BW-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX512BW-NEXT: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
; AVX512BW-NEXT: vpsrlw $1, %ymm1, %ymm1
; AVX512BW-NEXT: vpand {{.*}}(%rip), %ymm1, %ymm1
; AVX512BW-NEXT: vpmovzxbw {{.*#+}} zmm1 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero,ymm1[16],zero,ymm1[17],zero,ymm1[18],zero,ymm1[19],zero,ymm1[20],zero,ymm1[21],zero,ymm1[22],zero,ymm1[23],zero,ymm1[24],zero,ymm1[25],zero,ymm1[26],zero,ymm1[27],zero,ymm1[28],zero,ymm1[29],zero,ymm1[30],zero,ymm1[31],zero
-; AVX512BW-NEXT: vpandn {{.*}}(%rip), %ymm3, %ymm2
-; AVX512BW-NEXT: vpmovzxbw {{.*#+}} zmm2 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero,ymm2[16],zero,ymm2[17],zero,ymm2[18],zero,ymm2[19],zero,ymm2[20],zero,ymm2[21],zero,ymm2[22],zero,ymm2[23],zero,ymm2[24],zero,ymm2[25],zero,ymm2[26],zero,ymm2[27],zero,ymm2[28],zero,ymm2[29],zero,ymm2[30],zero,ymm2[31],zero
-; AVX512BW-NEXT: vpsrlvw %zmm2, %zmm1, %zmm1
+; AVX512BW-NEXT: vpsrlw %xmm4, %zmm1, %zmm1
+; AVX512BW-NEXT: vpand %xmm3, %xmm2, %xmm2
+; AVX512BW-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512BW-NEXT: vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
+; AVX512BW-NEXT: vpsllw %xmm2, %zmm0, %zmm0
; AVX512BW-NEXT: vporq %zmm1, %zmm0, %zmm0
; AVX512BW-NEXT: vpmovwb %zmm0, %ymm0
; AVX512BW-NEXT: retq
;
; AVX512VBMI2-LABEL: splatvar_funnnel_v32i8:
; AVX512VBMI2: # %bb.0:
-; AVX512VBMI2-NEXT: vpbroadcastb %xmm2, %ymm3
-; AVX512VBMI2-NEXT: vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512VBMI2-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX512VBMI2-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512VBMI2-NEXT: vpsllw %xmm2, %zmm0, %zmm0
+; AVX512VBMI2-NEXT: vpbroadcastb %xmm2, %xmm2
+; AVX512VBMI2-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512VBMI2-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX512VBMI2-NEXT: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
; AVX512VBMI2-NEXT: vpsrlw $1, %ymm1, %ymm1
; AVX512VBMI2-NEXT: vpand {{.*}}(%rip), %ymm1, %ymm1
; AVX512VBMI2-NEXT: vpmovzxbw {{.*#+}} zmm1 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero,ymm1[16],zero,ymm1[17],zero,ymm1[18],zero,ymm1[19],zero,ymm1[20],zero,ymm1[21],zero,ymm1[22],zero,ymm1[23],zero,ymm1[24],zero,ymm1[25],zero,ymm1[26],zero,ymm1[27],zero,ymm1[28],zero,ymm1[29],zero,ymm1[30],zero,ymm1[31],zero
-; AVX512VBMI2-NEXT: vpandn {{.*}}(%rip), %ymm3, %ymm2
-; AVX512VBMI2-NEXT: vpmovzxbw {{.*#+}} zmm2 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero,ymm2[16],zero,ymm2[17],zero,ymm2[18],zero,ymm2[19],zero,ymm2[20],zero,ymm2[21],zero,ymm2[22],zero,ymm2[23],zero,ymm2[24],zero,ymm2[25],zero,ymm2[26],zero,ymm2[27],zero,ymm2[28],zero,ymm2[29],zero,ymm2[30],zero,ymm2[31],zero
-; AVX512VBMI2-NEXT: vpsrlvw %zmm2, %zmm1, %zmm1
+; AVX512VBMI2-NEXT: vpsrlw %xmm4, %zmm1, %zmm1
+; AVX512VBMI2-NEXT: vpand %xmm3, %xmm2, %xmm2
+; AVX512VBMI2-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512VBMI2-NEXT: vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
+; AVX512VBMI2-NEXT: vpsllw %xmm2, %zmm0, %zmm0
; AVX512VBMI2-NEXT: vporq %zmm1, %zmm0, %zmm0
; AVX512VBMI2-NEXT: vpmovwb %zmm0, %ymm0
; AVX512VBMI2-NEXT: retq
;
; AVX512VLBW-LABEL: splatvar_funnnel_v32i8:
; AVX512VLBW: # %bb.0:
-; AVX512VLBW-NEXT: vpbroadcastb %xmm2, %ymm3
-; AVX512VLBW-NEXT: vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512VLBW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX512VLBW-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512VLBW-NEXT: vpsllw %xmm2, %zmm0, %zmm0
+; 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: vpandn %xmm3, %xmm2, %xmm4
+; AVX512VLBW-NEXT: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
; AVX512VLBW-NEXT: vpsrlw $1, %ymm1, %ymm1
; AVX512VLBW-NEXT: vpand {{.*}}(%rip), %ymm1, %ymm1
; AVX512VLBW-NEXT: vpmovzxbw {{.*#+}} zmm1 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero,ymm1[16],zero,ymm1[17],zero,ymm1[18],zero,ymm1[19],zero,ymm1[20],zero,ymm1[21],zero,ymm1[22],zero,ymm1[23],zero,ymm1[24],zero,ymm1[25],zero,ymm1[26],zero,ymm1[27],zero,ymm1[28],zero,ymm1[29],zero,ymm1[30],zero,ymm1[31],zero
-; AVX512VLBW-NEXT: vpandn {{.*}}(%rip), %ymm3, %ymm2
-; AVX512VLBW-NEXT: vpmovzxbw {{.*#+}} zmm2 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero,ymm2[16],zero,ymm2[17],zero,ymm2[18],zero,ymm2[19],zero,ymm2[20],zero,ymm2[21],zero,ymm2[22],zero,ymm2[23],zero,ymm2[24],zero,ymm2[25],zero,ymm2[26],zero,ymm2[27],zero,ymm2[28],zero,ymm2[29],zero,ymm2[30],zero,ymm2[31],zero
-; AVX512VLBW-NEXT: vpsrlvw %zmm2, %zmm1, %zmm1
+; AVX512VLBW-NEXT: vpsrlw %xmm4, %zmm1, %zmm1
+; AVX512VLBW-NEXT: vpand %xmm3, %xmm2, %xmm2
+; AVX512VLBW-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512VLBW-NEXT: vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
+; AVX512VLBW-NEXT: vpsllw %xmm2, %zmm0, %zmm0
; AVX512VLBW-NEXT: vporq %zmm1, %zmm0, %zmm0
; AVX512VLBW-NEXT: vpmovwb %zmm0, %ymm0
; AVX512VLBW-NEXT: retq
;
; AVX512VLVBMI2-LABEL: splatvar_funnnel_v32i8:
; AVX512VLVBMI2: # %bb.0:
-; AVX512VLVBMI2-NEXT: vpbroadcastb %xmm2, %ymm3
-; AVX512VLVBMI2-NEXT: vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512VLVBMI2-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX512VLVBMI2-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512VLVBMI2-NEXT: vpsllw %xmm2, %zmm0, %zmm0
+; 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: vpandn %xmm3, %xmm2, %xmm4
+; AVX512VLVBMI2-NEXT: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
; AVX512VLVBMI2-NEXT: vpsrlw $1, %ymm1, %ymm1
; AVX512VLVBMI2-NEXT: vpand {{.*}}(%rip), %ymm1, %ymm1
; AVX512VLVBMI2-NEXT: vpmovzxbw {{.*#+}} zmm1 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero,ymm1[16],zero,ymm1[17],zero,ymm1[18],zero,ymm1[19],zero,ymm1[20],zero,ymm1[21],zero,ymm1[22],zero,ymm1[23],zero,ymm1[24],zero,ymm1[25],zero,ymm1[26],zero,ymm1[27],zero,ymm1[28],zero,ymm1[29],zero,ymm1[30],zero,ymm1[31],zero
-; AVX512VLVBMI2-NEXT: vpandn {{.*}}(%rip), %ymm3, %ymm2
-; AVX512VLVBMI2-NEXT: vpmovzxbw {{.*#+}} zmm2 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero,ymm2[16],zero,ymm2[17],zero,ymm2[18],zero,ymm2[19],zero,ymm2[20],zero,ymm2[21],zero,ymm2[22],zero,ymm2[23],zero,ymm2[24],zero,ymm2[25],zero,ymm2[26],zero,ymm2[27],zero,ymm2[28],zero,ymm2[29],zero,ymm2[30],zero,ymm2[31],zero
-; AVX512VLVBMI2-NEXT: vpsrlvw %zmm2, %zmm1, %zmm1
+; AVX512VLVBMI2-NEXT: vpsrlw %xmm4, %zmm1, %zmm1
+; AVX512VLVBMI2-NEXT: vpand %xmm3, %xmm2, %xmm2
+; AVX512VLVBMI2-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512VLVBMI2-NEXT: vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
+; AVX512VLVBMI2-NEXT: vpsllw %xmm2, %zmm0, %zmm0
; AVX512VLVBMI2-NEXT: vporq %zmm1, %zmm0, %zmm0
; AVX512VLVBMI2-NEXT: vpmovwb %zmm0, %ymm0
; AVX512VLVBMI2-NEXT: retq
@@ -1521,19 +1413,17 @@ define <32 x i8> @splatvar_funnnel_v32i8(<32 x i8> %x, <32 x i8> %y, <32 x i8> %
; XOPAVX1: # %bb.0:
; XOPAVX1-NEXT: vpxor %xmm3, %xmm3, %xmm3
; XOPAVX1-NEXT: vpshufb %xmm3, %xmm2, %xmm2
-; XOPAVX1-NEXT: vinsertf128 $1, %xmm2, %ymm2, %ymm4
-; XOPAVX1-NEXT: vandnps {{.*}}(%rip), %ymm4, %ymm4
-; XOPAVX1-NEXT: vextractf128 $1, %ymm4, %xmm5
-; XOPAVX1-NEXT: vpsubb %xmm5, %xmm3, %xmm5
-; XOPAVX1-NEXT: vextractf128 $1, %ymm1, %xmm6
-; XOPAVX1-NEXT: vpcmpeqd %xmm7, %xmm7, %xmm7
-; XOPAVX1-NEXT: vpshlb %xmm7, %xmm6, %xmm6
-; XOPAVX1-NEXT: vpshlb %xmm5, %xmm6, %xmm5
-; XOPAVX1-NEXT: vpsubb %xmm4, %xmm3, %xmm3
-; XOPAVX1-NEXT: vpshlb %xmm7, %xmm1, %xmm1
+; XOPAVX1-NEXT: vextractf128 $1, %ymm1, %xmm4
+; XOPAVX1-NEXT: vpcmpeqd %xmm5, %xmm5, %xmm5
+; XOPAVX1-NEXT: vpshlb %xmm5, %xmm4, %xmm4
+; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm6 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; XOPAVX1-NEXT: vpandn %xmm6, %xmm2, %xmm7
+; XOPAVX1-NEXT: vpsubb %xmm7, %xmm3, %xmm3
+; XOPAVX1-NEXT: vpshlb %xmm3, %xmm4, %xmm4
+; XOPAVX1-NEXT: vpshlb %xmm5, %xmm1, %xmm1
; XOPAVX1-NEXT: vpshlb %xmm3, %xmm1, %xmm1
-; XOPAVX1-NEXT: vinsertf128 $1, %xmm5, %ymm1, %ymm1
-; XOPAVX1-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; XOPAVX1-NEXT: vinsertf128 $1, %xmm4, %ymm1, %ymm1
+; XOPAVX1-NEXT: vpand %xmm6, %xmm2, %xmm2
; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
; XOPAVX1-NEXT: vpshlb %xmm2, %xmm3, %xmm3
; XOPAVX1-NEXT: vpshlb %xmm2, %xmm0, %xmm0
@@ -1543,21 +1433,20 @@ define <32 x i8> @splatvar_funnnel_v32i8(<32 x i8> %x, <32 x i8> %y, <32 x i8> %
;
; XOPAVX2-LABEL: splatvar_funnnel_v32i8:
; XOPAVX2: # %bb.0:
-; XOPAVX2-NEXT: vpbroadcastb %xmm2, %ymm2
-; XOPAVX2-NEXT: vextracti128 $1, %ymm0, %xmm3
-; XOPAVX2-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm4
-; XOPAVX2-NEXT: vpshlb %xmm4, %xmm3, %xmm3
+; XOPAVX2-NEXT: vpbroadcastb %xmm2, %xmm2
+; XOPAVX2-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; XOPAVX2-NEXT: vpand %xmm3, %xmm2, %xmm4
+; XOPAVX2-NEXT: vextracti128 $1, %ymm0, %xmm5
+; XOPAVX2-NEXT: vpshlb %xmm4, %xmm5, %xmm5
; XOPAVX2-NEXT: vpshlb %xmm4, %xmm0, %xmm0
-; XOPAVX2-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm0
-; XOPAVX2-NEXT: vpandn {{.*}}(%rip), %ymm2, %ymm2
-; XOPAVX2-NEXT: vextracti128 $1, %ymm2, %xmm3
-; XOPAVX2-NEXT: vpxor %xmm4, %xmm4, %xmm4
-; XOPAVX2-NEXT: vpsubb %xmm3, %xmm4, %xmm3
+; XOPAVX2-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm0
+; XOPAVX2-NEXT: vpandn %xmm3, %xmm2, %xmm2
+; XOPAVX2-NEXT: vpxor %xmm3, %xmm3, %xmm3
+; XOPAVX2-NEXT: vpsubb %xmm2, %xmm3, %xmm2
; XOPAVX2-NEXT: vpsrlw $1, %ymm1, %ymm1
; XOPAVX2-NEXT: vpand {{.*}}(%rip), %ymm1, %ymm1
-; XOPAVX2-NEXT: vextracti128 $1, %ymm1, %xmm5
-; XOPAVX2-NEXT: vpshlb %xmm3, %xmm5, %xmm3
-; XOPAVX2-NEXT: vpsubb %xmm2, %xmm4, %xmm2
+; XOPAVX2-NEXT: vextracti128 $1, %ymm1, %xmm3
+; XOPAVX2-NEXT: vpshlb %xmm2, %xmm3, %xmm3
; XOPAVX2-NEXT: vpshlb %xmm2, %xmm1, %xmm1
; XOPAVX2-NEXT: vinserti128 $1, %xmm3, %ymm1, %ymm1
; XOPAVX2-NEXT: vpor %ymm1, %ymm0, %ymm0
diff --git a/llvm/test/CodeGen/X86/vector-fshl-512.ll b/llvm/test/CodeGen/X86/vector-fshl-512.ll
index 4e3fcca69514..9acde0137d89 100644
--- a/llvm/test/CodeGen/X86/vector-fshl-512.ll
+++ b/llvm/test/CodeGen/X86/vector-fshl-512.ll
@@ -536,34 +536,34 @@ define <64 x i8> @var_funnnel_v64i8(<64 x i8> %x, <64 x i8> %y, <64 x i8> %amt)
define <8 x i64> @splatvar_funnnel_v8i64(<8 x i64> %x, <8 x i64> %y, <8 x i64> %amt) nounwind {
; AVX512F-LABEL: splatvar_funnnel_v8i64:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpbroadcastq %xmm2, %zmm3
-; AVX512F-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX512F-NEXT: vpsllq %xmm2, %zmm0, %zmm0
+; AVX512F-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; AVX512F-NEXT: vpandn %xmm3, %xmm2, %xmm4
; AVX512F-NEXT: vpsrlq $1, %zmm1, %zmm1
-; AVX512F-NEXT: vpandnq {{.*}}(%rip){1to8}, %zmm3, %zmm2
-; AVX512F-NEXT: vpsrlvq %zmm2, %zmm1, %zmm1
+; AVX512F-NEXT: vpsrlq %xmm4, %zmm1, %zmm1
+; AVX512F-NEXT: vpand %xmm3, %xmm2, %xmm2
+; AVX512F-NEXT: vpsllq %xmm2, %zmm0, %zmm0
; AVX512F-NEXT: vporq %zmm1, %zmm0, %zmm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: splatvar_funnnel_v8i64:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastq %xmm2, %zmm3
-; AVX512VL-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX512VL-NEXT: vpsllq %xmm2, %zmm0, %zmm0
+; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; AVX512VL-NEXT: vpandn %xmm3, %xmm2, %xmm4
; AVX512VL-NEXT: vpsrlq $1, %zmm1, %zmm1
-; AVX512VL-NEXT: vpandnq {{.*}}(%rip){1to8}, %zmm3, %zmm2
-; AVX512VL-NEXT: vpsrlvq %zmm2, %zmm1, %zmm1
+; AVX512VL-NEXT: vpsrlq %xmm4, %zmm1, %zmm1
+; AVX512VL-NEXT: vpand %xmm3, %xmm2, %xmm2
+; AVX512VL-NEXT: vpsllq %xmm2, %zmm0, %zmm0
; AVX512VL-NEXT: vporq %zmm1, %zmm0, %zmm0
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: splatvar_funnnel_v8i64:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpbroadcastq %xmm2, %zmm3
-; AVX512BW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX512BW-NEXT: vpsllq %xmm2, %zmm0, %zmm0
+; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; AVX512BW-NEXT: vpandn %xmm3, %xmm2, %xmm4
; AVX512BW-NEXT: vpsrlq $1, %zmm1, %zmm1
-; AVX512BW-NEXT: vpandnq {{.*}}(%rip){1to8}, %zmm3, %zmm2
-; AVX512BW-NEXT: vpsrlvq %zmm2, %zmm1, %zmm1
+; AVX512BW-NEXT: vpsrlq %xmm4, %zmm1, %zmm1
+; AVX512BW-NEXT: vpand %xmm3, %xmm2, %xmm2
+; AVX512BW-NEXT: vpsllq %xmm2, %zmm0, %zmm0
; AVX512BW-NEXT: vporq %zmm1, %zmm0, %zmm0
; AVX512BW-NEXT: retq
;
@@ -575,12 +575,12 @@ define <8 x i64> @splatvar_funnnel_v8i64(<8 x i64> %x, <8 x i64> %y, <8 x i64> %
;
; AVX512VLBW-LABEL: splatvar_funnnel_v8i64:
; AVX512VLBW: # %bb.0:
-; AVX512VLBW-NEXT: vpbroadcastq %xmm2, %zmm3
-; AVX512VLBW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX512VLBW-NEXT: vpsllq %xmm2, %zmm0, %zmm0
+; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; AVX512VLBW-NEXT: vpandn %xmm3, %xmm2, %xmm4
; AVX512VLBW-NEXT: vpsrlq $1, %zmm1, %zmm1
-; AVX512VLBW-NEXT: vpandnq {{.*}}(%rip){1to8}, %zmm3, %zmm2
-; AVX512VLBW-NEXT: vpsrlvq %zmm2, %zmm1, %zmm1
+; AVX512VLBW-NEXT: vpsrlq %xmm4, %zmm1, %zmm1
+; AVX512VLBW-NEXT: vpand %xmm3, %xmm2, %xmm2
+; AVX512VLBW-NEXT: vpsllq %xmm2, %zmm0, %zmm0
; AVX512VLBW-NEXT: vporq %zmm1, %zmm0, %zmm0
; AVX512VLBW-NEXT: retq
;
@@ -597,11 +597,12 @@ define <8 x i64> @splatvar_funnnel_v8i64(<8 x i64> %x, <8 x i64> %y, <8 x i64> %
define <16 x i32> @splatvar_funnnel_v16i32(<16 x i32> %x, <16 x i32> %y, <16 x i32> %amt) nounwind {
; AVX512F-LABEL: splatvar_funnnel_v16i32:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpbroadcastd %xmm2, %zmm2
-; AVX512F-NEXT: vpbroadcastd {{.*#+}} zmm3 = [31,31,31,31,31,31,31,31,31,31,31,31,31,31,31,31]
-; AVX512F-NEXT: vpandnd %zmm3, %zmm2, %zmm4
+; AVX512F-NEXT: vpbroadcastd %xmm2, %xmm2
+; AVX512F-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31]
+; AVX512F-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX512F-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
; AVX512F-NEXT: vpsrld $1, %zmm1, %zmm1
-; AVX512F-NEXT: vpsrlvd %zmm4, %zmm1, %zmm1
+; AVX512F-NEXT: vpsrld %xmm4, %zmm1, %zmm1
; AVX512F-NEXT: vpand %xmm3, %xmm2, %xmm2
; AVX512F-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
; AVX512F-NEXT: vpslld %xmm2, %zmm0, %zmm0
@@ -610,11 +611,12 @@ define <16 x i32> @splatvar_funnnel_v16i32(<16 x i32> %x, <16 x i32> %y, <16 x i
;
; AVX512VL-LABEL: splatvar_funnnel_v16i32:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastd %xmm2, %zmm2
-; AVX512VL-NEXT: vpbroadcastd {{.*#+}} zmm3 = [31,31,31,31,31,31,31,31,31,31,31,31,31,31,31,31]
-; AVX512VL-NEXT: vpandnd %zmm3, %zmm2, %zmm4
+; AVX512VL-NEXT: vpbroadcastd %xmm2, %xmm2
+; AVX512VL-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31]
+; AVX512VL-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX512VL-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
; AVX512VL-NEXT: vpsrld $1, %zmm1, %zmm1
-; AVX512VL-NEXT: vpsrlvd %zmm4, %zmm1, %zmm1
+; AVX512VL-NEXT: vpsrld %xmm4, %zmm1, %zmm1
; AVX512VL-NEXT: vpand %xmm3, %xmm2, %xmm2
; AVX512VL-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
; AVX512VL-NEXT: vpslld %xmm2, %zmm0, %zmm0
@@ -623,11 +625,12 @@ define <16 x i32> @splatvar_funnnel_v16i32(<16 x i32> %x, <16 x i32> %y, <16 x i
;
; AVX512BW-LABEL: splatvar_funnnel_v16i32:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpbroadcastd %xmm2, %zmm2
-; AVX512BW-NEXT: vpbroadcastd {{.*#+}} zmm3 = [31,31,31,31,31,31,31,31,31,31,31,31,31,31,31,31]
-; AVX512BW-NEXT: vpandnd %zmm3, %zmm2, %zmm4
+; AVX512BW-NEXT: vpbroadcastd %xmm2, %xmm2
+; AVX512BW-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31]
+; AVX512BW-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX512BW-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
; AVX512BW-NEXT: vpsrld $1, %zmm1, %zmm1
-; AVX512BW-NEXT: vpsrlvd %zmm4, %zmm1, %zmm1
+; AVX512BW-NEXT: vpsrld %xmm4, %zmm1, %zmm1
; AVX512BW-NEXT: vpand %xmm3, %xmm2, %xmm2
; AVX512BW-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
; AVX512BW-NEXT: vpslld %xmm2, %zmm0, %zmm0
@@ -642,11 +645,12 @@ define <16 x i32> @splatvar_funnnel_v16i32(<16 x i32> %x, <16 x i32> %y, <16 x i
;
; AVX512VLBW-LABEL: splatvar_funnnel_v16i32:
; AVX512VLBW: # %bb.0:
-; AVX512VLBW-NEXT: vpbroadcastd %xmm2, %zmm2
-; AVX512VLBW-NEXT: vpbroadcastd {{.*#+}} zmm3 = [31,31,31,31,31,31,31,31,31,31,31,31,31,31,31,31]
-; AVX512VLBW-NEXT: vpandnd %zmm3, %zmm2, %zmm4
+; AVX512VLBW-NEXT: vpbroadcastd %xmm2, %xmm2
+; AVX512VLBW-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31]
+; AVX512VLBW-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX512VLBW-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
; AVX512VLBW-NEXT: vpsrld $1, %zmm1, %zmm1
-; AVX512VLBW-NEXT: vpsrlvd %zmm4, %zmm1, %zmm1
+; AVX512VLBW-NEXT: vpsrld %xmm4, %zmm1, %zmm1
; AVX512VLBW-NEXT: vpand %xmm3, %xmm2, %xmm2
; AVX512VLBW-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
; AVX512VLBW-NEXT: vpslld %xmm2, %zmm0, %zmm0
@@ -666,65 +670,55 @@ define <16 x i32> @splatvar_funnnel_v16i32(<16 x i32> %x, <16 x i32> %y, <16 x i
define <32 x i16> @splatvar_funnnel_v32i16(<32 x i16> %x, <32 x i16> %y, <32 x i16> %amt) nounwind {
; AVX512F-LABEL: splatvar_funnnel_v32i16:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpbroadcastw %xmm2, %ymm2
-; AVX512F-NEXT: vinserti64x4 $1, %ymm2, %zmm2, %zmm2
-; AVX512F-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm3
-; AVX512F-NEXT: vpmovzxwq {{.*#+}} xmm3 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero
-; AVX512F-NEXT: vextracti64x4 $1, %zmm0, %ymm4
-; AVX512F-NEXT: vpsllw %xmm3, %ymm4, %ymm4
-; AVX512F-NEXT: vpsllw %xmm3, %ymm0, %ymm0
-; AVX512F-NEXT: vinserti64x4 $1, %ymm4, %zmm0, %zmm0
-; AVX512F-NEXT: vpandnq {{.*}}(%rip), %zmm2, %zmm2
-; AVX512F-NEXT: vpmovzxwd {{.*#+}} zmm3 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero
-; AVX512F-NEXT: vpsrlw $1, %ymm1, %ymm4
-; AVX512F-NEXT: vpmovzxwd {{.*#+}} zmm4 = ymm4[0],zero,ymm4[1],zero,ymm4[2],zero,ymm4[3],zero,ymm4[4],zero,ymm4[5],zero,ymm4[6],zero,ymm4[7],zero,ymm4[8],zero,ymm4[9],zero,ymm4[10],zero,ymm4[11],zero,ymm4[12],zero,ymm4[13],zero,ymm4[14],zero,ymm4[15],zero
-; AVX512F-NEXT: vpsrlvd %zmm3, %zmm4, %zmm3
-; AVX512F-NEXT: vpmovdw %zmm3, %ymm3
-; AVX512F-NEXT: vextracti64x4 $1, %zmm2, %ymm2
-; AVX512F-NEXT: vpmovzxwd {{.*#+}} zmm2 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero
-; AVX512F-NEXT: vextracti64x4 $1, %zmm1, %ymm1
+; AVX512F-NEXT: vpbroadcastw %xmm2, %xmm2
+; AVX512F-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; AVX512F-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX512F-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
+; AVX512F-NEXT: vextracti64x4 $1, %zmm1, %ymm5
+; AVX512F-NEXT: vpsrlw $1, %ymm5, %ymm5
+; AVX512F-NEXT: vpsrlw %xmm4, %ymm5, %ymm5
; AVX512F-NEXT: vpsrlw $1, %ymm1, %ymm1
-; AVX512F-NEXT: vpmovzxwd {{.*#+}} zmm1 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero
-; AVX512F-NEXT: vpsrlvd %zmm2, %zmm1, %zmm1
-; AVX512F-NEXT: vpmovdw %zmm1, %ymm1
-; AVX512F-NEXT: vinserti64x4 $1, %ymm1, %zmm3, %zmm1
+; AVX512F-NEXT: vpsrlw %xmm4, %ymm1, %ymm1
+; AVX512F-NEXT: vinserti64x4 $1, %ymm5, %zmm1, %zmm1
+; AVX512F-NEXT: vpand %xmm3, %xmm2, %xmm2
+; AVX512F-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
+; AVX512F-NEXT: vextracti64x4 $1, %zmm0, %ymm3
+; AVX512F-NEXT: vpsllw %xmm2, %ymm3, %ymm3
+; AVX512F-NEXT: vpsllw %xmm2, %ymm0, %ymm0
+; AVX512F-NEXT: vinserti64x4 $1, %ymm3, %zmm0, %zmm0
; AVX512F-NEXT: vporq %zmm1, %zmm0, %zmm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: splatvar_funnnel_v32i16:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastw %xmm2, %ymm2
-; AVX512VL-NEXT: vinserti64x4 $1, %ymm2, %zmm2, %zmm2
-; AVX512VL-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm3
-; AVX512VL-NEXT: vpmovzxwq {{.*#+}} xmm3 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero
-; AVX512VL-NEXT: vextracti64x4 $1, %zmm0, %ymm4
-; AVX512VL-NEXT: vpsllw %xmm3, %ymm4, %ymm4
-; AVX512VL-NEXT: vpsllw %xmm3, %ymm0, %ymm0
-; AVX512VL-NEXT: vinserti64x4 $1, %ymm4, %zmm0, %zmm0
-; AVX512VL-NEXT: vpandnq {{.*}}(%rip), %zmm2, %zmm2
-; AVX512VL-NEXT: vpmovzxwd {{.*#+}} zmm3 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero
-; AVX512VL-NEXT: vpsrlw $1, %ymm1, %ymm4
-; AVX512VL-NEXT: vpmovzxwd {{.*#+}} zmm4 = ymm4[0],zero,ymm4[1],zero,ymm4[2],zero,ymm4[3],zero,ymm4[4],zero,ymm4[5],zero,ymm4[6],zero,ymm4[7],zero,ymm4[8],zero,ymm4[9],zero,ymm4[10],zero,ymm4[11],zero,ymm4[12],zero,ymm4[13],zero,ymm4[14],zero,ymm4[15],zero
-; AVX512VL-NEXT: vpsrlvd %zmm3, %zmm4, %zmm3
-; AVX512VL-NEXT: vpmovdw %zmm3, %ymm3
-; AVX512VL-NEXT: vextracti64x4 $1, %zmm2, %ymm2
-; AVX512VL-NEXT: vpmovzxwd {{.*#+}} zmm2 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero
-; AVX512VL-NEXT: vextracti64x4 $1, %zmm1, %ymm1
+; AVX512VL-NEXT: vpbroadcastw %xmm2, %xmm2
+; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; AVX512VL-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX512VL-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
+; AVX512VL-NEXT: vextracti64x4 $1, %zmm1, %ymm5
+; AVX512VL-NEXT: vpsrlw $1, %ymm5, %ymm5
+; AVX512VL-NEXT: vpsrlw %xmm4, %ymm5, %ymm5
; AVX512VL-NEXT: vpsrlw $1, %ymm1, %ymm1
-; AVX512VL-NEXT: vpmovzxwd {{.*#+}} zmm1 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero
-; AVX512VL-NEXT: vpsrlvd %zmm2, %zmm1, %zmm1
-; AVX512VL-NEXT: vpmovdw %zmm1, %ymm1
-; AVX512VL-NEXT: vinserti64x4 $1, %ymm1, %zmm3, %zmm1
+; AVX512VL-NEXT: vpsrlw %xmm4, %ymm1, %ymm1
+; AVX512VL-NEXT: vinserti64x4 $1, %ymm5, %zmm1, %zmm1
+; AVX512VL-NEXT: vpand %xmm3, %xmm2, %xmm2
+; AVX512VL-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
+; AVX512VL-NEXT: vextracti64x4 $1, %zmm0, %ymm3
+; AVX512VL-NEXT: vpsllw %xmm2, %ymm3, %ymm3
+; AVX512VL-NEXT: vpsllw %xmm2, %ymm0, %ymm0
+; AVX512VL-NEXT: vinserti64x4 $1, %ymm3, %zmm0, %zmm0
; AVX512VL-NEXT: vporq %zmm1, %zmm0, %zmm0
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: splatvar_funnnel_v32i16:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpbroadcastw %xmm2, %zmm2
+; AVX512BW-NEXT: vpbroadcastw %xmm2, %xmm2
+; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX512BW-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
; AVX512BW-NEXT: vpsrlw $1, %zmm1, %zmm1
-; AVX512BW-NEXT: vpandnq {{.*}}(%rip), %zmm2, %zmm3
-; AVX512BW-NEXT: vpsrlvw %zmm3, %zmm1, %zmm1
-; AVX512BW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; AVX512BW-NEXT: vpsrlw %xmm4, %zmm1, %zmm1
+; AVX512BW-NEXT: vpand %xmm3, %xmm2, %xmm2
; AVX512BW-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
; AVX512BW-NEXT: vpsllw %xmm2, %zmm0, %zmm0
; AVX512BW-NEXT: vporq %zmm1, %zmm0, %zmm0
@@ -738,11 +732,13 @@ define <32 x i16> @splatvar_funnnel_v32i16(<32 x i16> %x, <32 x i16> %y, <32 x i
;
; AVX512VLBW-LABEL: splatvar_funnnel_v32i16:
; AVX512VLBW: # %bb.0:
-; AVX512VLBW-NEXT: vpbroadcastw %xmm2, %zmm2
+; AVX512VLBW-NEXT: vpbroadcastw %xmm2, %xmm2
+; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; AVX512VLBW-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX512VLBW-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
; AVX512VLBW-NEXT: vpsrlw $1, %zmm1, %zmm1
-; AVX512VLBW-NEXT: vpandnq {{.*}}(%rip), %zmm2, %zmm3
-; AVX512VLBW-NEXT: vpsrlvw %zmm3, %zmm1, %zmm1
-; AVX512VLBW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; AVX512VLBW-NEXT: vpsrlw %xmm4, %zmm1, %zmm1
+; AVX512VLBW-NEXT: vpand %xmm3, %xmm2, %xmm2
; AVX512VLBW-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
; AVX512VLBW-NEXT: vpsllw %xmm2, %zmm0, %zmm0
; AVX512VLBW-NEXT: vporq %zmm1, %zmm0, %zmm0
@@ -761,230 +757,156 @@ define <32 x i16> @splatvar_funnnel_v32i16(<32 x i16> %x, <32 x i16> %y, <32 x i
define <64 x i8> @splatvar_funnnel_v64i8(<64 x i8> %x, <64 x i8> %y, <64 x i8> %amt) nounwind {
; AVX512F-LABEL: splatvar_funnnel_v64i8:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpbroadcastb %xmm2, %ymm2
-; AVX512F-NEXT: vinserti64x4 $1, %ymm2, %zmm2, %zmm2
+; AVX512F-NEXT: vpbroadcastb %xmm2, %xmm2
; AVX512F-NEXT: vextracti64x4 $1, %zmm1, %ymm3
; AVX512F-NEXT: vpsrlw $1, %ymm3, %ymm3
; AVX512F-NEXT: vmovdqa {{.*#+}} ymm4 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127]
; AVX512F-NEXT: vpand %ymm4, %ymm3, %ymm3
-; AVX512F-NEXT: vpsrlw $4, %ymm3, %ymm5
-; AVX512F-NEXT: vmovdqa {{.*#+}} ymm6 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
-; AVX512F-NEXT: vpand %ymm6, %ymm5, %ymm5
-; AVX512F-NEXT: vpandnq {{.*}}(%rip), %zmm2, %zmm7
-; AVX512F-NEXT: vextracti64x4 $1, %zmm7, %ymm8
-; AVX512F-NEXT: vpsllw $5, %ymm8, %ymm8
-; AVX512F-NEXT: vpblendvb %ymm8, %ymm5, %ymm3, %ymm3
-; AVX512F-NEXT: vpsrlw $2, %ymm3, %ymm5
-; AVX512F-NEXT: vmovdqa {{.*#+}} ymm9 = [63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63]
-; AVX512F-NEXT: vpand %ymm5, %ymm9, %ymm5
-; AVX512F-NEXT: vpaddb %ymm8, %ymm8, %ymm8
-; AVX512F-NEXT: vpblendvb %ymm8, %ymm5, %ymm3, %ymm3
-; AVX512F-NEXT: vpsrlw $1, %ymm3, %ymm5
-; AVX512F-NEXT: vpand %ymm4, %ymm5, %ymm5
-; AVX512F-NEXT: vpaddb %ymm8, %ymm8, %ymm8
-; AVX512F-NEXT: vpblendvb %ymm8, %ymm5, %ymm3, %ymm3
+; AVX512F-NEXT: vmovdqa {{.*#+}} xmm5 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512F-NEXT: vpandn %xmm5, %xmm2, %xmm6
+; AVX512F-NEXT: vpmovzxbq {{.*#+}} xmm6 = xmm6[0],zero,zero,zero,zero,zero,zero,zero,xmm6[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512F-NEXT: vpsrlw %xmm6, %ymm3, %ymm3
; AVX512F-NEXT: vpsrlw $1, %ymm1, %ymm1
; AVX512F-NEXT: vpand %ymm4, %ymm1, %ymm1
-; AVX512F-NEXT: vpsrlw $4, %ymm1, %ymm5
-; AVX512F-NEXT: vpand %ymm6, %ymm5, %ymm5
-; AVX512F-NEXT: vpsllw $5, %ymm7, %ymm6
-; AVX512F-NEXT: vpblendvb %ymm6, %ymm5, %ymm1, %ymm1
-; AVX512F-NEXT: vpsrlw $2, %ymm1, %ymm5
-; AVX512F-NEXT: vpand %ymm5, %ymm9, %ymm5
-; AVX512F-NEXT: vpaddb %ymm6, %ymm6, %ymm6
-; AVX512F-NEXT: vpblendvb %ymm6, %ymm5, %ymm1, %ymm1
-; AVX512F-NEXT: vpsrlw $1, %ymm1, %ymm5
-; AVX512F-NEXT: vpand %ymm4, %ymm5, %ymm4
-; AVX512F-NEXT: vpaddb %ymm6, %ymm6, %ymm5
-; AVX512F-NEXT: vpblendvb %ymm5, %ymm4, %ymm1, %ymm1
+; AVX512F-NEXT: vpsrlw %xmm6, %ymm1, %ymm1
; AVX512F-NEXT: vinserti64x4 $1, %ymm3, %zmm1, %zmm1
-; AVX512F-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; AVX512F-NEXT: vpand %xmm5, %xmm2, %xmm2
; AVX512F-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
; AVX512F-NEXT: vextracti64x4 $1, %zmm0, %ymm3
; AVX512F-NEXT: vpsllw %xmm2, %ymm3, %ymm3
; AVX512F-NEXT: vpsllw %xmm2, %ymm0, %ymm0
-; AVX512F-NEXT: vinserti64x4 $1, %ymm3, %zmm0, %zmm3
-; AVX512F-NEXT: vpcmpeqd %xmm0, %xmm0, %xmm0
-; AVX512F-NEXT: vpsllw %xmm2, %xmm0, %xmm0
+; AVX512F-NEXT: vinserti64x4 $1, %ymm3, %zmm0, %zmm0
+; AVX512F-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT: vpsllw %xmm2, %xmm3, %xmm2
+; AVX512F-NEXT: vpbroadcastb %xmm2, %ymm2
+; AVX512F-NEXT: vinserti64x4 $1, %ymm2, %zmm2, %zmm2
+; AVX512F-NEXT: vpandq %zmm2, %zmm0, %zmm2
+; AVX512F-NEXT: vpsrlw %xmm6, %xmm3, %xmm0
+; AVX512F-NEXT: vpsrlw $8, %xmm0, %xmm0
; AVX512F-NEXT: vpbroadcastb %xmm0, %ymm0
; AVX512F-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm0
-; AVX512F-NEXT: vpternlogq $236, %zmm3, %zmm1, %zmm0
+; AVX512F-NEXT: vpternlogq $234, %zmm2, %zmm1, %zmm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: splatvar_funnnel_v64i8:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastb %xmm2, %ymm2
-; AVX512VL-NEXT: vinserti64x4 $1, %ymm2, %zmm2, %zmm2
+; AVX512VL-NEXT: vpbroadcastb %xmm2, %xmm2
; AVX512VL-NEXT: vextracti64x4 $1, %zmm1, %ymm3
; AVX512VL-NEXT: vpsrlw $1, %ymm3, %ymm3
; AVX512VL-NEXT: vmovdqa {{.*#+}} ymm4 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127]
; AVX512VL-NEXT: vpand %ymm4, %ymm3, %ymm3
-; AVX512VL-NEXT: vpsrlw $4, %ymm3, %ymm5
-; AVX512VL-NEXT: vmovdqa {{.*#+}} ymm6 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
-; AVX512VL-NEXT: vpand %ymm6, %ymm5, %ymm5
-; AVX512VL-NEXT: vpandnq {{.*}}(%rip), %zmm2, %zmm7
-; AVX512VL-NEXT: vextracti64x4 $1, %zmm7, %ymm8
-; AVX512VL-NEXT: vpsllw $5, %ymm8, %ymm8
-; AVX512VL-NEXT: vpblendvb %ymm8, %ymm5, %ymm3, %ymm3
-; AVX512VL-NEXT: vpsrlw $2, %ymm3, %ymm5
-; AVX512VL-NEXT: vmovdqa {{.*#+}} ymm9 = [63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63]
-; AVX512VL-NEXT: vpand %ymm5, %ymm9, %ymm5
-; AVX512VL-NEXT: vpaddb %ymm8, %ymm8, %ymm8
-; AVX512VL-NEXT: vpblendvb %ymm8, %ymm5, %ymm3, %ymm3
-; AVX512VL-NEXT: vpsrlw $1, %ymm3, %ymm5
-; AVX512VL-NEXT: vpand %ymm4, %ymm5, %ymm5
-; AVX512VL-NEXT: vpaddb %ymm8, %ymm8, %ymm8
-; AVX512VL-NEXT: vpblendvb %ymm8, %ymm5, %ymm3, %ymm3
+; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm5 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512VL-NEXT: vpandn %xmm5, %xmm2, %xmm6
+; AVX512VL-NEXT: vpmovzxbq {{.*#+}} xmm6 = xmm6[0],zero,zero,zero,zero,zero,zero,zero,xmm6[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512VL-NEXT: vpsrlw %xmm6, %ymm3, %ymm3
; AVX512VL-NEXT: vpsrlw $1, %ymm1, %ymm1
; AVX512VL-NEXT: vpand %ymm4, %ymm1, %ymm1
-; AVX512VL-NEXT: vpsrlw $4, %ymm1, %ymm5
-; AVX512VL-NEXT: vpand %ymm6, %ymm5, %ymm5
-; AVX512VL-NEXT: vpsllw $5, %ymm7, %ymm6
-; AVX512VL-NEXT: vpblendvb %ymm6, %ymm5, %ymm1, %ymm1
-; AVX512VL-NEXT: vpsrlw $2, %ymm1, %ymm5
-; AVX512VL-NEXT: vpand %ymm5, %ymm9, %ymm5
-; AVX512VL-NEXT: vpaddb %ymm6, %ymm6, %ymm6
-; AVX512VL-NEXT: vpblendvb %ymm6, %ymm5, %ymm1, %ymm1
-; AVX512VL-NEXT: vpsrlw $1, %ymm1, %ymm5
-; AVX512VL-NEXT: vpand %ymm4, %ymm5, %ymm4
-; AVX512VL-NEXT: vpaddb %ymm6, %ymm6, %ymm5
-; AVX512VL-NEXT: vpblendvb %ymm5, %ymm4, %ymm1, %ymm1
+; AVX512VL-NEXT: vpsrlw %xmm6, %ymm1, %ymm1
; AVX512VL-NEXT: vinserti64x4 $1, %ymm3, %zmm1, %zmm1
-; AVX512VL-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; AVX512VL-NEXT: vpand %xmm5, %xmm2, %xmm2
; AVX512VL-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
; AVX512VL-NEXT: vextracti64x4 $1, %zmm0, %ymm3
; AVX512VL-NEXT: vpsllw %xmm2, %ymm3, %ymm3
; AVX512VL-NEXT: vpsllw %xmm2, %ymm0, %ymm0
-; AVX512VL-NEXT: vinserti64x4 $1, %ymm3, %zmm0, %zmm3
-; AVX512VL-NEXT: vpcmpeqd %xmm0, %xmm0, %xmm0
-; AVX512VL-NEXT: vpsllw %xmm2, %xmm0, %xmm0
+; AVX512VL-NEXT: vinserti64x4 $1, %ymm3, %zmm0, %zmm0
+; AVX512VL-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
+; AVX512VL-NEXT: vpsllw %xmm2, %xmm3, %xmm2
+; AVX512VL-NEXT: vpbroadcastb %xmm2, %ymm2
+; AVX512VL-NEXT: vinserti64x4 $1, %ymm2, %zmm2, %zmm2
+; AVX512VL-NEXT: vpandq %zmm2, %zmm0, %zmm2
+; AVX512VL-NEXT: vpsrlw %xmm6, %xmm3, %xmm0
+; AVX512VL-NEXT: vpsrlw $8, %xmm0, %xmm0
; AVX512VL-NEXT: vpbroadcastb %xmm0, %ymm0
; AVX512VL-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm0
-; AVX512VL-NEXT: vpternlogq $236, %zmm3, %zmm1, %zmm0
+; AVX512VL-NEXT: vpternlogq $234, %zmm2, %zmm1, %zmm0
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: splatvar_funnnel_v64i8:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpbroadcastb %xmm2, %zmm2
+; AVX512BW-NEXT: vpbroadcastb %xmm2, %xmm2
+; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512BW-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512BW-NEXT: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512BW-NEXT: vpsllw %xmm4, %zmm0, %zmm0
+; AVX512BW-NEXT: vpcmpeqd %xmm5, %xmm5, %xmm5
+; AVX512BW-NEXT: vpsllw %xmm4, %xmm5, %xmm4
+; AVX512BW-NEXT: vpbroadcastb %xmm4, %zmm4
+; AVX512BW-NEXT: vpandq %zmm4, %zmm0, %zmm4
+; AVX512BW-NEXT: vpandn %xmm3, %xmm2, %xmm0
+; AVX512BW-NEXT: vpmovzxbq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero
; AVX512BW-NEXT: vpsrlw $1, %zmm1, %zmm1
-; AVX512BW-NEXT: vmovdqa64 {{.*#+}} zmm3 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127]
-; AVX512BW-NEXT: vpandq %zmm3, %zmm1, %zmm1
-; AVX512BW-NEXT: vpsrlw $4, %zmm1, %zmm4
-; AVX512BW-NEXT: vpandq {{.*}}(%rip), %zmm4, %zmm4
-; AVX512BW-NEXT: vpandnq {{.*}}(%rip), %zmm2, %zmm5
-; AVX512BW-NEXT: vpsllw $5, %zmm5, %zmm5
-; AVX512BW-NEXT: vpmovb2m %zmm5, %k1
-; AVX512BW-NEXT: vmovdqu8 %zmm4, %zmm1 {%k1}
-; AVX512BW-NEXT: vpsrlw $2, %zmm1, %zmm4
-; AVX512BW-NEXT: vpandq {{.*}}(%rip), %zmm4, %zmm4
-; AVX512BW-NEXT: vpaddb %zmm5, %zmm5, %zmm5
-; AVX512BW-NEXT: vpmovb2m %zmm5, %k1
-; AVX512BW-NEXT: vmovdqu8 %zmm4, %zmm1 {%k1}
-; AVX512BW-NEXT: vpsrlw $1, %zmm1, %zmm4
-; AVX512BW-NEXT: vpandq %zmm3, %zmm4, %zmm3
-; AVX512BW-NEXT: vpaddb %zmm5, %zmm5, %zmm4
-; AVX512BW-NEXT: vpmovb2m %zmm4, %k1
-; AVX512BW-NEXT: vmovdqu8 %zmm3, %zmm1 {%k1}
-; AVX512BW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX512BW-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512BW-NEXT: vpsllw %xmm2, %zmm0, %zmm3
-; AVX512BW-NEXT: vpcmpeqd %xmm0, %xmm0, %xmm0
-; AVX512BW-NEXT: vpsllw %xmm2, %xmm0, %xmm0
+; AVX512BW-NEXT: vpandq {{.*}}(%rip), %zmm1, %zmm1
+; AVX512BW-NEXT: vpsrlw %xmm0, %zmm1, %zmm1
+; AVX512BW-NEXT: vpsrlw %xmm0, %xmm5, %xmm0
+; AVX512BW-NEXT: vpsrlw $8, %xmm0, %xmm0
; AVX512BW-NEXT: vpbroadcastb %xmm0, %zmm0
-; AVX512BW-NEXT: vpternlogq $236, %zmm3, %zmm1, %zmm0
+; AVX512BW-NEXT: vpternlogq $236, %zmm1, %zmm4, %zmm0
; AVX512BW-NEXT: retq
;
; AVX512VBMI2-LABEL: splatvar_funnnel_v64i8:
; AVX512VBMI2: # %bb.0:
-; AVX512VBMI2-NEXT: vpbroadcastb %xmm2, %zmm2
+; AVX512VBMI2-NEXT: vpbroadcastb %xmm2, %xmm2
+; AVX512VBMI2-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512VBMI2-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512VBMI2-NEXT: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512VBMI2-NEXT: vpsllw %xmm4, %zmm0, %zmm0
+; AVX512VBMI2-NEXT: vpcmpeqd %xmm5, %xmm5, %xmm5
+; AVX512VBMI2-NEXT: vpsllw %xmm4, %xmm5, %xmm4
+; AVX512VBMI2-NEXT: vpbroadcastb %xmm4, %zmm4
+; AVX512VBMI2-NEXT: vpandq %zmm4, %zmm0, %zmm4
+; AVX512VBMI2-NEXT: vpandn %xmm3, %xmm2, %xmm0
+; AVX512VBMI2-NEXT: vpmovzxbq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero
; AVX512VBMI2-NEXT: vpsrlw $1, %zmm1, %zmm1
-; AVX512VBMI2-NEXT: vmovdqa64 {{.*#+}} zmm3 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127]
-; AVX512VBMI2-NEXT: vpandq %zmm3, %zmm1, %zmm1
-; AVX512VBMI2-NEXT: vpsrlw $4, %zmm1, %zmm4
-; AVX512VBMI2-NEXT: vpandq {{.*}}(%rip), %zmm4, %zmm4
-; AVX512VBMI2-NEXT: vpandnq {{.*}}(%rip), %zmm2, %zmm5
-; AVX512VBMI2-NEXT: vpsllw $5, %zmm5, %zmm5
-; AVX512VBMI2-NEXT: vpmovb2m %zmm5, %k1
-; AVX512VBMI2-NEXT: vmovdqu8 %zmm4, %zmm1 {%k1}
-; AVX512VBMI2-NEXT: vpsrlw $2, %zmm1, %zmm4
-; AVX512VBMI2-NEXT: vpandq {{.*}}(%rip), %zmm4, %zmm4
-; AVX512VBMI2-NEXT: vpaddb %zmm5, %zmm5, %zmm5
-; AVX512VBMI2-NEXT: vpmovb2m %zmm5, %k1
-; AVX512VBMI2-NEXT: vmovdqu8 %zmm4, %zmm1 {%k1}
-; AVX512VBMI2-NEXT: vpsrlw $1, %zmm1, %zmm4
-; AVX512VBMI2-NEXT: vpandq %zmm3, %zmm4, %zmm3
-; AVX512VBMI2-NEXT: vpaddb %zmm5, %zmm5, %zmm4
-; AVX512VBMI2-NEXT: vpmovb2m %zmm4, %k1
-; AVX512VBMI2-NEXT: vmovdqu8 %zmm3, %zmm1 {%k1}
-; AVX512VBMI2-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX512VBMI2-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512VBMI2-NEXT: vpsllw %xmm2, %zmm0, %zmm3
-; AVX512VBMI2-NEXT: vpcmpeqd %xmm0, %xmm0, %xmm0
-; AVX512VBMI2-NEXT: vpsllw %xmm2, %xmm0, %xmm0
+; AVX512VBMI2-NEXT: vpandq {{.*}}(%rip), %zmm1, %zmm1
+; AVX512VBMI2-NEXT: vpsrlw %xmm0, %zmm1, %zmm1
+; AVX512VBMI2-NEXT: vpsrlw %xmm0, %xmm5, %xmm0
+; AVX512VBMI2-NEXT: vpsrlw $8, %xmm0, %xmm0
; AVX512VBMI2-NEXT: vpbroadcastb %xmm0, %zmm0
-; AVX512VBMI2-NEXT: vpternlogq $236, %zmm3, %zmm1, %zmm0
+; AVX512VBMI2-NEXT: vpternlogq $236, %zmm1, %zmm4, %zmm0
; AVX512VBMI2-NEXT: retq
;
; AVX512VLBW-LABEL: splatvar_funnnel_v64i8:
; AVX512VLBW: # %bb.0:
-; AVX512VLBW-NEXT: vpbroadcastb %xmm2, %zmm2
+; 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: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512VLBW-NEXT: vpsllw %xmm4, %zmm0, %zmm0
+; AVX512VLBW-NEXT: vpcmpeqd %xmm5, %xmm5, %xmm5
+; AVX512VLBW-NEXT: vpsllw %xmm4, %xmm5, %xmm4
+; AVX512VLBW-NEXT: vpbroadcastb %xmm4, %zmm4
+; AVX512VLBW-NEXT: vpandq %zmm4, %zmm0, %zmm4
+; AVX512VLBW-NEXT: vpandn %xmm3, %xmm2, %xmm0
+; AVX512VLBW-NEXT: vpmovzxbq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero
; AVX512VLBW-NEXT: vpsrlw $1, %zmm1, %zmm1
-; AVX512VLBW-NEXT: vmovdqa64 {{.*#+}} zmm3 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127]
-; AVX512VLBW-NEXT: vpandq %zmm3, %zmm1, %zmm1
-; AVX512VLBW-NEXT: vpsrlw $4, %zmm1, %zmm4
-; AVX512VLBW-NEXT: vpandq {{.*}}(%rip), %zmm4, %zmm4
-; AVX512VLBW-NEXT: vpandnq {{.*}}(%rip), %zmm2, %zmm5
-; AVX512VLBW-NEXT: vpsllw $5, %zmm5, %zmm5
-; AVX512VLBW-NEXT: vpmovb2m %zmm5, %k1
-; AVX512VLBW-NEXT: vmovdqu8 %zmm4, %zmm1 {%k1}
-; AVX512VLBW-NEXT: vpsrlw $2, %zmm1, %zmm4
-; AVX512VLBW-NEXT: vpandq {{.*}}(%rip), %zmm4, %zmm4
-; AVX512VLBW-NEXT: vpaddb %zmm5, %zmm5, %zmm5
-; AVX512VLBW-NEXT: vpmovb2m %zmm5, %k1
-; AVX512VLBW-NEXT: vmovdqu8 %zmm4, %zmm1 {%k1}
-; AVX512VLBW-NEXT: vpsrlw $1, %zmm1, %zmm4
-; AVX512VLBW-NEXT: vpandq %zmm3, %zmm4, %zmm3
-; AVX512VLBW-NEXT: vpaddb %zmm5, %zmm5, %zmm4
-; AVX512VLBW-NEXT: vpmovb2m %zmm4, %k1
-; AVX512VLBW-NEXT: vmovdqu8 %zmm3, %zmm1 {%k1}
-; AVX512VLBW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX512VLBW-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512VLBW-NEXT: vpsllw %xmm2, %zmm0, %zmm3
-; AVX512VLBW-NEXT: vpcmpeqd %xmm0, %xmm0, %xmm0
-; AVX512VLBW-NEXT: vpsllw %xmm2, %xmm0, %xmm0
+; AVX512VLBW-NEXT: vpandq {{.*}}(%rip), %zmm1, %zmm1
+; AVX512VLBW-NEXT: vpsrlw %xmm0, %zmm1, %zmm1
+; AVX512VLBW-NEXT: vpsrlw %xmm0, %xmm5, %xmm0
+; AVX512VLBW-NEXT: vpsrlw $8, %xmm0, %xmm0
; AVX512VLBW-NEXT: vpbroadcastb %xmm0, %zmm0
-; AVX512VLBW-NEXT: vpternlogq $236, %zmm3, %zmm1, %zmm0
+; AVX512VLBW-NEXT: vpternlogq $236, %zmm1, %zmm4, %zmm0
; AVX512VLBW-NEXT: retq
;
; AVX512VLVBMI2-LABEL: splatvar_funnnel_v64i8:
; AVX512VLVBMI2: # %bb.0:
-; AVX512VLVBMI2-NEXT: vpbroadcastb %xmm2, %zmm2
+; 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: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512VLVBMI2-NEXT: vpsllw %xmm4, %zmm0, %zmm0
+; AVX512VLVBMI2-NEXT: vpcmpeqd %xmm5, %xmm5, %xmm5
+; AVX512VLVBMI2-NEXT: vpsllw %xmm4, %xmm5, %xmm4
+; AVX512VLVBMI2-NEXT: vpbroadcastb %xmm4, %zmm4
+; AVX512VLVBMI2-NEXT: vpandq %zmm4, %zmm0, %zmm4
+; AVX512VLVBMI2-NEXT: vpandn %xmm3, %xmm2, %xmm0
+; AVX512VLVBMI2-NEXT: vpmovzxbq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero
; AVX512VLVBMI2-NEXT: vpsrlw $1, %zmm1, %zmm1
-; AVX512VLVBMI2-NEXT: vmovdqa64 {{.*#+}} zmm3 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127]
-; AVX512VLVBMI2-NEXT: vpandq %zmm3, %zmm1, %zmm1
-; AVX512VLVBMI2-NEXT: vpsrlw $4, %zmm1, %zmm4
-; AVX512VLVBMI2-NEXT: vpandq {{.*}}(%rip), %zmm4, %zmm4
-; AVX512VLVBMI2-NEXT: vpandnq {{.*}}(%rip), %zmm2, %zmm5
-; AVX512VLVBMI2-NEXT: vpsllw $5, %zmm5, %zmm5
-; AVX512VLVBMI2-NEXT: vpmovb2m %zmm5, %k1
-; AVX512VLVBMI2-NEXT: vmovdqu8 %zmm4, %zmm1 {%k1}
-; AVX512VLVBMI2-NEXT: vpsrlw $2, %zmm1, %zmm4
-; AVX512VLVBMI2-NEXT: vpandq {{.*}}(%rip), %zmm4, %zmm4
-; AVX512VLVBMI2-NEXT: vpaddb %zmm5, %zmm5, %zmm5
-; AVX512VLVBMI2-NEXT: vpmovb2m %zmm5, %k1
-; AVX512VLVBMI2-NEXT: vmovdqu8 %zmm4, %zmm1 {%k1}
-; AVX512VLVBMI2-NEXT: vpsrlw $1, %zmm1, %zmm4
-; AVX512VLVBMI2-NEXT: vpandq %zmm3, %zmm4, %zmm3
-; AVX512VLVBMI2-NEXT: vpaddb %zmm5, %zmm5, %zmm4
-; AVX512VLVBMI2-NEXT: vpmovb2m %zmm4, %k1
-; AVX512VLVBMI2-NEXT: vmovdqu8 %zmm3, %zmm1 {%k1}
-; AVX512VLVBMI2-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX512VLVBMI2-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512VLVBMI2-NEXT: vpsllw %xmm2, %zmm0, %zmm3
-; AVX512VLVBMI2-NEXT: vpcmpeqd %xmm0, %xmm0, %xmm0
-; AVX512VLVBMI2-NEXT: vpsllw %xmm2, %xmm0, %xmm0
+; AVX512VLVBMI2-NEXT: vpandq {{.*}}(%rip), %zmm1, %zmm1
+; AVX512VLVBMI2-NEXT: vpsrlw %xmm0, %zmm1, %zmm1
+; AVX512VLVBMI2-NEXT: vpsrlw %xmm0, %xmm5, %xmm0
+; AVX512VLVBMI2-NEXT: vpsrlw $8, %xmm0, %xmm0
; AVX512VLVBMI2-NEXT: vpbroadcastb %xmm0, %zmm0
-; AVX512VLVBMI2-NEXT: vpternlogq $236, %zmm3, %zmm1, %zmm0
+; AVX512VLVBMI2-NEXT: vpternlogq $236, %zmm1, %zmm4, %zmm0
; AVX512VLVBMI2-NEXT: retq
%splat = shufflevector <64 x i8> %amt, <64 x i8> undef, <64 x i32> zeroinitializer
%res = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %x, <64 x i8> %y, <64 x i8> %splat)
diff --git a/llvm/test/CodeGen/X86/vector-fshr-128.ll b/llvm/test/CodeGen/X86/vector-fshr-128.ll
index bee498617069..6e7d036d9267 100644
--- a/llvm/test/CodeGen/X86/vector-fshr-128.ll
+++ b/llvm/test/CodeGen/X86/vector-fshr-128.ll
@@ -1062,98 +1062,59 @@ define <16 x i8> @var_funnnel_v16i8(<16 x i8> %x, <16 x i8> %y, <16 x i8> %amt)
;
define <2 x i64> @splatvar_funnnel_v2i64(<2 x i64> %x, <2 x i64> %y, <2 x i64> %amt) nounwind {
-; SSE2-LABEL: splatvar_funnnel_v2i64:
-; SSE2: # %bb.0:
-; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm2[0,1,0,1]
-; SSE2-NEXT: movdqa {{.*#+}} xmm4 = [63,63]
-; SSE2-NEXT: pandn %xmm4, %xmm3
-; SSE2-NEXT: psllq $1, %xmm0
-; SSE2-NEXT: movdqa %xmm0, %xmm5
-; SSE2-NEXT: psllq %xmm3, %xmm5
-; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm3[2,3,2,3]
-; SSE2-NEXT: psllq %xmm3, %xmm0
-; SSE2-NEXT: movsd {{.*#+}} xmm0 = xmm5[0],xmm0[1]
-; SSE2-NEXT: pand %xmm4, %xmm2
-; SSE2-NEXT: psrlq %xmm2, %xmm1
-; SSE2-NEXT: orpd %xmm1, %xmm0
-; SSE2-NEXT: retq
-;
-; SSE41-LABEL: splatvar_funnnel_v2i64:
-; SSE41: # %bb.0:
-; SSE41-NEXT: pshufd {{.*#+}} xmm3 = xmm2[0,1,0,1]
-; SSE41-NEXT: movdqa {{.*#+}} xmm4 = [63,63]
-; SSE41-NEXT: pandn %xmm4, %xmm3
-; SSE41-NEXT: psllq $1, %xmm0
-; SSE41-NEXT: movdqa %xmm0, %xmm5
-; SSE41-NEXT: psllq %xmm3, %xmm5
-; SSE41-NEXT: pshufd {{.*#+}} xmm3 = xmm3[2,3,2,3]
-; SSE41-NEXT: psllq %xmm3, %xmm0
-; SSE41-NEXT: pblendw {{.*#+}} xmm0 = xmm5[0,1,2,3],xmm0[4,5,6,7]
-; SSE41-NEXT: pand %xmm4, %xmm2
-; SSE41-NEXT: psrlq %xmm2, %xmm1
-; SSE41-NEXT: por %xmm1, %xmm0
-; SSE41-NEXT: retq
-;
-; AVX1-LABEL: splatvar_funnnel_v2i64:
-; AVX1: # %bb.0:
-; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm2[0,1,0,1]
-; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [63,63]
-; AVX1-NEXT: vpandn %xmm4, %xmm3, %xmm3
-; AVX1-NEXT: vpsllq $1, %xmm0, %xmm0
-; AVX1-NEXT: vpsllq %xmm3, %xmm0, %xmm5
-; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[2,3,2,3]
-; AVX1-NEXT: vpsllq %xmm3, %xmm0, %xmm0
-; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm5[0,1,2,3],xmm0[4,5,6,7]
-; AVX1-NEXT: vpand %xmm4, %xmm2, %xmm2
-; AVX1-NEXT: vpsrlq %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpor %xmm1, %xmm0, %xmm0
-; AVX1-NEXT: retq
+; SSE-LABEL: splatvar_funnnel_v2i64:
+; SSE: # %bb.0:
+; SSE-NEXT: movdqa {{.*#+}} xmm3 = [63,63]
+; SSE-NEXT: movdqa %xmm2, %xmm4
+; SSE-NEXT: pand %xmm3, %xmm4
+; SSE-NEXT: psrlq %xmm4, %xmm1
+; SSE-NEXT: pandn %xmm3, %xmm2
+; SSE-NEXT: psllq $1, %xmm0
+; SSE-NEXT: psllq %xmm2, %xmm0
+; SSE-NEXT: por %xmm1, %xmm0
+; SSE-NEXT: retq
;
-; AVX2-LABEL: splatvar_funnnel_v2i64:
-; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastq %xmm2, %xmm3
-; AVX2-NEXT: vmovdqa {{.*#+}} xmm4 = [63,63]
-; AVX2-NEXT: vpand %xmm4, %xmm2, %xmm2
-; AVX2-NEXT: vpsrlq %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpandn %xmm4, %xmm3, %xmm2
-; AVX2-NEXT: vpsllq $1, %xmm0, %xmm0
-; AVX2-NEXT: vpsllvq %xmm2, %xmm0, %xmm0
-; AVX2-NEXT: vpor %xmm1, %xmm0, %xmm0
-; AVX2-NEXT: retq
+; AVX-LABEL: splatvar_funnnel_v2i64:
+; AVX: # %bb.0:
+; AVX-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; AVX-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX-NEXT: vpsrlq %xmm4, %xmm1, %xmm1
+; AVX-NEXT: vpandn %xmm3, %xmm2, %xmm2
+; AVX-NEXT: vpsllq $1, %xmm0, %xmm0
+; AVX-NEXT: vpsllq %xmm2, %xmm0, %xmm0
+; AVX-NEXT: vpor %xmm1, %xmm0, %xmm0
+; AVX-NEXT: retq
;
; AVX512F-LABEL: splatvar_funnnel_v2i64:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpbroadcastq %xmm2, %xmm3
-; AVX512F-NEXT: vmovdqa {{.*#+}} xmm4 = [63,63]
-; AVX512F-NEXT: vpand %xmm4, %xmm2, %xmm2
-; AVX512F-NEXT: vpsrlq %xmm2, %xmm1, %xmm1
-; AVX512F-NEXT: vpandn %xmm4, %xmm3, %xmm2
+; AVX512F-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; AVX512F-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512F-NEXT: vpsrlq %xmm4, %xmm1, %xmm1
+; AVX512F-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512F-NEXT: vpsllq $1, %xmm0, %xmm0
-; AVX512F-NEXT: vpsllvq %xmm2, %xmm0, %xmm0
+; AVX512F-NEXT: vpsllq %xmm2, %xmm0, %xmm0
; AVX512F-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: splatvar_funnnel_v2i64:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastq %xmm2, %xmm3
-; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm4 = [63,63]
-; AVX512VL-NEXT: vpand %xmm4, %xmm2, %xmm2
-; AVX512VL-NEXT: vpsrlq %xmm2, %xmm1, %xmm1
-; AVX512VL-NEXT: vpandn %xmm4, %xmm3, %xmm2
+; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; AVX512VL-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512VL-NEXT: vpsrlq %xmm4, %xmm1, %xmm1
+; AVX512VL-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512VL-NEXT: vpsllq $1, %xmm0, %xmm0
-; AVX512VL-NEXT: vpsllvq %xmm2, %xmm0, %xmm0
+; AVX512VL-NEXT: vpsllq %xmm2, %xmm0, %xmm0
; AVX512VL-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: splatvar_funnnel_v2i64:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpbroadcastq %xmm2, %xmm3
-; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm4 = [63,63]
-; AVX512BW-NEXT: vpand %xmm4, %xmm2, %xmm2
-; AVX512BW-NEXT: vpsrlq %xmm2, %xmm1, %xmm1
-; AVX512BW-NEXT: vpandn %xmm4, %xmm3, %xmm2
+; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; AVX512BW-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512BW-NEXT: vpsrlq %xmm4, %xmm1, %xmm1
+; AVX512BW-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512BW-NEXT: vpsllq $1, %xmm0, %xmm0
-; AVX512BW-NEXT: vpsllvq %xmm2, %xmm0, %xmm0
+; AVX512BW-NEXT: vpsllq %xmm2, %xmm0, %xmm0
; AVX512BW-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512BW-NEXT: retq
;
@@ -1169,13 +1130,12 @@ 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, %xmm3
-; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm4 = [63,63]
-; AVX512VLBW-NEXT: vpand %xmm4, %xmm2, %xmm2
-; AVX512VLBW-NEXT: vpsrlq %xmm2, %xmm1, %xmm1
-; AVX512VLBW-NEXT: vpandn %xmm4, %xmm3, %xmm2
+; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; AVX512VLBW-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512VLBW-NEXT: vpsrlq %xmm4, %xmm1, %xmm1
+; AVX512VLBW-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512VLBW-NEXT: vpsllq $1, %xmm0, %xmm0
-; AVX512VLBW-NEXT: vpsllvq %xmm2, %xmm0, %xmm0
+; AVX512VLBW-NEXT: vpsllq %xmm2, %xmm0, %xmm0
; AVX512VLBW-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512VLBW-NEXT: retq
;
@@ -1186,29 +1146,16 @@ define <2 x i64> @splatvar_funnnel_v2i64(<2 x i64> %x, <2 x i64> %y, <2 x i64> %
; AVX512VLVBMI2-NEXT: vmovdqa %xmm1, %xmm0
; AVX512VLVBMI2-NEXT: retq
;
-; XOPAVX1-LABEL: splatvar_funnnel_v2i64:
-; XOPAVX1: # %bb.0:
-; XOPAVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm2[0,1,0,1]
-; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [63,63]
-; XOPAVX1-NEXT: vpand %xmm4, %xmm2, %xmm2
-; XOPAVX1-NEXT: vpsrlq %xmm2, %xmm1, %xmm1
-; XOPAVX1-NEXT: vpandn %xmm4, %xmm3, %xmm2
-; XOPAVX1-NEXT: vpsllq $1, %xmm0, %xmm0
-; XOPAVX1-NEXT: vpshlq %xmm2, %xmm0, %xmm0
-; XOPAVX1-NEXT: vpor %xmm1, %xmm0, %xmm0
-; XOPAVX1-NEXT: retq
-;
-; XOPAVX2-LABEL: splatvar_funnnel_v2i64:
-; XOPAVX2: # %bb.0:
-; XOPAVX2-NEXT: vpbroadcastq %xmm2, %xmm3
-; XOPAVX2-NEXT: vmovdqa {{.*#+}} xmm4 = [63,63]
-; XOPAVX2-NEXT: vpand %xmm4, %xmm2, %xmm2
-; XOPAVX2-NEXT: vpsrlq %xmm2, %xmm1, %xmm1
-; XOPAVX2-NEXT: vpandn %xmm4, %xmm3, %xmm2
-; XOPAVX2-NEXT: vpsllq $1, %xmm0, %xmm0
-; XOPAVX2-NEXT: vpsllvq %xmm2, %xmm0, %xmm0
-; XOPAVX2-NEXT: vpor %xmm1, %xmm0, %xmm0
-; XOPAVX2-NEXT: retq
+; XOP-LABEL: splatvar_funnnel_v2i64:
+; XOP: # %bb.0:
+; XOP-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; XOP-NEXT: vpand %xmm3, %xmm2, %xmm4
+; XOP-NEXT: vpsrlq %xmm4, %xmm1, %xmm1
+; XOP-NEXT: vpandn %xmm3, %xmm2, %xmm2
+; XOP-NEXT: vpsllq $1, %xmm0, %xmm0
+; XOP-NEXT: vpsllq %xmm2, %xmm0, %xmm0
+; XOP-NEXT: vpor %xmm1, %xmm0, %xmm0
+; XOP-NEXT: retq
;
; X86-SSE2-LABEL: splatvar_funnnel_v2i64:
; X86-SSE2: # %bb.0:
@@ -1238,107 +1185,95 @@ define <2 x i64> @splatvar_funnnel_v2i64(<2 x i64> %x, <2 x i64> %y, <2 x i64> %
define <4 x i32> @splatvar_funnnel_v4i32(<4 x i32> %x, <4 x i32> %y, <4 x i32> %amt) nounwind {
; SSE2-LABEL: splatvar_funnnel_v4i32:
; SSE2: # %bb.0:
-; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm2[0,0,0,0]
-; SSE2-NEXT: pslld $1, %xmm0
-; SSE2-NEXT: pandn {{.*}}(%rip), %xmm3
-; SSE2-NEXT: pslld $23, %xmm3
-; SSE2-NEXT: paddd {{.*}}(%rip), %xmm3
-; SSE2-NEXT: cvttps2dq %xmm3, %xmm3
-; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm0[1,1,3,3]
-; SSE2-NEXT: pmuludq %xmm3, %xmm0
-; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm3[1,1,3,3]
-; SSE2-NEXT: pmuludq %xmm4, %xmm3
-; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]
-; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; SSE2-NEXT: movd %xmm2, %eax
+; SSE2-NEXT: movl %eax, %ecx
+; SSE2-NEXT: andl $31, %ecx
+; SSE2-NEXT: movd %ecx, %xmm2
+; SSE2-NEXT: psrld %xmm2, %xmm1
+; SSE2-NEXT: pslld $1, %xmm0
+; SSE2-NEXT: notl %eax
; SSE2-NEXT: andl $31, %eax
; SSE2-NEXT: movd %eax, %xmm2
-; SSE2-NEXT: psrld %xmm2, %xmm1
+; SSE2-NEXT: pslld %xmm2, %xmm0
; SSE2-NEXT: por %xmm1, %xmm0
; SSE2-NEXT: retq
;
; SSE41-LABEL: splatvar_funnnel_v4i32:
; SSE41: # %bb.0:
-; SSE41-NEXT: pshufd {{.*#+}} xmm3 = xmm2[0,0,0,0]
-; SSE41-NEXT: movdqa {{.*#+}} xmm4 = [31,31,31,31]
-; SSE41-NEXT: pand %xmm4, %xmm2
+; SSE41-NEXT: movdqa {{.*#+}} xmm3 = [31,31,31,31]
+; SSE41-NEXT: movdqa %xmm2, %xmm4
+; SSE41-NEXT: pand %xmm3, %xmm4
+; SSE41-NEXT: pmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
+; SSE41-NEXT: psrld %xmm4, %xmm1
+; SSE41-NEXT: pandn %xmm3, %xmm2
; SSE41-NEXT: pmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
-; SSE41-NEXT: psrld %xmm2, %xmm1
-; SSE41-NEXT: pandn %xmm4, %xmm3
-; SSE41-NEXT: pslld $23, %xmm3
-; SSE41-NEXT: paddd {{.*}}(%rip), %xmm3
-; SSE41-NEXT: cvttps2dq %xmm3, %xmm2
; SSE41-NEXT: pslld $1, %xmm0
-; SSE41-NEXT: pmulld %xmm2, %xmm0
+; SSE41-NEXT: pslld %xmm2, %xmm0
; SSE41-NEXT: por %xmm1, %xmm0
; SSE41-NEXT: retq
;
; AVX1-LABEL: splatvar_funnnel_v4i32:
; AVX1: # %bb.0:
-; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm2[0,0,0,0]
-; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [31,31,31,31]
-; AVX1-NEXT: vpand %xmm4, %xmm2, %xmm2
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [31,31,31,31]
+; AVX1-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX1-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
+; AVX1-NEXT: vpsrld %xmm4, %xmm1, %xmm1
+; AVX1-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX1-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
-; AVX1-NEXT: vpsrld %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpandn %xmm4, %xmm3, %xmm2
-; AVX1-NEXT: vpslld $23, %xmm2, %xmm2
-; AVX1-NEXT: vpaddd {{.*}}(%rip), %xmm2, %xmm2
-; AVX1-NEXT: vcvttps2dq %xmm2, %xmm2
; AVX1-NEXT: vpslld $1, %xmm0, %xmm0
-; AVX1-NEXT: vpmulld %xmm2, %xmm0, %xmm0
+; AVX1-NEXT: vpslld %xmm2, %xmm0, %xmm0
; AVX1-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: splatvar_funnnel_v4i32:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastd %xmm2, %xmm3
-; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm4 = [31,31,31,31]
-; AVX2-NEXT: vpand %xmm4, %xmm2, %xmm2
+; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31]
+; AVX2-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX2-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
+; AVX2-NEXT: vpsrld %xmm4, %xmm1, %xmm1
+; AVX2-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
-; AVX2-NEXT: vpsrld %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpandn %xmm4, %xmm3, %xmm2
; AVX2-NEXT: vpslld $1, %xmm0, %xmm0
-; AVX2-NEXT: vpsllvd %xmm2, %xmm0, %xmm0
+; AVX2-NEXT: vpslld %xmm2, %xmm0, %xmm0
; AVX2-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX2-NEXT: retq
;
; AVX512F-LABEL: splatvar_funnnel_v4i32:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpbroadcastd %xmm2, %xmm3
-; AVX512F-NEXT: vpbroadcastd {{.*#+}} xmm4 = [31,31,31,31]
-; AVX512F-NEXT: vpand %xmm4, %xmm2, %xmm2
+; AVX512F-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31]
+; AVX512F-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512F-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
+; AVX512F-NEXT: vpsrld %xmm4, %xmm1, %xmm1
+; AVX512F-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512F-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
-; AVX512F-NEXT: vpsrld %xmm2, %xmm1, %xmm1
-; AVX512F-NEXT: vpandn %xmm4, %xmm3, %xmm2
; AVX512F-NEXT: vpslld $1, %xmm0, %xmm0
-; AVX512F-NEXT: vpsllvd %xmm2, %xmm0, %xmm0
+; AVX512F-NEXT: vpslld %xmm2, %xmm0, %xmm0
; AVX512F-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: splatvar_funnnel_v4i32:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastd %xmm2, %xmm3
-; AVX512VL-NEXT: vpbroadcastd {{.*#+}} xmm4 = [31,31,31,31]
-; AVX512VL-NEXT: vpand %xmm4, %xmm2, %xmm2
+; AVX512VL-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31]
+; AVX512VL-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512VL-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
+; AVX512VL-NEXT: vpsrld %xmm4, %xmm1, %xmm1
+; AVX512VL-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512VL-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
-; AVX512VL-NEXT: vpsrld %xmm2, %xmm1, %xmm1
-; AVX512VL-NEXT: vpandn %xmm4, %xmm3, %xmm2
; AVX512VL-NEXT: vpslld $1, %xmm0, %xmm0
-; AVX512VL-NEXT: vpsllvd %xmm2, %xmm0, %xmm0
+; AVX512VL-NEXT: vpslld %xmm2, %xmm0, %xmm0
; AVX512VL-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: splatvar_funnnel_v4i32:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpbroadcastd %xmm2, %xmm3
-; AVX512BW-NEXT: vpbroadcastd {{.*#+}} xmm4 = [31,31,31,31]
-; AVX512BW-NEXT: vpand %xmm4, %xmm2, %xmm2
+; AVX512BW-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31]
+; AVX512BW-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512BW-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
+; AVX512BW-NEXT: vpsrld %xmm4, %xmm1, %xmm1
+; AVX512BW-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512BW-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
-; AVX512BW-NEXT: vpsrld %xmm2, %xmm1, %xmm1
-; AVX512BW-NEXT: vpandn %xmm4, %xmm3, %xmm2
; AVX512BW-NEXT: vpslld $1, %xmm0, %xmm0
-; AVX512BW-NEXT: vpsllvd %xmm2, %xmm0, %xmm0
+; AVX512BW-NEXT: vpslld %xmm2, %xmm0, %xmm0
; AVX512BW-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512BW-NEXT: retq
;
@@ -1354,14 +1289,14 @@ 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, %xmm3
-; AVX512VLBW-NEXT: vpbroadcastd {{.*#+}} xmm4 = [31,31,31,31]
-; AVX512VLBW-NEXT: vpand %xmm4, %xmm2, %xmm2
+; AVX512VLBW-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31]
+; AVX512VLBW-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512VLBW-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
+; AVX512VLBW-NEXT: vpsrld %xmm4, %xmm1, %xmm1
+; AVX512VLBW-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512VLBW-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
-; AVX512VLBW-NEXT: vpsrld %xmm2, %xmm1, %xmm1
-; AVX512VLBW-NEXT: vpandn %xmm4, %xmm3, %xmm2
; AVX512VLBW-NEXT: vpslld $1, %xmm0, %xmm0
-; AVX512VLBW-NEXT: vpsllvd %xmm2, %xmm0, %xmm0
+; AVX512VLBW-NEXT: vpslld %xmm2, %xmm0, %xmm0
; AVX512VLBW-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512VLBW-NEXT: retq
;
@@ -1374,49 +1309,42 @@ define <4 x i32> @splatvar_funnnel_v4i32(<4 x i32> %x, <4 x i32> %y, <4 x i32> %
;
; XOPAVX1-LABEL: splatvar_funnnel_v4i32:
; XOPAVX1: # %bb.0:
-; XOPAVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm2[0,0,0,0]
-; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [31,31,31,31]
-; XOPAVX1-NEXT: vpand %xmm4, %xmm2, %xmm2
+; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [31,31,31,31]
+; XOPAVX1-NEXT: vpand %xmm3, %xmm2, %xmm4
+; XOPAVX1-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
+; XOPAVX1-NEXT: vpsrld %xmm4, %xmm1, %xmm1
+; XOPAVX1-NEXT: vpandn %xmm3, %xmm2, %xmm2
; XOPAVX1-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
-; XOPAVX1-NEXT: vpsrld %xmm2, %xmm1, %xmm1
-; XOPAVX1-NEXT: vpandn %xmm4, %xmm3, %xmm2
; XOPAVX1-NEXT: vpslld $1, %xmm0, %xmm0
-; XOPAVX1-NEXT: vpshld %xmm2, %xmm0, %xmm0
+; XOPAVX1-NEXT: vpslld %xmm2, %xmm0, %xmm0
; XOPAVX1-NEXT: vpor %xmm1, %xmm0, %xmm0
; XOPAVX1-NEXT: retq
;
; XOPAVX2-LABEL: splatvar_funnnel_v4i32:
; XOPAVX2: # %bb.0:
-; XOPAVX2-NEXT: vpbroadcastd %xmm2, %xmm3
-; XOPAVX2-NEXT: vpbroadcastd {{.*#+}} xmm4 = [31,31,31,31]
-; XOPAVX2-NEXT: vpand %xmm4, %xmm2, %xmm2
+; XOPAVX2-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31]
+; XOPAVX2-NEXT: vpand %xmm3, %xmm2, %xmm4
+; XOPAVX2-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
+; XOPAVX2-NEXT: vpsrld %xmm4, %xmm1, %xmm1
+; XOPAVX2-NEXT: vpandn %xmm3, %xmm2, %xmm2
; XOPAVX2-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
-; XOPAVX2-NEXT: vpsrld %xmm2, %xmm1, %xmm1
-; XOPAVX2-NEXT: vpandn %xmm4, %xmm3, %xmm2
; XOPAVX2-NEXT: vpslld $1, %xmm0, %xmm0
-; XOPAVX2-NEXT: vpsllvd %xmm2, %xmm0, %xmm0
+; XOPAVX2-NEXT: vpslld %xmm2, %xmm0, %xmm0
; XOPAVX2-NEXT: vpor %xmm1, %xmm0, %xmm0
; XOPAVX2-NEXT: retq
;
; X86-SSE2-LABEL: splatvar_funnnel_v4i32:
; X86-SSE2: # %bb.0:
-; X86-SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm2[0,0,0,0]
-; X86-SSE2-NEXT: pslld $1, %xmm0
-; X86-SSE2-NEXT: pandn {{\.LCPI.*}}, %xmm3
-; X86-SSE2-NEXT: pslld $23, %xmm3
-; X86-SSE2-NEXT: paddd {{\.LCPI.*}}, %xmm3
-; X86-SSE2-NEXT: cvttps2dq %xmm3, %xmm3
-; X86-SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm0[1,1,3,3]
-; X86-SSE2-NEXT: pmuludq %xmm3, %xmm0
-; X86-SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; X86-SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm3[1,1,3,3]
-; X86-SSE2-NEXT: pmuludq %xmm4, %xmm3
-; X86-SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]
-; X86-SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; X86-SSE2-NEXT: movd %xmm2, %eax
+; X86-SSE2-NEXT: movl %eax, %ecx
+; X86-SSE2-NEXT: andl $31, %ecx
+; X86-SSE2-NEXT: movd %ecx, %xmm2
+; X86-SSE2-NEXT: psrld %xmm2, %xmm1
+; X86-SSE2-NEXT: pslld $1, %xmm0
+; X86-SSE2-NEXT: notl %eax
; X86-SSE2-NEXT: andl $31, %eax
; X86-SSE2-NEXT: movd %eax, %xmm2
-; X86-SSE2-NEXT: psrld %xmm2, %xmm1
+; X86-SSE2-NEXT: pslld %xmm2, %xmm0
; X86-SSE2-NEXT: por %xmm1, %xmm0
; X86-SSE2-NEXT: retl
%splat = shufflevector <4 x i32> %amt, <4 x i32> undef, <4 x i32> zeroinitializer
@@ -1427,148 +1355,84 @@ define <4 x i32> @splatvar_funnnel_v4i32(<4 x i32> %x, <4 x i32> %y, <4 x i32> %
define <8 x i16> @splatvar_funnnel_v8i16(<8 x i16> %x, <8 x i16> %y, <8 x i16> %amt) nounwind {
; SSE2-LABEL: splatvar_funnnel_v8i16:
; SSE2: # %bb.0:
-; SSE2-NEXT: pshuflw {{.*#+}} xmm3 = xmm2[0,0,0,0,4,5,6,7]
-; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,0,0,0]
-; SSE2-NEXT: movdqa {{.*#+}} xmm4 = [15,15,15,15,15,15,15,15]
-; SSE2-NEXT: pandn %xmm4, %xmm3
-; SSE2-NEXT: movdqa %xmm3, %xmm5
-; SSE2-NEXT: punpckhwd {{.*#+}} xmm5 = xmm5[4,4,5,5,6,6,7,7]
-; SSE2-NEXT: pslld $23, %xmm5
-; SSE2-NEXT: movdqa {{.*#+}} xmm6 = [1065353216,1065353216,1065353216,1065353216]
-; SSE2-NEXT: paddd %xmm6, %xmm5
-; SSE2-NEXT: cvttps2dq %xmm5, %xmm5
-; SSE2-NEXT: pshuflw {{.*#+}} xmm5 = xmm5[0,2,2,3,4,5,6,7]
-; SSE2-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,6,6,7]
-; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm5[0,2,2,3]
-; SSE2-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3]
-; SSE2-NEXT: pslld $23, %xmm3
-; SSE2-NEXT: paddd %xmm6, %xmm3
-; SSE2-NEXT: cvttps2dq %xmm3, %xmm3
-; SSE2-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,2,2,3,4,5,6,7]
-; SSE2-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,6,6,7]
-; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]
-; SSE2-NEXT: punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm5[0]
-; SSE2-NEXT: psllw $1, %xmm0
-; SSE2-NEXT: pmullw %xmm3, %xmm0
-; SSE2-NEXT: pand %xmm4, %xmm2
+; SSE2-NEXT: movdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; SSE2-NEXT: movdqa %xmm2, %xmm4
+; SSE2-NEXT: pand %xmm3, %xmm4
+; SSE2-NEXT: pslldq {{.*#+}} xmm4 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm4[0,1]
+; SSE2-NEXT: psrldq {{.*#+}} xmm4 = xmm4[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE2-NEXT: psrlw %xmm4, %xmm1
+; SSE2-NEXT: pandn %xmm3, %xmm2
; SSE2-NEXT: pslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1]
; SSE2-NEXT: psrldq {{.*#+}} xmm2 = xmm2[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE2-NEXT: psrlw %xmm2, %xmm1
+; SSE2-NEXT: psllw $1, %xmm0
+; SSE2-NEXT: psllw %xmm2, %xmm0
; SSE2-NEXT: por %xmm1, %xmm0
; SSE2-NEXT: retq
;
; SSE41-LABEL: splatvar_funnnel_v8i16:
; SSE41: # %bb.0:
-; SSE41-NEXT: pshuflw {{.*#+}} xmm3 = xmm2[0,0,0,0,4,5,6,7]
-; SSE41-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,0,0,0]
-; SSE41-NEXT: movdqa {{.*#+}} xmm4 = [15,15,15,15,15,15,15,15]
-; SSE41-NEXT: pandn %xmm4, %xmm3
-; SSE41-NEXT: pmovzxwd {{.*#+}} xmm5 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero
-; SSE41-NEXT: punpckhwd {{.*#+}} xmm3 = xmm3[4,4,5,5,6,6,7,7]
-; SSE41-NEXT: pslld $23, %xmm3
-; SSE41-NEXT: movdqa {{.*#+}} xmm6 = [1065353216,1065353216,1065353216,1065353216]
-; SSE41-NEXT: paddd %xmm6, %xmm3
-; SSE41-NEXT: cvttps2dq %xmm3, %xmm3
-; SSE41-NEXT: pslld $23, %xmm5
-; SSE41-NEXT: paddd %xmm6, %xmm5
-; SSE41-NEXT: cvttps2dq %xmm5, %xmm5
-; SSE41-NEXT: packusdw %xmm3, %xmm5
-; SSE41-NEXT: psllw $1, %xmm0
-; SSE41-NEXT: pmullw %xmm5, %xmm0
-; SSE41-NEXT: pand %xmm4, %xmm2
+; SSE41-NEXT: movdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; SSE41-NEXT: movdqa %xmm2, %xmm4
+; SSE41-NEXT: pand %xmm3, %xmm4
+; SSE41-NEXT: pmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
+; SSE41-NEXT: psrlw %xmm4, %xmm1
+; SSE41-NEXT: pandn %xmm3, %xmm2
; SSE41-NEXT: pmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
-; SSE41-NEXT: psrlw %xmm2, %xmm1
+; SSE41-NEXT: psllw $1, %xmm0
+; SSE41-NEXT: psllw %xmm2, %xmm0
; SSE41-NEXT: por %xmm1, %xmm0
; SSE41-NEXT: retq
;
-; AVX1-LABEL: splatvar_funnnel_v8i16:
-; AVX1: # %bb.0:
-; AVX1-NEXT: vpshuflw {{.*#+}} xmm3 = xmm2[0,0,0,0,4,5,6,7]
-; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,0,0,0]
-; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [15,15,15,15,15,15,15,15]
-; AVX1-NEXT: vpandn %xmm4, %xmm3, %xmm3
-; AVX1-NEXT: vpunpckhwd {{.*#+}} xmm5 = xmm3[4,4,5,5,6,6,7,7]
-; AVX1-NEXT: vpslld $23, %xmm5, %xmm5
-; AVX1-NEXT: vmovdqa {{.*#+}} xmm6 = [1065353216,1065353216,1065353216,1065353216]
-; AVX1-NEXT: vpaddd %xmm6, %xmm5, %xmm5
-; AVX1-NEXT: vcvttps2dq %xmm5, %xmm5
-; AVX1-NEXT: vpmovzxwd {{.*#+}} xmm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero
-; AVX1-NEXT: vpslld $23, %xmm3, %xmm3
-; AVX1-NEXT: vpaddd %xmm6, %xmm3, %xmm3
-; AVX1-NEXT: vcvttps2dq %xmm3, %xmm3
-; AVX1-NEXT: vpackusdw %xmm5, %xmm3, %xmm3
-; AVX1-NEXT: vpsllw $1, %xmm0, %xmm0
-; AVX1-NEXT: vpmullw %xmm3, %xmm0, %xmm0
-; AVX1-NEXT: vpand %xmm4, %xmm2, %xmm2
-; AVX1-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
-; AVX1-NEXT: vpsrlw %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpor %xmm1, %xmm0, %xmm0
-; AVX1-NEXT: retq
-;
-; AVX2-LABEL: splatvar_funnnel_v8i16:
-; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastw %xmm2, %xmm3
-; AVX2-NEXT: vmovdqa {{.*#+}} xmm4 = [15,15,15,15,15,15,15,15]
-; AVX2-NEXT: vpand %xmm4, %xmm2, %xmm2
-; AVX2-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
-; AVX2-NEXT: vpsrlw %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpandn %xmm4, %xmm3, %xmm2
-; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero
-; AVX2-NEXT: vpsllw $1, %xmm0, %xmm0
-; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
-; AVX2-NEXT: vpsllvd %ymm2, %ymm0, %ymm0
-; AVX2-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u,16,17,20,21,24,25,28,29,u,u,u,u,u,u,u,u]
-; AVX2-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-NEXT: vpor %xmm1, %xmm0, %xmm0
-; AVX2-NEXT: vzeroupper
-; AVX2-NEXT: retq
+; AVX-LABEL: splatvar_funnnel_v8i16:
+; AVX: # %bb.0:
+; AVX-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; AVX-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
+; AVX-NEXT: vpsrlw %xmm4, %xmm1, %xmm1
+; AVX-NEXT: vpandn %xmm3, %xmm2, %xmm2
+; AVX-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
+; AVX-NEXT: vpsllw $1, %xmm0, %xmm0
+; AVX-NEXT: vpsllw %xmm2, %xmm0, %xmm0
+; AVX-NEXT: vpor %xmm1, %xmm0, %xmm0
+; AVX-NEXT: retq
;
; AVX512F-LABEL: splatvar_funnnel_v8i16:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpbroadcastw %xmm2, %xmm3
-; AVX512F-NEXT: vmovdqa {{.*#+}} xmm4 = [15,15,15,15,15,15,15,15]
-; AVX512F-NEXT: vpand %xmm4, %xmm2, %xmm2
+; AVX512F-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; AVX512F-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512F-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
+; AVX512F-NEXT: vpsrlw %xmm4, %xmm1, %xmm1
+; AVX512F-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512F-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
-; AVX512F-NEXT: vpsrlw %xmm2, %xmm1, %xmm1
-; AVX512F-NEXT: vpandn %xmm4, %xmm3, %xmm2
-; AVX512F-NEXT: vpmovzxwd {{.*#+}} ymm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero
; AVX512F-NEXT: vpsllw $1, %xmm0, %xmm0
-; AVX512F-NEXT: vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
-; AVX512F-NEXT: vpsllvd %ymm2, %ymm0, %ymm0
-; AVX512F-NEXT: vpmovdw %zmm0, %ymm0
+; AVX512F-NEXT: vpsllw %xmm2, %xmm0, %xmm0
; AVX512F-NEXT: vpor %xmm1, %xmm0, %xmm0
-; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: splatvar_funnnel_v8i16:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastw %xmm2, %xmm3
-; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm4 = [15,15,15,15,15,15,15,15]
-; AVX512VL-NEXT: vpand %xmm4, %xmm2, %xmm2
+; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; AVX512VL-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512VL-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
+; AVX512VL-NEXT: vpsrlw %xmm4, %xmm1, %xmm1
+; AVX512VL-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512VL-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
-; AVX512VL-NEXT: vpsrlw %xmm2, %xmm1, %xmm1
-; AVX512VL-NEXT: vpandn %xmm4, %xmm3, %xmm2
-; AVX512VL-NEXT: vpmovzxwd {{.*#+}} ymm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero
; AVX512VL-NEXT: vpsllw $1, %xmm0, %xmm0
-; AVX512VL-NEXT: vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
-; AVX512VL-NEXT: vpsllvd %ymm2, %ymm0, %ymm0
-; AVX512VL-NEXT: vpmovdw %ymm0, %xmm0
+; AVX512VL-NEXT: vpsllw %xmm2, %xmm0, %xmm0
; AVX512VL-NEXT: vpor %xmm1, %xmm0, %xmm0
-; AVX512VL-NEXT: vzeroupper
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: splatvar_funnnel_v8i16:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpbroadcastw %xmm2, %xmm3
-; AVX512BW-NEXT: vpsllw $1, %xmm0, %xmm0
-; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm4 = [15,15,15,15,15,15,15,15]
-; AVX512BW-NEXT: vpandn %xmm4, %xmm3, %xmm3
-; AVX512BW-NEXT: vpsllvw %zmm3, %zmm0, %zmm0
-; AVX512BW-NEXT: vpand %xmm4, %xmm2, %xmm2
+; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512BW-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
+; AVX512BW-NEXT: vpsrlw %xmm4, %xmm1, %xmm1
+; AVX512BW-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512BW-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
-; AVX512BW-NEXT: vpsrlw %xmm2, %xmm1, %xmm1
+; AVX512BW-NEXT: vpsllw $1, %xmm0, %xmm0
+; AVX512BW-NEXT: vpsllw %xmm2, %xmm0, %xmm0
; AVX512BW-NEXT: vpor %xmm1, %xmm0, %xmm0
-; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
;
; AVX512VBMI2-LABEL: splatvar_funnnel_v8i16:
@@ -1583,14 +1447,14 @@ 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, %xmm3
-; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm4 = [15,15,15,15,15,15,15,15]
-; AVX512VLBW-NEXT: vpand %xmm4, %xmm2, %xmm2
+; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; AVX512VLBW-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512VLBW-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
+; AVX512VLBW-NEXT: vpsrlw %xmm4, %xmm1, %xmm1
+; AVX512VLBW-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512VLBW-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
-; AVX512VLBW-NEXT: vpsrlw %xmm2, %xmm1, %xmm1
-; AVX512VLBW-NEXT: vpandn %xmm4, %xmm3, %xmm2
; AVX512VLBW-NEXT: vpsllw $1, %xmm0, %xmm0
-; AVX512VLBW-NEXT: vpsllvw %xmm2, %xmm0, %xmm0
+; AVX512VLBW-NEXT: vpsllw %xmm2, %xmm0, %xmm0
; AVX512VLBW-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512VLBW-NEXT: retq
;
@@ -1601,62 +1465,32 @@ define <8 x i16> @splatvar_funnnel_v8i16(<8 x i16> %x, <8 x i16> %y, <8 x i16> %
; AVX512VLVBMI2-NEXT: vmovdqa %xmm1, %xmm0
; AVX512VLVBMI2-NEXT: retq
;
-; XOPAVX1-LABEL: splatvar_funnnel_v8i16:
-; XOPAVX1: # %bb.0:
-; XOPAVX1-NEXT: vpshuflw {{.*#+}} xmm3 = xmm2[0,0,0,0,4,5,6,7]
-; XOPAVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,0,0,0]
-; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [15,15,15,15,15,15,15,15]
-; XOPAVX1-NEXT: vpand %xmm4, %xmm2, %xmm2
-; XOPAVX1-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
-; XOPAVX1-NEXT: vpsrlw %xmm2, %xmm1, %xmm1
-; XOPAVX1-NEXT: vpandn %xmm4, %xmm3, %xmm2
-; XOPAVX1-NEXT: vpsllw $1, %xmm0, %xmm0
-; XOPAVX1-NEXT: vpshlw %xmm2, %xmm0, %xmm0
-; XOPAVX1-NEXT: vpor %xmm1, %xmm0, %xmm0
-; XOPAVX1-NEXT: retq
-;
-; XOPAVX2-LABEL: splatvar_funnnel_v8i16:
-; XOPAVX2: # %bb.0:
-; XOPAVX2-NEXT: vpbroadcastw %xmm2, %xmm3
-; XOPAVX2-NEXT: vmovdqa {{.*#+}} xmm4 = [15,15,15,15,15,15,15,15]
-; XOPAVX2-NEXT: vpand %xmm4, %xmm2, %xmm2
-; XOPAVX2-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
-; XOPAVX2-NEXT: vpsrlw %xmm2, %xmm1, %xmm1
-; XOPAVX2-NEXT: vpandn %xmm4, %xmm3, %xmm2
-; XOPAVX2-NEXT: vpsllw $1, %xmm0, %xmm0
-; XOPAVX2-NEXT: vpshlw %xmm2, %xmm0, %xmm0
-; XOPAVX2-NEXT: vpor %xmm1, %xmm0, %xmm0
-; XOPAVX2-NEXT: retq
+; XOP-LABEL: splatvar_funnnel_v8i16:
+; XOP: # %bb.0:
+; XOP-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; XOP-NEXT: vpand %xmm3, %xmm2, %xmm4
+; XOP-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
+; XOP-NEXT: vpsrlw %xmm4, %xmm1, %xmm1
+; XOP-NEXT: vpandn %xmm3, %xmm2, %xmm2
+; XOP-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
+; XOP-NEXT: vpsllw $1, %xmm0, %xmm0
+; XOP-NEXT: vpsllw %xmm2, %xmm0, %xmm0
+; XOP-NEXT: vpor %xmm1, %xmm0, %xmm0
+; XOP-NEXT: retq
;
; X86-SSE2-LABEL: splatvar_funnnel_v8i16:
; X86-SSE2: # %bb.0:
-; X86-SSE2-NEXT: pshuflw {{.*#+}} xmm3 = xmm2[0,0,0,0,4,5,6,7]
-; X86-SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,0,0,0]
-; X86-SSE2-NEXT: movdqa {{.*#+}} xmm4 = [15,15,15,15,15,15,15,15]
-; X86-SSE2-NEXT: pandn %xmm4, %xmm3
-; X86-SSE2-NEXT: movdqa %xmm3, %xmm5
-; X86-SSE2-NEXT: punpckhwd {{.*#+}} xmm5 = xmm5[4,4,5,5,6,6,7,7]
-; X86-SSE2-NEXT: pslld $23, %xmm5
-; X86-SSE2-NEXT: movdqa {{.*#+}} xmm6 = [1065353216,1065353216,1065353216,1065353216]
-; X86-SSE2-NEXT: paddd %xmm6, %xmm5
-; X86-SSE2-NEXT: cvttps2dq %xmm5, %xmm5
-; X86-SSE2-NEXT: pshuflw {{.*#+}} xmm5 = xmm5[0,2,2,3,4,5,6,7]
-; X86-SSE2-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,6,6,7]
-; X86-SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm5[0,2,2,3]
-; X86-SSE2-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3]
-; X86-SSE2-NEXT: pslld $23, %xmm3
-; X86-SSE2-NEXT: paddd %xmm6, %xmm3
-; X86-SSE2-NEXT: cvttps2dq %xmm3, %xmm3
-; X86-SSE2-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,2,2,3,4,5,6,7]
-; X86-SSE2-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,6,6,7]
-; X86-SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]
-; X86-SSE2-NEXT: punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm5[0]
-; X86-SSE2-NEXT: psllw $1, %xmm0
-; X86-SSE2-NEXT: pmullw %xmm3, %xmm0
-; X86-SSE2-NEXT: pand %xmm4, %xmm2
+; X86-SSE2-NEXT: movdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; X86-SSE2-NEXT: movdqa %xmm2, %xmm4
+; X86-SSE2-NEXT: pand %xmm3, %xmm4
+; X86-SSE2-NEXT: pslldq {{.*#+}} xmm4 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm4[0,1]
+; X86-SSE2-NEXT: psrldq {{.*#+}} xmm4 = xmm4[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-SSE2-NEXT: psrlw %xmm4, %xmm1
+; X86-SSE2-NEXT: pandn %xmm3, %xmm2
; X86-SSE2-NEXT: pslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1]
; X86-SSE2-NEXT: psrldq {{.*#+}} xmm2 = xmm2[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; X86-SSE2-NEXT: psrlw %xmm2, %xmm1
+; X86-SSE2-NEXT: psllw $1, %xmm0
+; X86-SSE2-NEXT: psllw %xmm2, %xmm0
; X86-SSE2-NEXT: por %xmm1, %xmm0
; X86-SSE2-NEXT: retl
%splat = shufflevector <8 x i16> %amt, <8 x i16> undef, <8 x i32> zeroinitializer
@@ -1667,140 +1501,92 @@ define <8 x i16> @splatvar_funnnel_v8i16(<8 x i16> %x, <8 x i16> %y, <8 x i16> %
define <16 x i8> @splatvar_funnnel_v16i8(<16 x i8> %x, <16 x i8> %y, <16 x i8> %amt) nounwind {
; SSE2-LABEL: splatvar_funnnel_v16i8:
; SSE2: # %bb.0:
-; SSE2-NEXT: movdqa %xmm2, %xmm3
-; SSE2-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3],xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
-; SSE2-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,0,0,0,4,5,6,7]
-; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm3[0,0,0,0]
; SSE2-NEXT: movdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; SSE2-NEXT: movdqa %xmm2, %xmm4
; SSE2-NEXT: pandn %xmm3, %xmm4
-; SSE2-NEXT: psllw $5, %xmm4
-; SSE2-NEXT: pxor %xmm5, %xmm5
-; SSE2-NEXT: pxor %xmm6, %xmm6
-; SSE2-NEXT: pcmpgtb %xmm4, %xmm6
-; SSE2-NEXT: paddb %xmm0, %xmm0
-; SSE2-NEXT: movdqa %xmm6, %xmm7
-; SSE2-NEXT: pandn %xmm0, %xmm7
-; SSE2-NEXT: psllw $4, %xmm0
-; SSE2-NEXT: pand %xmm6, %xmm0
-; SSE2-NEXT: pand {{.*}}(%rip), %xmm0
-; SSE2-NEXT: por %xmm7, %xmm0
-; SSE2-NEXT: paddb %xmm4, %xmm4
-; SSE2-NEXT: pxor %xmm6, %xmm6
-; SSE2-NEXT: pcmpgtb %xmm4, %xmm6
-; SSE2-NEXT: movdqa %xmm6, %xmm7
-; SSE2-NEXT: pandn %xmm0, %xmm7
-; SSE2-NEXT: psllw $2, %xmm0
-; SSE2-NEXT: pand %xmm6, %xmm0
-; SSE2-NEXT: pand {{.*}}(%rip), %xmm0
-; SSE2-NEXT: por %xmm7, %xmm0
-; SSE2-NEXT: paddb %xmm4, %xmm4
-; SSE2-NEXT: pcmpgtb %xmm4, %xmm5
-; SSE2-NEXT: movdqa %xmm5, %xmm4
-; SSE2-NEXT: pandn %xmm0, %xmm4
+; SSE2-NEXT: pslldq {{.*#+}} xmm4 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm4[0]
+; SSE2-NEXT: psrldq {{.*#+}} xmm4 = xmm4[15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; SSE2-NEXT: paddb %xmm0, %xmm0
-; SSE2-NEXT: pand %xmm5, %xmm0
+; SSE2-NEXT: psllw %xmm4, %xmm0
+; SSE2-NEXT: pcmpeqd %xmm5, %xmm5
+; SSE2-NEXT: pcmpeqd %xmm6, %xmm6
+; SSE2-NEXT: psllw %xmm4, %xmm6
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm6 = xmm6[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE2-NEXT: pshuflw {{.*#+}} xmm4 = xmm6[0,0,0,0,4,5,6,7]
+; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm4[0,0,0,0]
+; SSE2-NEXT: pand %xmm4, %xmm0
; SSE2-NEXT: pand %xmm3, %xmm2
; SSE2-NEXT: pslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0]
; SSE2-NEXT: psrldq {{.*#+}} xmm2 = xmm2[15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; SSE2-NEXT: psrlw %xmm2, %xmm1
-; SSE2-NEXT: pcmpeqd %xmm3, %xmm3
-; SSE2-NEXT: psrlw %xmm2, %xmm3
-; SSE2-NEXT: psrlw $8, %xmm3
-; SSE2-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT: pshuflw {{.*#+}} xmm2 = xmm3[0,0,0,0,4,5,6,7]
+; SSE2-NEXT: psrlw %xmm2, %xmm5
+; SSE2-NEXT: psrlw $8, %xmm5
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm5 = xmm5[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE2-NEXT: pshuflw {{.*#+}} xmm2 = xmm5[0,0,0,0,4,5,6,7]
; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
; SSE2-NEXT: pand %xmm1, %xmm2
-; SSE2-NEXT: por %xmm4, %xmm2
; SSE2-NEXT: por %xmm2, %xmm0
; SSE2-NEXT: retq
;
; SSE41-LABEL: splatvar_funnnel_v16i8:
; SSE41: # %bb.0:
-; SSE41-NEXT: movdqa %xmm0, %xmm3
-; SSE41-NEXT: pxor %xmm4, %xmm4
-; SSE41-NEXT: movdqa %xmm2, %xmm0
-; SSE41-NEXT: pshufb %xmm4, %xmm0
-; SSE41-NEXT: movdqa {{.*#+}} xmm5 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; SSE41-NEXT: pandn %xmm5, %xmm0
-; SSE41-NEXT: psllw $5, %xmm0
-; SSE41-NEXT: movdqa %xmm0, %xmm4
-; SSE41-NEXT: paddb %xmm0, %xmm4
-; SSE41-NEXT: paddb %xmm3, %xmm3
-; SSE41-NEXT: movdqa %xmm3, %xmm6
-; SSE41-NEXT: psllw $4, %xmm6
-; SSE41-NEXT: pand {{.*}}(%rip), %xmm6
-; SSE41-NEXT: pblendvb %xmm0, %xmm6, %xmm3
-; SSE41-NEXT: movdqa %xmm3, %xmm6
-; SSE41-NEXT: psllw $2, %xmm6
-; SSE41-NEXT: pand {{.*}}(%rip), %xmm6
-; SSE41-NEXT: movdqa %xmm4, %xmm0
-; SSE41-NEXT: pblendvb %xmm0, %xmm6, %xmm3
-; SSE41-NEXT: movdqa %xmm3, %xmm6
-; SSE41-NEXT: paddb %xmm3, %xmm6
-; SSE41-NEXT: paddb %xmm4, %xmm4
-; SSE41-NEXT: movdqa %xmm4, %xmm0
-; SSE41-NEXT: pblendvb %xmm0, %xmm6, %xmm3
-; SSE41-NEXT: pand %xmm5, %xmm2
-; SSE41-NEXT: pmovzxbq {{.*#+}} xmm0 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
-; SSE41-NEXT: psrlw %xmm0, %xmm1
-; SSE41-NEXT: pcmpeqd %xmm2, %xmm2
-; SSE41-NEXT: psrlw %xmm0, %xmm2
-; SSE41-NEXT: pshufb {{.*#+}} xmm2 = xmm2[1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1]
-; SSE41-NEXT: pand %xmm1, %xmm2
-; SSE41-NEXT: por %xmm2, %xmm3
-; SSE41-NEXT: movdqa %xmm3, %xmm0
+; SSE41-NEXT: movdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; SSE41-NEXT: movdqa %xmm2, %xmm4
+; SSE41-NEXT: pand %xmm3, %xmm4
+; SSE41-NEXT: pmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
+; SSE41-NEXT: psrlw %xmm4, %xmm1
+; SSE41-NEXT: pcmpeqd %xmm5, %xmm5
+; SSE41-NEXT: pcmpeqd %xmm6, %xmm6
+; SSE41-NEXT: psrlw %xmm4, %xmm6
+; SSE41-NEXT: pshufb {{.*#+}} xmm6 = xmm6[1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1]
+; SSE41-NEXT: pand %xmm1, %xmm6
+; SSE41-NEXT: pandn %xmm3, %xmm2
+; SSE41-NEXT: pmovzxbq {{.*#+}} xmm1 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
+; SSE41-NEXT: paddb %xmm0, %xmm0
+; SSE41-NEXT: psllw %xmm1, %xmm0
+; SSE41-NEXT: psllw %xmm1, %xmm5
+; SSE41-NEXT: pxor %xmm1, %xmm1
+; SSE41-NEXT: pshufb %xmm1, %xmm5
+; SSE41-NEXT: pand %xmm5, %xmm0
+; SSE41-NEXT: por %xmm6, %xmm0
; SSE41-NEXT: retq
;
; AVX1-LABEL: splatvar_funnnel_v16i8:
; AVX1: # %bb.0:
-; AVX1-NEXT: vpxor %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpshufb %xmm3, %xmm2, %xmm3
-; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; AVX1-NEXT: vpandn %xmm4, %xmm3, %xmm3
-; AVX1-NEXT: vpsllw $5, %xmm3, %xmm3
-; AVX1-NEXT: vpaddb %xmm3, %xmm3, %xmm5
-; AVX1-NEXT: vpaddb %xmm0, %xmm0, %xmm0
-; AVX1-NEXT: vpsllw $4, %xmm0, %xmm6
-; AVX1-NEXT: vpand {{.*}}(%rip), %xmm6, %xmm6
-; AVX1-NEXT: vpblendvb %xmm3, %xmm6, %xmm0, %xmm0
-; AVX1-NEXT: vpsllw $2, %xmm0, %xmm3
-; AVX1-NEXT: vpand {{.*}}(%rip), %xmm3, %xmm3
-; AVX1-NEXT: vpblendvb %xmm5, %xmm3, %xmm0, %xmm0
-; AVX1-NEXT: vpaddb %xmm0, %xmm0, %xmm3
-; AVX1-NEXT: vpaddb %xmm5, %xmm5, %xmm5
-; AVX1-NEXT: vpblendvb %xmm5, %xmm3, %xmm0, %xmm0
-; AVX1-NEXT: vpand %xmm4, %xmm2, %xmm2
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX1-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX1-NEXT: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
+; AVX1-NEXT: vpsrlw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT: vpcmpeqd %xmm5, %xmm5, %xmm5
+; AVX1-NEXT: vpsrlw %xmm4, %xmm5, %xmm4
+; AVX1-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1]
+; AVX1-NEXT: vpand %xmm4, %xmm1, %xmm1
+; AVX1-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX1-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
-; AVX1-NEXT: vpsrlw %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpsrlw %xmm2, %xmm3, %xmm2
-; AVX1-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1]
-; AVX1-NEXT: vpand %xmm2, %xmm1, %xmm1
+; AVX1-NEXT: vpaddb %xmm0, %xmm0, %xmm0
+; AVX1-NEXT: vpsllw %xmm2, %xmm0, %xmm0
+; AVX1-NEXT: vpsllw %xmm2, %xmm5, %xmm2
+; AVX1-NEXT: vpxor %xmm3, %xmm3, %xmm3
+; AVX1-NEXT: vpshufb %xmm3, %xmm2, %xmm2
+; AVX1-NEXT: vpand %xmm2, %xmm0, %xmm0
; AVX1-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: splatvar_funnnel_v16i8:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastb %xmm2, %xmm3
-; AVX2-NEXT: vmovdqa {{.*#+}} xmm4 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; AVX2-NEXT: vpandn %xmm4, %xmm3, %xmm3
-; AVX2-NEXT: vpsllw $5, %xmm3, %xmm3
-; AVX2-NEXT: vpaddb %xmm3, %xmm3, %xmm5
+; AVX2-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX2-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX2-NEXT: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
; AVX2-NEXT: vpaddb %xmm0, %xmm0, %xmm0
-; AVX2-NEXT: vpsllw $4, %xmm0, %xmm6
-; AVX2-NEXT: vpand {{.*}}(%rip), %xmm6, %xmm6
-; AVX2-NEXT: vpblendvb %xmm3, %xmm6, %xmm0, %xmm0
-; AVX2-NEXT: vpsllw $2, %xmm0, %xmm3
-; AVX2-NEXT: vpand {{.*}}(%rip), %xmm3, %xmm3
-; AVX2-NEXT: vpblendvb %xmm5, %xmm3, %xmm0, %xmm0
-; AVX2-NEXT: vpaddb %xmm0, %xmm0, %xmm3
-; AVX2-NEXT: vpaddb %xmm5, %xmm5, %xmm5
-; AVX2-NEXT: vpblendvb %xmm5, %xmm3, %xmm0, %xmm0
-; AVX2-NEXT: vpand %xmm4, %xmm2, %xmm2
+; AVX2-NEXT: vpsllw %xmm4, %xmm0, %xmm0
+; AVX2-NEXT: vpcmpeqd %xmm5, %xmm5, %xmm5
+; AVX2-NEXT: vpsllw %xmm4, %xmm5, %xmm4
+; AVX2-NEXT: vpbroadcastb %xmm4, %xmm4
+; AVX2-NEXT: vpand %xmm4, %xmm0, %xmm0
+; AVX2-NEXT: vpand %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
; AVX2-NEXT: vpsrlw %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX2-NEXT: vpsrlw %xmm2, %xmm3, %xmm2
+; AVX2-NEXT: vpsrlw %xmm2, %xmm5, %xmm2
; AVX2-NEXT: vpsrlw $8, %xmm2, %xmm2
; AVX2-NEXT: vpbroadcastb %xmm2, %xmm2
; AVX2-NEXT: vpand %xmm2, %xmm1, %xmm1
@@ -1809,17 +1595,16 @@ define <16 x i8> @splatvar_funnnel_v16i8(<16 x i8> %x, <16 x i8> %y, <16 x i8> %
;
; AVX512F-LABEL: splatvar_funnnel_v16i8:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpbroadcastb %xmm2, %xmm3
-; AVX512F-NEXT: vmovdqa {{.*#+}} xmm4 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; AVX512F-NEXT: vpand %xmm4, %xmm2, %xmm2
-; AVX512F-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512F-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512F-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512F-NEXT: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
; AVX512F-NEXT: vpmovzxbd {{.*#+}} zmm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero,xmm1[2],zero,zero,zero,xmm1[3],zero,zero,zero,xmm1[4],zero,zero,zero,xmm1[5],zero,zero,zero,xmm1[6],zero,zero,zero,xmm1[7],zero,zero,zero,xmm1[8],zero,zero,zero,xmm1[9],zero,zero,zero,xmm1[10],zero,zero,zero,xmm1[11],zero,zero,zero,xmm1[12],zero,zero,zero,xmm1[13],zero,zero,zero,xmm1[14],zero,zero,zero,xmm1[15],zero,zero,zero
-; AVX512F-NEXT: vpsrld %xmm2, %zmm1, %zmm1
-; AVX512F-NEXT: vpandn %xmm4, %xmm3, %xmm2
-; AVX512F-NEXT: vpmovzxbd {{.*#+}} zmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero,xmm2[2],zero,zero,zero,xmm2[3],zero,zero,zero,xmm2[4],zero,zero,zero,xmm2[5],zero,zero,zero,xmm2[6],zero,zero,zero,xmm2[7],zero,zero,zero,xmm2[8],zero,zero,zero,xmm2[9],zero,zero,zero,xmm2[10],zero,zero,zero,xmm2[11],zero,zero,zero,xmm2[12],zero,zero,zero,xmm2[13],zero,zero,zero,xmm2[14],zero,zero,zero,xmm2[15],zero,zero,zero
+; AVX512F-NEXT: vpsrld %xmm4, %zmm1, %zmm1
+; AVX512F-NEXT: vpandn %xmm3, %xmm2, %xmm2
+; AVX512F-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
; AVX512F-NEXT: vpaddb %xmm0, %xmm0, %xmm0
; AVX512F-NEXT: vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
-; AVX512F-NEXT: vpsllvd %zmm2, %zmm0, %zmm0
+; AVX512F-NEXT: vpslld %xmm2, %zmm0, %zmm0
; AVX512F-NEXT: vpord %zmm1, %zmm0, %zmm0
; AVX512F-NEXT: vpmovdb %zmm0, %xmm0
; AVX512F-NEXT: vzeroupper
@@ -1827,17 +1612,16 @@ define <16 x i8> @splatvar_funnnel_v16i8(<16 x i8> %x, <16 x i8> %y, <16 x i8> %
;
; AVX512VL-LABEL: splatvar_funnnel_v16i8:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastb %xmm2, %xmm3
-; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm4 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; AVX512VL-NEXT: vpand %xmm4, %xmm2, %xmm2
-; AVX512VL-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512VL-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512VL-NEXT: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
; AVX512VL-NEXT: vpmovzxbd {{.*#+}} zmm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero,xmm1[2],zero,zero,zero,xmm1[3],zero,zero,zero,xmm1[4],zero,zero,zero,xmm1[5],zero,zero,zero,xmm1[6],zero,zero,zero,xmm1[7],zero,zero,zero,xmm1[8],zero,zero,zero,xmm1[9],zero,zero,zero,xmm1[10],zero,zero,zero,xmm1[11],zero,zero,zero,xmm1[12],zero,zero,zero,xmm1[13],zero,zero,zero,xmm1[14],zero,zero,zero,xmm1[15],zero,zero,zero
-; AVX512VL-NEXT: vpsrld %xmm2, %zmm1, %zmm1
-; AVX512VL-NEXT: vpandn %xmm4, %xmm3, %xmm2
-; AVX512VL-NEXT: vpmovzxbd {{.*#+}} zmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero,xmm2[2],zero,zero,zero,xmm2[3],zero,zero,zero,xmm2[4],zero,zero,zero,xmm2[5],zero,zero,zero,xmm2[6],zero,zero,zero,xmm2[7],zero,zero,zero,xmm2[8],zero,zero,zero,xmm2[9],zero,zero,zero,xmm2[10],zero,zero,zero,xmm2[11],zero,zero,zero,xmm2[12],zero,zero,zero,xmm2[13],zero,zero,zero,xmm2[14],zero,zero,zero,xmm2[15],zero,zero,zero
+; AVX512VL-NEXT: vpsrld %xmm4, %zmm1, %zmm1
+; AVX512VL-NEXT: vpandn %xmm3, %xmm2, %xmm2
+; AVX512VL-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
; AVX512VL-NEXT: vpaddb %xmm0, %xmm0, %xmm0
; AVX512VL-NEXT: vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
-; AVX512VL-NEXT: vpsllvd %zmm2, %zmm0, %zmm0
+; AVX512VL-NEXT: vpslld %xmm2, %zmm0, %zmm0
; AVX512VL-NEXT: vpord %zmm1, %zmm0, %zmm0
; AVX512VL-NEXT: vpmovdb %zmm0, %xmm0
; AVX512VL-NEXT: vzeroupper
@@ -1845,17 +1629,16 @@ define <16 x i8> @splatvar_funnnel_v16i8(<16 x i8> %x, <16 x i8> %y, <16 x i8> %
;
; AVX512BW-LABEL: splatvar_funnnel_v16i8:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpbroadcastb %xmm2, %xmm3
+; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512BW-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512BW-NEXT: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
+; 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: vpsrlw %xmm4, %ymm1, %ymm1
+; AVX512BW-NEXT: vpandn %xmm3, %xmm2, %xmm2
+; AVX512BW-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
; AVX512BW-NEXT: vpaddb %xmm0, %xmm0, %xmm0
; AVX512BW-NEXT: vpmovzxbw {{.*#+}} ymm0 = 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: vpandn %xmm4, %xmm3, %xmm3
-; AVX512BW-NEXT: vpmovzxbw {{.*#+}} ymm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero,xmm3[8],zero,xmm3[9],zero,xmm3[10],zero,xmm3[11],zero,xmm3[12],zero,xmm3[13],zero,xmm3[14],zero,xmm3[15],zero
-; AVX512BW-NEXT: vpsllvw %zmm3, %zmm0, %zmm0
-; AVX512BW-NEXT: vpand %xmm4, %xmm2, %xmm2
-; AVX512BW-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
-; 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: vpsrlw %xmm2, %ymm1, %ymm1
+; AVX512BW-NEXT: vpsllw %xmm2, %ymm0, %ymm0
; AVX512BW-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX512BW-NEXT: vpmovwb %zmm0, %ymm0
; AVX512BW-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0
@@ -1864,17 +1647,16 @@ define <16 x i8> @splatvar_funnnel_v16i8(<16 x i8> %x, <16 x i8> %y, <16 x i8> %
;
; AVX512VBMI2-LABEL: splatvar_funnnel_v16i8:
; AVX512VBMI2: # %bb.0:
-; AVX512VBMI2-NEXT: vpbroadcastb %xmm2, %xmm3
+; AVX512VBMI2-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512VBMI2-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512VBMI2-NEXT: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
+; 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: vpsrlw %xmm4, %ymm1, %ymm1
+; AVX512VBMI2-NEXT: vpandn %xmm3, %xmm2, %xmm2
+; AVX512VBMI2-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
; AVX512VBMI2-NEXT: vpaddb %xmm0, %xmm0, %xmm0
; AVX512VBMI2-NEXT: vpmovzxbw {{.*#+}} ymm0 = 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: vpandn %xmm4, %xmm3, %xmm3
-; AVX512VBMI2-NEXT: vpmovzxbw {{.*#+}} ymm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero,xmm3[8],zero,xmm3[9],zero,xmm3[10],zero,xmm3[11],zero,xmm3[12],zero,xmm3[13],zero,xmm3[14],zero,xmm3[15],zero
-; AVX512VBMI2-NEXT: vpsllvw %zmm3, %zmm0, %zmm0
-; AVX512VBMI2-NEXT: vpand %xmm4, %xmm2, %xmm2
-; AVX512VBMI2-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
-; 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: vpsrlw %xmm2, %ymm1, %ymm1
+; AVX512VBMI2-NEXT: vpsllw %xmm2, %ymm0, %ymm0
; AVX512VBMI2-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX512VBMI2-NEXT: vpmovwb %zmm0, %ymm0
; AVX512VBMI2-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0
@@ -1883,17 +1665,16 @@ 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, %xmm3
-; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm4 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; AVX512VLBW-NEXT: vpand %xmm4, %xmm2, %xmm2
-; AVX512VLBW-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
+; 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: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,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: vpsrlw %xmm2, %ymm1, %ymm1
-; AVX512VLBW-NEXT: vpandn %xmm4, %xmm3, %xmm2
-; AVX512VLBW-NEXT: vpmovzxbw {{.*#+}} ymm2 = 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: vpsrlw %xmm4, %ymm1, %ymm1
+; AVX512VLBW-NEXT: vpandn %xmm3, %xmm2, %xmm2
+; AVX512VLBW-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
; AVX512VLBW-NEXT: vpaddb %xmm0, %xmm0, %xmm0
; AVX512VLBW-NEXT: vpmovzxbw {{.*#+}} ymm0 = 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 %ymm2, %ymm0, %ymm0
+; AVX512VLBW-NEXT: vpsllw %xmm2, %ymm0, %ymm0
; AVX512VLBW-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX512VLBW-NEXT: vpmovwb %ymm0, %xmm0
; AVX512VLBW-NEXT: vzeroupper
@@ -1901,17 +1682,16 @@ 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, %xmm3
-; AVX512VLVBMI2-NEXT: vmovdqa {{.*#+}} xmm4 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; AVX512VLVBMI2-NEXT: vpand %xmm4, %xmm2, %xmm2
-; AVX512VLVBMI2-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
+; 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: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,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: vpsrlw %xmm2, %ymm1, %ymm1
-; AVX512VLVBMI2-NEXT: vpandn %xmm4, %xmm3, %xmm2
-; AVX512VLVBMI2-NEXT: vpmovzxbw {{.*#+}} ymm2 = 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: vpsrlw %xmm4, %ymm1, %ymm1
+; AVX512VLVBMI2-NEXT: vpandn %xmm3, %xmm2, %xmm2
+; AVX512VLVBMI2-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
; AVX512VLVBMI2-NEXT: vpaddb %xmm0, %xmm0, %xmm0
; AVX512VLVBMI2-NEXT: vpmovzxbw {{.*#+}} ymm0 = 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 %ymm2, %ymm0, %ymm0
+; AVX512VLVBMI2-NEXT: vpsllw %xmm2, %ymm0, %ymm0
; AVX512VLVBMI2-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX512VLVBMI2-NEXT: vpmovwb %ymm0, %xmm0
; AVX512VLVBMI2-NEXT: vzeroupper
@@ -1947,50 +1727,30 @@ define <16 x i8> @splatvar_funnnel_v16i8(<16 x i8> %x, <16 x i8> %y, <16 x i8> %
;
; X86-SSE2-LABEL: splatvar_funnnel_v16i8:
; X86-SSE2: # %bb.0:
-; X86-SSE2-NEXT: movdqa %xmm2, %xmm3
-; X86-SSE2-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3],xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
-; X86-SSE2-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,0,0,0,4,5,6,7]
-; X86-SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm3[0,0,0,0]
; X86-SSE2-NEXT: movdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; X86-SSE2-NEXT: movdqa %xmm2, %xmm4
; X86-SSE2-NEXT: pandn %xmm3, %xmm4
-; X86-SSE2-NEXT: psllw $5, %xmm4
-; X86-SSE2-NEXT: pxor %xmm5, %xmm5
-; X86-SSE2-NEXT: pxor %xmm6, %xmm6
-; X86-SSE2-NEXT: pcmpgtb %xmm4, %xmm6
+; X86-SSE2-NEXT: pslldq {{.*#+}} xmm4 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm4[0]
+; X86-SSE2-NEXT: psrldq {{.*#+}} xmm4 = xmm4[15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; X86-SSE2-NEXT: paddb %xmm0, %xmm0
-; X86-SSE2-NEXT: movdqa %xmm6, %xmm7
-; X86-SSE2-NEXT: pandn %xmm0, %xmm7
-; X86-SSE2-NEXT: psllw $4, %xmm0
-; X86-SSE2-NEXT: pand %xmm6, %xmm0
-; X86-SSE2-NEXT: pand {{\.LCPI.*}}, %xmm0
-; X86-SSE2-NEXT: por %xmm7, %xmm0
-; X86-SSE2-NEXT: paddb %xmm4, %xmm4
-; X86-SSE2-NEXT: pxor %xmm6, %xmm6
-; X86-SSE2-NEXT: pcmpgtb %xmm4, %xmm6
-; X86-SSE2-NEXT: movdqa %xmm6, %xmm7
-; X86-SSE2-NEXT: pandn %xmm0, %xmm7
-; X86-SSE2-NEXT: psllw $2, %xmm0
-; X86-SSE2-NEXT: pand %xmm6, %xmm0
-; X86-SSE2-NEXT: pand {{\.LCPI.*}}, %xmm0
-; X86-SSE2-NEXT: por %xmm7, %xmm0
-; X86-SSE2-NEXT: paddb %xmm4, %xmm4
-; X86-SSE2-NEXT: pcmpgtb %xmm4, %xmm5
-; X86-SSE2-NEXT: movdqa %xmm5, %xmm4
-; X86-SSE2-NEXT: pandn %xmm0, %xmm4
-; X86-SSE2-NEXT: paddb %xmm0, %xmm0
-; X86-SSE2-NEXT: pand %xmm5, %xmm0
+; X86-SSE2-NEXT: psllw %xmm4, %xmm0
+; X86-SSE2-NEXT: pcmpeqd %xmm5, %xmm5
+; X86-SSE2-NEXT: pcmpeqd %xmm6, %xmm6
+; X86-SSE2-NEXT: psllw %xmm4, %xmm6
+; X86-SSE2-NEXT: punpcklbw {{.*#+}} xmm6 = xmm6[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; X86-SSE2-NEXT: pshuflw {{.*#+}} xmm4 = xmm6[0,0,0,0,4,5,6,7]
+; X86-SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm4[0,0,0,0]
+; X86-SSE2-NEXT: pand %xmm4, %xmm0
; X86-SSE2-NEXT: pand %xmm3, %xmm2
; X86-SSE2-NEXT: pslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0]
; X86-SSE2-NEXT: psrldq {{.*#+}} xmm2 = xmm2[15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; X86-SSE2-NEXT: psrlw %xmm2, %xmm1
-; X86-SSE2-NEXT: pcmpeqd %xmm3, %xmm3
-; X86-SSE2-NEXT: psrlw %xmm2, %xmm3
-; X86-SSE2-NEXT: psrlw $8, %xmm3
-; X86-SSE2-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; X86-SSE2-NEXT: pshuflw {{.*#+}} xmm2 = xmm3[0,0,0,0,4,5,6,7]
+; X86-SSE2-NEXT: psrlw %xmm2, %xmm5
+; X86-SSE2-NEXT: psrlw $8, %xmm5
+; X86-SSE2-NEXT: punpcklbw {{.*#+}} xmm5 = xmm5[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; X86-SSE2-NEXT: pshuflw {{.*#+}} xmm2 = xmm5[0,0,0,0,4,5,6,7]
; X86-SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
; X86-SSE2-NEXT: pand %xmm1, %xmm2
-; X86-SSE2-NEXT: por %xmm4, %xmm2
; X86-SSE2-NEXT: por %xmm2, %xmm0
; X86-SSE2-NEXT: retl
%splat = shufflevector <16 x i8> %amt, <16 x i8> undef, <16 x i32> zeroinitializer
diff --git a/llvm/test/CodeGen/X86/vector-fshr-256.ll b/llvm/test/CodeGen/X86/vector-fshr-256.ll
index 01e9af278e7f..cae30cc6493e 100644
--- a/llvm/test/CodeGen/X86/vector-fshr-256.ll
+++ b/llvm/test/CodeGen/X86/vector-fshr-256.ll
@@ -810,74 +810,63 @@ define <32 x i8> @var_funnnel_v32i8(<32 x i8> %x, <32 x i8> %y, <32 x i8> %amt)
define <4 x i64> @splatvar_funnnel_v4i64(<4 x i64> %x, <4 x i64> %y, <4 x i64> %amt) nounwind {
; AVX1-LABEL: splatvar_funnnel_v4i64:
; AVX1: # %bb.0:
-; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm2[0,1,0,1]
-; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm3, %ymm3
-; AVX1-NEXT: vandnps {{.*}}(%rip), %ymm3, %ymm3
-; AVX1-NEXT: vextractf128 $1, %ymm3, %xmm4
-; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm5
-; AVX1-NEXT: vpsllq $1, %xmm5, %xmm5
-; AVX1-NEXT: vpsllq %xmm4, %xmm5, %xmm6
-; AVX1-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[2,3,2,3]
-; AVX1-NEXT: vpsllq %xmm4, %xmm5, %xmm4
-; AVX1-NEXT: vpblendw {{.*#+}} xmm4 = xmm6[0,1,2,3],xmm4[4,5,6,7]
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; AVX1-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm5
+; AVX1-NEXT: vpsrlq %xmm4, %xmm5, %xmm5
+; AVX1-NEXT: vpsrlq %xmm4, %xmm1, %xmm1
+; AVX1-NEXT: vinsertf128 $1, %xmm5, %ymm1, %ymm1
+; AVX1-NEXT: vpandn %xmm3, %xmm2, %xmm2
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
+; AVX1-NEXT: vpsllq $1, %xmm3, %xmm3
+; AVX1-NEXT: vpsllq %xmm2, %xmm3, %xmm3
; AVX1-NEXT: vpsllq $1, %xmm0, %xmm0
-; AVX1-NEXT: vpsllq %xmm3, %xmm0, %xmm5
-; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[2,3,2,3]
-; AVX1-NEXT: vpsllq %xmm3, %xmm0, %xmm0
-; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm5[0,1,2,3],xmm0[4,5,6,7]
-; AVX1-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm0
-; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm3
-; AVX1-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX1-NEXT: vpsrlq %xmm2, %xmm3, %xmm3
-; AVX1-NEXT: vpsrlq %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm1, %ymm1
+; AVX1-NEXT: vpsllq %xmm2, %xmm0, %xmm0
+; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; AVX1-NEXT: vorps %ymm1, %ymm0, %ymm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: splatvar_funnnel_v4i64:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastq %xmm2, %ymm3
-; AVX2-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX2-NEXT: vpsrlq %xmm2, %ymm1, %ymm1
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm2 = [63,63,63,63]
-; AVX2-NEXT: vpandn %ymm2, %ymm3, %ymm2
+; AVX2-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; AVX2-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX2-NEXT: vpsrlq %xmm4, %ymm1, %ymm1
+; AVX2-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vpsllq $1, %ymm0, %ymm0
-; AVX2-NEXT: vpsllvq %ymm2, %ymm0, %ymm0
+; AVX2-NEXT: vpsllq %xmm2, %ymm0, %ymm0
; AVX2-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX2-NEXT: retq
;
; AVX512F-LABEL: splatvar_funnnel_v4i64:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpbroadcastq %xmm2, %ymm3
-; AVX512F-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX512F-NEXT: vpsrlq %xmm2, %ymm1, %ymm1
-; AVX512F-NEXT: vpbroadcastq {{.*#+}} ymm2 = [63,63,63,63]
-; AVX512F-NEXT: vpandn %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; AVX512F-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512F-NEXT: vpsrlq %xmm4, %ymm1, %ymm1
+; AVX512F-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512F-NEXT: vpsllq $1, %ymm0, %ymm0
-; AVX512F-NEXT: vpsllvq %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT: vpsllq %xmm2, %ymm0, %ymm0
; AVX512F-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: splatvar_funnnel_v4i64:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastq %xmm2, %ymm3
-; AVX512VL-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX512VL-NEXT: vpsrlq %xmm2, %ymm1, %ymm1
+; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; AVX512VL-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512VL-NEXT: vpsrlq %xmm4, %ymm1, %ymm1
+; AVX512VL-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512VL-NEXT: vpsllq $1, %ymm0, %ymm0
-; AVX512VL-NEXT: vpandnq {{.*}}(%rip){1to4}, %ymm3, %ymm2
-; AVX512VL-NEXT: vpsllvq %ymm2, %ymm0, %ymm0
+; AVX512VL-NEXT: vpsllq %xmm2, %ymm0, %ymm0
; AVX512VL-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: splatvar_funnnel_v4i64:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpbroadcastq %xmm2, %ymm3
-; AVX512BW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX512BW-NEXT: vpsrlq %xmm2, %ymm1, %ymm1
-; AVX512BW-NEXT: vpbroadcastq {{.*#+}} ymm2 = [63,63,63,63]
-; AVX512BW-NEXT: vpandn %ymm2, %ymm3, %ymm2
+; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; AVX512BW-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512BW-NEXT: vpsrlq %xmm4, %ymm1, %ymm1
+; AVX512BW-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512BW-NEXT: vpsllq $1, %ymm0, %ymm0
-; AVX512BW-NEXT: vpsllvq %ymm2, %ymm0, %ymm0
+; AVX512BW-NEXT: vpsllq %xmm2, %ymm0, %ymm0
; AVX512BW-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX512BW-NEXT: retq
;
@@ -892,12 +881,12 @@ define <4 x i64> @splatvar_funnnel_v4i64(<4 x i64> %x, <4 x i64> %y, <4 x i64> %
;
; AVX512VLBW-LABEL: splatvar_funnnel_v4i64:
; AVX512VLBW: # %bb.0:
-; AVX512VLBW-NEXT: vpbroadcastq %xmm2, %ymm3
-; AVX512VLBW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX512VLBW-NEXT: vpsrlq %xmm2, %ymm1, %ymm1
+; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; AVX512VLBW-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512VLBW-NEXT: vpsrlq %xmm4, %ymm1, %ymm1
+; AVX512VLBW-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512VLBW-NEXT: vpsllq $1, %ymm0, %ymm0
-; AVX512VLBW-NEXT: vpandnq {{.*}}(%rip){1to4}, %ymm3, %ymm2
-; AVX512VLBW-NEXT: vpsllvq %ymm2, %ymm0, %ymm0
+; AVX512VLBW-NEXT: vpsllq %xmm2, %ymm0, %ymm0
; AVX512VLBW-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX512VLBW-NEXT: retq
;
@@ -910,33 +899,30 @@ define <4 x i64> @splatvar_funnnel_v4i64(<4 x i64> %x, <4 x i64> %y, <4 x i64> %
;
; XOPAVX1-LABEL: splatvar_funnnel_v4i64:
; XOPAVX1: # %bb.0:
-; XOPAVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm2[0,1,0,1]
-; XOPAVX1-NEXT: vinsertf128 $1, %xmm3, %ymm3, %ymm3
-; XOPAVX1-NEXT: vextractf128 $1, %ymm1, %xmm4
-; XOPAVX1-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
-; XOPAVX1-NEXT: vpsrlq %xmm2, %xmm4, %xmm4
-; XOPAVX1-NEXT: vpsrlq %xmm2, %xmm1, %xmm1
-; XOPAVX1-NEXT: vinsertf128 $1, %xmm4, %ymm1, %ymm1
-; XOPAVX1-NEXT: vandnps {{.*}}(%rip), %ymm3, %ymm2
-; XOPAVX1-NEXT: vextractf128 $1, %ymm2, %xmm3
-; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm4
-; XOPAVX1-NEXT: vpsllq $1, %xmm4, %xmm4
-; XOPAVX1-NEXT: vpshlq %xmm3, %xmm4, %xmm3
+; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; XOPAVX1-NEXT: vpand %xmm3, %xmm2, %xmm4
+; XOPAVX1-NEXT: vextractf128 $1, %ymm1, %xmm5
+; XOPAVX1-NEXT: vpsrlq %xmm4, %xmm5, %xmm5
+; XOPAVX1-NEXT: vpsrlq %xmm4, %xmm1, %xmm1
+; XOPAVX1-NEXT: vinsertf128 $1, %xmm5, %ymm1, %ymm1
+; XOPAVX1-NEXT: vpandn %xmm3, %xmm2, %xmm2
+; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
+; XOPAVX1-NEXT: vpsllq $1, %xmm3, %xmm3
+; XOPAVX1-NEXT: vpsllq %xmm2, %xmm3, %xmm3
; XOPAVX1-NEXT: vpsllq $1, %xmm0, %xmm0
-; XOPAVX1-NEXT: vpshlq %xmm2, %xmm0, %xmm0
+; XOPAVX1-NEXT: vpsllq %xmm2, %xmm0, %xmm0
; XOPAVX1-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; XOPAVX1-NEXT: vorps %ymm1, %ymm0, %ymm0
; XOPAVX1-NEXT: retq
;
; XOPAVX2-LABEL: splatvar_funnnel_v4i64:
; XOPAVX2: # %bb.0:
-; XOPAVX2-NEXT: vpbroadcastq %xmm2, %ymm3
-; XOPAVX2-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
-; XOPAVX2-NEXT: vpsrlq %xmm2, %ymm1, %ymm1
-; XOPAVX2-NEXT: vpbroadcastq {{.*#+}} ymm2 = [63,63,63,63]
-; XOPAVX2-NEXT: vpandn %ymm2, %ymm3, %ymm2
+; XOPAVX2-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; XOPAVX2-NEXT: vpand %xmm3, %xmm2, %xmm4
+; XOPAVX2-NEXT: vpsrlq %xmm4, %ymm1, %ymm1
+; XOPAVX2-NEXT: vpandn %xmm3, %xmm2, %xmm2
; XOPAVX2-NEXT: vpsllq $1, %ymm0, %ymm0
-; XOPAVX2-NEXT: vpsllvq %ymm2, %ymm0, %ymm0
+; XOPAVX2-NEXT: vpsllq %xmm2, %ymm0, %ymm0
; XOPAVX2-NEXT: vpor %ymm1, %ymm0, %ymm0
; XOPAVX2-NEXT: retq
%splat = shufflevector <4 x i64> %amt, <4 x i64> undef, <4 x i32> zeroinitializer
@@ -947,81 +933,78 @@ define <4 x i64> @splatvar_funnnel_v4i64(<4 x i64> %x, <4 x i64> %y, <4 x i64> %
define <8 x i32> @splatvar_funnnel_v8i32(<8 x i32> %x, <8 x i32> %y, <8 x i32> %amt) nounwind {
; AVX1-LABEL: splatvar_funnnel_v8i32:
; AVX1: # %bb.0:
-; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm2[0,0,0,0]
-; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm3, %ymm3
-; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm4
-; AVX1-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [31,31,31,31]
+; AVX1-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX1-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
+; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm5
+; AVX1-NEXT: vpsrld %xmm4, %xmm5, %xmm5
+; AVX1-NEXT: vpsrld %xmm4, %xmm1, %xmm1
+; AVX1-NEXT: vinsertf128 $1, %xmm5, %ymm1, %ymm1
+; AVX1-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX1-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
-; AVX1-NEXT: vpsrld %xmm2, %xmm4, %xmm4
-; AVX1-NEXT: vpsrld %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vinsertf128 $1, %xmm4, %ymm1, %ymm1
-; AVX1-NEXT: vandnps {{.*}}(%rip), %ymm3, %ymm2
-; AVX1-NEXT: vextractf128 $1, %ymm2, %xmm3
-; AVX1-NEXT: vpslld $23, %xmm3, %xmm3
-; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [1065353216,1065353216,1065353216,1065353216]
-; AVX1-NEXT: vpaddd %xmm4, %xmm3, %xmm3
-; AVX1-NEXT: vcvttps2dq %xmm3, %xmm3
-; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm5
-; AVX1-NEXT: vpslld $1, %xmm5, %xmm5
-; AVX1-NEXT: vpmulld %xmm3, %xmm5, %xmm3
-; AVX1-NEXT: vpslld $23, %xmm2, %xmm2
-; AVX1-NEXT: vpaddd %xmm4, %xmm2, %xmm2
-; AVX1-NEXT: vcvttps2dq %xmm2, %xmm2
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
+; AVX1-NEXT: vpslld $1, %xmm3, %xmm3
+; AVX1-NEXT: vpslld %xmm2, %xmm3, %xmm3
; AVX1-NEXT: vpslld $1, %xmm0, %xmm0
-; AVX1-NEXT: vpmulld %xmm2, %xmm0, %xmm0
+; AVX1-NEXT: vpslld %xmm2, %xmm0, %xmm0
; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; AVX1-NEXT: vorps %ymm1, %ymm0, %ymm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: splatvar_funnnel_v8i32:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastd %xmm2, %ymm2
-; AVX2-NEXT: vpbroadcastd {{.*#+}} ymm3 = [31,31,31,31,31,31,31,31]
-; AVX2-NEXT: vpandn %ymm3, %ymm2, %ymm4
-; AVX2-NEXT: vpslld $1, %ymm0, %ymm0
-; AVX2-NEXT: vpsllvd %ymm4, %ymm0, %ymm0
-; AVX2-NEXT: vpand %xmm3, %xmm2, %xmm2
+; AVX2-NEXT: vpbroadcastd %xmm2, %xmm2
+; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31]
+; AVX2-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX2-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
+; AVX2-NEXT: vpsrld %xmm4, %ymm1, %ymm1
+; AVX2-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
-; AVX2-NEXT: vpsrld %xmm2, %ymm1, %ymm1
+; AVX2-NEXT: vpslld $1, %ymm0, %ymm0
+; AVX2-NEXT: vpslld %xmm2, %ymm0, %ymm0
; AVX2-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX2-NEXT: retq
;
; AVX512F-LABEL: splatvar_funnnel_v8i32:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpbroadcastd %xmm2, %ymm2
-; AVX512F-NEXT: vpbroadcastd {{.*#+}} ymm3 = [31,31,31,31,31,31,31,31]
-; AVX512F-NEXT: vpandn %ymm3, %ymm2, %ymm4
-; AVX512F-NEXT: vpslld $1, %ymm0, %ymm0
-; AVX512F-NEXT: vpsllvd %ymm4, %ymm0, %ymm0
-; AVX512F-NEXT: vpand %xmm3, %xmm2, %xmm2
+; AVX512F-NEXT: vpbroadcastd %xmm2, %xmm2
+; AVX512F-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31]
+; AVX512F-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512F-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
+; AVX512F-NEXT: vpsrld %xmm4, %ymm1, %ymm1
+; AVX512F-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512F-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
-; AVX512F-NEXT: vpsrld %xmm2, %ymm1, %ymm1
+; AVX512F-NEXT: vpslld $1, %ymm0, %ymm0
+; AVX512F-NEXT: vpslld %xmm2, %ymm0, %ymm0
; AVX512F-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: splatvar_funnnel_v8i32:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastd %xmm2, %ymm2
-; AVX512VL-NEXT: vpbroadcastd {{.*#+}} ymm3 = [31,31,31,31,31,31,31,31]
-; AVX512VL-NEXT: vpandn %ymm3, %ymm2, %ymm4
-; AVX512VL-NEXT: vpslld $1, %ymm0, %ymm0
-; AVX512VL-NEXT: vpsllvd %ymm4, %ymm0, %ymm0
-; AVX512VL-NEXT: vpand %xmm3, %xmm2, %xmm2
+; AVX512VL-NEXT: vpbroadcastd %xmm2, %xmm2
+; AVX512VL-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31]
+; AVX512VL-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512VL-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
+; AVX512VL-NEXT: vpsrld %xmm4, %ymm1, %ymm1
+; AVX512VL-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512VL-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
-; AVX512VL-NEXT: vpsrld %xmm2, %ymm1, %ymm1
+; AVX512VL-NEXT: vpslld $1, %ymm0, %ymm0
+; AVX512VL-NEXT: vpslld %xmm2, %ymm0, %ymm0
; AVX512VL-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: splatvar_funnnel_v8i32:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpbroadcastd %xmm2, %ymm2
-; AVX512BW-NEXT: vpbroadcastd {{.*#+}} ymm3 = [31,31,31,31,31,31,31,31]
-; AVX512BW-NEXT: vpandn %ymm3, %ymm2, %ymm4
-; AVX512BW-NEXT: vpslld $1, %ymm0, %ymm0
-; AVX512BW-NEXT: vpsllvd %ymm4, %ymm0, %ymm0
-; AVX512BW-NEXT: vpand %xmm3, %xmm2, %xmm2
+; AVX512BW-NEXT: vpbroadcastd %xmm2, %xmm2
+; AVX512BW-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31]
+; AVX512BW-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512BW-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
+; AVX512BW-NEXT: vpsrld %xmm4, %ymm1, %ymm1
+; AVX512BW-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512BW-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
-; AVX512BW-NEXT: vpsrld %xmm2, %ymm1, %ymm1
+; AVX512BW-NEXT: vpslld $1, %ymm0, %ymm0
+; AVX512BW-NEXT: vpslld %xmm2, %ymm0, %ymm0
; AVX512BW-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX512BW-NEXT: retq
;
@@ -1036,14 +1019,15 @@ define <8 x i32> @splatvar_funnnel_v8i32(<8 x i32> %x, <8 x i32> %y, <8 x i32> %
;
; AVX512VLBW-LABEL: splatvar_funnnel_v8i32:
; AVX512VLBW: # %bb.0:
-; AVX512VLBW-NEXT: vpbroadcastd %xmm2, %ymm2
-; AVX512VLBW-NEXT: vpbroadcastd {{.*#+}} ymm3 = [31,31,31,31,31,31,31,31]
-; AVX512VLBW-NEXT: vpandn %ymm3, %ymm2, %ymm4
-; AVX512VLBW-NEXT: vpslld $1, %ymm0, %ymm0
-; AVX512VLBW-NEXT: vpsllvd %ymm4, %ymm0, %ymm0
-; AVX512VLBW-NEXT: vpand %xmm3, %xmm2, %xmm2
+; AVX512VLBW-NEXT: vpbroadcastd %xmm2, %xmm2
+; AVX512VLBW-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31]
+; AVX512VLBW-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512VLBW-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
+; AVX512VLBW-NEXT: vpsrld %xmm4, %ymm1, %ymm1
+; AVX512VLBW-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512VLBW-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
-; AVX512VLBW-NEXT: vpsrld %xmm2, %ymm1, %ymm1
+; AVX512VLBW-NEXT: vpslld $1, %ymm0, %ymm0
+; AVX512VLBW-NEXT: vpslld %xmm2, %ymm0, %ymm0
; AVX512VLBW-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX512VLBW-NEXT: retq
;
@@ -1056,35 +1040,36 @@ define <8 x i32> @splatvar_funnnel_v8i32(<8 x i32> %x, <8 x i32> %y, <8 x i32> %
;
; XOPAVX1-LABEL: splatvar_funnnel_v8i32:
; XOPAVX1: # %bb.0:
-; XOPAVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm2[0,0,0,0]
-; XOPAVX1-NEXT: vinsertf128 $1, %xmm3, %ymm3, %ymm3
-; XOPAVX1-NEXT: vextractf128 $1, %ymm1, %xmm4
-; XOPAVX1-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; XOPAVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
+; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [31,31,31,31]
+; XOPAVX1-NEXT: vpand %xmm3, %xmm2, %xmm4
+; XOPAVX1-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
+; XOPAVX1-NEXT: vextractf128 $1, %ymm1, %xmm5
+; XOPAVX1-NEXT: vpsrld %xmm4, %xmm5, %xmm5
+; XOPAVX1-NEXT: vpsrld %xmm4, %xmm1, %xmm1
+; XOPAVX1-NEXT: vinsertf128 $1, %xmm5, %ymm1, %ymm1
+; XOPAVX1-NEXT: vpandn %xmm3, %xmm2, %xmm2
; XOPAVX1-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
-; XOPAVX1-NEXT: vpsrld %xmm2, %xmm4, %xmm4
-; XOPAVX1-NEXT: vpsrld %xmm2, %xmm1, %xmm1
-; XOPAVX1-NEXT: vinsertf128 $1, %xmm4, %ymm1, %ymm1
-; XOPAVX1-NEXT: vandnps {{.*}}(%rip), %ymm3, %ymm2
-; XOPAVX1-NEXT: vextractf128 $1, %ymm2, %xmm3
-; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm4
-; XOPAVX1-NEXT: vpslld $1, %xmm4, %xmm4
-; XOPAVX1-NEXT: vpshld %xmm3, %xmm4, %xmm3
+; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
+; XOPAVX1-NEXT: vpslld $1, %xmm3, %xmm3
+; XOPAVX1-NEXT: vpslld %xmm2, %xmm3, %xmm3
; XOPAVX1-NEXT: vpslld $1, %xmm0, %xmm0
-; XOPAVX1-NEXT: vpshld %xmm2, %xmm0, %xmm0
+; XOPAVX1-NEXT: vpslld %xmm2, %xmm0, %xmm0
; XOPAVX1-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; XOPAVX1-NEXT: vorps %ymm1, %ymm0, %ymm0
; XOPAVX1-NEXT: retq
;
; XOPAVX2-LABEL: splatvar_funnnel_v8i32:
; XOPAVX2: # %bb.0:
-; XOPAVX2-NEXT: vpbroadcastd %xmm2, %ymm2
-; XOPAVX2-NEXT: vpbroadcastd {{.*#+}} ymm3 = [31,31,31,31,31,31,31,31]
-; XOPAVX2-NEXT: vpandn %ymm3, %ymm2, %ymm4
-; XOPAVX2-NEXT: vpslld $1, %ymm0, %ymm0
-; XOPAVX2-NEXT: vpsllvd %ymm4, %ymm0, %ymm0
-; XOPAVX2-NEXT: vpand %xmm3, %xmm2, %xmm2
+; XOPAVX2-NEXT: vpbroadcastd %xmm2, %xmm2
+; XOPAVX2-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31]
+; XOPAVX2-NEXT: vpand %xmm3, %xmm2, %xmm4
+; XOPAVX2-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
+; XOPAVX2-NEXT: vpsrld %xmm4, %ymm1, %ymm1
+; XOPAVX2-NEXT: vpandn %xmm3, %xmm2, %xmm2
; XOPAVX2-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
-; XOPAVX2-NEXT: vpsrld %xmm2, %ymm1, %ymm1
+; XOPAVX2-NEXT: vpslld $1, %ymm0, %ymm0
+; XOPAVX2-NEXT: vpslld %xmm2, %ymm0, %ymm0
; XOPAVX2-NEXT: vpor %ymm1, %ymm0, %ymm0
; XOPAVX2-NEXT: retq
%splat = shufflevector <8 x i32> %amt, <8 x i32> undef, <8 x i32> zeroinitializer
@@ -1096,104 +1081,78 @@ define <16 x i16> @splatvar_funnnel_v16i16(<16 x i16> %x, <16 x i16> %y, <16 x i
; AVX1-LABEL: splatvar_funnnel_v16i16:
; AVX1: # %bb.0:
; AVX1-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,0,0,0,4,5,6,7]
-; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm2[0,0,0,0]
-; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm3, %ymm3
-; AVX1-NEXT: vandnps {{.*}}(%rip), %ymm3, %ymm3
-; AVX1-NEXT: vextractf128 $1, %ymm3, %xmm4
-; AVX1-NEXT: vpunpckhwd {{.*#+}} xmm5 = xmm4[4,4,5,5,6,6,7,7]
-; AVX1-NEXT: vpslld $23, %xmm5, %xmm5
-; AVX1-NEXT: vmovdqa {{.*#+}} xmm6 = [1065353216,1065353216,1065353216,1065353216]
-; AVX1-NEXT: vpaddd %xmm6, %xmm5, %xmm5
-; AVX1-NEXT: vcvttps2dq %xmm5, %xmm5
-; AVX1-NEXT: vpmovzxwd {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero
-; AVX1-NEXT: vpslld $23, %xmm4, %xmm4
-; AVX1-NEXT: vpaddd %xmm6, %xmm4, %xmm4
-; AVX1-NEXT: vcvttps2dq %xmm4, %xmm4
-; AVX1-NEXT: vpackusdw %xmm5, %xmm4, %xmm4
-; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm5
-; AVX1-NEXT: vpsllw $1, %xmm5, %xmm5
-; AVX1-NEXT: vpmullw %xmm4, %xmm5, %xmm4
-; AVX1-NEXT: vpunpckhwd {{.*#+}} xmm5 = xmm3[4,4,5,5,6,6,7,7]
-; AVX1-NEXT: vpslld $23, %xmm5, %xmm5
-; AVX1-NEXT: vpaddd %xmm6, %xmm5, %xmm5
-; AVX1-NEXT: vcvttps2dq %xmm5, %xmm5
-; AVX1-NEXT: vpmovzxwd {{.*#+}} xmm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero
-; AVX1-NEXT: vpslld $23, %xmm3, %xmm3
-; AVX1-NEXT: vpaddd %xmm6, %xmm3, %xmm3
-; AVX1-NEXT: vcvttps2dq %xmm3, %xmm3
-; AVX1-NEXT: vpackusdw %xmm5, %xmm3, %xmm3
-; AVX1-NEXT: vpsllw $1, %xmm0, %xmm0
-; AVX1-NEXT: vpmullw %xmm3, %xmm0, %xmm0
-; AVX1-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm0
-; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm3
-; AVX1-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; AVX1-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX1-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
+; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm5
+; AVX1-NEXT: vpsrlw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT: vpsrlw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT: vinsertf128 $1, %xmm5, %ymm1, %ymm1
+; AVX1-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX1-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
-; AVX1-NEXT: vpsrlw %xmm2, %xmm3, %xmm3
-; AVX1-NEXT: vpsrlw %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm1, %ymm1
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
+; AVX1-NEXT: vpsllw $1, %xmm3, %xmm3
+; AVX1-NEXT: vpsllw %xmm2, %xmm3, %xmm3
+; AVX1-NEXT: vpsllw $1, %xmm0, %xmm0
+; AVX1-NEXT: vpsllw %xmm2, %xmm0, %xmm0
+; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; AVX1-NEXT: vorps %ymm1, %ymm0, %ymm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: splatvar_funnnel_v16i16:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastw %xmm2, %ymm2
-; AVX2-NEXT: vpsllw $1, %ymm0, %ymm0
-; AVX2-NEXT: vpxor %xmm3, %xmm3, %xmm3
-; AVX2-NEXT: vpunpckhwd {{.*#+}} ymm4 = ymm3[4],ymm0[4],ymm3[5],ymm0[5],ymm3[6],ymm0[6],ymm3[7],ymm0[7],ymm3[12],ymm0[12],ymm3[13],ymm0[13],ymm3[14],ymm0[14],ymm3[15],ymm0[15]
-; AVX2-NEXT: vpandn {{.*}}(%rip), %ymm2, %ymm5
-; AVX2-NEXT: vpunpckhwd {{.*#+}} ymm6 = ymm5[4],ymm3[4],ymm5[5],ymm3[5],ymm5[6],ymm3[6],ymm5[7],ymm3[7],ymm5[12],ymm3[12],ymm5[13],ymm3[13],ymm5[14],ymm3[14],ymm5[15],ymm3[15]
-; AVX2-NEXT: vpsllvd %ymm6, %ymm4, %ymm4
-; AVX2-NEXT: vpsrld $16, %ymm4, %ymm4
-; AVX2-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[1],ymm0[1],ymm3[2],ymm0[2],ymm3[3],ymm0[3],ymm3[8],ymm0[8],ymm3[9],ymm0[9],ymm3[10],ymm0[10],ymm3[11],ymm0[11]
-; AVX2-NEXT: vpunpcklwd {{.*#+}} ymm3 = ymm5[0],ymm3[0],ymm5[1],ymm3[1],ymm5[2],ymm3[2],ymm5[3],ymm3[3],ymm5[8],ymm3[8],ymm5[9],ymm3[9],ymm5[10],ymm3[10],ymm5[11],ymm3[11]
-; AVX2-NEXT: vpsllvd %ymm3, %ymm0, %ymm0
-; AVX2-NEXT: vpsrld $16, %ymm0, %ymm0
-; AVX2-NEXT: vpackusdw %ymm4, %ymm0, %ymm0
-; AVX2-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; AVX2-NEXT: vpbroadcastw %xmm2, %xmm2
+; AVX2-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; AVX2-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX2-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
+; AVX2-NEXT: vpsrlw %xmm4, %ymm1, %ymm1
+; AVX2-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
-; AVX2-NEXT: vpsrlw %xmm2, %ymm1, %ymm1
+; AVX2-NEXT: vpsllw $1, %ymm0, %ymm0
+; AVX2-NEXT: vpsllw %xmm2, %ymm0, %ymm0
; AVX2-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX2-NEXT: retq
;
; AVX512F-LABEL: splatvar_funnnel_v16i16:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpbroadcastw %xmm2, %ymm2
-; AVX512F-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm3
-; AVX512F-NEXT: vpmovzxwq {{.*#+}} xmm3 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero
-; AVX512F-NEXT: vpsrlw %xmm3, %ymm1, %ymm1
+; AVX512F-NEXT: vpbroadcastw %xmm2, %xmm2
+; AVX512F-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; AVX512F-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512F-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
+; AVX512F-NEXT: vpsrlw %xmm4, %ymm1, %ymm1
+; AVX512F-NEXT: vpandn %xmm3, %xmm2, %xmm2
+; AVX512F-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
; AVX512F-NEXT: vpsllw $1, %ymm0, %ymm0
-; AVX512F-NEXT: vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
-; AVX512F-NEXT: vpandn {{.*}}(%rip), %ymm2, %ymm2
-; AVX512F-NEXT: vpmovzxwd {{.*#+}} zmm2 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero
-; AVX512F-NEXT: vpsllvd %zmm2, %zmm0, %zmm0
-; AVX512F-NEXT: vpmovdw %zmm0, %ymm0
+; AVX512F-NEXT: vpsllw %xmm2, %ymm0, %ymm0
; AVX512F-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: splatvar_funnnel_v16i16:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastw %xmm2, %ymm2
-; AVX512VL-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm3
-; AVX512VL-NEXT: vpmovzxwq {{.*#+}} xmm3 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero
-; AVX512VL-NEXT: vpsrlw %xmm3, %ymm1, %ymm1
+; AVX512VL-NEXT: vpbroadcastw %xmm2, %xmm2
+; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; AVX512VL-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512VL-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
+; AVX512VL-NEXT: vpsrlw %xmm4, %ymm1, %ymm1
+; AVX512VL-NEXT: vpandn %xmm3, %xmm2, %xmm2
+; AVX512VL-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
; AVX512VL-NEXT: vpsllw $1, %ymm0, %ymm0
-; AVX512VL-NEXT: vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
-; AVX512VL-NEXT: vpandn {{.*}}(%rip), %ymm2, %ymm2
-; AVX512VL-NEXT: vpmovzxwd {{.*#+}} zmm2 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero
-; AVX512VL-NEXT: vpsllvd %zmm2, %zmm0, %zmm0
-; AVX512VL-NEXT: vpmovdw %zmm0, %ymm0
+; AVX512VL-NEXT: vpsllw %xmm2, %ymm0, %ymm0
; AVX512VL-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: splatvar_funnnel_v16i16:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpbroadcastw %xmm2, %ymm2
-; AVX512BW-NEXT: vpsllw $1, %ymm0, %ymm0
-; AVX512BW-NEXT: vpandn {{.*}}(%rip), %ymm2, %ymm3
-; AVX512BW-NEXT: vpsllvw %zmm3, %zmm0, %zmm0
-; AVX512BW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; 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 {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
+; AVX512BW-NEXT: vpsrlw %xmm4, %ymm1, %ymm1
+; AVX512BW-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512BW-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
-; AVX512BW-NEXT: vpsrlw %xmm2, %ymm1, %ymm1
+; AVX512BW-NEXT: vpsllw $1, %ymm0, %ymm0
+; AVX512BW-NEXT: vpsllw %xmm2, %ymm0, %ymm0
; AVX512BW-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX512BW-NEXT: retq
;
@@ -1208,13 +1167,15 @@ define <16 x i16> @splatvar_funnnel_v16i16(<16 x i16> %x, <16 x i16> %y, <16 x i
;
; AVX512VLBW-LABEL: splatvar_funnnel_v16i16:
; AVX512VLBW: # %bb.0:
-; AVX512VLBW-NEXT: vpbroadcastw %xmm2, %ymm2
-; AVX512VLBW-NEXT: vpsllw $1, %ymm0, %ymm0
-; AVX512VLBW-NEXT: vpandn {{.*}}(%rip), %ymm2, %ymm3
-; AVX512VLBW-NEXT: vpsllvw %ymm3, %ymm0, %ymm0
-; AVX512VLBW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; 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 {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
+; AVX512VLBW-NEXT: vpsrlw %xmm4, %ymm1, %ymm1
+; AVX512VLBW-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512VLBW-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
-; AVX512VLBW-NEXT: vpsrlw %xmm2, %ymm1, %ymm1
+; AVX512VLBW-NEXT: vpsllw $1, %ymm0, %ymm0
+; AVX512VLBW-NEXT: vpsllw %xmm2, %ymm0, %ymm0
; AVX512VLBW-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX512VLBW-NEXT: retq
;
@@ -1228,38 +1189,36 @@ define <16 x i16> @splatvar_funnnel_v16i16(<16 x i16> %x, <16 x i16> %y, <16 x i
; XOPAVX1-LABEL: splatvar_funnnel_v16i16:
; XOPAVX1: # %bb.0:
; XOPAVX1-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,0,0,0,4,5,6,7]
-; XOPAVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm2[0,0,0,0]
-; XOPAVX1-NEXT: vinsertf128 $1, %xmm3, %ymm3, %ymm3
-; XOPAVX1-NEXT: vextractf128 $1, %ymm1, %xmm4
-; XOPAVX1-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; XOPAVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
+; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; XOPAVX1-NEXT: vpand %xmm3, %xmm2, %xmm4
+; XOPAVX1-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
+; XOPAVX1-NEXT: vextractf128 $1, %ymm1, %xmm5
+; XOPAVX1-NEXT: vpsrlw %xmm4, %xmm5, %xmm5
+; XOPAVX1-NEXT: vpsrlw %xmm4, %xmm1, %xmm1
+; XOPAVX1-NEXT: vinsertf128 $1, %xmm5, %ymm1, %ymm1
+; XOPAVX1-NEXT: vpandn %xmm3, %xmm2, %xmm2
; XOPAVX1-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
-; XOPAVX1-NEXT: vpsrlw %xmm2, %xmm4, %xmm4
-; XOPAVX1-NEXT: vpsrlw %xmm2, %xmm1, %xmm1
-; XOPAVX1-NEXT: vinsertf128 $1, %xmm4, %ymm1, %ymm1
-; XOPAVX1-NEXT: vandnps {{.*}}(%rip), %ymm3, %ymm2
-; XOPAVX1-NEXT: vextractf128 $1, %ymm2, %xmm3
-; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm4
-; XOPAVX1-NEXT: vpsllw $1, %xmm4, %xmm4
-; XOPAVX1-NEXT: vpshlw %xmm3, %xmm4, %xmm3
+; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
+; XOPAVX1-NEXT: vpsllw $1, %xmm3, %xmm3
+; XOPAVX1-NEXT: vpsllw %xmm2, %xmm3, %xmm3
; XOPAVX1-NEXT: vpsllw $1, %xmm0, %xmm0
-; XOPAVX1-NEXT: vpshlw %xmm2, %xmm0, %xmm0
+; XOPAVX1-NEXT: vpsllw %xmm2, %xmm0, %xmm0
; XOPAVX1-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; XOPAVX1-NEXT: vorps %ymm1, %ymm0, %ymm0
; XOPAVX1-NEXT: retq
;
; XOPAVX2-LABEL: splatvar_funnnel_v16i16:
; XOPAVX2: # %bb.0:
-; XOPAVX2-NEXT: vpbroadcastw %xmm2, %ymm2
-; XOPAVX2-NEXT: vpsllw $1, %ymm0, %ymm0
-; XOPAVX2-NEXT: vextracti128 $1, %ymm0, %xmm3
-; XOPAVX2-NEXT: vpandn {{.*}}(%rip), %ymm2, %ymm4
-; XOPAVX2-NEXT: vextracti128 $1, %ymm4, %xmm5
-; XOPAVX2-NEXT: vpshlw %xmm5, %xmm3, %xmm3
-; XOPAVX2-NEXT: vpshlw %xmm4, %xmm0, %xmm0
-; XOPAVX2-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm0
-; XOPAVX2-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; XOPAVX2-NEXT: vpbroadcastw %xmm2, %xmm2
+; XOPAVX2-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; XOPAVX2-NEXT: vpand %xmm3, %xmm2, %xmm4
+; XOPAVX2-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
+; XOPAVX2-NEXT: vpsrlw %xmm4, %ymm1, %ymm1
+; XOPAVX2-NEXT: vpandn %xmm3, %xmm2, %xmm2
; XOPAVX2-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
-; XOPAVX2-NEXT: vpsrlw %xmm2, %ymm1, %ymm1
+; XOPAVX2-NEXT: vpsllw $1, %ymm0, %ymm0
+; XOPAVX2-NEXT: vpsllw %xmm2, %ymm0, %ymm0
; XOPAVX2-NEXT: vpor %ymm1, %ymm0, %ymm0
; XOPAVX2-NEXT: retq
%splat = shufflevector <16 x i16> %amt, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -1270,74 +1229,51 @@ define <16 x i16> @splatvar_funnnel_v16i16(<16 x i16> %x, <16 x i16> %y, <16 x i
define <32 x i8> @splatvar_funnnel_v32i8(<32 x i8> %x, <32 x i8> %y, <32 x i8> %amt) nounwind {
; AVX1-LABEL: splatvar_funnnel_v32i8:
; AVX1: # %bb.0:
-; AVX1-NEXT: vpxor %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpshufb %xmm3, %xmm2, %xmm3
-; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm3, %ymm3
-; AVX1-NEXT: vandnps {{.*}}(%rip), %ymm3, %ymm3
-; AVX1-NEXT: vextractf128 $1, %ymm3, %xmm4
-; AVX1-NEXT: vpsllw $5, %xmm4, %xmm4
-; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm5
-; AVX1-NEXT: vpaddb %xmm5, %xmm5, %xmm5
-; AVX1-NEXT: vpsllw $4, %xmm5, %xmm6
-; AVX1-NEXT: vmovdqa {{.*#+}} xmm8 = [240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240]
-; AVX1-NEXT: vpand %xmm6, %xmm8, %xmm6
-; AVX1-NEXT: vpblendvb %xmm4, %xmm6, %xmm5, %xmm5
-; AVX1-NEXT: vpsllw $2, %xmm5, %xmm6
-; AVX1-NEXT: vmovdqa {{.*#+}} xmm7 = [252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252]
-; AVX1-NEXT: vpand %xmm7, %xmm6, %xmm6
-; AVX1-NEXT: vpaddb %xmm4, %xmm4, %xmm4
-; AVX1-NEXT: vpblendvb %xmm4, %xmm6, %xmm5, %xmm5
-; AVX1-NEXT: vpaddb %xmm5, %xmm5, %xmm6
-; AVX1-NEXT: vpaddb %xmm4, %xmm4, %xmm4
-; AVX1-NEXT: vpblendvb %xmm4, %xmm6, %xmm5, %xmm4
-; AVX1-NEXT: vpsllw $5, %xmm3, %xmm3
-; AVX1-NEXT: vpaddb %xmm0, %xmm0, %xmm0
-; AVX1-NEXT: vpsllw $4, %xmm0, %xmm5
-; AVX1-NEXT: vpand %xmm5, %xmm8, %xmm5
-; AVX1-NEXT: vpblendvb %xmm3, %xmm5, %xmm0, %xmm0
-; AVX1-NEXT: vpsllw $2, %xmm0, %xmm5
-; AVX1-NEXT: vpand %xmm7, %xmm5, %xmm5
-; AVX1-NEXT: vpaddb %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpblendvb %xmm3, %xmm5, %xmm0, %xmm0
-; AVX1-NEXT: vpaddb %xmm0, %xmm0, %xmm5
-; AVX1-NEXT: vpaddb %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpblendvb %xmm3, %xmm5, %xmm0, %xmm0
-; AVX1-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm0
-; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm3
-; AVX1-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; AVX1-NEXT: vpxor %xmm8, %xmm8, %xmm8
+; AVX1-NEXT: vpshufb %xmm8, %xmm2, %xmm2
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX1-NEXT: vpand %xmm4, %xmm2, %xmm5
+; AVX1-NEXT: vpmovzxbq {{.*#+}} xmm5 = xmm5[0],zero,zero,zero,zero,zero,zero,zero,xmm5[1],zero,zero,zero,zero,zero,zero,zero
+; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm6
+; AVX1-NEXT: vpsrlw %xmm5, %xmm6, %xmm6
+; AVX1-NEXT: vpcmpeqd %xmm7, %xmm7, %xmm7
+; AVX1-NEXT: vpsrlw %xmm5, %xmm7, %xmm3
+; AVX1-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1]
+; AVX1-NEXT: vpand %xmm3, %xmm6, %xmm6
+; AVX1-NEXT: vpsrlw %xmm5, %xmm1, %xmm1
+; AVX1-NEXT: vpand %xmm3, %xmm1, %xmm1
+; AVX1-NEXT: vinsertf128 $1, %xmm6, %ymm1, %ymm1
+; AVX1-NEXT: vpandn %xmm4, %xmm2, %xmm2
; AVX1-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
-; AVX1-NEXT: vpsrlw %xmm2, %xmm3, %xmm3
-; AVX1-NEXT: vpcmpeqd %xmm4, %xmm4, %xmm4
-; AVX1-NEXT: vpsrlw %xmm2, %xmm4, %xmm4
-; AVX1-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1]
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
+; AVX1-NEXT: vpaddb %xmm3, %xmm3, %xmm3
+; AVX1-NEXT: vpsllw %xmm2, %xmm3, %xmm3
+; AVX1-NEXT: vpsllw %xmm2, %xmm7, %xmm4
+; AVX1-NEXT: vpshufb %xmm8, %xmm4, %xmm4
; AVX1-NEXT: vpand %xmm4, %xmm3, %xmm3
-; AVX1-NEXT: vpsrlw %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpand %xmm4, %xmm1, %xmm1
-; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm1, %ymm1
+; AVX1-NEXT: vpaddb %xmm0, %xmm0, %xmm0
+; AVX1-NEXT: vpsllw %xmm2, %xmm0, %xmm0
+; AVX1-NEXT: vpand %xmm4, %xmm0, %xmm0
+; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; AVX1-NEXT: vorps %ymm1, %ymm0, %ymm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: splatvar_funnnel_v32i8:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastb %xmm2, %ymm2
+; AVX2-NEXT: vpbroadcastb %xmm2, %xmm2
+; AVX2-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX2-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX2-NEXT: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
; AVX2-NEXT: vpaddb %ymm0, %ymm0, %ymm0
-; AVX2-NEXT: vpsllw $4, %ymm0, %ymm3
-; AVX2-NEXT: vpand {{.*}}(%rip), %ymm3, %ymm3
-; AVX2-NEXT: vpandn {{.*}}(%rip), %ymm2, %ymm4
-; AVX2-NEXT: vpsllw $5, %ymm4, %ymm4
-; AVX2-NEXT: vpblendvb %ymm4, %ymm3, %ymm0, %ymm0
-; AVX2-NEXT: vpsllw $2, %ymm0, %ymm3
-; AVX2-NEXT: vpand {{.*}}(%rip), %ymm3, %ymm3
-; AVX2-NEXT: vpaddb %ymm4, %ymm4, %ymm4
-; AVX2-NEXT: vpblendvb %ymm4, %ymm3, %ymm0, %ymm0
-; AVX2-NEXT: vpaddb %ymm0, %ymm0, %ymm3
-; AVX2-NEXT: vpaddb %ymm4, %ymm4, %ymm4
-; AVX2-NEXT: vpblendvb %ymm4, %ymm3, %ymm0, %ymm0
-; AVX2-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; AVX2-NEXT: vpsllw %xmm4, %ymm0, %ymm0
+; AVX2-NEXT: vpcmpeqd %xmm5, %xmm5, %xmm5
+; AVX2-NEXT: vpsllw %xmm4, %xmm5, %xmm4
+; AVX2-NEXT: vpbroadcastb %xmm4, %ymm4
+; AVX2-NEXT: vpand %ymm4, %ymm0, %ymm0
+; AVX2-NEXT: vpand %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
; AVX2-NEXT: vpsrlw %xmm2, %ymm1, %ymm1
-; AVX2-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX2-NEXT: vpsrlw %xmm2, %xmm3, %xmm2
+; AVX2-NEXT: vpsrlw %xmm2, %xmm5, %xmm2
; AVX2-NEXT: vpsrlw $8, %xmm2, %xmm2
; AVX2-NEXT: vpbroadcastb %xmm2, %ymm2
; AVX2-NEXT: vpand %ymm2, %ymm1, %ymm1
@@ -1346,25 +1282,20 @@ define <32 x i8> @splatvar_funnnel_v32i8(<32 x i8> %x, <32 x i8> %y, <32 x i8> %
;
; AVX512F-LABEL: splatvar_funnnel_v32i8:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpbroadcastb %xmm2, %ymm2
+; AVX512F-NEXT: vpbroadcastb %xmm2, %xmm2
+; AVX512F-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512F-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX512F-NEXT: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
; AVX512F-NEXT: vpaddb %ymm0, %ymm0, %ymm0
-; AVX512F-NEXT: vpsllw $4, %ymm0, %ymm3
-; AVX512F-NEXT: vpand {{.*}}(%rip), %ymm3, %ymm3
-; AVX512F-NEXT: vpandn {{.*}}(%rip), %ymm2, %ymm4
-; AVX512F-NEXT: vpsllw $5, %ymm4, %ymm4
-; AVX512F-NEXT: vpblendvb %ymm4, %ymm3, %ymm0, %ymm0
-; AVX512F-NEXT: vpsllw $2, %ymm0, %ymm3
-; AVX512F-NEXT: vpand {{.*}}(%rip), %ymm3, %ymm3
-; AVX512F-NEXT: vpaddb %ymm4, %ymm4, %ymm4
-; AVX512F-NEXT: vpblendvb %ymm4, %ymm3, %ymm0, %ymm0
-; AVX512F-NEXT: vpaddb %ymm0, %ymm0, %ymm3
-; AVX512F-NEXT: vpaddb %ymm4, %ymm4, %ymm4
-; AVX512F-NEXT: vpblendvb %ymm4, %ymm3, %ymm0, %ymm0
-; AVX512F-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; AVX512F-NEXT: vpsllw %xmm4, %ymm0, %ymm0
+; AVX512F-NEXT: vpcmpeqd %xmm5, %xmm5, %xmm5
+; AVX512F-NEXT: vpsllw %xmm4, %xmm5, %xmm4
+; AVX512F-NEXT: vpbroadcastb %xmm4, %ymm4
+; AVX512F-NEXT: vpand %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT: vpand %xmm3, %xmm2, %xmm2
; AVX512F-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
; AVX512F-NEXT: vpsrlw %xmm2, %ymm1, %ymm1
-; AVX512F-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX512F-NEXT: vpsrlw %xmm2, %xmm3, %xmm2
+; AVX512F-NEXT: vpsrlw %xmm2, %xmm5, %xmm2
; AVX512F-NEXT: vpsrlw $8, %xmm2, %xmm2
; AVX512F-NEXT: vpbroadcastb %xmm2, %ymm2
; AVX512F-NEXT: vpand %ymm2, %ymm1, %ymm1
@@ -1373,90 +1304,89 @@ define <32 x i8> @splatvar_funnnel_v32i8(<32 x i8> %x, <32 x i8> %y, <32 x i8> %
;
; AVX512VL-LABEL: splatvar_funnnel_v32i8:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastb %xmm2, %ymm2
+; AVX512VL-NEXT: vpbroadcastb %xmm2, %xmm2
+; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512VL-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX512VL-NEXT: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
; AVX512VL-NEXT: vpaddb %ymm0, %ymm0, %ymm0
-; AVX512VL-NEXT: vpsllw $4, %ymm0, %ymm3
-; AVX512VL-NEXT: vpand {{.*}}(%rip), %ymm3, %ymm3
-; AVX512VL-NEXT: vpandn {{.*}}(%rip), %ymm2, %ymm4
-; AVX512VL-NEXT: vpsllw $5, %ymm4, %ymm4
-; AVX512VL-NEXT: vpblendvb %ymm4, %ymm3, %ymm0, %ymm0
-; AVX512VL-NEXT: vpsllw $2, %ymm0, %ymm3
-; AVX512VL-NEXT: vpand {{.*}}(%rip), %ymm3, %ymm3
-; AVX512VL-NEXT: vpaddb %ymm4, %ymm4, %ymm4
-; AVX512VL-NEXT: vpblendvb %ymm4, %ymm3, %ymm0, %ymm0
-; AVX512VL-NEXT: vpaddb %ymm0, %ymm0, %ymm3
-; AVX512VL-NEXT: vpaddb %ymm4, %ymm4, %ymm4
-; AVX512VL-NEXT: vpblendvb %ymm4, %ymm3, %ymm0, %ymm3
-; AVX512VL-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm0
+; AVX512VL-NEXT: vpsllw %xmm4, %ymm0, %ymm0
+; AVX512VL-NEXT: vpcmpeqd %xmm5, %xmm5, %xmm5
+; AVX512VL-NEXT: vpsllw %xmm4, %xmm5, %xmm4
+; AVX512VL-NEXT: vpbroadcastb %xmm4, %ymm4
+; AVX512VL-NEXT: vpand %ymm4, %ymm0, %ymm4
+; AVX512VL-NEXT: vpand %xmm3, %xmm2, %xmm0
; AVX512VL-NEXT: vpmovzxbq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero
; AVX512VL-NEXT: vpsrlw %xmm0, %ymm1, %ymm1
-; AVX512VL-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX512VL-NEXT: vpsrlw %xmm0, %xmm2, %xmm0
+; AVX512VL-NEXT: vpsrlw %xmm0, %xmm5, %xmm0
; AVX512VL-NEXT: vpsrlw $8, %xmm0, %xmm0
; AVX512VL-NEXT: vpbroadcastb %xmm0, %ymm0
-; AVX512VL-NEXT: vpternlogq $236, %ymm1, %ymm3, %ymm0
+; AVX512VL-NEXT: vpternlogq $236, %ymm1, %ymm4, %ymm0
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: splatvar_funnnel_v32i8:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpbroadcastb %xmm2, %ymm3
+; AVX512BW-NEXT: vpbroadcastb %xmm2, %xmm2
+; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512BW-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512BW-NEXT: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
; AVX512BW-NEXT: vpmovzxbw {{.*#+}} zmm1 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero,ymm1[16],zero,ymm1[17],zero,ymm1[18],zero,ymm1[19],zero,ymm1[20],zero,ymm1[21],zero,ymm1[22],zero,ymm1[23],zero,ymm1[24],zero,ymm1[25],zero,ymm1[26],zero,ymm1[27],zero,ymm1[28],zero,ymm1[29],zero,ymm1[30],zero,ymm1[31],zero
-; AVX512BW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; AVX512BW-NEXT: vpsrlw %xmm4, %zmm1, %zmm1
+; AVX512BW-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512BW-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512BW-NEXT: vpsrlw %xmm2, %zmm1, %zmm1
; AVX512BW-NEXT: vpaddb %ymm0, %ymm0, %ymm0
; AVX512BW-NEXT: vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512BW-NEXT: vpandn {{.*}}(%rip), %ymm3, %ymm2
-; AVX512BW-NEXT: vpmovzxbw {{.*#+}} zmm2 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero,ymm2[16],zero,ymm2[17],zero,ymm2[18],zero,ymm2[19],zero,ymm2[20],zero,ymm2[21],zero,ymm2[22],zero,ymm2[23],zero,ymm2[24],zero,ymm2[25],zero,ymm2[26],zero,ymm2[27],zero,ymm2[28],zero,ymm2[29],zero,ymm2[30],zero,ymm2[31],zero
-; AVX512BW-NEXT: vpsllvw %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT: vpsllw %xmm2, %zmm0, %zmm0
; AVX512BW-NEXT: vporq %zmm1, %zmm0, %zmm0
; AVX512BW-NEXT: vpmovwb %zmm0, %ymm0
; AVX512BW-NEXT: retq
;
; AVX512VBMI2-LABEL: splatvar_funnnel_v32i8:
; AVX512VBMI2: # %bb.0:
-; AVX512VBMI2-NEXT: vpbroadcastb %xmm2, %ymm3
+; AVX512VBMI2-NEXT: vpbroadcastb %xmm2, %xmm2
+; AVX512VBMI2-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512VBMI2-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512VBMI2-NEXT: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
; AVX512VBMI2-NEXT: vpmovzxbw {{.*#+}} zmm1 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero,ymm1[16],zero,ymm1[17],zero,ymm1[18],zero,ymm1[19],zero,ymm1[20],zero,ymm1[21],zero,ymm1[22],zero,ymm1[23],zero,ymm1[24],zero,ymm1[25],zero,ymm1[26],zero,ymm1[27],zero,ymm1[28],zero,ymm1[29],zero,ymm1[30],zero,ymm1[31],zero
-; AVX512VBMI2-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; AVX512VBMI2-NEXT: vpsrlw %xmm4, %zmm1, %zmm1
+; AVX512VBMI2-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512VBMI2-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512VBMI2-NEXT: vpsrlw %xmm2, %zmm1, %zmm1
; AVX512VBMI2-NEXT: vpaddb %ymm0, %ymm0, %ymm0
; AVX512VBMI2-NEXT: vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512VBMI2-NEXT: vpandn {{.*}}(%rip), %ymm3, %ymm2
-; AVX512VBMI2-NEXT: vpmovzxbw {{.*#+}} zmm2 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero,ymm2[16],zero,ymm2[17],zero,ymm2[18],zero,ymm2[19],zero,ymm2[20],zero,ymm2[21],zero,ymm2[22],zero,ymm2[23],zero,ymm2[24],zero,ymm2[25],zero,ymm2[26],zero,ymm2[27],zero,ymm2[28],zero,ymm2[29],zero,ymm2[30],zero,ymm2[31],zero
-; AVX512VBMI2-NEXT: vpsllvw %zmm2, %zmm0, %zmm0
+; AVX512VBMI2-NEXT: vpsllw %xmm2, %zmm0, %zmm0
; AVX512VBMI2-NEXT: vporq %zmm1, %zmm0, %zmm0
; AVX512VBMI2-NEXT: vpmovwb %zmm0, %ymm0
; AVX512VBMI2-NEXT: retq
;
; AVX512VLBW-LABEL: splatvar_funnnel_v32i8:
; AVX512VLBW: # %bb.0:
-; AVX512VLBW-NEXT: vpbroadcastb %xmm2, %ymm3
+; 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: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
; AVX512VLBW-NEXT: vpmovzxbw {{.*#+}} zmm1 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero,ymm1[16],zero,ymm1[17],zero,ymm1[18],zero,ymm1[19],zero,ymm1[20],zero,ymm1[21],zero,ymm1[22],zero,ymm1[23],zero,ymm1[24],zero,ymm1[25],zero,ymm1[26],zero,ymm1[27],zero,ymm1[28],zero,ymm1[29],zero,ymm1[30],zero,ymm1[31],zero
-; AVX512VLBW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; AVX512VLBW-NEXT: vpsrlw %xmm4, %zmm1, %zmm1
+; AVX512VLBW-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512VLBW-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512VLBW-NEXT: vpsrlw %xmm2, %zmm1, %zmm1
; AVX512VLBW-NEXT: vpaddb %ymm0, %ymm0, %ymm0
; AVX512VLBW-NEXT: vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512VLBW-NEXT: vpandn {{.*}}(%rip), %ymm3, %ymm2
-; AVX512VLBW-NEXT: vpmovzxbw {{.*#+}} zmm2 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero,ymm2[16],zero,ymm2[17],zero,ymm2[18],zero,ymm2[19],zero,ymm2[20],zero,ymm2[21],zero,ymm2[22],zero,ymm2[23],zero,ymm2[24],zero,ymm2[25],zero,ymm2[26],zero,ymm2[27],zero,ymm2[28],zero,ymm2[29],zero,ymm2[30],zero,ymm2[31],zero
-; AVX512VLBW-NEXT: vpsllvw %zmm2, %zmm0, %zmm0
+; AVX512VLBW-NEXT: vpsllw %xmm2, %zmm0, %zmm0
; AVX512VLBW-NEXT: vporq %zmm1, %zmm0, %zmm0
; AVX512VLBW-NEXT: vpmovwb %zmm0, %ymm0
; AVX512VLBW-NEXT: retq
;
; AVX512VLVBMI2-LABEL: splatvar_funnnel_v32i8:
; AVX512VLVBMI2: # %bb.0:
-; AVX512VLVBMI2-NEXT: vpbroadcastb %xmm2, %ymm3
+; 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: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
; AVX512VLVBMI2-NEXT: vpmovzxbw {{.*#+}} zmm1 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero,ymm1[16],zero,ymm1[17],zero,ymm1[18],zero,ymm1[19],zero,ymm1[20],zero,ymm1[21],zero,ymm1[22],zero,ymm1[23],zero,ymm1[24],zero,ymm1[25],zero,ymm1[26],zero,ymm1[27],zero,ymm1[28],zero,ymm1[29],zero,ymm1[30],zero,ymm1[31],zero
-; AVX512VLVBMI2-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; AVX512VLVBMI2-NEXT: vpsrlw %xmm4, %zmm1, %zmm1
+; AVX512VLVBMI2-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512VLVBMI2-NEXT: vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512VLVBMI2-NEXT: vpsrlw %xmm2, %zmm1, %zmm1
; AVX512VLVBMI2-NEXT: vpaddb %ymm0, %ymm0, %ymm0
; AVX512VLVBMI2-NEXT: vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512VLVBMI2-NEXT: vpandn {{.*}}(%rip), %ymm3, %ymm2
-; AVX512VLVBMI2-NEXT: vpmovzxbw {{.*#+}} zmm2 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero,ymm2[16],zero,ymm2[17],zero,ymm2[18],zero,ymm2[19],zero,ymm2[20],zero,ymm2[21],zero,ymm2[22],zero,ymm2[23],zero,ymm2[24],zero,ymm2[25],zero,ymm2[26],zero,ymm2[27],zero,ymm2[28],zero,ymm2[29],zero,ymm2[30],zero,ymm2[31],zero
-; AVX512VLVBMI2-NEXT: vpsllvw %zmm2, %zmm0, %zmm0
+; AVX512VLVBMI2-NEXT: vpsllw %xmm2, %zmm0, %zmm0
; AVX512VLVBMI2-NEXT: vporq %zmm1, %zmm0, %zmm0
; AVX512VLVBMI2-NEXT: vpmovwb %zmm0, %ymm0
; AVX512VLVBMI2-NEXT: retq
@@ -1465,16 +1395,15 @@ define <32 x i8> @splatvar_funnnel_v32i8(<32 x i8> %x, <32 x i8> %y, <32 x i8> %
; XOPAVX1: # %bb.0:
; XOPAVX1-NEXT: vpxor %xmm3, %xmm3, %xmm3
; XOPAVX1-NEXT: vpshufb %xmm3, %xmm2, %xmm2
-; XOPAVX1-NEXT: vinsertf128 $1, %xmm2, %ymm2, %ymm4
-; XOPAVX1-NEXT: vandnps {{.*}}(%rip), %ymm4, %ymm4
-; XOPAVX1-NEXT: vextractf128 $1, %ymm4, %xmm5
+; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; XOPAVX1-NEXT: vpandn %xmm4, %xmm2, %xmm5
; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm6
; XOPAVX1-NEXT: vpaddb %xmm6, %xmm6, %xmm6
-; XOPAVX1-NEXT: vpshlb %xmm5, %xmm6, %xmm5
+; XOPAVX1-NEXT: vpshlb %xmm5, %xmm6, %xmm6
; XOPAVX1-NEXT: vpaddb %xmm0, %xmm0, %xmm0
-; XOPAVX1-NEXT: vpshlb %xmm4, %xmm0, %xmm0
-; XOPAVX1-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm0
-; XOPAVX1-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; XOPAVX1-NEXT: vpshlb %xmm5, %xmm0, %xmm0
+; XOPAVX1-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm0
+; XOPAVX1-NEXT: vpand %xmm4, %xmm2, %xmm2
; XOPAVX1-NEXT: vpsubb %xmm2, %xmm3, %xmm2
; XOPAVX1-NEXT: vextractf128 $1, %ymm1, %xmm3
; XOPAVX1-NEXT: vpshlb %xmm2, %xmm3, %xmm3
@@ -1485,15 +1414,15 @@ define <32 x i8> @splatvar_funnnel_v32i8(<32 x i8> %x, <32 x i8> %y, <32 x i8> %
;
; XOPAVX2-LABEL: splatvar_funnnel_v32i8:
; XOPAVX2: # %bb.0:
-; XOPAVX2-NEXT: vpbroadcastb %xmm2, %ymm2
+; XOPAVX2-NEXT: vpbroadcastb %xmm2, %xmm2
+; XOPAVX2-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; XOPAVX2-NEXT: vpandn %xmm3, %xmm2, %xmm4
; XOPAVX2-NEXT: vpaddb %ymm0, %ymm0, %ymm0
-; XOPAVX2-NEXT: vextracti128 $1, %ymm0, %xmm3
-; XOPAVX2-NEXT: vpandn {{.*}}(%rip), %ymm2, %ymm4
-; XOPAVX2-NEXT: vextracti128 $1, %ymm4, %xmm5
-; XOPAVX2-NEXT: vpshlb %xmm5, %xmm3, %xmm3
+; XOPAVX2-NEXT: vextracti128 $1, %ymm0, %xmm5
+; XOPAVX2-NEXT: vpshlb %xmm4, %xmm5, %xmm5
; XOPAVX2-NEXT: vpshlb %xmm4, %xmm0, %xmm0
-; XOPAVX2-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm0
-; XOPAVX2-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; XOPAVX2-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm0
+; XOPAVX2-NEXT: vpand %xmm3, %xmm2, %xmm2
; XOPAVX2-NEXT: vpxor %xmm3, %xmm3, %xmm3
; XOPAVX2-NEXT: vpsubb %xmm2, %xmm3, %xmm2
; XOPAVX2-NEXT: vextracti128 $1, %ymm1, %xmm3
diff --git a/llvm/test/CodeGen/X86/vector-fshr-512.ll b/llvm/test/CodeGen/X86/vector-fshr-512.ll
index 2e19d36d6d86..4e1f73d44f82 100644
--- a/llvm/test/CodeGen/X86/vector-fshr-512.ll
+++ b/llvm/test/CodeGen/X86/vector-fshr-512.ll
@@ -530,34 +530,34 @@ define <64 x i8> @var_funnnel_v64i8(<64 x i8> %x, <64 x i8> %y, <64 x i8> %amt)
define <8 x i64> @splatvar_funnnel_v8i64(<8 x i64> %x, <8 x i64> %y, <8 x i64> %amt) nounwind {
; AVX512F-LABEL: splatvar_funnnel_v8i64:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpbroadcastq %xmm2, %zmm3
-; AVX512F-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX512F-NEXT: vpsrlq %xmm2, %zmm1, %zmm1
+; AVX512F-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; AVX512F-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512F-NEXT: vpsrlq %xmm4, %zmm1, %zmm1
+; AVX512F-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512F-NEXT: vpsllq $1, %zmm0, %zmm0
-; AVX512F-NEXT: vpandnq {{.*}}(%rip){1to8}, %zmm3, %zmm2
-; AVX512F-NEXT: vpsllvq %zmm2, %zmm0, %zmm0
+; AVX512F-NEXT: vpsllq %xmm2, %zmm0, %zmm0
; AVX512F-NEXT: vporq %zmm1, %zmm0, %zmm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: splatvar_funnnel_v8i64:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastq %xmm2, %zmm3
-; AVX512VL-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX512VL-NEXT: vpsrlq %xmm2, %zmm1, %zmm1
+; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; AVX512VL-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512VL-NEXT: vpsrlq %xmm4, %zmm1, %zmm1
+; AVX512VL-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512VL-NEXT: vpsllq $1, %zmm0, %zmm0
-; AVX512VL-NEXT: vpandnq {{.*}}(%rip){1to8}, %zmm3, %zmm2
-; AVX512VL-NEXT: vpsllvq %zmm2, %zmm0, %zmm0
+; AVX512VL-NEXT: vpsllq %xmm2, %zmm0, %zmm0
; AVX512VL-NEXT: vporq %zmm1, %zmm0, %zmm0
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: splatvar_funnnel_v8i64:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpbroadcastq %xmm2, %zmm3
-; AVX512BW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX512BW-NEXT: vpsrlq %xmm2, %zmm1, %zmm1
+; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; AVX512BW-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512BW-NEXT: vpsrlq %xmm4, %zmm1, %zmm1
+; AVX512BW-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512BW-NEXT: vpsllq $1, %zmm0, %zmm0
-; AVX512BW-NEXT: vpandnq {{.*}}(%rip){1to8}, %zmm3, %zmm2
-; AVX512BW-NEXT: vpsllvq %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT: vpsllq %xmm2, %zmm0, %zmm0
; AVX512BW-NEXT: vporq %zmm1, %zmm0, %zmm0
; AVX512BW-NEXT: retq
;
@@ -570,12 +570,12 @@ define <8 x i64> @splatvar_funnnel_v8i64(<8 x i64> %x, <8 x i64> %y, <8 x i64> %
;
; AVX512VLBW-LABEL: splatvar_funnnel_v8i64:
; AVX512VLBW: # %bb.0:
-; AVX512VLBW-NEXT: vpbroadcastq %xmm2, %zmm3
-; AVX512VLBW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX512VLBW-NEXT: vpsrlq %xmm2, %zmm1, %zmm1
+; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm3 = [63,63]
+; AVX512VLBW-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512VLBW-NEXT: vpsrlq %xmm4, %zmm1, %zmm1
+; AVX512VLBW-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512VLBW-NEXT: vpsllq $1, %zmm0, %zmm0
-; AVX512VLBW-NEXT: vpandnq {{.*}}(%rip){1to8}, %zmm3, %zmm2
-; AVX512VLBW-NEXT: vpsllvq %zmm2, %zmm0, %zmm0
+; AVX512VLBW-NEXT: vpsllq %xmm2, %zmm0, %zmm0
; AVX512VLBW-NEXT: vporq %zmm1, %zmm0, %zmm0
; AVX512VLBW-NEXT: retq
;
@@ -593,40 +593,43 @@ define <8 x i64> @splatvar_funnnel_v8i64(<8 x i64> %x, <8 x i64> %y, <8 x i64> %
define <16 x i32> @splatvar_funnnel_v16i32(<16 x i32> %x, <16 x i32> %y, <16 x i32> %amt) nounwind {
; AVX512F-LABEL: splatvar_funnnel_v16i32:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpbroadcastd %xmm2, %zmm2
-; AVX512F-NEXT: vpbroadcastd {{.*#+}} zmm3 = [31,31,31,31,31,31,31,31,31,31,31,31,31,31,31,31]
-; AVX512F-NEXT: vpandnd %zmm3, %zmm2, %zmm4
-; AVX512F-NEXT: vpslld $1, %zmm0, %zmm0
-; AVX512F-NEXT: vpsllvd %zmm4, %zmm0, %zmm0
-; AVX512F-NEXT: vpand %xmm3, %xmm2, %xmm2
+; AVX512F-NEXT: vpbroadcastd %xmm2, %xmm2
+; AVX512F-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31]
+; AVX512F-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512F-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
+; AVX512F-NEXT: vpsrld %xmm4, %zmm1, %zmm1
+; AVX512F-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512F-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
-; AVX512F-NEXT: vpsrld %xmm2, %zmm1, %zmm1
+; AVX512F-NEXT: vpslld $1, %zmm0, %zmm0
+; AVX512F-NEXT: vpslld %xmm2, %zmm0, %zmm0
; AVX512F-NEXT: vpord %zmm1, %zmm0, %zmm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: splatvar_funnnel_v16i32:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastd %xmm2, %zmm2
-; AVX512VL-NEXT: vpbroadcastd {{.*#+}} zmm3 = [31,31,31,31,31,31,31,31,31,31,31,31,31,31,31,31]
-; AVX512VL-NEXT: vpandnd %zmm3, %zmm2, %zmm4
-; AVX512VL-NEXT: vpslld $1, %zmm0, %zmm0
-; AVX512VL-NEXT: vpsllvd %zmm4, %zmm0, %zmm0
-; AVX512VL-NEXT: vpand %xmm3, %xmm2, %xmm2
+; AVX512VL-NEXT: vpbroadcastd %xmm2, %xmm2
+; AVX512VL-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31]
+; AVX512VL-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512VL-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
+; AVX512VL-NEXT: vpsrld %xmm4, %zmm1, %zmm1
+; AVX512VL-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512VL-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
-; AVX512VL-NEXT: vpsrld %xmm2, %zmm1, %zmm1
+; AVX512VL-NEXT: vpslld $1, %zmm0, %zmm0
+; AVX512VL-NEXT: vpslld %xmm2, %zmm0, %zmm0
; AVX512VL-NEXT: vpord %zmm1, %zmm0, %zmm0
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: splatvar_funnnel_v16i32:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpbroadcastd %xmm2, %zmm2
-; AVX512BW-NEXT: vpbroadcastd {{.*#+}} zmm3 = [31,31,31,31,31,31,31,31,31,31,31,31,31,31,31,31]
-; AVX512BW-NEXT: vpandnd %zmm3, %zmm2, %zmm4
-; AVX512BW-NEXT: vpslld $1, %zmm0, %zmm0
-; AVX512BW-NEXT: vpsllvd %zmm4, %zmm0, %zmm0
-; AVX512BW-NEXT: vpand %xmm3, %xmm2, %xmm2
+; AVX512BW-NEXT: vpbroadcastd %xmm2, %xmm2
+; AVX512BW-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31]
+; AVX512BW-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512BW-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
+; AVX512BW-NEXT: vpsrld %xmm4, %zmm1, %zmm1
+; AVX512BW-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512BW-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
-; AVX512BW-NEXT: vpsrld %xmm2, %zmm1, %zmm1
+; AVX512BW-NEXT: vpslld $1, %zmm0, %zmm0
+; AVX512BW-NEXT: vpslld %xmm2, %zmm0, %zmm0
; AVX512BW-NEXT: vpord %zmm1, %zmm0, %zmm0
; AVX512BW-NEXT: retq
;
@@ -639,14 +642,15 @@ define <16 x i32> @splatvar_funnnel_v16i32(<16 x i32> %x, <16 x i32> %y, <16 x i
;
; AVX512VLBW-LABEL: splatvar_funnnel_v16i32:
; AVX512VLBW: # %bb.0:
-; AVX512VLBW-NEXT: vpbroadcastd %xmm2, %zmm2
-; AVX512VLBW-NEXT: vpbroadcastd {{.*#+}} zmm3 = [31,31,31,31,31,31,31,31,31,31,31,31,31,31,31,31]
-; AVX512VLBW-NEXT: vpandnd %zmm3, %zmm2, %zmm4
-; AVX512VLBW-NEXT: vpslld $1, %zmm0, %zmm0
-; AVX512VLBW-NEXT: vpsllvd %zmm4, %zmm0, %zmm0
-; AVX512VLBW-NEXT: vpand %xmm3, %xmm2, %xmm2
+; AVX512VLBW-NEXT: vpbroadcastd %xmm2, %xmm2
+; AVX512VLBW-NEXT: vpbroadcastd {{.*#+}} xmm3 = [31,31,31,31]
+; AVX512VLBW-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512VLBW-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
+; AVX512VLBW-NEXT: vpsrld %xmm4, %zmm1, %zmm1
+; AVX512VLBW-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512VLBW-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
-; AVX512VLBW-NEXT: vpsrld %xmm2, %zmm1, %zmm1
+; AVX512VLBW-NEXT: vpslld $1, %zmm0, %zmm0
+; AVX512VLBW-NEXT: vpslld %xmm2, %zmm0, %zmm0
; AVX512VLBW-NEXT: vpord %zmm1, %zmm0, %zmm0
; AVX512VLBW-NEXT: retq
;
@@ -664,67 +668,57 @@ define <16 x i32> @splatvar_funnnel_v16i32(<16 x i32> %x, <16 x i32> %y, <16 x i
define <32 x i16> @splatvar_funnnel_v32i16(<32 x i16> %x, <32 x i16> %y, <32 x i16> %amt) nounwind {
; AVX512F-LABEL: splatvar_funnnel_v32i16:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpbroadcastw %xmm2, %ymm2
-; AVX512F-NEXT: vinserti64x4 $1, %ymm2, %zmm2, %zmm2
-; AVX512F-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm3
-; AVX512F-NEXT: vpmovzxwq {{.*#+}} xmm3 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero
-; AVX512F-NEXT: vextracti64x4 $1, %zmm1, %ymm4
-; AVX512F-NEXT: vpsrlw %xmm3, %ymm4, %ymm4
-; AVX512F-NEXT: vpsrlw %xmm3, %ymm1, %ymm1
-; AVX512F-NEXT: vinserti64x4 $1, %ymm4, %zmm1, %zmm1
-; AVX512F-NEXT: vpandnq {{.*}}(%rip), %zmm2, %zmm2
-; AVX512F-NEXT: vpmovzxwd {{.*#+}} zmm3 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero
-; AVX512F-NEXT: vpsllw $1, %ymm0, %ymm4
-; AVX512F-NEXT: vpmovzxwd {{.*#+}} zmm4 = ymm4[0],zero,ymm4[1],zero,ymm4[2],zero,ymm4[3],zero,ymm4[4],zero,ymm4[5],zero,ymm4[6],zero,ymm4[7],zero,ymm4[8],zero,ymm4[9],zero,ymm4[10],zero,ymm4[11],zero,ymm4[12],zero,ymm4[13],zero,ymm4[14],zero,ymm4[15],zero
-; AVX512F-NEXT: vpsllvd %zmm3, %zmm4, %zmm3
-; AVX512F-NEXT: vpmovdw %zmm3, %ymm3
-; AVX512F-NEXT: vextracti64x4 $1, %zmm2, %ymm2
-; AVX512F-NEXT: vpmovzxwd {{.*#+}} zmm2 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero
-; AVX512F-NEXT: vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT: vpbroadcastw %xmm2, %xmm2
+; AVX512F-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; AVX512F-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512F-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
+; AVX512F-NEXT: vextracti64x4 $1, %zmm1, %ymm5
+; AVX512F-NEXT: vpsrlw %xmm4, %ymm5, %ymm5
+; AVX512F-NEXT: vpsrlw %xmm4, %ymm1, %ymm1
+; AVX512F-NEXT: vinserti64x4 $1, %ymm5, %zmm1, %zmm1
+; AVX512F-NEXT: vpandn %xmm3, %xmm2, %xmm2
+; AVX512F-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
+; AVX512F-NEXT: vextracti64x4 $1, %zmm0, %ymm3
+; AVX512F-NEXT: vpsllw $1, %ymm3, %ymm3
+; AVX512F-NEXT: vpsllw %xmm2, %ymm3, %ymm3
; AVX512F-NEXT: vpsllw $1, %ymm0, %ymm0
-; AVX512F-NEXT: vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
-; AVX512F-NEXT: vpsllvd %zmm2, %zmm0, %zmm0
-; AVX512F-NEXT: vpmovdw %zmm0, %ymm0
-; AVX512F-NEXT: vinserti64x4 $1, %ymm0, %zmm3, %zmm0
+; AVX512F-NEXT: vpsllw %xmm2, %ymm0, %ymm0
+; AVX512F-NEXT: vinserti64x4 $1, %ymm3, %zmm0, %zmm0
; AVX512F-NEXT: vporq %zmm1, %zmm0, %zmm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: splatvar_funnnel_v32i16:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastw %xmm2, %ymm2
-; AVX512VL-NEXT: vinserti64x4 $1, %ymm2, %zmm2, %zmm2
-; AVX512VL-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm3
-; AVX512VL-NEXT: vpmovzxwq {{.*#+}} xmm3 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero
-; AVX512VL-NEXT: vextracti64x4 $1, %zmm1, %ymm4
-; AVX512VL-NEXT: vpsrlw %xmm3, %ymm4, %ymm4
-; AVX512VL-NEXT: vpsrlw %xmm3, %ymm1, %ymm1
-; AVX512VL-NEXT: vinserti64x4 $1, %ymm4, %zmm1, %zmm1
-; AVX512VL-NEXT: vpandnq {{.*}}(%rip), %zmm2, %zmm2
-; AVX512VL-NEXT: vpmovzxwd {{.*#+}} zmm3 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero
-; AVX512VL-NEXT: vpsllw $1, %ymm0, %ymm4
-; AVX512VL-NEXT: vpmovzxwd {{.*#+}} zmm4 = ymm4[0],zero,ymm4[1],zero,ymm4[2],zero,ymm4[3],zero,ymm4[4],zero,ymm4[5],zero,ymm4[6],zero,ymm4[7],zero,ymm4[8],zero,ymm4[9],zero,ymm4[10],zero,ymm4[11],zero,ymm4[12],zero,ymm4[13],zero,ymm4[14],zero,ymm4[15],zero
-; AVX512VL-NEXT: vpsllvd %zmm3, %zmm4, %zmm3
-; AVX512VL-NEXT: vpmovdw %zmm3, %ymm3
-; AVX512VL-NEXT: vextracti64x4 $1, %zmm2, %ymm2
-; AVX512VL-NEXT: vpmovzxwd {{.*#+}} zmm2 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero
-; AVX512VL-NEXT: vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VL-NEXT: vpbroadcastw %xmm2, %xmm2
+; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15]
+; AVX512VL-NEXT: vpand %xmm3, %xmm2, %xmm4
+; AVX512VL-NEXT: vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
+; AVX512VL-NEXT: vextracti64x4 $1, %zmm1, %ymm5
+; AVX512VL-NEXT: vpsrlw %xmm4, %ymm5, %ymm5
+; AVX512VL-NEXT: vpsrlw %xmm4, %ymm1, %ymm1
+; AVX512VL-NEXT: vinserti64x4 $1, %ymm5, %zmm1, %zmm1
+; AVX512VL-NEXT: vpandn %xmm3, %xmm2, %xmm2
+; AVX512VL-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
+; AVX512VL-NEXT: vextracti64x4 $1, %zmm0, %ymm3
+; AVX512VL-NEXT: vpsllw $1, %ymm3, %ymm3
+; AVX512VL-NEXT: vpsllw %xmm2, %ymm3, %ymm3
; AVX512VL-NEXT: vpsllw $1, %ymm0, %ymm0
-; AVX512VL-NEXT: vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
-; AVX512VL-NEXT: vpsllvd %zmm2, %zmm0, %zmm0
-; AVX512VL-NEXT: vpmovdw %zmm0, %ymm0
-; AVX512VL-NEXT: vinserti64x4 $1, %ymm0, %zmm3, %zmm0
+; AVX512VL-NEXT: vpsllw %xmm2, %ymm0, %ymm0
+; AVX512VL-NEXT: vinserti64x4 $1, %ymm3, %zmm0, %zmm0
; AVX512VL-NEXT: vporq %zmm1, %zmm0, %zmm0
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: splatvar_funnnel_v32i16:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpbroadcastw %xmm2, %zmm2
-; AVX512BW-NEXT: vpsllw $1, %zmm0, %zmm0
-; AVX512BW-NEXT: vpandnq {{.*}}(%rip), %zmm2, %zmm3
-; AVX512BW-NEXT: vpsllvw %zmm3, %zmm0, %zmm0
-; AVX512BW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; 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 {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
+; AVX512BW-NEXT: vpsrlw %xmm4, %zmm1, %zmm1
+; AVX512BW-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512BW-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
-; AVX512BW-NEXT: vpsrlw %xmm2, %zmm1, %zmm1
+; AVX512BW-NEXT: vpsllw $1, %zmm0, %zmm0
+; AVX512BW-NEXT: vpsllw %xmm2, %zmm0, %zmm0
; AVX512BW-NEXT: vporq %zmm1, %zmm0, %zmm0
; AVX512BW-NEXT: retq
;
@@ -737,13 +731,15 @@ define <32 x i16> @splatvar_funnnel_v32i16(<32 x i16> %x, <32 x i16> %y, <32 x i
;
; AVX512VLBW-LABEL: splatvar_funnnel_v32i16:
; AVX512VLBW: # %bb.0:
-; AVX512VLBW-NEXT: vpbroadcastw %xmm2, %zmm2
-; AVX512VLBW-NEXT: vpsllw $1, %zmm0, %zmm0
-; AVX512VLBW-NEXT: vpandnq {{.*}}(%rip), %zmm2, %zmm3
-; AVX512VLBW-NEXT: vpsllvw %zmm3, %zmm0, %zmm0
-; AVX512VLBW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; 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 {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
+; AVX512VLBW-NEXT: vpsrlw %xmm4, %zmm1, %zmm1
+; AVX512VLBW-NEXT: vpandn %xmm3, %xmm2, %xmm2
; AVX512VLBW-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
-; AVX512VLBW-NEXT: vpsrlw %xmm2, %zmm1, %zmm1
+; AVX512VLBW-NEXT: vpsllw $1, %zmm0, %zmm0
+; AVX512VLBW-NEXT: vpsllw %xmm2, %zmm0, %zmm0
; AVX512VLBW-NEXT: vporq %zmm1, %zmm0, %zmm0
; AVX512VLBW-NEXT: retq
;
@@ -761,210 +757,146 @@ define <32 x i16> @splatvar_funnnel_v32i16(<32 x i16> %x, <32 x i16> %y, <32 x i
define <64 x i8> @splatvar_funnnel_v64i8(<64 x i8> %x, <64 x i8> %y, <64 x i8> %amt) nounwind {
; AVX512F-LABEL: splatvar_funnnel_v64i8:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpbroadcastb %xmm2, %ymm2
-; AVX512F-NEXT: vinserti64x4 $1, %ymm2, %zmm2, %zmm2
-; AVX512F-NEXT: vextracti64x4 $1, %zmm0, %ymm3
-; AVX512F-NEXT: vpaddb %ymm3, %ymm3, %ymm3
-; AVX512F-NEXT: vpsllw $4, %ymm3, %ymm4
-; AVX512F-NEXT: vmovdqa {{.*#+}} ymm5 = [240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240]
-; AVX512F-NEXT: vpand %ymm5, %ymm4, %ymm4
-; AVX512F-NEXT: vpandnq {{.*}}(%rip), %zmm2, %zmm6
-; AVX512F-NEXT: vextracti64x4 $1, %zmm6, %ymm7
-; AVX512F-NEXT: vpsllw $5, %ymm7, %ymm7
-; AVX512F-NEXT: vpblendvb %ymm7, %ymm4, %ymm3, %ymm3
-; AVX512F-NEXT: vpsllw $2, %ymm3, %ymm4
-; AVX512F-NEXT: vmovdqa {{.*#+}} ymm8 = [252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252]
-; AVX512F-NEXT: vpand %ymm4, %ymm8, %ymm4
-; AVX512F-NEXT: vpaddb %ymm7, %ymm7, %ymm7
-; AVX512F-NEXT: vpblendvb %ymm7, %ymm4, %ymm3, %ymm3
-; AVX512F-NEXT: vpaddb %ymm3, %ymm3, %ymm4
-; AVX512F-NEXT: vpaddb %ymm7, %ymm7, %ymm7
-; AVX512F-NEXT: vpblendvb %ymm7, %ymm4, %ymm3, %ymm3
-; AVX512F-NEXT: vpaddb %ymm0, %ymm0, %ymm0
-; AVX512F-NEXT: vpsllw $4, %ymm0, %ymm4
-; AVX512F-NEXT: vpand %ymm5, %ymm4, %ymm4
-; AVX512F-NEXT: vpsllw $5, %ymm6, %ymm5
-; AVX512F-NEXT: vpblendvb %ymm5, %ymm4, %ymm0, %ymm0
-; AVX512F-NEXT: vpsllw $2, %ymm0, %ymm4
-; AVX512F-NEXT: vpand %ymm4, %ymm8, %ymm4
-; AVX512F-NEXT: vpaddb %ymm5, %ymm5, %ymm5
-; AVX512F-NEXT: vpblendvb %ymm5, %ymm4, %ymm0, %ymm0
-; AVX512F-NEXT: vpaddb %ymm0, %ymm0, %ymm4
+; AVX512F-NEXT: vpbroadcastb %xmm2, %xmm2
+; AVX512F-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512F-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX512F-NEXT: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512F-NEXT: vextracti64x4 $1, %zmm0, %ymm5
; AVX512F-NEXT: vpaddb %ymm5, %ymm5, %ymm5
-; AVX512F-NEXT: vpblendvb %ymm5, %ymm4, %ymm0, %ymm0
-; AVX512F-NEXT: vinserti64x4 $1, %ymm3, %zmm0, %zmm3
-; AVX512F-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm0
+; AVX512F-NEXT: vpsllw %xmm4, %ymm5, %ymm5
+; AVX512F-NEXT: vpaddb %ymm0, %ymm0, %ymm0
+; AVX512F-NEXT: vpsllw %xmm4, %ymm0, %ymm0
+; AVX512F-NEXT: vinserti64x4 $1, %ymm5, %zmm0, %zmm0
+; AVX512F-NEXT: vpcmpeqd %xmm5, %xmm5, %xmm5
+; AVX512F-NEXT: vpsllw %xmm4, %xmm5, %xmm4
+; AVX512F-NEXT: vpbroadcastb %xmm4, %ymm4
+; AVX512F-NEXT: vinserti64x4 $1, %ymm4, %zmm4, %zmm4
+; AVX512F-NEXT: vpandq %zmm4, %zmm0, %zmm4
+; AVX512F-NEXT: vpand %xmm3, %xmm2, %xmm0
; AVX512F-NEXT: vpmovzxbq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero
; AVX512F-NEXT: vextracti64x4 $1, %zmm1, %ymm2
; AVX512F-NEXT: vpsrlw %xmm0, %ymm2, %ymm2
; AVX512F-NEXT: vpsrlw %xmm0, %ymm1, %ymm1
; AVX512F-NEXT: vinserti64x4 $1, %ymm2, %zmm1, %zmm1
-; AVX512F-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX512F-NEXT: vpsrlw %xmm0, %xmm2, %xmm0
+; AVX512F-NEXT: vpsrlw %xmm0, %xmm5, %xmm0
; AVX512F-NEXT: vpsrlw $8, %xmm0, %xmm0
; AVX512F-NEXT: vpbroadcastb %xmm0, %ymm0
; AVX512F-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm0
-; AVX512F-NEXT: vpternlogq $236, %zmm1, %zmm3, %zmm0
+; AVX512F-NEXT: vpternlogq $236, %zmm1, %zmm4, %zmm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: splatvar_funnnel_v64i8:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastb %xmm2, %ymm2
-; AVX512VL-NEXT: vinserti64x4 $1, %ymm2, %zmm2, %zmm2
-; AVX512VL-NEXT: vextracti64x4 $1, %zmm0, %ymm3
-; AVX512VL-NEXT: vpaddb %ymm3, %ymm3, %ymm3
-; AVX512VL-NEXT: vpsllw $4, %ymm3, %ymm4
-; AVX512VL-NEXT: vmovdqa {{.*#+}} ymm5 = [240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240]
-; AVX512VL-NEXT: vpand %ymm5, %ymm4, %ymm4
-; AVX512VL-NEXT: vpandnq {{.*}}(%rip), %zmm2, %zmm6
-; AVX512VL-NEXT: vextracti64x4 $1, %zmm6, %ymm7
-; AVX512VL-NEXT: vpsllw $5, %ymm7, %ymm7
-; AVX512VL-NEXT: vpblendvb %ymm7, %ymm4, %ymm3, %ymm3
-; AVX512VL-NEXT: vpsllw $2, %ymm3, %ymm4
-; AVX512VL-NEXT: vmovdqa {{.*#+}} ymm8 = [252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252]
-; AVX512VL-NEXT: vpand %ymm4, %ymm8, %ymm4
-; AVX512VL-NEXT: vpaddb %ymm7, %ymm7, %ymm7
-; AVX512VL-NEXT: vpblendvb %ymm7, %ymm4, %ymm3, %ymm3
-; AVX512VL-NEXT: vpaddb %ymm3, %ymm3, %ymm4
-; AVX512VL-NEXT: vpaddb %ymm7, %ymm7, %ymm7
-; AVX512VL-NEXT: vpblendvb %ymm7, %ymm4, %ymm3, %ymm3
-; AVX512VL-NEXT: vpaddb %ymm0, %ymm0, %ymm0
-; AVX512VL-NEXT: vpsllw $4, %ymm0, %ymm4
-; AVX512VL-NEXT: vpand %ymm5, %ymm4, %ymm4
-; AVX512VL-NEXT: vpsllw $5, %ymm6, %ymm5
-; AVX512VL-NEXT: vpblendvb %ymm5, %ymm4, %ymm0, %ymm0
-; AVX512VL-NEXT: vpsllw $2, %ymm0, %ymm4
-; AVX512VL-NEXT: vpand %ymm4, %ymm8, %ymm4
+; AVX512VL-NEXT: vpbroadcastb %xmm2, %xmm2
+; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512VL-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX512VL-NEXT: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512VL-NEXT: vextracti64x4 $1, %zmm0, %ymm5
; AVX512VL-NEXT: vpaddb %ymm5, %ymm5, %ymm5
-; AVX512VL-NEXT: vpblendvb %ymm5, %ymm4, %ymm0, %ymm0
-; AVX512VL-NEXT: vpaddb %ymm0, %ymm0, %ymm4
-; AVX512VL-NEXT: vpaddb %ymm5, %ymm5, %ymm5
-; AVX512VL-NEXT: vpblendvb %ymm5, %ymm4, %ymm0, %ymm0
-; AVX512VL-NEXT: vinserti64x4 $1, %ymm3, %zmm0, %zmm3
-; AVX512VL-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm0
+; AVX512VL-NEXT: vpsllw %xmm4, %ymm5, %ymm5
+; AVX512VL-NEXT: vpaddb %ymm0, %ymm0, %ymm0
+; AVX512VL-NEXT: vpsllw %xmm4, %ymm0, %ymm0
+; AVX512VL-NEXT: vinserti64x4 $1, %ymm5, %zmm0, %zmm0
+; AVX512VL-NEXT: vpcmpeqd %xmm5, %xmm5, %xmm5
+; AVX512VL-NEXT: vpsllw %xmm4, %xmm5, %xmm4
+; AVX512VL-NEXT: vpbroadcastb %xmm4, %ymm4
+; AVX512VL-NEXT: vinserti64x4 $1, %ymm4, %zmm4, %zmm4
+; AVX512VL-NEXT: vpandq %zmm4, %zmm0, %zmm4
+; AVX512VL-NEXT: vpand %xmm3, %xmm2, %xmm0
; AVX512VL-NEXT: vpmovzxbq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero
; AVX512VL-NEXT: vextracti64x4 $1, %zmm1, %ymm2
; AVX512VL-NEXT: vpsrlw %xmm0, %ymm2, %ymm2
; AVX512VL-NEXT: vpsrlw %xmm0, %ymm1, %ymm1
; AVX512VL-NEXT: vinserti64x4 $1, %ymm2, %zmm1, %zmm1
-; AVX512VL-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX512VL-NEXT: vpsrlw %xmm0, %xmm2, %xmm0
+; AVX512VL-NEXT: vpsrlw %xmm0, %xmm5, %xmm0
; AVX512VL-NEXT: vpsrlw $8, %xmm0, %xmm0
; AVX512VL-NEXT: vpbroadcastb %xmm0, %ymm0
; AVX512VL-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm0
-; AVX512VL-NEXT: vpternlogq $236, %zmm1, %zmm3, %zmm0
+; AVX512VL-NEXT: vpternlogq $236, %zmm1, %zmm4, %zmm0
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: splatvar_funnnel_v64i8:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpbroadcastb %xmm2, %zmm2
-; AVX512BW-NEXT: vpaddb %zmm0, %zmm0, %zmm3
-; AVX512BW-NEXT: vpsllw $4, %zmm3, %zmm0
-; AVX512BW-NEXT: vpandq {{.*}}(%rip), %zmm0, %zmm0
-; AVX512BW-NEXT: vpandnq {{.*}}(%rip), %zmm2, %zmm4
-; AVX512BW-NEXT: vpsllw $5, %zmm4, %zmm4
-; AVX512BW-NEXT: vpmovb2m %zmm4, %k1
-; AVX512BW-NEXT: vmovdqu8 %zmm0, %zmm3 {%k1}
-; AVX512BW-NEXT: vpsllw $2, %zmm3, %zmm0
-; AVX512BW-NEXT: vpandq {{.*}}(%rip), %zmm0, %zmm0
-; AVX512BW-NEXT: vpaddb %zmm4, %zmm4, %zmm4
-; AVX512BW-NEXT: vpmovb2m %zmm4, %k1
-; AVX512BW-NEXT: vmovdqu8 %zmm0, %zmm3 {%k1}
-; AVX512BW-NEXT: vpaddb %zmm4, %zmm4, %zmm0
-; AVX512BW-NEXT: vpmovb2m %zmm0, %k1
-; AVX512BW-NEXT: vpaddb %zmm3, %zmm3, %zmm3 {%k1}
-; AVX512BW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm0
+; AVX512BW-NEXT: vpbroadcastb %xmm2, %xmm2
+; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512BW-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX512BW-NEXT: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512BW-NEXT: vpaddb %zmm0, %zmm0, %zmm0
+; AVX512BW-NEXT: vpsllw %xmm4, %zmm0, %zmm0
+; AVX512BW-NEXT: vpcmpeqd %xmm5, %xmm5, %xmm5
+; AVX512BW-NEXT: vpsllw %xmm4, %xmm5, %xmm4
+; AVX512BW-NEXT: vpbroadcastb %xmm4, %zmm4
+; AVX512BW-NEXT: vpandq %zmm4, %zmm0, %zmm4
+; AVX512BW-NEXT: vpand %xmm3, %xmm2, %xmm0
; AVX512BW-NEXT: vpmovzxbq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero
; AVX512BW-NEXT: vpsrlw %xmm0, %zmm1, %zmm1
-; AVX512BW-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX512BW-NEXT: vpsrlw %xmm0, %xmm2, %xmm0
+; AVX512BW-NEXT: vpsrlw %xmm0, %xmm5, %xmm0
; AVX512BW-NEXT: vpsrlw $8, %xmm0, %xmm0
; AVX512BW-NEXT: vpbroadcastb %xmm0, %zmm0
-; AVX512BW-NEXT: vpternlogq $236, %zmm1, %zmm3, %zmm0
+; AVX512BW-NEXT: vpternlogq $236, %zmm1, %zmm4, %zmm0
; AVX512BW-NEXT: retq
;
; AVX512VBMI2-LABEL: splatvar_funnnel_v64i8:
; AVX512VBMI2: # %bb.0:
-; AVX512VBMI2-NEXT: vpbroadcastb %xmm2, %zmm2
-; AVX512VBMI2-NEXT: vpaddb %zmm0, %zmm0, %zmm3
-; AVX512VBMI2-NEXT: vpsllw $4, %zmm3, %zmm0
-; AVX512VBMI2-NEXT: vpandq {{.*}}(%rip), %zmm0, %zmm0
-; AVX512VBMI2-NEXT: vpandnq {{.*}}(%rip), %zmm2, %zmm4
-; AVX512VBMI2-NEXT: vpsllw $5, %zmm4, %zmm4
-; AVX512VBMI2-NEXT: vpmovb2m %zmm4, %k1
-; AVX512VBMI2-NEXT: vmovdqu8 %zmm0, %zmm3 {%k1}
-; AVX512VBMI2-NEXT: vpsllw $2, %zmm3, %zmm0
-; AVX512VBMI2-NEXT: vpandq {{.*}}(%rip), %zmm0, %zmm0
-; AVX512VBMI2-NEXT: vpaddb %zmm4, %zmm4, %zmm4
-; AVX512VBMI2-NEXT: vpmovb2m %zmm4, %k1
-; AVX512VBMI2-NEXT: vmovdqu8 %zmm0, %zmm3 {%k1}
-; AVX512VBMI2-NEXT: vpaddb %zmm4, %zmm4, %zmm0
-; AVX512VBMI2-NEXT: vpmovb2m %zmm0, %k1
-; AVX512VBMI2-NEXT: vpaddb %zmm3, %zmm3, %zmm3 {%k1}
-; AVX512VBMI2-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm0
+; AVX512VBMI2-NEXT: vpbroadcastb %xmm2, %xmm2
+; AVX512VBMI2-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512VBMI2-NEXT: vpandn %xmm3, %xmm2, %xmm4
+; AVX512VBMI2-NEXT: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512VBMI2-NEXT: vpaddb %zmm0, %zmm0, %zmm0
+; AVX512VBMI2-NEXT: vpsllw %xmm4, %zmm0, %zmm0
+; AVX512VBMI2-NEXT: vpcmpeqd %xmm5, %xmm5, %xmm5
+; AVX512VBMI2-NEXT: vpsllw %xmm4, %xmm5, %xmm4
+; AVX512VBMI2-NEXT: vpbroadcastb %xmm4, %zmm4
+; AVX512VBMI2-NEXT: vpandq %zmm4, %zmm0, %zmm4
+; AVX512VBMI2-NEXT: vpand %xmm3, %xmm2, %xmm0
; AVX512VBMI2-NEXT: vpmovzxbq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero
; AVX512VBMI2-NEXT: vpsrlw %xmm0, %zmm1, %zmm1
-; AVX512VBMI2-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX512VBMI2-NEXT: vpsrlw %xmm0, %xmm2, %xmm0
+; AVX512VBMI2-NEXT: vpsrlw %xmm0, %xmm5, %xmm0
; AVX512VBMI2-NEXT: vpsrlw $8, %xmm0, %xmm0
; AVX512VBMI2-NEXT: vpbroadcastb %xmm0, %zmm0
-; AVX512VBMI2-NEXT: vpternlogq $236, %zmm1, %zmm3, %zmm0
+; AVX512VBMI2-NEXT: vpternlogq $236, %zmm1, %zmm4, %zmm0
; AVX512VBMI2-NEXT: retq
;
; AVX512VLBW-LABEL: splatvar_funnnel_v64i8:
; AVX512VLBW: # %bb.0:
-; AVX512VLBW-NEXT: vpbroadcastb %xmm2, %zmm2
-; AVX512VLBW-NEXT: vpaddb %zmm0, %zmm0, %zmm3
-; AVX512VLBW-NEXT: vpsllw $4, %zmm3, %zmm0
-; AVX512VLBW-NEXT: vpandq {{.*}}(%rip), %zmm0, %zmm0
-; AVX512VLBW-NEXT: vpandnq {{.*}}(%rip), %zmm2, %zmm4
-; AVX512VLBW-NEXT: vpsllw $5, %zmm4, %zmm4
-; AVX512VLBW-NEXT: vpmovb2m %zmm4, %k1
-; AVX512VLBW-NEXT: vmovdqu8 %zmm0, %zmm3 {%k1}
-; AVX512VLBW-NEXT: vpsllw $2, %zmm3, %zmm0
-; AVX512VLBW-NEXT: vpandq {{.*}}(%rip), %zmm0, %zmm0
-; AVX512VLBW-NEXT: vpaddb %zmm4, %zmm4, %zmm4
-; AVX512VLBW-NEXT: vpmovb2m %zmm4, %k1
-; AVX512VLBW-NEXT: vmovdqu8 %zmm0, %zmm3 {%k1}
-; AVX512VLBW-NEXT: vpaddb %zmm4, %zmm4, %zmm0
-; AVX512VLBW-NEXT: vpmovb2m %zmm0, %k1
-; AVX512VLBW-NEXT: vpaddb %zmm3, %zmm3, %zmm3 {%k1}
-; AVX512VLBW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm0
+; 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: vpandn %xmm3, %xmm2, %xmm4
+; AVX512VLBW-NEXT: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512VLBW-NEXT: vpaddb %zmm0, %zmm0, %zmm0
+; AVX512VLBW-NEXT: vpsllw %xmm4, %zmm0, %zmm0
+; AVX512VLBW-NEXT: vpcmpeqd %xmm5, %xmm5, %xmm5
+; AVX512VLBW-NEXT: vpsllw %xmm4, %xmm5, %xmm4
+; AVX512VLBW-NEXT: vpbroadcastb %xmm4, %zmm4
+; AVX512VLBW-NEXT: vpandq %zmm4, %zmm0, %zmm4
+; AVX512VLBW-NEXT: vpand %xmm3, %xmm2, %xmm0
; AVX512VLBW-NEXT: vpmovzxbq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero
; AVX512VLBW-NEXT: vpsrlw %xmm0, %zmm1, %zmm1
-; AVX512VLBW-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX512VLBW-NEXT: vpsrlw %xmm0, %xmm2, %xmm0
+; AVX512VLBW-NEXT: vpsrlw %xmm0, %xmm5, %xmm0
; AVX512VLBW-NEXT: vpsrlw $8, %xmm0, %xmm0
; AVX512VLBW-NEXT: vpbroadcastb %xmm0, %zmm0
-; AVX512VLBW-NEXT: vpternlogq $236, %zmm1, %zmm3, %zmm0
+; AVX512VLBW-NEXT: vpternlogq $236, %zmm1, %zmm4, %zmm0
; AVX512VLBW-NEXT: retq
;
; AVX512VLVBMI2-LABEL: splatvar_funnnel_v64i8:
; AVX512VLVBMI2: # %bb.0:
-; AVX512VLVBMI2-NEXT: vpbroadcastb %xmm2, %zmm2
-; AVX512VLVBMI2-NEXT: vpaddb %zmm0, %zmm0, %zmm3
-; AVX512VLVBMI2-NEXT: vpsllw $4, %zmm3, %zmm0
-; AVX512VLVBMI2-NEXT: vpandq {{.*}}(%rip), %zmm0, %zmm0
-; AVX512VLVBMI2-NEXT: vpandnq {{.*}}(%rip), %zmm2, %zmm4
-; AVX512VLVBMI2-NEXT: vpsllw $5, %zmm4, %zmm4
-; AVX512VLVBMI2-NEXT: vpmovb2m %zmm4, %k1
-; AVX512VLVBMI2-NEXT: vmovdqu8 %zmm0, %zmm3 {%k1}
-; AVX512VLVBMI2-NEXT: vpsllw $2, %zmm3, %zmm0
-; AVX512VLVBMI2-NEXT: vpandq {{.*}}(%rip), %zmm0, %zmm0
-; AVX512VLVBMI2-NEXT: vpaddb %zmm4, %zmm4, %zmm4
-; AVX512VLVBMI2-NEXT: vpmovb2m %zmm4, %k1
-; AVX512VLVBMI2-NEXT: vmovdqu8 %zmm0, %zmm3 {%k1}
-; AVX512VLVBMI2-NEXT: vpaddb %zmm4, %zmm4, %zmm0
-; AVX512VLVBMI2-NEXT: vpmovb2m %zmm0, %k1
-; AVX512VLVBMI2-NEXT: vpaddb %zmm3, %zmm3, %zmm3 {%k1}
-; AVX512VLVBMI2-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm0
+; 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: vpandn %xmm3, %xmm2, %xmm4
+; AVX512VLVBMI2-NEXT: vpmovzxbq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,zero,zero,zero,zero,xmm4[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512VLVBMI2-NEXT: vpaddb %zmm0, %zmm0, %zmm0
+; AVX512VLVBMI2-NEXT: vpsllw %xmm4, %zmm0, %zmm0
+; AVX512VLVBMI2-NEXT: vpcmpeqd %xmm5, %xmm5, %xmm5
+; AVX512VLVBMI2-NEXT: vpsllw %xmm4, %xmm5, %xmm4
+; AVX512VLVBMI2-NEXT: vpbroadcastb %xmm4, %zmm4
+; AVX512VLVBMI2-NEXT: vpandq %zmm4, %zmm0, %zmm4
+; AVX512VLVBMI2-NEXT: vpand %xmm3, %xmm2, %xmm0
; AVX512VLVBMI2-NEXT: vpmovzxbq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero
; AVX512VLVBMI2-NEXT: vpsrlw %xmm0, %zmm1, %zmm1
-; AVX512VLVBMI2-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX512VLVBMI2-NEXT: vpsrlw %xmm0, %xmm2, %xmm0
+; AVX512VLVBMI2-NEXT: vpsrlw %xmm0, %xmm5, %xmm0
; AVX512VLVBMI2-NEXT: vpsrlw $8, %xmm0, %xmm0
; AVX512VLVBMI2-NEXT: vpbroadcastb %xmm0, %zmm0
-; AVX512VLVBMI2-NEXT: vpternlogq $236, %zmm1, %zmm3, %zmm0
+; AVX512VLVBMI2-NEXT: vpternlogq $236, %zmm1, %zmm4, %zmm0
; AVX512VLVBMI2-NEXT: retq
%splat = shufflevector <64 x i8> %amt, <64 x i8> undef, <64 x i32> zeroinitializer
%res = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %x, <64 x i8> %y, <64 x i8> %splat)
More information about the llvm-commits
mailing list