Commit 4762c077 authored by Simon Pilgrim's avatar Simon Pilgrim
Browse files

[X86] LowerFunnelShift - always lower vXi8 fshl by constant amounts as unpack(y,x) << zext(z)

This can always be lowered as PMULLW+PSRLWI+PACKUSWB
parent e59964b6
......@@ -29831,6 +29831,7 @@ static SDValue LowerFunnelShift(SDValue Op, const X86Subtarget &Subtarget,
 
SDValue AmtMask = DAG.getConstant(EltSizeInBits - 1, DL, VT);
SDValue AmtMod = DAG.getNode(ISD::AND, DL, VT, Amt, AmtMask);
bool IsCst = ISD::isBuildVectorOfConstantSDNodes(AmtMod.getNode());
 
unsigned ShiftOpc = IsFSHR ? ISD::SRL : ISD::SHL;
unsigned NumElts = VT.getVectorNumElements();
......@@ -29867,7 +29868,7 @@ static SDValue LowerFunnelShift(SDValue Op, const X86Subtarget &Subtarget,
MVT WideVT = MVT::getVectorVT(WideSVT, NumElts);
 
// If per-element shifts are legal, fallback to generic expansion.
if (supportedVectorVarShift(VT, Subtarget, ShiftOpc))
if (supportedVectorVarShift(VT, Subtarget, ShiftOpc) || Subtarget.hasXOP())
return SDValue();
 
// Attempt to fold as:
......@@ -29889,7 +29890,8 @@ static SDValue LowerFunnelShift(SDValue Op, const X86Subtarget &Subtarget,
}
 
// Attempt to fold per-element (ExtVT) shift as unpack(y,x) << zext(z)
if (supportedVectorVarShift(ExtVT, Subtarget, ShiftOpc)) {
if ((IsCst && !IsFSHR && EltSizeInBits == 8) ||
supportedVectorVarShift(ExtVT, Subtarget, ShiftOpc)) {
SDValue Z = DAG.getConstant(0, DL, VT);
SDValue RLo = DAG.getBitcast(ExtVT, getUnpackl(DAG, DL, VT, Op1, Op0));
SDValue RHi = DAG.getBitcast(ExtVT, getUnpackh(DAG, DL, VT, Op1, Op0));
......@@ -2182,119 +2182,50 @@ define <8 x i16> @constant_funnnel_v8i16(<8 x i16> %x, <8 x i16> %y) nounwind {
}
define <16 x i8> @constant_funnnel_v16i8(<16 x i8> %x, <16 x i8> %y) nounwind {
; SSE2-LABEL: constant_funnnel_v16i8:
; SSE2: # %bb.0:
; SSE2-NEXT: movdqa %xmm0, %xmm2
; SSE2-NEXT: punpckhbw {{.*#+}} xmm2 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
; SSE2-NEXT: pmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2
; SSE2-NEXT: movdqa {{.*#+}} xmm3 = [255,255,255,255,255,255,255,255]
; SSE2-NEXT: pand %xmm3, %xmm2
; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
; SSE2-NEXT: pmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; SSE2-NEXT: pand %xmm3, %xmm0
; SSE2-NEXT: packuswb %xmm2, %xmm0
; SSE2-NEXT: psrlw $1, %xmm1
; SSE2-NEXT: pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
; SSE2-NEXT: pxor %xmm2, %xmm2
; SSE2-NEXT: movdqa %xmm1, %xmm3
; SSE2-NEXT: punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm2[8],xmm3[9],xmm2[9],xmm3[10],xmm2[10],xmm3[11],xmm2[11],xmm3[12],xmm2[12],xmm3[13],xmm2[13],xmm3[14],xmm2[14],xmm3[15],xmm2[15]
; SSE2-NEXT: pmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm3
; SSE2-NEXT: psrlw $8, %xmm3
; SSE2-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
; SSE2-NEXT: pmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
; SSE2-NEXT: psrlw $8, %xmm1
; SSE2-NEXT: packuswb %xmm3, %xmm1
; SSE2-NEXT: por %xmm1, %xmm0
; SSE2-NEXT: retq
;
; SSE41-LABEL: constant_funnnel_v16i8:
; SSE41: # %bb.0:
; SSE41-NEXT: pmovzxbw {{.*#+}} xmm2 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
; SSE41-NEXT: punpckhbw {{.*#+}} xmm0 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
; SSE41-NEXT: pmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; SSE41-NEXT: movdqa {{.*#+}} xmm3 = [255,255,255,255,255,255,255,255]
; SSE41-NEXT: pand %xmm3, %xmm0
; SSE41-NEXT: pmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2
; SSE41-NEXT: pand %xmm3, %xmm2
; SSE41-NEXT: packuswb %xmm0, %xmm2
; SSE41-NEXT: psrlw $1, %xmm1
; SSE41-NEXT: pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
; SSE41-NEXT: pxor %xmm3, %xmm3
; SSE41-NEXT: pmovzxbw {{.*#+}} xmm0 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
; SSE41-NEXT: punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm3[8],xmm1[9],xmm3[9],xmm1[10],xmm3[10],xmm1[11],xmm3[11],xmm1[12],xmm3[12],xmm1[13],xmm3[13],xmm1[14],xmm3[14],xmm1[15],xmm3[15]
; SSE41-NEXT: pmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
; SSE41-NEXT: psrlw $8, %xmm1
; SSE41-NEXT: pmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; SSE41-NEXT: psrlw $8, %xmm0
; SSE41-NEXT: packuswb %xmm1, %xmm0
; SSE41-NEXT: por %xmm2, %xmm0
; SSE41-NEXT: retq
;
; AVX1-LABEL: constant_funnnel_v16i8:
; AVX1: # %bb.0:
; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm2 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
; AVX1-NEXT: vpmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2
; AVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [255,255,255,255,255,255,255,255]
; AVX1-NEXT: vpand %xmm3, %xmm2, %xmm2
; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
; AVX1-NEXT: vpmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; AVX1-NEXT: vpand %xmm3, %xmm0, %xmm0
; AVX1-NEXT: vpackuswb %xmm2, %xmm0, %xmm0
; AVX1-NEXT: vpsrlw $1, %xmm1, %xmm1
; AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm2 = xmm1[8],xmm2[8],xmm1[9],xmm2[9],xmm1[10],xmm2[10],xmm1[11],xmm2[11],xmm1[12],xmm2[12],xmm1[13],xmm2[13],xmm1[14],xmm2[14],xmm1[15],xmm2[15]
; AVX1-NEXT: vpmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2
; AVX1-NEXT: vpsrlw $8, %xmm2, %xmm2
; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
; AVX1-NEXT: vpmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
; AVX1-NEXT: vpsrlw $8, %xmm1, %xmm1
; AVX1-NEXT: vpackuswb %xmm2, %xmm1, %xmm1
; AVX1-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX1-NEXT: retq
; SSE-LABEL: constant_funnnel_v16i8:
; SSE: # %bb.0:
; SSE-NEXT: movdqa %xmm1, %xmm2
; SSE-NEXT: punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm0[8],xmm2[9],xmm0[9],xmm2[10],xmm0[10],xmm2[11],xmm0[11],xmm2[12],xmm0[12],xmm2[13],xmm0[13],xmm2[14],xmm0[14],xmm2[15],xmm0[15]
; SSE-NEXT: pmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2
; SSE-NEXT: psrlw $8, %xmm2
; SSE-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
; SSE-NEXT: pmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
; SSE-NEXT: psrlw $8, %xmm1
; SSE-NEXT: packuswb %xmm2, %xmm1
; SSE-NEXT: movdqa %xmm1, %xmm0
; SSE-NEXT: retq
;
; AVX2-LABEL: constant_funnnel_v16i8:
; AVX2: # %bb.0:
; AVX2-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
; AVX2-NEXT: vpmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
; AVX2-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm2
; AVX2-NEXT: vpackuswb %xmm2, %xmm0, %xmm0
; AVX2-NEXT: vpsrlw $1, %xmm1, %xmm1
; AVX2-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
; AVX2-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
; AVX2-NEXT: vpmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
; AVX2-NEXT: vpsrlw $8, %ymm1, %ymm1
; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-NEXT: vpackuswb %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
; AVX-LABEL: constant_funnnel_v16i8:
; AVX: # %bb.0:
; AVX-NEXT: vpunpckhbw {{.*#+}} xmm2 = xmm1[8],xmm0[8],xmm1[9],xmm0[9],xmm1[10],xmm0[10],xmm1[11],xmm0[11],xmm1[12],xmm0[12],xmm1[13],xmm0[13],xmm1[14],xmm0[14],xmm1[15],xmm0[15]
; AVX-NEXT: vpmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2
; AVX-NEXT: vpsrlw $8, %xmm2, %xmm2
; AVX-NEXT: vpunpcklbw {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
; AVX-NEXT: vpmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; AVX-NEXT: vpsrlw $8, %xmm0, %xmm0
; AVX-NEXT: vpackuswb %xmm2, %xmm0, %xmm0
; AVX-NEXT: retq
;
; AVX512F-LABEL: constant_funnnel_v16i8:
; AVX512F: # %bb.0:
; 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 {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm0
; AVX512F-NEXT: vpsrlw $1, %xmm1, %xmm1
; AVX512F-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%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 {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm1, %zmm1
; AVX512F-NEXT: vpord %zmm1, %zmm0, %zmm0
; AVX512F-NEXT: vpmovdb %zmm0, %xmm0
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: vpunpckhbw {{.*#+}} xmm2 = xmm1[8],xmm0[8],xmm1[9],xmm0[9],xmm1[10],xmm0[10],xmm1[11],xmm0[11],xmm1[12],xmm0[12],xmm1[13],xmm0[13],xmm1[14],xmm0[14],xmm1[15],xmm0[15]
; AVX512F-NEXT: vpmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2
; AVX512F-NEXT: vpsrlw $8, %xmm2, %xmm2
; AVX512F-NEXT: vpunpcklbw {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
; AVX512F-NEXT: vpmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; AVX512F-NEXT: vpsrlw $8, %xmm0, %xmm0
; AVX512F-NEXT: vpackuswb %xmm2, %xmm0, %xmm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: constant_funnnel_v16i8:
; AVX512VL: # %bb.0:
; 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 {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm0
; AVX512VL-NEXT: vpsrlw $1, %xmm1, %xmm1
; AVX512VL-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%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 {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm1, %zmm1
; AVX512VL-NEXT: vpord %zmm1, %zmm0, %zmm0
; AVX512VL-NEXT: vpmovdb %zmm0, %xmm0
; AVX512VL-NEXT: vzeroupper
; AVX512VL-NEXT: vpunpckhbw {{.*#+}} xmm2 = xmm1[8],xmm0[8],xmm1[9],xmm0[9],xmm1[10],xmm0[10],xmm1[11],xmm0[11],xmm1[12],xmm0[12],xmm1[13],xmm0[13],xmm1[14],xmm0[14],xmm1[15],xmm0[15]
; AVX512VL-NEXT: vpmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2
; AVX512VL-NEXT: vpsrlw $8, %xmm2, %xmm2
; AVX512VL-NEXT: vpunpcklbw {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
; AVX512VL-NEXT: vpmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; AVX512VL-NEXT: vpsrlw $8, %xmm0, %xmm0
; AVX512VL-NEXT: vpackuswb %xmm2, %xmm0, %xmm0
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: constant_funnnel_v16i8:
......@@ -2360,27 +2291,15 @@ define <16 x i8> @constant_funnnel_v16i8(<16 x i8> %x, <16 x i8> %y) nounwind {
;
; X86-SSE2-LABEL: constant_funnnel_v16i8:
; X86-SSE2: # %bb.0:
; X86-SSE2-NEXT: movdqa %xmm0, %xmm2
; X86-SSE2-NEXT: punpckhbw {{.*#+}} xmm2 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
; X86-SSE2-NEXT: movdqa %xmm1, %xmm2
; X86-SSE2-NEXT: punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm0[8],xmm2[9],xmm0[9],xmm2[10],xmm0[10],xmm2[11],xmm0[11],xmm2[12],xmm0[12],xmm2[13],xmm0[13],xmm2[14],xmm0[14],xmm2[15],xmm0[15]
; X86-SSE2-NEXT: pmullw {{\.?LCPI[0-9]+_[0-9]+}}, %xmm2
; X86-SSE2-NEXT: movdqa {{.*#+}} xmm3 = [255,255,255,255,255,255,255,255]
; X86-SSE2-NEXT: pand %xmm3, %xmm2
; X86-SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
; X86-SSE2-NEXT: pmullw {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
; X86-SSE2-NEXT: pand %xmm3, %xmm0
; X86-SSE2-NEXT: packuswb %xmm2, %xmm0
; X86-SSE2-NEXT: psrlw $1, %xmm1
; X86-SSE2-NEXT: pand {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
; X86-SSE2-NEXT: pxor %xmm2, %xmm2
; X86-SSE2-NEXT: movdqa %xmm1, %xmm3
; X86-SSE2-NEXT: punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm2[8],xmm3[9],xmm2[9],xmm3[10],xmm2[10],xmm3[11],xmm2[11],xmm3[12],xmm2[12],xmm3[13],xmm2[13],xmm3[14],xmm2[14],xmm3[15],xmm2[15]
; X86-SSE2-NEXT: pmullw {{\.?LCPI[0-9]+_[0-9]+}}, %xmm3
; X86-SSE2-NEXT: psrlw $8, %xmm3
; X86-SSE2-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
; X86-SSE2-NEXT: psrlw $8, %xmm2
; X86-SSE2-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
; X86-SSE2-NEXT: pmullw {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
; X86-SSE2-NEXT: psrlw $8, %xmm1
; X86-SSE2-NEXT: packuswb %xmm3, %xmm1
; X86-SSE2-NEXT: por %xmm1, %xmm0
; X86-SSE2-NEXT: packuswb %xmm2, %xmm1
; X86-SSE2-NEXT: movdqa %xmm1, %xmm0
; X86-SSE2-NEXT: retl
%res = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %x, <16 x i8> %y, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1>)
ret <16 x i8> %res
......
......@@ -1935,118 +1935,57 @@ define <32 x i8> @constant_funnnel_v32i8(<32 x i8> %x, <32 x i8> %y) nounwind {
; AVX1-LABEL: constant_funnnel_v32i8:
; AVX1: # %bb.0:
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm2
; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm3 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
; AVX1-NEXT: vmovdqa {{.*#+}} xmm8 = [1,128,64,32,16,8,4,2]
; AVX1-NEXT: vpmullw %xmm3, %xmm8, %xmm3
; AVX1-NEXT: vmovdqa {{.*#+}} xmm5 = [255,255,255,255,255,255,255,255]
; AVX1-NEXT: vpand %xmm5, %xmm3, %xmm3
; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero
; AVX1-NEXT: vmovdqa {{.*#+}} xmm9 = [1,2,4,8,16,32,64,128]
; AVX1-NEXT: vpmullw %xmm2, %xmm9, %xmm2
; AVX1-NEXT: vpand %xmm5, %xmm2, %xmm2
; AVX1-NEXT: vpackuswb %xmm3, %xmm2, %xmm2
; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm3
; AVX1-NEXT: vpsrlw $1, %xmm3, %xmm3
; AVX1-NEXT: vmovdqa {{.*#+}} xmm10 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127]
; AVX1-NEXT: vpand %xmm3, %xmm10, %xmm3
; AVX1-NEXT: vpxor %xmm11, %xmm11, %xmm11
; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm3[8],xmm11[8],xmm3[9],xmm11[9],xmm3[10],xmm11[10],xmm3[11],xmm11[11],xmm3[12],xmm11[12],xmm3[13],xmm11[13],xmm3[14],xmm11[14],xmm3[15],xmm11[15]
; AVX1-NEXT: vmovdqa {{.*#+}} xmm7 = [2,256,128,64,32,16,8,4]
; AVX1-NEXT: vpmullw %xmm7, %xmm6, %xmm6
; AVX1-NEXT: vpsrlw $8, %xmm6, %xmm6
; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero
; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [2,4,8,16,32,64,128,256]
; AVX1-NEXT: vpmullw %xmm4, %xmm3, %xmm3
; AVX1-NEXT: vpsrlw $8, %xmm3, %xmm3
; AVX1-NEXT: vpackuswb %xmm6, %xmm3, %xmm3
; AVX1-NEXT: vpor %xmm3, %xmm2, %xmm2
; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm3 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
; AVX1-NEXT: vpmullw %xmm3, %xmm8, %xmm3
; AVX1-NEXT: vpand %xmm5, %xmm3, %xmm3
; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
; AVX1-NEXT: vpmullw %xmm0, %xmm9, %xmm0
; AVX1-NEXT: vpand %xmm5, %xmm0, %xmm0
; AVX1-NEXT: vpackuswb %xmm3, %xmm0, %xmm0
; AVX1-NEXT: vpsrlw $1, %xmm1, %xmm1
; AVX1-NEXT: vpand %xmm1, %xmm10, %xmm1
; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm3 = xmm1[8],xmm11[8],xmm1[9],xmm11[9],xmm1[10],xmm11[10],xmm1[11],xmm11[11],xmm1[12],xmm11[12],xmm1[13],xmm11[13],xmm1[14],xmm11[14],xmm1[15],xmm11[15]
; AVX1-NEXT: vpmullw %xmm7, %xmm3, %xmm3
; AVX1-NEXT: vpsrlw $8, %xmm3, %xmm3
; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
; AVX1-NEXT: vpmullw %xmm4, %xmm1, %xmm1
; AVX1-NEXT: vpsrlw $8, %xmm1, %xmm1
; AVX1-NEXT: vpackuswb %xmm3, %xmm1, %xmm1
; AVX1-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm4 = xmm3[8],xmm2[8],xmm3[9],xmm2[9],xmm3[10],xmm2[10],xmm3[11],xmm2[11],xmm3[12],xmm2[12],xmm3[13],xmm2[13],xmm3[14],xmm2[14],xmm3[15],xmm2[15]
; AVX1-NEXT: vmovdqa {{.*#+}} xmm5 = [1,128,64,32,16,8,4,2]
; AVX1-NEXT: vpmullw %xmm5, %xmm4, %xmm4
; AVX1-NEXT: vpsrlw $8, %xmm4, %xmm4
; AVX1-NEXT: vpunpcklbw {{.*#+}} xmm2 = 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]
; AVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [1,2,4,8,16,32,64,128]
; AVX1-NEXT: vpmullw %xmm3, %xmm2, %xmm2
; AVX1-NEXT: vpsrlw $8, %xmm2, %xmm2
; AVX1-NEXT: vpackuswb %xmm4, %xmm2, %xmm2
; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm4 = xmm1[8],xmm0[8],xmm1[9],xmm0[9],xmm1[10],xmm0[10],xmm1[11],xmm0[11],xmm1[12],xmm0[12],xmm1[13],xmm0[13],xmm1[14],xmm0[14],xmm1[15],xmm0[15]
; AVX1-NEXT: vpmullw %xmm5, %xmm4, %xmm4
; AVX1-NEXT: vpsrlw $8, %xmm4, %xmm4
; AVX1-NEXT: vpunpcklbw {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
; AVX1-NEXT: vpmullw %xmm3, %xmm0, %xmm0
; AVX1-NEXT: vpsrlw $8, %xmm0, %xmm0
; AVX1-NEXT: vpackuswb %xmm4, %xmm0, %xmm0
; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: constant_funnnel_v32i8:
; AVX2: # %bb.0:
; AVX2-NEXT: vpunpckhbw {{.*#+}} ymm2 = ymm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
; AVX2-NEXT: vpunpckhbw {{.*#+}} ymm2 = ymm1[8],ymm0[8],ymm1[9],ymm0[9],ymm1[10],ymm0[10],ymm1[11],ymm0[11],ymm1[12],ymm0[12],ymm1[13],ymm0[13],ymm1[14],ymm0[14],ymm1[15],ymm0[15],ymm1[24],ymm0[24],ymm1[25],ymm0[25],ymm1[26],ymm0[26],ymm1[27],ymm0[27],ymm1[28],ymm0[28],ymm1[29],ymm0[29],ymm1[30],ymm0[30],ymm1[31],ymm0[31]
; AVX2-NEXT: vpmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm2, %ymm2
; AVX2-NEXT: vmovdqa {{.*#+}} ymm3 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
; AVX2-NEXT: vpand %ymm3, %ymm2, %ymm2
; AVX2-NEXT: vpunpcklbw {{.*#+}} ymm0 = ymm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
; AVX2-NEXT: vpsrlw $8, %ymm2, %ymm2
; AVX2-NEXT: vpunpcklbw {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[4],ymm0[4],ymm1[5],ymm0[5],ymm1[6],ymm0[6],ymm1[7],ymm0[7],ymm1[16],ymm0[16],ymm1[17],ymm0[17],ymm1[18],ymm0[18],ymm1[19],ymm0[19],ymm1[20],ymm0[20],ymm1[21],ymm0[21],ymm1[22],ymm0[22],ymm1[23],ymm0[23]
; AVX2-NEXT: vpmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
; AVX2-NEXT: vpand %ymm3, %ymm0, %ymm0
; AVX2-NEXT: vpsrlw $8, %ymm0, %ymm0
; AVX2-NEXT: vpackuswb %ymm2, %ymm0, %ymm0
; AVX2-NEXT: vpsrlw $1, %ymm1, %ymm1
; AVX2-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
; AVX2-NEXT: vpunpckhbw {{.*#+}} ymm3 = ymm1[8],ymm2[8],ymm1[9],ymm2[9],ymm1[10],ymm2[10],ymm1[11],ymm2[11],ymm1[12],ymm2[12],ymm1[13],ymm2[13],ymm1[14],ymm2[14],ymm1[15],ymm2[15],ymm1[24],ymm2[24],ymm1[25],ymm2[25],ymm1[26],ymm2[26],ymm1[27],ymm2[27],ymm1[28],ymm2[28],ymm1[29],ymm2[29],ymm1[30],ymm2[30],ymm1[31],ymm2[31]
; AVX2-NEXT: vpmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm3, %ymm3
; AVX2-NEXT: vpsrlw $8, %ymm3, %ymm3
; AVX2-NEXT: vpunpcklbw {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[4],ymm2[4],ymm1[5],ymm2[5],ymm1[6],ymm2[6],ymm1[7],ymm2[7],ymm1[16],ymm2[16],ymm1[17],ymm2[17],ymm1[18],ymm2[18],ymm1[19],ymm2[19],ymm1[20],ymm2[20],ymm1[21],ymm2[21],ymm1[22],ymm2[22],ymm1[23],ymm2[23]
; AVX2-NEXT: vpmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
; AVX2-NEXT: vpsrlw $8, %ymm1, %ymm1
; AVX2-NEXT: vpackuswb %ymm3, %ymm1, %ymm1
; AVX2-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX2-NEXT: retq
;
; AVX512F-LABEL: constant_funnnel_v32i8:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vpunpckhbw {{.*#+}} ymm2 = ymm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
; AVX512F-NEXT: vpunpckhbw {{.*#+}} ymm2 = ymm1[8],ymm0[8],ymm1[9],ymm0[9],ymm1[10],ymm0[10],ymm1[11],ymm0[11],ymm1[12],ymm0[12],ymm1[13],ymm0[13],ymm1[14],ymm0[14],ymm1[15],ymm0[15],ymm1[24],ymm0[24],ymm1[25],ymm0[25],ymm1[26],ymm0[26],ymm1[27],ymm0[27],ymm1[28],ymm0[28],ymm1[29],ymm0[29],ymm1[30],ymm0[30],ymm1[31],ymm0[31]
; AVX512F-NEXT: vpmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm2, %ymm2
; AVX512F-NEXT: vmovdqa {{.*#+}} ymm3 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
; AVX512F-NEXT: vpand %ymm3, %ymm2, %ymm2
; AVX512F-NEXT: vpunpcklbw {{.*#+}} ymm0 = ymm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
; AVX512F-NEXT: vpsrlw $8, %ymm2, %ymm2
; AVX512F-NEXT: vpunpcklbw {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[4],ymm0[4],ymm1[5],ymm0[5],ymm1[6],ymm0[6],ymm1[7],ymm0[7],ymm1[16],ymm0[16],ymm1[17],ymm0[17],ymm1[18],ymm0[18],ymm1[19],ymm0[19],ymm1[20],ymm0[20],ymm1[21],ymm0[21],ymm1[22],ymm0[22],ymm1[23],ymm0[23]
; AVX512F-NEXT: vpmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
; AVX512F-NEXT: vpand %ymm3, %ymm0, %ymm0
; AVX512F-NEXT: vpsrlw $8, %ymm0, %ymm0
; AVX512F-NEXT: vpackuswb %ymm2, %ymm0, %ymm0
; AVX512F-NEXT: vpsrlw $1, %ymm1, %ymm1
; AVX512F-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
; AVX512F-NEXT: vpxor %xmm2, %xmm2, %xmm2
; AVX512F-NEXT: vpunpckhbw {{.*#+}} ymm3 = ymm1[8],ymm2[8],ymm1[9],ymm2[9],ymm1[10],ymm2[10],ymm1[11],ymm2[11],ymm1[12],ymm2[12],ymm1[13],ymm2[13],ymm1[14],ymm2[14],ymm1[15],ymm2[15],ymm1[24],ymm2[24],ymm1[25],ymm2[25],ymm1[26],ymm2[26],ymm1[27],ymm2[27],ymm1[28],ymm2[28],ymm1[29],ymm2[29],ymm1[30],ymm2[30],ymm1[31],ymm2[31]
; AVX512F-NEXT: vpmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm3, %ymm3
; AVX512F-NEXT: vpsrlw $8, %ymm3, %ymm3
; AVX512F-NEXT: vpunpcklbw {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[4],ymm2[4],ymm1[5],ymm2[5],ymm1[6],ymm2[6],ymm1[7],ymm2[7],ymm1[16],ymm2[16],ymm1[17],ymm2[17],ymm1[18],ymm2[18],ymm1[19],ymm2[19],ymm1[20],ymm2[20],ymm1[21],ymm2[21],ymm1[22],ymm2[22],ymm1[23],ymm2[23]
; AVX512F-NEXT: vpmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
; AVX512F-NEXT: vpsrlw $8, %ymm1, %ymm1
; AVX512F-NEXT: vpackuswb %ymm3, %ymm1, %ymm1
; AVX512F-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: constant_funnnel_v32i8:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpunpckhbw {{.*#+}} ymm2 = ymm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
; AVX512VL-NEXT: vpunpckhbw {{.*#+}} ymm2 = ymm1[8],ymm0[8],ymm1[9],ymm0[9],ymm1[10],ymm0[10],ymm1[11],ymm0[11],ymm1[12],ymm0[12],ymm1[13],ymm0[13],ymm1[14],ymm0[14],ymm1[15],ymm0[15],ymm1[24],ymm0[24],ymm1[25],ymm0[25],ymm1[26],ymm0[26],ymm1[27],ymm0[27],ymm1[28],ymm0[28],ymm1[29],ymm0[29],ymm1[30],ymm0[30],ymm1[31],ymm0[31]
; AVX512VL-NEXT: vpmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm2, %ymm2
; AVX512VL-NEXT: vmovdqa {{.*#+}} ymm3 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
; AVX512VL-NEXT: vpand %ymm3, %ymm2, %ymm2
; AVX512VL-NEXT: vpunpcklbw {{.*#+}} ymm0 = ymm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
; AVX512VL-NEXT: vpsrlw $8, %ymm2, %ymm2
; AVX512VL-NEXT: vpunpcklbw {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[4],ymm0[4],ymm1[5],ymm0[5],ymm1[6],ymm0[6],ymm1[7],ymm0[7],ymm1[16],ymm0[16],ymm1[17],ymm0[17],ymm1[18],ymm0[18],ymm1[19],ymm0[19],ymm1[20],ymm0[20],ymm1[21],ymm0[21],ymm1[22],ymm0[22],ymm1[23],ymm0[23]
; AVX512VL-NEXT: vpmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
; AVX512VL-NEXT: vpand %ymm3, %ymm0, %ymm0
; AVX512VL-NEXT: vpsrlw $8, %ymm0, %ymm0
; AVX512VL-NEXT: vpackuswb %ymm2, %ymm0, %ymm0
; AVX512VL-NEXT: vpsrlw $1, %ymm1, %ymm1
; AVX512VL-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
; AVX512VL-NEXT: vpxor %xmm2, %xmm2, %xmm2
; AVX512VL-NEXT: vpunpckhbw {{.*#+}} ymm3 = ymm1[8],ymm2[8],ymm1[9],ymm2[9],ymm1[10],ymm2[10],ymm1[11],ymm2[11],ymm1[12],ymm2[12],ymm1[13],ymm2[13],ymm1[14],ymm2[14],ymm1[15],ymm2[15],ymm1[24],ymm2[24],ymm1[25],ymm2[25],ymm1[26],ymm2[26],ymm1[27],ymm2[27],ymm1[28],ymm2[28],ymm1[29],ymm2[29],ymm1[30],ymm2[30],ymm1[31],ymm2[31]
; AVX512VL-NEXT: vpmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm3, %ymm3
; AVX512VL-NEXT: vpsrlw $8, %ymm3, %ymm3
; AVX512VL-NEXT: vpunpcklbw {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[4],ymm2[4],ymm1[5],ymm2[5],ymm1[6],ymm2[6],ymm1[7],ymm2[7],ymm1[16],ymm2[16],ymm1[17],ymm2[17],ymm1[18],ymm2[18],ymm1[19],ymm2[19],ymm1[20],ymm2[20],ymm1[21],ymm2[21],ymm1[22],ymm2[22],ymm1[23],ymm2[23]
; AVX512VL-NEXT: vpmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
; AVX512VL-NEXT: vpsrlw $8, %ymm1, %ymm1
; AVX512VL-NEXT: vpackuswb %ymm3, %ymm1, %ymm1
; AVX512VL-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: constant_funnnel_v32i8:
......
......@@ -911,105 +911,51 @@ define <64 x i8> @constant_funnnel_v64i8(<64 x i8> %x, <64 x i8> %y) nounwind {
; AVX512F-LABEL: constant_funnnel_v64i8:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vextracti64x4 $1, %zmm0, %ymm2
; AVX512F-NEXT: vpunpckhbw {{.*#+}} ymm3 = ymm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
; AVX512F-NEXT: vbroadcasti128 {{.*#+}} ymm4 = [1,128,64,32,16,8,4,2,1,128,64,32,16,8,4,2]
; AVX512F-NEXT: # ymm4 = mem[0,1,0,1]
; AVX512F-NEXT: vpmullw %ymm4, %ymm3, %ymm3
; AVX512F-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
; AVX512F-NEXT: vpand %ymm5, %ymm3, %ymm3
; AVX512F-NEXT: vpunpcklbw {{.*#+}} ymm2 = ymm2[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
; AVX512F-NEXT: vbroadcasti128 {{.*#+}} ymm6 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
; AVX512F-NEXT: # ymm6 = mem[0,1,0,1]
; AVX512F-NEXT: vpmullw %ymm6, %ymm2, %ymm2
; AVX512F-NEXT: vpand %ymm5, %ymm2, %ymm2
; AVX512F-NEXT: vpackuswb %ymm3, %ymm2, %ymm2
; AVX512F-NEXT: vpunpckhbw {{.*#+}} ymm3 = ymm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
; AVX512F-NEXT: vpmullw %ymm4, %ymm3, %ymm3
; AVX512F-NEXT: vpand %ymm5, %ymm3, %ymm3
; AVX512F-NEXT: vpunpcklbw {{.*#+}} ymm0 = ymm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
; AVX512F-NEXT: vpmullw %ymm6, %ymm0, %ymm0
; AVX512F-NEXT: vpand %ymm5, %ymm0, %ymm0
; AVX512F-NEXT: vpackuswb %ymm3, %ymm0, %ymm0
; AVX512F-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm0
; AVX512F-NEXT: vextracti64x4 $1, %zmm1, %ymm2
; AVX512F-NEXT: vpsrlw $1, %ymm2, %ymm2
; 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, %ymm2, %ymm2
; AVX512F-NEXT: vpxor %xmm4, %xmm4, %xmm4
; AVX512F-NEXT: vpunpckhbw {{.*#+}} ymm5 = ymm2[8],ymm4[8],ymm2[9],ymm4[9],ymm2[10],ymm4[10],ymm2[11],ymm4[11],ymm2[12],ymm4[12],ymm2[13],ymm4[13],ymm2[14],ymm4[14],ymm2[15],ymm4[15],ymm2[24],ymm4[24],ymm2[25],ymm4[25],ymm2[26],ymm4[26],ymm2[27],ymm4[27],ymm2[28],ymm4[28],ymm2[29],ymm4[29],ymm2[30],ymm4[30],ymm2[31],ymm4[31]
; AVX512F-NEXT: vbroadcasti128 {{.*#+}} ymm6 = [2,256,128,64,32,16,8,4,2,256,128,64,32,16,8,4]
; AVX512F-NEXT: # ymm6 = mem[0,1,0,1]
; AVX512F-NEXT: vpmullw %ymm6, %ymm5, %ymm5
; AVX512F-NEXT: vpsrlw $8, %ymm5, %ymm5
; AVX512F-NEXT: vpunpcklbw {{.*#+}} ymm2 = ymm2[0],ymm4[0],ymm2[1],ymm4[1],ymm2[2],ymm4[2],ymm2[3],ymm4[3],ymm2[4],ymm4[4],ymm2[5],ymm4[5],ymm2[6],ymm4[6],ymm2[7],ymm4[7],ymm2[16],ymm4[16],ymm2[17],ymm4[17],ymm2[18],ymm4[18],ymm2[19],ymm4[19],ymm2[20],ymm4[20],ymm2[21],ymm4[21],ymm2[22],ymm4[22],ymm2[23],ymm4[23]
; AVX512F-NEXT: vbroadcasti128 {{.*#+}} ymm7 = [2,4,8,16,32,64,128,256,2,4,8,16,32,64,128,256]
; AVX512F-NEXT: # ymm7 = mem[0,1,0,1]
; AVX512F-NEXT: vpmullw %ymm7, %ymm2, %ymm2
; AVX512F-NEXT: vextracti64x4 $1, %zmm1, %ymm3
; AVX512F-NEXT: vpunpckhbw {{.*#+}} ymm4 = ymm3[8],ymm2[8],ymm3[9],ymm2[9],ymm3[10],ymm2[10],ymm3[11],ymm2[11],ymm3[12],ymm2[12],ymm3[13],ymm2[13],ymm3[14],ymm2[14],ymm3[15],ymm2[15],ymm3[24],ymm2[24],ymm3[25],ymm2[25],ymm3[26],ymm2[26],ymm3[27],ymm2[27],ymm3[28],ymm2[28],ymm3[29],ymm2[29],ymm3[30],ymm2[30],ymm3[31],ymm2[31]
; AVX512F-NEXT: vbroadcasti128 {{.*#+}} ymm5 = [1,128,64,32,16,8,4,2,1,128,64,32,16,8,4,2]
; AVX512F-NEXT: # ymm5 = mem[0,1,0,1]
; AVX512F-NEXT: vpmullw %ymm5, %ymm4, %ymm4
; AVX512F-NEXT: vpsrlw $8, %ymm4, %ymm4
; AVX512F-NEXT: vpunpcklbw {{.*#+}} ymm2 = ymm3[0],ymm2[0],ymm3[1],ymm2[1],ymm3[2],ymm2[2],ymm3[3],ymm2[3],ymm3[4],ymm2[4],ymm3[5],ymm2[5],ymm3[6],ymm2[6],ymm3[7],ymm2[7],ymm3[16],ymm2[16],ymm3[17],ymm2[17],ymm3[18],ymm2[18],ymm3[19],ymm2[19],ymm3[20],ymm2[20],ymm3[21],ymm2[21],ymm3[22],ymm2[22],ymm3[23],ymm2[23]
; AVX512F-NEXT: vbroadcasti128 {{.*#+}} ymm3 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
; AVX512F-NEXT: # ymm3 = mem[0,1,0,1]
; AVX512F-NEXT: vpmullw %ymm3, %ymm2, %ymm2
; AVX512F-NEXT: vpsrlw $8, %ymm2, %ymm2
; AVX512F-NEXT: vpackuswb %ymm5, %ymm2, %ymm2
; AVX512F-NEXT: vpsrlw $1, %ymm1, %ymm1
; AVX512F-NEXT: vpand %ymm3, %ymm1, %ymm1
; AVX512F-NEXT: vpunpckhbw {{.*#+}} ymm3 = ymm1[8],ymm4[8],ymm1[9],ymm4[9],ymm1[10],ymm4[10],ymm1[11],ymm4[11],ymm1[12],ymm4[12],ymm1[13],ymm4[13],ymm1[14],ymm4[14],ymm1[15],ymm4[15],ymm1[24],ymm4[24],ymm1[25],ymm4[25],ymm1[26],ymm4[26],ymm1[27],ymm4[27],ymm1[28],ymm4[28],ymm1[29],ymm4[29],ymm1[30],ymm4[30],ymm1[31],ymm4[31]
; AVX512F-NEXT: vpmullw %ymm6, %ymm3, %ymm3
; AVX512F-NEXT: vpsrlw $8, %ymm3, %ymm3
; AVX512F-NEXT: vpunpcklbw {{.*#+}} ymm1 = ymm1[0],ymm4[0],ymm1[1],ymm4[1],ymm1[2],ymm4[2],ymm1[3],ymm4[3],ymm1[4],ymm4[4],ymm1[5],ymm4[5],ymm1[6],ymm4[6],ymm1[7],ymm4[7],ymm1[16],ymm4[16],ymm1[17],ymm4[17],ymm1[18],ymm4[18],ymm1[19],ymm4[19],ymm1[20],ymm4[20],ymm1[21],ymm4[21],ymm1[22],ymm4[22],ymm1[23],ymm4[23]
; AVX512F-NEXT: vpmullw %ymm7, %ymm1, %ymm1
; AVX512F-NEXT: vpsrlw $8, %ymm1, %ymm1
; AVX512F-NEXT: vpackuswb %ymm3, %ymm1, %ymm1
; AVX512F-NEXT: vinserti64x4 $1, %ymm2, %zmm1, %zmm1
; AVX512F-NEXT: vporq %zmm1, %zmm0, %zmm0
; AVX512F-NEXT: vpackuswb %ymm4, %ymm2, %ymm2
; AVX512F-NEXT: vpunpckhbw {{.*#+}} ymm4 = ymm1[8],ymm0[8],ymm1[9],ymm0[9],ymm1[10],ymm0[10],ymm1[11],ymm0[11],ymm1[12],ymm0[12],ymm1[13],ymm0[13],ymm1[14],ymm0[14],ymm1[15],ymm0[15],ymm1[24],ymm0[24],ymm1[25],ymm0[25],ymm1[26],ymm0[26],ymm1[27],ymm0[27],ymm1[28],ymm0[28],ymm1[29],ymm0[29],ymm1[30],ymm0[30],ymm1[31],ymm0[31]
; AVX512F-NEXT: vpmullw %ymm5, %ymm4, %ymm4
; AVX512F-NEXT: vpsrlw $8, %ymm4, %ymm4
; AVX512F-NEXT: vpunpcklbw {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[4],ymm0[4],ymm1[5],ymm0[5],ymm1[6],ymm0[6],ymm1[7],ymm0[7],ymm1[16],ymm0[16],ymm1[17],ymm0[17],ymm1[18],ymm0[18],ymm1[19],ymm0[19],ymm1[20],ymm0[20],ymm1[21],ymm0[21],ymm1[22],ymm0[22],ymm1[23],ymm0[23]
; AVX512F-NEXT: vpmullw %ymm3, %ymm0, %ymm0
; AVX512F-NEXT: vpsrlw $8, %ymm0, %ymm0
; AVX512F-NEXT: vpackuswb %ymm4, %ymm0, %ymm0
; AVX512F-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: constant_funnnel_v64i8:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vextracti64x4 $1, %zmm0, %ymm2
; AVX512VL-NEXT: vpunpckhbw {{.*#+}} ymm3 = ymm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
; AVX512VL-NEXT: vbroadcasti128 {{.*#+}} ymm4 = [1,128,64,32,16,8,4,2,1,128,64,32,16,8,4,2]
; AVX512VL-NEXT: # ymm4 = mem[0,1,0,1]
; AVX512VL-NEXT: vpmullw %ymm4, %ymm3, %ymm3
; AVX512VL-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
; AVX512VL-NEXT: vpand %ymm5, %ymm3, %ymm3
; AVX512VL-NEXT: vpunpcklbw {{.*#+}} ymm2 = ymm2[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
; AVX512VL-NEXT: vbroadcasti128 {{.*#+}} ymm6 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
; AVX512VL-NEXT: # ymm6 = mem[0,1,0,1]
; AVX512VL-NEXT: vpmullw %ymm6, %ymm2, %ymm2
; AVX512VL-NEXT: vpand %ymm5, %ymm2, %ymm2
; AVX512VL-NEXT: vpackuswb %ymm3, %ymm2, %ymm2
; AVX512VL-NEXT: vpunpckhbw {{.*#+}} ymm3 = ymm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
; AVX512VL-NEXT: vpmullw %ymm4, %ymm3, %ymm3
; AVX512VL-NEXT: vpand %ymm5, %ymm3, %ymm3
; AVX512VL-NEXT: vpunpcklbw {{.*#+}} ymm0 = ymm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
; AVX512VL-NEXT: vpmullw %ymm6, %ymm0, %ymm0
; AVX512VL-NEXT: vpand %ymm5, %ymm0, %ymm0
; AVX512VL-NEXT: vpackuswb %ymm3, %ymm0, %ymm0
; AVX512VL-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm0
; AVX512VL-NEXT: vextracti64x4 $1, %zmm1, %ymm2
; AVX512VL-NEXT: vpsrlw $1, %ymm2, %ymm2
; 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, %ymm2, %ymm2
; AVX512VL-NEXT: vpxor %xmm4, %xmm4, %xmm4
; AVX512VL-NEXT: vpunpckhbw {{.*#+}} ymm5 = ymm2[8],ymm4[8],ymm2[9],ymm4[9],ymm2[10],ymm4[10],ymm2[11],ymm4[11],ymm2[12],ymm4[12],ymm2[13],ymm4[13],ymm2[14],ymm4[14],ymm2[15],ymm4[15],ymm2[24],ymm4[24],ymm2[25],ymm4[25],ymm2[26],ymm4[26],ymm2[27],ymm4[27],ymm2[28],ymm4[28],ymm2[29],ymm4[29],ymm2[30],ymm4[30],ymm2[31],ymm4[31]
; AVX512VL-NEXT: vbroadcasti128 {{.*#+}} ymm6 = [2,256,128,64,32,16,8,4,2,256,128,64,32,16,8,4]
; AVX512VL-NEXT: # ymm6 = mem[0,1,0,1]
; AVX512VL-NEXT: vpmullw %ymm6, %ymm5, %ymm5
; AVX512VL-NEXT: vpsrlw $8, %ymm5, %ymm5
; AVX512VL-NEXT: vpunpcklbw {{.*#+}} ymm2 = ymm2[0],ymm4[0],ymm2[1],ymm4[1],ymm2[2],ymm4[2],ymm2[3],ymm4[3],ymm2[4],ymm4[4],ymm2[5],ymm4[5],ymm2[6],ymm4[6],ymm2[7],ymm4[7],ymm2[16],ymm4[16],ymm2[17],ymm4[17],ymm2[18],ymm4[18],ymm2[19],ymm4[19],ymm2[20],ymm4[20],ymm2[21],ymm4[21],ymm2[22],ymm4[22],ymm2[23],ymm4[23]
; AVX512VL-NEXT: vbroadcasti128 {{.*#+}} ymm7 = [2,4,8,16,32,64,128,256,2,4,8,16,32,64,128,256]
; AVX512VL-NEXT: # ymm7 = mem[0,1,0,1]
; AVX512VL-NEXT: vpmullw %ymm7, %ymm2, %ymm2
; AVX512VL-NEXT: vextracti64x4 $1, %zmm1, %ymm3
; AVX512VL-NEXT: vpunpckhbw {{.*#+}} ymm4 = ymm3[8],ymm2[8],ymm3[9],ymm2[9],ymm3[10],ymm2[10],ymm3[11],ymm2[11],ymm3[12],ymm2[12],ymm3[13],ymm2[13],ymm3[14],ymm2[14],ymm3[15],ymm2[15],ymm3[24],ymm2[24],ymm3[25],ymm2[25],ymm3[26],ymm2[26],ymm3[27],ymm2[27],ymm3[28],ymm2[28],ymm3[29],ymm2[29],ymm3[30],ymm2[30],ymm3[31],ymm2[31]
; AVX512VL-NEXT: vbroadcasti128 {{.*#+}} ymm5 = [1,128,64,32,16,8,4,2,1,128,64,32,16,8,4,2]
; AVX512VL-NEXT: # ymm5 = mem[0,1,0,1]
; AVX512VL-NEXT: vpmullw %ymm5, %ymm4, %ymm4
; AVX512VL-NEXT: vpsrlw $8, %ymm4, %ymm4
; AVX512VL-NEXT: vpunpcklbw {{.*#+}} ymm2 = ymm3[0],ymm2[0],ymm3[1],ymm2[1],ymm3[2],ymm2[2],ymm3[3],ymm2[3],ymm3[4],ymm2[4],ymm3[5],ymm2[5],ymm3[6],ymm2[6],ymm3[7],ymm2[7],ymm3[16],ymm2[16],ymm3[17],ymm2[17],ymm3[18],ymm2[18],ymm3[19],ymm2[19],ymm3[20],ymm2[20],ymm3[21],ymm2[21],ymm3[22],ymm2[22],ymm3[23],ymm2[23]
; AVX512VL-NEXT: vbroadcasti128 {{.*#+}} ymm3 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
; AVX512VL-NEXT: # ymm3 = mem[0,1,0,1]
; AVX512VL-NEXT: vpmullw %ymm3, %ymm2, %ymm2
; AVX512VL-NEXT: vpsrlw $8, %ymm2, %ymm2
; AVX512VL-NEXT: vpackuswb %ymm5, %ymm2, %ymm2
; AVX512VL-NEXT: vpsrlw $1, %ymm1, %ymm1
; AVX512VL-NEXT: vpand %ymm3, %ymm1, %ymm1
; AVX512VL-NEXT: vpunpckhbw {{.*#+}} ymm3 = ymm1[8],ymm4[8],ymm1[9],ymm4[9],ymm1[10],ymm4[10],ymm1[11],ymm4[11],ymm1[12],ymm4[12],ymm1[13],ymm4[13],ymm1[14],ymm4[14],ymm1[15],ymm4[15],ymm1[24],ymm4[24],ymm1[25],ymm4[25],ymm1[26],ymm4[26],ymm1[27],ymm4[27],ymm1[28],ymm4[28],ymm1[29],ymm4[29],ymm1[30],ymm4[30],ymm1[31],ymm4[31]
; AVX512VL-NEXT: vpmullw %ymm6, %ymm3, %ymm3
; AVX512VL-NEXT: vpsrlw $8, %ymm3, %ymm3
; AVX512VL-NEXT: vpunpcklbw {{.*#+}} ymm1 = ymm1[0],ymm4[0],ymm1[1],ymm4[1],ymm1[2],ymm4[2],ymm1[3],ymm4[3],ymm1[4],ymm4[4],ymm1[5],ymm4[5],ymm1[6],ymm4[6],ymm1[7],ymm4[7],ymm1[16],ymm4[16],ymm1[17],ymm4[17],ymm1[18],ymm4[18],ymm1[19],ymm4[19],ymm1[20],ymm4[20],ymm1[21],ymm4[21],ymm1[22],ymm4[22],ymm1[23],ymm4[23]
; AVX512VL-NEXT: vpmullw %ymm7, %ymm1, %ymm1
; AVX512VL-NEXT: vpsrlw $8, %ymm1, %ymm1
; AVX512VL-NEXT: vpackuswb %ymm3, %ymm1, %ymm1
; AVX512VL-NEXT: vinserti64x4 $1, %ymm2, %zmm1, %zmm1
; AVX512VL-NEXT: vporq %zmm1, %zmm0, %zmm0
; AVX512VL-NEXT: vpackuswb %ymm4, %ymm2, %ymm2
; AVX512VL-NEXT: vpunpckhbw {{.*#+}} ymm4 = ymm1[8],ymm0[8],ymm1[9],ymm0[9],ymm1[10],ymm0[10],ymm1[11],ymm0[11],ymm1[12],ymm0[12],ymm1[13],ymm0[13],ymm1[14],ymm0[14],ymm1[15],ymm0[15],ymm1[24],ymm0[24],ymm1[25],ymm0[25],ymm1[26],ymm0[26],ymm1[27],ymm0[27],ymm1[28],ymm0[28],ymm1[29],ymm0[29],ymm1[30],ymm0[30],ymm1[31],ymm0[31]
; AVX512VL-NEXT: vpmullw %ymm5, %ymm4, %ymm4
; AVX512VL-NEXT: vpsrlw $8, %ymm4, %ymm4
; AVX512VL-NEXT: vpunpcklbw {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[4],ymm0[4],ymm1[5],ymm0[5],ymm1[6],ymm0[6],ymm1[7],ymm0[7],ymm1[16],ymm0[16],ymm1[17],ymm0[17],ymm1[18],ymm0[18],ymm1[19],ymm0[19],ymm1[20],ymm0[20],ymm1[21],ymm0[21],ymm1[22],ymm0[22],ymm1[23],ymm0[23]
; AVX512VL-NEXT: vpmullw %ymm3, %ymm0, %ymm0
; AVX512VL-NEXT: vpsrlw $8, %ymm0, %ymm0
; AVX512VL-NEXT: vpackuswb %ymm4, %ymm0, %ymm0
; AVX512VL-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm0
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: constant_funnnel_v64i8:
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment