There's still a couple of minor SimplifyDemandedElts regressions in some of the shift amount splats that will be fixed in future patches.
llvm-svn: 349052
return SignBitSelect(VT, Amt, M, R);
}
+ // ISD::ROT* uses modulo rotate amounts.
+ Amt = DAG.getNode(ISD::AND, DL, VT, Amt,
+ DAG.getConstant(EltSizeInBits - 1, DL, VT));
+
bool ConstantAmt = ISD::isBuildVectorOfConstantSDNodes(Amt.getNode());
bool LegalVarShifts = SupportedVectorVarShift(VT, Subtarget, ISD::SHL) &&
SupportedVectorVarShift(VT, Subtarget, ISD::SRL);
- // Rotate by splat - expand back to shifts.
- // Best to fallback for all supported variable shifts.
- // AVX2 - best to fallback for non-constants as well.
- // TODO - legalizers should be able to handle this.
- // TODO - We need explicit modulo rotation amounts.
+ // Fallback for splats + all supported variable shifts.
+ // Fallback for non-constants AVX2 vXi16 as well.
if (LegalVarShifts || (Subtarget.hasAVX2() && !ConstantAmt) ||
DAG.isSplatValue(Amt)) {
SDValue AmtR = DAG.getConstant(EltSizeInBits, DL, VT);
return DAG.getNode(ISD::OR, DL, VT, SHL, SRL);
}
- // ISD::ROT* uses modulo rotate amounts.
- Amt = DAG.getNode(ISD::AND, DL, VT, Amt,
- DAG.getConstant(EltSizeInBits - 1, DL, VT));
-
// As with shifts, convert the rotation amount to a multiplication factor.
SDValue Scale = convertShiftLeftToScale(Amt, DL, Subtarget, DAG);
assert(Scale && "Failed to convert ROTL amount to scale");
;
; AVX2-LABEL: var_rotate_v4i32:
; AVX2: # %bb.0:
+; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm2 = [31,31,31,31]
+; AVX2-NEXT: vpand %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vpsllvd %xmm1, %xmm0, %xmm2
; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm3 = [32,32,32,32]
; AVX2-NEXT: vpsubd %xmm1, %xmm3, %xmm1
;
; AVX2-LABEL: var_rotate_v8i16:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm2 = 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: 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: vpand {{.*}}(%rip), %xmm1, %xmm1
+; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm2 = 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: vpsllvd %ymm2, %ymm0, %ymm2
; AVX2-NEXT: vmovdqa {{.*#+}} ymm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
; AVX2-NEXT: vpshufb %ymm3, %ymm2, %ymm2
;
; AVX512F-LABEL: var_rotate_v8i16:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpmovzxwd {{.*#+}} ymm2 = 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: 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: vpand {{.*}}(%rip), %xmm1, %xmm1
+; AVX512F-NEXT: vpmovzxwd {{.*#+}} ymm2 = 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: vpsllvd %ymm2, %ymm0, %ymm2
; AVX512F-NEXT: vpmovdw %zmm2, %ymm2
; AVX512F-NEXT: vmovdqa {{.*#+}} xmm3 = [16,16,16,16,16,16,16,16]
;
; AVX512VL-LABEL: var_rotate_v8i16:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpmovzxwd {{.*#+}} ymm2 = 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: 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: vpand {{.*}}(%rip), %xmm1, %xmm1
+; AVX512VL-NEXT: vpmovzxwd {{.*#+}} ymm2 = 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: vpsllvd %ymm2, %ymm0, %ymm2
; AVX512VL-NEXT: vpmovdw %ymm2, %xmm2
; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm3 = [16,16,16,16,16,16,16,16]
;
; AVX512BW-LABEL: var_rotate_v8i16:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: # kill: def $xmm1 killed $xmm1 def $zmm1
; AVX512BW-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512BW-NEXT: vpand {{.*}}(%rip), %xmm1, %xmm1
; AVX512BW-NEXT: vpsllvw %zmm1, %zmm0, %zmm2
; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm3 = [16,16,16,16,16,16,16,16]
; AVX512BW-NEXT: vpsubw %xmm1, %xmm3, %xmm1
;
; AVX512VLBW-LABEL: var_rotate_v8i16:
; AVX512VLBW: # %bb.0:
+; AVX512VLBW-NEXT: vpand {{.*}}(%rip), %xmm1, %xmm1
; AVX512VLBW-NEXT: vpsllvw %xmm1, %xmm0, %xmm2
; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm3 = [16,16,16,16,16,16,16,16]
; AVX512VLBW-NEXT: vpsubw %xmm1, %xmm3, %xmm1
define <4 x i32> @splatvar_rotate_v4i32(<4 x i32> %a, <4 x i32> %b) nounwind {
; SSE2-LABEL: splatvar_rotate_v4i32:
; SSE2: # %bb.0:
+; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
+; SSE2-NEXT: pand {{.*}}(%rip), %xmm1
; SSE2-NEXT: xorps %xmm2, %xmm2
; SSE2-NEXT: xorps %xmm3, %xmm3
; SSE2-NEXT: movss {{.*#+}} xmm3 = xmm1[0],xmm3[1,2,3]
;
; SSE41-LABEL: splatvar_rotate_v4i32:
; SSE41: # %bb.0:
+; SSE41-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
+; SSE41-NEXT: pand {{.*}}(%rip), %xmm1
; SSE41-NEXT: pmovzxdq {{.*#+}} xmm2 = xmm1[0],zero,xmm1[1],zero
; SSE41-NEXT: movdqa %xmm0, %xmm3
; SSE41-NEXT: pslld %xmm2, %xmm3
;
; AVX1-LABEL: splatvar_rotate_v4i32:
; AVX1: # %bb.0:
+; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
+; AVX1-NEXT: vpand {{.*}}(%rip), %xmm1, %xmm1
; AVX1-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm1[0],zero,xmm1[1],zero
; AVX1-NEXT: vpslld %xmm2, %xmm0, %xmm2
; AVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [32,32,32,32]
;
; AVX2-LABEL: splatvar_rotate_v4i32:
; AVX2: # %bb.0:
+; AVX2-NEXT: vpbroadcastd %xmm1, %xmm1
+; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm2 = [31,31,31,31]
+; AVX2-NEXT: vpand %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm1[0],zero,xmm1[1],zero
; AVX2-NEXT: vpslld %xmm2, %xmm0, %xmm2
; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm3 = [32,32,32,32]
;
; X32-SSE-LABEL: splatvar_rotate_v4i32:
; X32-SSE: # %bb.0:
+; X32-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
+; X32-SSE-NEXT: pand {{\.LCPI.*}}, %xmm1
; X32-SSE-NEXT: xorps %xmm2, %xmm2
; X32-SSE-NEXT: xorps %xmm3, %xmm3
; X32-SSE-NEXT: movss {{.*#+}} xmm3 = xmm1[0],xmm3[1,2,3]
define <8 x i16> @splatvar_rotate_v8i16(<8 x i16> %a, <8 x i16> %b) nounwind {
; SSE2-LABEL: splatvar_rotate_v8i16:
; SSE2: # %bb.0:
+; SSE2-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,0,2,3,4,5,6,7]
+; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
+; SSE2-NEXT: pand {{.*}}(%rip), %xmm1
; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [16,16,16,16,16,16,16,16]
; SSE2-NEXT: psubw %xmm1, %xmm2
; SSE2-NEXT: pslldq {{.*#+}} xmm1 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm1[0,1]
;
; SSE41-LABEL: splatvar_rotate_v8i16:
; SSE41: # %bb.0:
+; SSE41-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,0,2,3,4,5,6,7]
+; SSE41-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
+; SSE41-NEXT: pand {{.*}}(%rip), %xmm1
; SSE41-NEXT: pmovzxwq {{.*#+}} xmm2 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero
; SSE41-NEXT: movdqa %xmm0, %xmm3
; SSE41-NEXT: psllw %xmm2, %xmm3
; SSE41-NEXT: por %xmm3, %xmm0
; SSE41-NEXT: retq
;
-; AVX-LABEL: splatvar_rotate_v8i16:
-; AVX: # %bb.0:
-; AVX-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero
-; AVX-NEXT: vpsllw %xmm2, %xmm0, %xmm2
-; AVX-NEXT: vmovdqa {{.*#+}} xmm3 = [16,16,16,16,16,16,16,16]
-; AVX-NEXT: vpsubw %xmm1, %xmm3, %xmm1
-; AVX-NEXT: vpmovzxwq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero
-; AVX-NEXT: vpsrlw %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpor %xmm0, %xmm2, %xmm0
-; AVX-NEXT: retq
+; AVX1-LABEL: splatvar_rotate_v8i16:
+; AVX1: # %bb.0:
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,0,2,3,4,5,6,7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
+; AVX1-NEXT: vpand {{.*}}(%rip), %xmm1, %xmm1
+; AVX1-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero
+; AVX1-NEXT: vpsllw %xmm2, %xmm0, %xmm2
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [16,16,16,16,16,16,16,16]
+; AVX1-NEXT: vpsubw %xmm1, %xmm3, %xmm1
+; AVX1-NEXT: vpmovzxwq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero
+; AVX1-NEXT: vpsrlw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT: vpor %xmm0, %xmm2, %xmm0
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: splatvar_rotate_v8i16:
+; AVX2: # %bb.0:
+; AVX2-NEXT: vpbroadcastw %xmm1, %xmm1
+; AVX2-NEXT: vpand {{.*}}(%rip), %xmm1, %xmm1
+; AVX2-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero
+; AVX2-NEXT: vpsllw %xmm2, %xmm0, %xmm2
+; AVX2-NEXT: vmovdqa {{.*#+}} xmm3 = [16,16,16,16,16,16,16,16]
+; AVX2-NEXT: vpsubw %xmm1, %xmm3, %xmm1
+; AVX2-NEXT: vpmovzxwq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero
+; AVX2-NEXT: vpsrlw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT: vpor %xmm0, %xmm2, %xmm0
+; AVX2-NEXT: retq
;
; AVX512-LABEL: splatvar_rotate_v8i16:
; AVX512: # %bb.0:
+; AVX512-NEXT: vpbroadcastw %xmm1, %xmm1
+; AVX512-NEXT: vpand {{.*}}(%rip), %xmm1, %xmm1
; AVX512-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero
; AVX512-NEXT: vpsllw %xmm2, %xmm0, %xmm2
; AVX512-NEXT: vmovdqa {{.*#+}} xmm3 = [16,16,16,16,16,16,16,16]
;
; X32-SSE-LABEL: splatvar_rotate_v8i16:
; X32-SSE: # %bb.0:
+; X32-SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,0,2,3,4,5,6,7]
+; X32-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
+; X32-SSE-NEXT: pand {{\.LCPI.*}}, %xmm1
; X32-SSE-NEXT: movdqa {{.*#+}} xmm2 = [16,16,16,16,16,16,16,16]
; X32-SSE-NEXT: psubw %xmm1, %xmm2
; X32-SSE-NEXT: pslldq {{.*#+}} xmm1 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm1[0,1]
;
; AVX2-LABEL: var_rotate_v8i32:
; AVX2: # %bb.0:
+; AVX2-NEXT: vpbroadcastd {{.*#+}} ymm2 = [31,31,31,31,31,31,31,31]
+; AVX2-NEXT: vpand %ymm2, %ymm1, %ymm1
; AVX2-NEXT: vpsllvd %ymm1, %ymm0, %ymm2
; AVX2-NEXT: vpbroadcastd {{.*#+}} ymm3 = [32,32,32,32,32,32,32,32]
; AVX2-NEXT: vpsubd %ymm1, %ymm3, %ymm1
; AVX2-LABEL: var_rotate_v16i16:
; AVX2: # %bb.0:
; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpunpckhwd {{.*#+}} ymm3 = ymm1[4],ymm2[4],ymm1[5],ymm2[5],ymm1[6],ymm2[6],ymm1[7],ymm2[7],ymm1[12],ymm2[12],ymm1[13],ymm2[13],ymm1[14],ymm2[14],ymm1[15],ymm2[15]
-; AVX2-NEXT: vpunpckhwd {{.*#+}} ymm4 = ymm2[4],ymm0[4],ymm2[5],ymm0[5],ymm2[6],ymm0[6],ymm2[7],ymm0[7],ymm2[12],ymm0[12],ymm2[13],ymm0[13],ymm2[14],ymm0[14],ymm2[15],ymm0[15]
-; AVX2-NEXT: vpsllvd %ymm3, %ymm4, %ymm3
-; AVX2-NEXT: vpsrld $16, %ymm3, %ymm3
-; AVX2-NEXT: vpunpcklwd {{.*#+}} ymm5 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[8],ymm2[8],ymm1[9],ymm2[9],ymm1[10],ymm2[10],ymm1[11],ymm2[11]
+; AVX2-NEXT: vpunpckhwd {{.*#+}} ymm3 = ymm2[4],ymm0[4],ymm2[5],ymm0[5],ymm2[6],ymm0[6],ymm2[7],ymm0[7],ymm2[12],ymm0[12],ymm2[13],ymm0[13],ymm2[14],ymm0[14],ymm2[15],ymm0[15]
+; AVX2-NEXT: vpand {{.*}}(%rip), %ymm1, %ymm1
+; AVX2-NEXT: vpunpckhwd {{.*#+}} ymm4 = ymm1[4],ymm2[4],ymm1[5],ymm2[5],ymm1[6],ymm2[6],ymm1[7],ymm2[7],ymm1[12],ymm2[12],ymm1[13],ymm2[13],ymm1[14],ymm2[14],ymm1[15],ymm2[15]
+; AVX2-NEXT: vpsllvd %ymm4, %ymm3, %ymm4
+; AVX2-NEXT: vpsrld $16, %ymm4, %ymm4
; AVX2-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm2[0],ymm0[0],ymm2[1],ymm0[1],ymm2[2],ymm0[2],ymm2[3],ymm0[3],ymm2[8],ymm0[8],ymm2[9],ymm0[9],ymm2[10],ymm0[10],ymm2[11],ymm0[11]
+; AVX2-NEXT: vpunpcklwd {{.*#+}} ymm5 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[8],ymm2[8],ymm1[9],ymm2[9],ymm1[10],ymm2[10],ymm1[11],ymm2[11]
; AVX2-NEXT: vpsllvd %ymm5, %ymm0, %ymm5
; AVX2-NEXT: vpsrld $16, %ymm5, %ymm5
-; AVX2-NEXT: vpackusdw %ymm3, %ymm5, %ymm3
+; AVX2-NEXT: vpackusdw %ymm4, %ymm5, %ymm4
; AVX2-NEXT: vmovdqa {{.*#+}} ymm5 = [16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
; AVX2-NEXT: vpsubw %ymm1, %ymm5, %ymm1
; AVX2-NEXT: vpunpckhwd {{.*#+}} ymm5 = ymm1[4],ymm2[4],ymm1[5],ymm2[5],ymm1[6],ymm2[6],ymm1[7],ymm2[7],ymm1[12],ymm2[12],ymm1[13],ymm2[13],ymm1[14],ymm2[14],ymm1[15],ymm2[15]
-; AVX2-NEXT: vpsrlvd %ymm5, %ymm4, %ymm4
-; AVX2-NEXT: vpsrld $16, %ymm4, %ymm4
+; AVX2-NEXT: vpsrlvd %ymm5, %ymm3, %ymm3
+; AVX2-NEXT: vpsrld $16, %ymm3, %ymm3
; AVX2-NEXT: vpunpcklwd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[8],ymm2[8],ymm1[9],ymm2[9],ymm1[10],ymm2[10],ymm1[11],ymm2[11]
; AVX2-NEXT: vpsrlvd %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vpsrld $16, %ymm0, %ymm0
-; AVX2-NEXT: vpackusdw %ymm4, %ymm0, %ymm0
-; AVX2-NEXT: vpor %ymm0, %ymm3, %ymm0
+; AVX2-NEXT: vpackusdw %ymm3, %ymm0, %ymm0
+; AVX2-NEXT: vpor %ymm0, %ymm4, %ymm0
; AVX2-NEXT: retq
;
; AVX512F-LABEL: var_rotate_v16i16:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpmovzxwd {{.*#+}} zmm2 = 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: 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: vpand {{.*}}(%rip), %ymm1, %ymm1
+; AVX512F-NEXT: vpmovzxwd {{.*#+}} zmm2 = 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: vpsllvd %zmm2, %zmm0, %zmm2
; AVX512F-NEXT: vpmovdw %zmm2, %ymm2
; AVX512F-NEXT: vmovdqa {{.*#+}} ymm3 = [16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
;
; AVX512VL-LABEL: var_rotate_v16i16:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpmovzxwd {{.*#+}} zmm2 = 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: 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: vpand {{.*}}(%rip), %ymm1, %ymm1
+; AVX512VL-NEXT: vpmovzxwd {{.*#+}} zmm2 = 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: vpsllvd %zmm2, %zmm0, %zmm2
; AVX512VL-NEXT: vpmovdw %zmm2, %ymm2
; AVX512VL-NEXT: vmovdqa {{.*#+}} ymm3 = [16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
;
; AVX512BW-LABEL: var_rotate_v16i16:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: # kill: def $ymm1 killed $ymm1 def $zmm1
; AVX512BW-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512BW-NEXT: vpand {{.*}}(%rip), %ymm1, %ymm1
; AVX512BW-NEXT: vpsllvw %zmm1, %zmm0, %zmm2
; AVX512BW-NEXT: vmovdqa {{.*#+}} ymm3 = [16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
; AVX512BW-NEXT: vpsubw %ymm1, %ymm3, %ymm1
;
; AVX512VLBW-LABEL: var_rotate_v16i16:
; AVX512VLBW: # %bb.0:
+; AVX512VLBW-NEXT: vpand {{.*}}(%rip), %ymm1, %ymm1
; AVX512VLBW-NEXT: vpsllvw %ymm1, %ymm0, %ymm2
; AVX512VLBW-NEXT: vmovdqa {{.*#+}} ymm3 = [16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
; AVX512VLBW-NEXT: vpsubw %ymm1, %ymm3, %ymm1
define <8 x i32> @splatvar_rotate_v8i32(<8 x i32> %a, <8 x i32> %b) nounwind {
; AVX1-LABEL: splatvar_rotate_v8i32:
; AVX1: # %bb.0:
-; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm1[0,0,0,0]
-; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
-; AVX1-NEXT: vpmovzxdq {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero
-; AVX1-NEXT: vpslld %xmm1, %xmm3, %xmm4
+; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm2
+; AVX1-NEXT: vpand {{.*}}(%rip), %xmm1, %xmm1
+; AVX1-NEXT: vpmovzxdq {{.*#+}} xmm3 = xmm1[0],zero,xmm1[1],zero
+; AVX1-NEXT: vpslld %xmm3, %xmm2, %xmm4
; AVX1-NEXT: vmovdqa {{.*#+}} xmm5 = [32,32,32,32]
-; AVX1-NEXT: vpsubd %xmm2, %xmm5, %xmm2
-; AVX1-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
-; AVX1-NEXT: vpsrld %xmm2, %xmm3, %xmm3
-; AVX1-NEXT: vpor %xmm3, %xmm4, %xmm3
-; AVX1-NEXT: vpslld %xmm1, %xmm0, %xmm1
-; AVX1-NEXT: vpsrld %xmm2, %xmm0, %xmm0
-; AVX1-NEXT: vpor %xmm0, %xmm1, %xmm0
-; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
+; AVX1-NEXT: vpsubd %xmm1, %xmm5, %xmm1
+; AVX1-NEXT: vpmovzxdq {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero
+; AVX1-NEXT: vpsrld %xmm1, %xmm2, %xmm2
+; AVX1-NEXT: vpor %xmm2, %xmm4, %xmm2
+; AVX1-NEXT: vpslld %xmm3, %xmm0, %xmm3
+; AVX1-NEXT: vpsrld %xmm1, %xmm0, %xmm0
+; AVX1-NEXT: vpor %xmm0, %xmm3, %xmm0
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: splatvar_rotate_v8i32:
; AVX2: # %bb.0:
+; AVX2-NEXT: vpbroadcastd %xmm1, %ymm1
+; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm2 = [31,31,31,31]
+; AVX2-NEXT: vpand %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm1[0],zero,xmm1[1],zero
; AVX2-NEXT: vpslld %xmm2, %ymm0, %ymm2
; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm3 = [32,32,32,32]
define <16 x i16> @splatvar_rotate_v16i16(<16 x i16> %a, <16 x i16> %b) nounwind {
; AVX1-LABEL: splatvar_rotate_v16i16:
; AVX1: # %bb.0:
-; AVX1-NEXT: vpshuflw {{.*#+}} xmm2 = xmm1[0,0,2,3,4,5,6,7]
-; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
-; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
-; AVX1-NEXT: vpmovzxwq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero
-; AVX1-NEXT: vpsllw %xmm1, %xmm3, %xmm4
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,0,2,3,4,5,6,7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm2
+; AVX1-NEXT: vpand {{.*}}(%rip), %xmm1, %xmm1
+; AVX1-NEXT: vpmovzxwq {{.*#+}} xmm3 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero
+; AVX1-NEXT: vpsllw %xmm3, %xmm2, %xmm4
; AVX1-NEXT: vmovdqa {{.*#+}} xmm5 = [16,16,16,16,16,16,16,16]
-; AVX1-NEXT: vpsubw %xmm2, %xmm5, %xmm2
-; AVX1-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
-; AVX1-NEXT: vpsrlw %xmm2, %xmm3, %xmm3
-; AVX1-NEXT: vpor %xmm3, %xmm4, %xmm3
-; AVX1-NEXT: vpsllw %xmm1, %xmm0, %xmm1
-; AVX1-NEXT: vpsrlw %xmm2, %xmm0, %xmm0
-; AVX1-NEXT: vpor %xmm0, %xmm1, %xmm0
-; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
+; AVX1-NEXT: vpsubw %xmm1, %xmm5, %xmm1
+; AVX1-NEXT: vpmovzxwq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero
+; AVX1-NEXT: vpsrlw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT: vpor %xmm2, %xmm4, %xmm2
+; AVX1-NEXT: vpsllw %xmm3, %xmm0, %xmm3
+; AVX1-NEXT: vpsrlw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT: vpor %xmm0, %xmm3, %xmm0
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: splatvar_rotate_v16i16:
; AVX2: # %bb.0:
+; AVX2-NEXT: vpbroadcastw %xmm1, %ymm1
+; AVX2-NEXT: vpand {{.*}}(%rip), %xmm1, %xmm1
; AVX2-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero
; AVX2-NEXT: vpsllw %xmm2, %ymm0, %ymm2
; AVX2-NEXT: vmovdqa {{.*#+}} xmm3 = [16,16,16,16,16,16,16,16]
;
; AVX512-LABEL: splatvar_rotate_v16i16:
; AVX512: # %bb.0:
+; AVX512-NEXT: vpbroadcastw %xmm1, %ymm1
+; AVX512-NEXT: vpand {{.*}}(%rip), %xmm1, %xmm1
; AVX512-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero
; AVX512-NEXT: vpsllw %xmm2, %ymm0, %ymm2
; AVX512-NEXT: vmovdqa {{.*#+}} xmm3 = [16,16,16,16,16,16,16,16]
define <32 x i16> @var_rotate_v32i16(<32 x i16> %a, <32 x i16> %b) nounwind {
; AVX512F-LABEL: var_rotate_v32i16:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpmovzxwd {{.*#+}} zmm4 = 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: vmovdqa {{.*#+}} ymm4 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT: vpand %ymm4, %ymm2, %ymm2
+; AVX512F-NEXT: vpmovzxwd {{.*#+}} zmm5 = 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: 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 %zmm4, %zmm0, %zmm4
-; AVX512F-NEXT: vpmovdw %zmm4, %ymm4
-; AVX512F-NEXT: vmovdqa {{.*#+}} ymm5 = [16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
-; AVX512F-NEXT: vpsubw %ymm2, %ymm5, %ymm2
+; AVX512F-NEXT: vpsllvd %zmm5, %zmm0, %zmm5
+; AVX512F-NEXT: vpmovdw %zmm5, %ymm5
+; AVX512F-NEXT: vmovdqa {{.*#+}} ymm6 = [16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
+; AVX512F-NEXT: vpsubw %ymm2, %ymm6, %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, %zmm0, %zmm0
; AVX512F-NEXT: vpmovdw %zmm0, %ymm0
-; AVX512F-NEXT: vpor %ymm0, %ymm4, %ymm0
-; AVX512F-NEXT: vpmovzxwd {{.*#+}} zmm2 = ymm3[0],zero,ymm3[1],zero,ymm3[2],zero,ymm3[3],zero,ymm3[4],zero,ymm3[5],zero,ymm3[6],zero,ymm3[7],zero,ymm3[8],zero,ymm3[9],zero,ymm3[10],zero,ymm3[11],zero,ymm3[12],zero,ymm3[13],zero,ymm3[14],zero,ymm3[15],zero
+; AVX512F-NEXT: vpor %ymm0, %ymm5, %ymm0
+; AVX512F-NEXT: vpand %ymm4, %ymm3, %ymm2
+; 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: 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: vpsllvd %zmm2, %zmm1, %zmm2
-; AVX512F-NEXT: vpmovdw %zmm2, %ymm2
-; AVX512F-NEXT: vpsubw %ymm3, %ymm5, %ymm3
-; AVX512F-NEXT: vpmovzxwd {{.*#+}} zmm3 = ymm3[0],zero,ymm3[1],zero,ymm3[2],zero,ymm3[3],zero,ymm3[4],zero,ymm3[5],zero,ymm3[6],zero,ymm3[7],zero,ymm3[8],zero,ymm3[9],zero,ymm3[10],zero,ymm3[11],zero,ymm3[12],zero,ymm3[13],zero,ymm3[14],zero,ymm3[15],zero
-; AVX512F-NEXT: vpsrlvd %zmm3, %zmm1, %zmm1
+; AVX512F-NEXT: vpsllvd %zmm3, %zmm1, %zmm3
+; AVX512F-NEXT: vpmovdw %zmm3, %ymm3
+; AVX512F-NEXT: vpsubw %ymm2, %ymm6, %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: vpor %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT: vpor %ymm1, %ymm3, %ymm1
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: var_rotate_v32i16:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpmovzxwd {{.*#+}} zmm4 = 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: vmovdqa {{.*#+}} ymm4 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VL-NEXT: vpand %ymm4, %ymm2, %ymm2
+; AVX512VL-NEXT: vpmovzxwd {{.*#+}} zmm5 = 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: 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 %zmm4, %zmm0, %zmm4
-; AVX512VL-NEXT: vpmovdw %zmm4, %ymm4
-; AVX512VL-NEXT: vmovdqa {{.*#+}} ymm5 = [16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
-; AVX512VL-NEXT: vpsubw %ymm2, %ymm5, %ymm2
+; AVX512VL-NEXT: vpsllvd %zmm5, %zmm0, %zmm5
+; AVX512VL-NEXT: vpmovdw %zmm5, %ymm5
+; AVX512VL-NEXT: vmovdqa {{.*#+}} ymm6 = [16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
+; AVX512VL-NEXT: vpsubw %ymm2, %ymm6, %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, %zmm0, %zmm0
; AVX512VL-NEXT: vpmovdw %zmm0, %ymm0
-; AVX512VL-NEXT: vpor %ymm0, %ymm4, %ymm0
-; AVX512VL-NEXT: vpmovzxwd {{.*#+}} zmm2 = ymm3[0],zero,ymm3[1],zero,ymm3[2],zero,ymm3[3],zero,ymm3[4],zero,ymm3[5],zero,ymm3[6],zero,ymm3[7],zero,ymm3[8],zero,ymm3[9],zero,ymm3[10],zero,ymm3[11],zero,ymm3[12],zero,ymm3[13],zero,ymm3[14],zero,ymm3[15],zero
+; AVX512VL-NEXT: vpor %ymm0, %ymm5, %ymm0
+; AVX512VL-NEXT: vpand %ymm4, %ymm3, %ymm2
+; 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: 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: vpsllvd %zmm2, %zmm1, %zmm2
-; AVX512VL-NEXT: vpmovdw %zmm2, %ymm2
-; AVX512VL-NEXT: vpsubw %ymm3, %ymm5, %ymm3
-; AVX512VL-NEXT: vpmovzxwd {{.*#+}} zmm3 = ymm3[0],zero,ymm3[1],zero,ymm3[2],zero,ymm3[3],zero,ymm3[4],zero,ymm3[5],zero,ymm3[6],zero,ymm3[7],zero,ymm3[8],zero,ymm3[9],zero,ymm3[10],zero,ymm3[11],zero,ymm3[12],zero,ymm3[13],zero,ymm3[14],zero,ymm3[15],zero
-; AVX512VL-NEXT: vpsrlvd %zmm3, %zmm1, %zmm1
+; AVX512VL-NEXT: vpsllvd %zmm3, %zmm1, %zmm3
+; AVX512VL-NEXT: vpmovdw %zmm3, %ymm3
+; AVX512VL-NEXT: vpsubw %ymm2, %ymm6, %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: vpor %ymm1, %ymm2, %ymm1
+; AVX512VL-NEXT: vpor %ymm1, %ymm3, %ymm1
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: var_rotate_v32i16:
define <32 x i16> @splatvar_rotate_v32i16(<32 x i16> %a, <32 x i16> %b) nounwind {
; AVX512F-LABEL: splatvar_rotate_v32i16:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpbroadcastw %xmm2, %ymm3
-; AVX512F-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
-; AVX512F-NEXT: vpsllw %xmm2, %ymm0, %ymm4
+; AVX512F-NEXT: vpbroadcastw %xmm2, %ymm2
+; AVX512F-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; AVX512F-NEXT: vpmovzxwq {{.*#+}} xmm3 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
+; AVX512F-NEXT: vpsllw %xmm3, %ymm0, %ymm4
; AVX512F-NEXT: vmovdqa {{.*#+}} xmm5 = [16,16,16,16,16,16,16,16]
-; AVX512F-NEXT: vpsubw %xmm3, %xmm5, %xmm3
-; AVX512F-NEXT: vpmovzxwq {{.*#+}} xmm3 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero
-; AVX512F-NEXT: vpsrlw %xmm3, %ymm0, %ymm0
+; AVX512F-NEXT: vpsubw %xmm2, %xmm5, %xmm2
+; AVX512F-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
+; AVX512F-NEXT: vpsrlw %xmm2, %ymm0, %ymm0
; AVX512F-NEXT: vpor %ymm0, %ymm4, %ymm0
-; AVX512F-NEXT: vpsllw %xmm2, %ymm1, %ymm2
-; AVX512F-NEXT: vpsrlw %xmm3, %ymm1, %ymm1
-; AVX512F-NEXT: vpor %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT: vpsllw %xmm3, %ymm1, %ymm3
+; AVX512F-NEXT: vpsrlw %xmm2, %ymm1, %ymm1
+; AVX512F-NEXT: vpor %ymm1, %ymm3, %ymm1
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: splatvar_rotate_v32i16:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastw %xmm2, %ymm3
-; AVX512VL-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
-; AVX512VL-NEXT: vpsllw %xmm2, %ymm0, %ymm4
+; AVX512VL-NEXT: vpbroadcastw %xmm2, %ymm2
+; AVX512VL-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; AVX512VL-NEXT: vpmovzxwq {{.*#+}} xmm3 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
+; AVX512VL-NEXT: vpsllw %xmm3, %ymm0, %ymm4
; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm5 = [16,16,16,16,16,16,16,16]
-; AVX512VL-NEXT: vpsubw %xmm3, %xmm5, %xmm3
-; AVX512VL-NEXT: vpmovzxwq {{.*#+}} xmm3 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero
-; AVX512VL-NEXT: vpsrlw %xmm3, %ymm0, %ymm0
+; AVX512VL-NEXT: vpsubw %xmm2, %xmm5, %xmm2
+; AVX512VL-NEXT: vpmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
+; AVX512VL-NEXT: vpsrlw %xmm2, %ymm0, %ymm0
; AVX512VL-NEXT: vpor %ymm0, %ymm4, %ymm0
-; AVX512VL-NEXT: vpsllw %xmm2, %ymm1, %ymm2
-; AVX512VL-NEXT: vpsrlw %xmm3, %ymm1, %ymm1
-; AVX512VL-NEXT: vpor %ymm1, %ymm2, %ymm1
+; AVX512VL-NEXT: vpsllw %xmm3, %ymm1, %ymm3
+; AVX512VL-NEXT: vpsrlw %xmm2, %ymm1, %ymm1
+; AVX512VL-NEXT: vpor %ymm1, %ymm3, %ymm1
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: splatvar_rotate_v32i16: