[X86][AVX512] Autoupgrade the VPERMILPD/VPERMILPS intrinsics
llvm-svn: 274498
This commit is contained in:
Simon Pilgrim
parent
e617adea12
commit
9fca300cbe
|
|
@ -225,6 +225,7 @@ static bool UpgradeIntrinsicFunction1(Function *F, Function *&NewFn) {
|
|||
Name.startswith("x86.avx512.mask.pshuf.d.") ||
|
||||
Name.startswith("x86.avx512.mask.pshufl.w.") ||
|
||||
Name.startswith("x86.avx512.mask.pshufh.w.") ||
|
||||
Name.startswith("x86.avx512.mask.vpermil.p") ||
|
||||
Name.startswith("x86.avx512.mask.punpckl") ||
|
||||
Name.startswith("x86.avx512.mask.punpckh") ||
|
||||
Name.startswith("x86.avx512.mask.unpckl.") ||
|
||||
|
|
@ -1007,12 +1008,13 @@ void llvm::UpgradeIntrinsicCall(CallInst *CI, Function *NewFn) {
|
|||
Rep = nullptr;
|
||||
} else if (Name.startswith("llvm.x86.avx.vpermil.") ||
|
||||
Name == "llvm.x86.sse2.pshuf.d" ||
|
||||
Name.startswith("llvm.x86.avx512.mask.vpermil.p") ||
|
||||
Name.startswith("llvm.x86.avx512.mask.pshuf.d.")) {
|
||||
Value *Op0 = CI->getArgOperand(0);
|
||||
unsigned Imm = cast<ConstantInt>(CI->getArgOperand(1))->getZExtValue();
|
||||
VectorType *VecTy = cast<VectorType>(CI->getType());
|
||||
unsigned NumElts = VecTy->getNumElements();
|
||||
// Calcuate the size of each index in the immediate.
|
||||
// Calculate the size of each index in the immediate.
|
||||
unsigned IdxSize = 64 / VecTy->getScalarSizeInBits();
|
||||
unsigned IdxMask = ((1 << IdxSize) - 1);
|
||||
|
||||
|
|
|
|||
|
|
@ -317,6 +317,46 @@ define <8 x i64> @test_mask_load_aligned_q(<8 x i64> %data, i8* %ptr, i8 %mask)
|
|||
ret <8 x i64> %res4
|
||||
}
|
||||
|
||||
declare <8 x double> @llvm.x86.avx512.mask.vpermil.pd.512(<8 x double>, i32, <8 x double>, i8)
|
||||
|
||||
define <8 x double>@test_int_x86_avx512_mask_vpermil_pd_512(<8 x double> %x0, <8 x double> %x2, i8 %x3) {
|
||||
; CHECK-LABEL: test_int_x86_avx512_mask_vpermil_pd_512:
|
||||
; CHECK: ## BB#0:
|
||||
; CHECK-NEXT: vpermilpd {{.*#+}} zmm2 = zmm0[0,1,3,2,5,4,6,6]
|
||||
; CHECK-NEXT: kmovw %edi, %k1
|
||||
; CHECK-NEXT: vpermilpd {{.*#+}} zmm1 {%k1} = zmm0[0,1,3,2,5,4,6,6]
|
||||
; CHECK-NEXT: vpermilpd {{.*#+}} zmm0 {%k1} {z} = zmm0[0,1,3,2,5,4,6,6]
|
||||
; CHECK-NEXT: vaddpd %zmm0, %zmm1, %zmm0
|
||||
; CHECK-NEXT: vaddpd %zmm2, %zmm0, %zmm0
|
||||
; CHECK-NEXT: retq
|
||||
%res = call <8 x double> @llvm.x86.avx512.mask.vpermil.pd.512(<8 x double> %x0, i32 22, <8 x double> %x2, i8 %x3)
|
||||
%res1 = call <8 x double> @llvm.x86.avx512.mask.vpermil.pd.512(<8 x double> %x0, i32 22, <8 x double> zeroinitializer, i8 %x3)
|
||||
%res2 = call <8 x double> @llvm.x86.avx512.mask.vpermil.pd.512(<8 x double> %x0, i32 22, <8 x double> %x2, i8 -1)
|
||||
%res3 = fadd <8 x double> %res, %res1
|
||||
%res4 = fadd <8 x double> %res3, %res2
|
||||
ret <8 x double> %res4
|
||||
}
|
||||
|
||||
declare <16 x float> @llvm.x86.avx512.mask.vpermil.ps.512(<16 x float>, i32, <16 x float>, i16)
|
||||
|
||||
define <16 x float>@test_int_x86_avx512_mask_vpermil_ps_512(<16 x float> %x0, <16 x float> %x2, i16 %x3) {
|
||||
; CHECK-LABEL: test_int_x86_avx512_mask_vpermil_ps_512:
|
||||
; CHECK: ## BB#0:
|
||||
; CHECK-NEXT: vpermilps {{.*#+}} zmm2 = zmm0[2,1,1,0,6,5,5,4,10,9,9,8,14,13,13,12]
|
||||
; CHECK-NEXT: kmovw %edi, %k1
|
||||
; CHECK-NEXT: vpermilps {{.*#+}} zmm1 {%k1} = zmm0[2,1,1,0,6,5,5,4,10,9,9,8,14,13,13,12]
|
||||
; CHECK-NEXT: vpermilps {{.*#+}} zmm0 {%k1} {z} = zmm0[2,1,1,0,6,5,5,4,10,9,9,8,14,13,13,12]
|
||||
; CHECK-NEXT: vaddps %zmm0, %zmm1, %zmm0
|
||||
; CHECK-NEXT: vaddps %zmm2, %zmm0, %zmm0
|
||||
; CHECK-NEXT: retq
|
||||
%res = call <16 x float> @llvm.x86.avx512.mask.vpermil.ps.512(<16 x float> %x0, i32 22, <16 x float> %x2, i16 %x3)
|
||||
%res1 = call <16 x float> @llvm.x86.avx512.mask.vpermil.ps.512(<16 x float> %x0, i32 22, <16 x float> zeroinitializer, i16 %x3)
|
||||
%res2 = call <16 x float> @llvm.x86.avx512.mask.vpermil.ps.512(<16 x float> %x0, i32 22, <16 x float> %x2, i16 -1)
|
||||
%res3 = fadd <16 x float> %res, %res1
|
||||
%res4 = fadd <16 x float> %res3, %res2
|
||||
ret <16 x float> %res4
|
||||
}
|
||||
|
||||
declare <16 x i32> @llvm.x86.avx512.mask.pshuf.d.512(<16 x i32>, i32, <16 x i32>, i16)
|
||||
|
||||
define <16 x i32>@test_int_x86_avx512_mask_pshuf_d_512(<16 x i32> %x0, i32 %x1, <16 x i32> %x2, i16 %x3) {
|
||||
|
|
|
|||
|
|
@ -5438,46 +5438,6 @@ define <16 x float>@test_int_x86_avx512_mask_shuf_ps_512(<16 x float> %x0, <16 x
|
|||
ret <16 x float> %res2
|
||||
}
|
||||
|
||||
declare <8 x double> @llvm.x86.avx512.mask.vpermil.pd.512(<8 x double>, i32, <8 x double>, i8)
|
||||
|
||||
define <8 x double>@test_int_x86_avx512_mask_vpermil_pd_512(<8 x double> %x0, <8 x double> %x2, i8 %x3) {
|
||||
; CHECK-LABEL: test_int_x86_avx512_mask_vpermil_pd_512:
|
||||
; CHECK: ## BB#0:
|
||||
; CHECK-NEXT: kmovw %edi, %k1
|
||||
; CHECK-NEXT: vpermilpd {{.*#+}} zmm1 {%k1} = zmm0[0,1,3,2,5,4,6,6]
|
||||
; CHECK-NEXT: vpermilpd {{.*#+}} zmm2 {%k1} {z} = zmm0[0,1,3,2,5,4,6,6]
|
||||
; CHECK-NEXT: vpermilpd {{.*#+}} zmm0 = zmm0[0,1,3,2,5,4,6,6]
|
||||
; CHECK-NEXT: vaddpd %zmm2, %zmm1, %zmm1
|
||||
; CHECK-NEXT: vaddpd %zmm0, %zmm1, %zmm0
|
||||
; CHECK-NEXT: retq
|
||||
%res = call <8 x double> @llvm.x86.avx512.mask.vpermil.pd.512(<8 x double> %x0, i32 22, <8 x double> %x2, i8 %x3)
|
||||
%res1 = call <8 x double> @llvm.x86.avx512.mask.vpermil.pd.512(<8 x double> %x0, i32 22, <8 x double> zeroinitializer, i8 %x3)
|
||||
%res2 = call <8 x double> @llvm.x86.avx512.mask.vpermil.pd.512(<8 x double> %x0, i32 22, <8 x double> %x2, i8 -1)
|
||||
%res3 = fadd <8 x double> %res, %res1
|
||||
%res4 = fadd <8 x double> %res3, %res2
|
||||
ret <8 x double> %res4
|
||||
}
|
||||
|
||||
declare <16 x float> @llvm.x86.avx512.mask.vpermil.ps.512(<16 x float>, i32, <16 x float>, i16)
|
||||
|
||||
define <16 x float>@test_int_x86_avx512_mask_vpermil_ps_512(<16 x float> %x0, <16 x float> %x2, i16 %x3) {
|
||||
; CHECK-LABEL: test_int_x86_avx512_mask_vpermil_ps_512:
|
||||
; CHECK: ## BB#0:
|
||||
; CHECK-NEXT: kmovw %edi, %k1
|
||||
; CHECK-NEXT: vpermilps {{.*#+}} zmm1 {%k1} = zmm0[2,1,1,0,6,5,5,4,10,9,9,8,14,13,13,12]
|
||||
; CHECK-NEXT: vpermilps {{.*#+}} zmm2 {%k1} {z} = zmm0[2,1,1,0,6,5,5,4,10,9,9,8,14,13,13,12]
|
||||
; CHECK-NEXT: vpermilps {{.*#+}} zmm0 = zmm0[2,1,1,0,6,5,5,4,10,9,9,8,14,13,13,12]
|
||||
; CHECK-NEXT: vaddps %zmm2, %zmm1, %zmm1
|
||||
; CHECK-NEXT: vaddps %zmm0, %zmm1, %zmm0
|
||||
; CHECK-NEXT: retq
|
||||
%res = call <16 x float> @llvm.x86.avx512.mask.vpermil.ps.512(<16 x float> %x0, i32 22, <16 x float> %x2, i16 %x3)
|
||||
%res1 = call <16 x float> @llvm.x86.avx512.mask.vpermil.ps.512(<16 x float> %x0, i32 22, <16 x float> zeroinitializer, i16 %x3)
|
||||
%res2 = call <16 x float> @llvm.x86.avx512.mask.vpermil.ps.512(<16 x float> %x0, i32 22, <16 x float> %x2, i16 -1)
|
||||
%res3 = fadd <16 x float> %res, %res1
|
||||
%res4 = fadd <16 x float> %res3, %res2
|
||||
ret <16 x float> %res4
|
||||
}
|
||||
|
||||
declare <8 x double> @llvm.x86.avx512.mask.vpermilvar.pd.512(<8 x double>, <8 x i64>, <8 x double>, i8)
|
||||
|
||||
define <8 x double>@test_int_x86_avx512_mask_vpermilvar_pd_512(<8 x double> %x0, <8 x i64> %x1, <8 x double> %x2, i8 %x3) {
|
||||
|
|
|
|||
|
|
@ -138,6 +138,98 @@ define <4 x double>@test_int_x86_avx512_mask_movddup_256(<4 x double> %x0, <4 x
|
|||
ret <4 x double> %res4
|
||||
}
|
||||
|
||||
declare <4 x double> @llvm.x86.avx512.mask.vpermil.pd.256(<4 x double>, i32, <4 x double>, i8)
|
||||
|
||||
define <4 x double>@test_int_x86_avx512_mask_vpermil_pd_256(<4 x double> %x0, <4 x double> %x2, i8 %x3) {
|
||||
; CHECK-LABEL: test_int_x86_avx512_mask_vpermil_pd_256:
|
||||
; CHECK: ## BB#0:
|
||||
; CHECK-NEXT: vpermilpd $6, %ymm0, %ymm2 ## encoding: [0x62,0xf3,0xfd,0x28,0x05,0xd0,0x06]
|
||||
; CHECK-NEXT: ## ymm2 = ymm0[0,1,3,2]
|
||||
; CHECK-NEXT: kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
|
||||
; CHECK-NEXT: vpermilpd $6, %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x05,0xc8,0x06]
|
||||
; CHECK-NEXT: ## ymm1 {%k1} = ymm0[0,1,3,2]
|
||||
; CHECK-NEXT: vpermilpd $6, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf3,0xfd,0xa9,0x05,0xc0,0x06]
|
||||
; CHECK-NEXT: ## ymm0 {%k1} {z} = ymm0[0,1,3,2]
|
||||
; CHECK-NEXT: vaddpd %ymm0, %ymm1, %ymm0 ## encoding: [0x62,0xf1,0xf5,0x28,0x58,0xc0]
|
||||
; CHECK-NEXT: vaddpd %ymm0, %ymm2, %ymm0 ## encoding: [0x62,0xf1,0xed,0x28,0x58,0xc0]
|
||||
; CHECK-NEXT: retq ## encoding: [0xc3]
|
||||
%res = call <4 x double> @llvm.x86.avx512.mask.vpermil.pd.256(<4 x double> %x0, i32 22, <4 x double> %x2, i8 %x3)
|
||||
%res1 = call <4 x double> @llvm.x86.avx512.mask.vpermil.pd.256(<4 x double> %x0, i32 22, <4 x double> zeroinitializer, i8 %x3)
|
||||
%res2 = call <4 x double> @llvm.x86.avx512.mask.vpermil.pd.256(<4 x double> %x0, i32 22, <4 x double> %x2, i8 -1)
|
||||
%res3 = fadd <4 x double> %res, %res1
|
||||
%res4 = fadd <4 x double> %res2, %res3
|
||||
ret <4 x double> %res4
|
||||
}
|
||||
|
||||
declare <2 x double> @llvm.x86.avx512.mask.vpermil.pd.128(<2 x double>, i32, <2 x double>, i8)
|
||||
|
||||
define <2 x double>@test_int_x86_avx512_mask_vpermil_pd_128(<2 x double> %x0, <2 x double> %x2, i8 %x3) {
|
||||
; CHECK-LABEL: test_int_x86_avx512_mask_vpermil_pd_128:
|
||||
; CHECK: ## BB#0:
|
||||
; CHECK-NEXT: vpermilpd $1, %xmm0, %xmm2 ## encoding: [0x62,0xf3,0xfd,0x08,0x05,0xd0,0x01]
|
||||
; CHECK-NEXT: ## xmm2 = xmm0[1,0]
|
||||
; CHECK-NEXT: kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
|
||||
; CHECK-NEXT: vpermilpd $1, %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x05,0xc8,0x01]
|
||||
; CHECK-NEXT: ## xmm1 {%k1} = xmm0[1,0]
|
||||
; CHECK-NEXT: vpermilpd $1, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf3,0xfd,0x89,0x05,0xc0,0x01]
|
||||
; CHECK-NEXT: ## xmm0 {%k1} {z} = xmm0[1,0]
|
||||
; CHECK-NEXT: vaddpd %xmm0, %xmm1, %xmm0 ## encoding: [0x62,0xf1,0xf5,0x08,0x58,0xc0]
|
||||
; CHECK-NEXT: vaddpd %xmm2, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0xfd,0x08,0x58,0xc2]
|
||||
; CHECK-NEXT: retq ## encoding: [0xc3]
|
||||
%res = call <2 x double> @llvm.x86.avx512.mask.vpermil.pd.128(<2 x double> %x0, i32 1, <2 x double> %x2, i8 %x3)
|
||||
%res1 = call <2 x double> @llvm.x86.avx512.mask.vpermil.pd.128(<2 x double> %x0, i32 1, <2 x double> zeroinitializer, i8 %x3)
|
||||
%res2 = call <2 x double> @llvm.x86.avx512.mask.vpermil.pd.128(<2 x double> %x0, i32 1, <2 x double> %x2, i8 -1)
|
||||
%res3 = fadd <2 x double> %res, %res1
|
||||
%res4 = fadd <2 x double> %res3, %res2
|
||||
ret <2 x double> %res4
|
||||
}
|
||||
|
||||
declare <8 x float> @llvm.x86.avx512.mask.vpermil.ps.256(<8 x float>, i32, <8 x float>, i8)
|
||||
|
||||
define <8 x float>@test_int_x86_avx512_mask_vpermil_ps_256(<8 x float> %x0, <8 x float> %x2, i8 %x3) {
|
||||
; CHECK-LABEL: test_int_x86_avx512_mask_vpermil_ps_256:
|
||||
; CHECK: ## BB#0:
|
||||
; CHECK-NEXT: vpermilps $22, %ymm0, %ymm2 ## encoding: [0x62,0xf3,0x7d,0x28,0x04,0xd0,0x16]
|
||||
; CHECK-NEXT: ## ymm2 = ymm0[2,1,1,0,6,5,5,4]
|
||||
; CHECK-NEXT: kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
|
||||
; CHECK-NEXT: vpermilps $22, %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x04,0xc8,0x16]
|
||||
; CHECK-NEXT: ## ymm1 {%k1} = ymm0[2,1,1,0,6,5,5,4]
|
||||
; CHECK-NEXT: vpermilps $22, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf3,0x7d,0xa9,0x04,0xc0,0x16]
|
||||
; CHECK-NEXT: ## ymm0 {%k1} {z} = ymm0[2,1,1,0,6,5,5,4]
|
||||
; CHECK-NEXT: vaddps %ymm0, %ymm1, %ymm0 ## encoding: [0x62,0xf1,0x74,0x28,0x58,0xc0]
|
||||
; CHECK-NEXT: vaddps %ymm2, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7c,0x28,0x58,0xc2]
|
||||
; CHECK-NEXT: retq ## encoding: [0xc3]
|
||||
%res = call <8 x float> @llvm.x86.avx512.mask.vpermil.ps.256(<8 x float> %x0, i32 22, <8 x float> %x2, i8 %x3)
|
||||
%res1 = call <8 x float> @llvm.x86.avx512.mask.vpermil.ps.256(<8 x float> %x0, i32 22, <8 x float> zeroinitializer, i8 %x3)
|
||||
%res2 = call <8 x float> @llvm.x86.avx512.mask.vpermil.ps.256(<8 x float> %x0, i32 22, <8 x float> %x2, i8 -1)
|
||||
%res3 = fadd <8 x float> %res, %res1
|
||||
%res4 = fadd <8 x float> %res3, %res2
|
||||
ret <8 x float> %res4
|
||||
}
|
||||
|
||||
declare <4 x float> @llvm.x86.avx512.mask.vpermil.ps.128(<4 x float>, i32, <4 x float>, i8)
|
||||
|
||||
define <4 x float>@test_int_x86_avx512_mask_vpermil_ps_128(<4 x float> %x0, <4 x float> %x2, i8 %x3) {
|
||||
; CHECK-LABEL: test_int_x86_avx512_mask_vpermil_ps_128:
|
||||
; CHECK: ## BB#0:
|
||||
; CHECK-NEXT: vpermilps $22, %xmm0, %xmm2 ## encoding: [0x62,0xf3,0x7d,0x08,0x04,0xd0,0x16]
|
||||
; CHECK-NEXT: ## xmm2 = xmm0[2,1,1,0]
|
||||
; CHECK-NEXT: kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
|
||||
; CHECK-NEXT: vpermilps $22, %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x04,0xc8,0x16]
|
||||
; CHECK-NEXT: ## xmm1 {%k1} = xmm0[2,1,1,0]
|
||||
; CHECK-NEXT: vpermilps $22, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf3,0x7d,0x89,0x04,0xc0,0x16]
|
||||
; CHECK-NEXT: ## xmm0 {%k1} {z} = xmm0[2,1,1,0]
|
||||
; CHECK-NEXT: vaddps %xmm0, %xmm1, %xmm0 ## encoding: [0x62,0xf1,0x74,0x08,0x58,0xc0]
|
||||
; CHECK-NEXT: vaddps %xmm0, %xmm2, %xmm0 ## encoding: [0x62,0xf1,0x6c,0x08,0x58,0xc0]
|
||||
; CHECK-NEXT: retq ## encoding: [0xc3]
|
||||
%res = call <4 x float> @llvm.x86.avx512.mask.vpermil.ps.128(<4 x float> %x0, i32 22, <4 x float> %x2, i8 %x3)
|
||||
%res1 = call <4 x float> @llvm.x86.avx512.mask.vpermil.ps.128(<4 x float> %x0, i32 22, <4 x float> zeroinitializer, i8 %x3)
|
||||
%res2 = call <4 x float> @llvm.x86.avx512.mask.vpermil.ps.128(<4 x float> %x0, i32 22, <4 x float> %x2, i8 -1)
|
||||
%res3 = fadd <4 x float> %res, %res1
|
||||
%res4 = fadd <4 x float> %res2, %res3
|
||||
ret <4 x float> %res4
|
||||
}
|
||||
|
||||
declare void @llvm.x86.avx512.mask.store.pd.128(i8*, <2 x double>, i8)
|
||||
|
||||
define void@test_int_x86_avx512_mask_store_pd_128(i8* %ptr1, i8* %ptr2, <2 x double> %x1, i8 %x2) {
|
||||
|
|
|
|||
|
|
@ -5722,98 +5722,6 @@ define <4 x i64>@test_int_x86_avx512_mask_valign_q_256(<4 x i64> %x0, <4 x i64>
|
|||
ret <4 x i64> %res2
|
||||
}
|
||||
|
||||
declare <4 x double> @llvm.x86.avx512.mask.vpermil.pd.256(<4 x double>, i32, <4 x double>, i8)
|
||||
|
||||
define <4 x double>@test_int_x86_avx512_mask_vpermil_pd_256(<4 x double> %x0, <4 x double> %x2, i8 %x3) {
|
||||
; CHECK-LABEL: test_int_x86_avx512_mask_vpermil_pd_256:
|
||||
; CHECK: ## BB#0:
|
||||
; CHECK-NEXT: kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
|
||||
; CHECK-NEXT: vpermilpd $22, %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x05,0xc8,0x16]
|
||||
; CHECK-NEXT: ## ymm1 {%k1} = ymm0[0,1,3,2]
|
||||
; CHECK-NEXT: vpermilpd $22, %ymm0, %ymm2 {%k1} {z} ## encoding: [0x62,0xf3,0xfd,0xa9,0x05,0xd0,0x16]
|
||||
; CHECK-NEXT: ## ymm2 {%k1} {z} = ymm0[0,1,3,2]
|
||||
; CHECK-NEXT: vpermilpd $22, %ymm0, %ymm0 ## encoding: [0x62,0xf3,0xfd,0x28,0x05,0xc0,0x16]
|
||||
; CHECK-NEXT: ## ymm0 = ymm0[0,1,3,2]
|
||||
; CHECK-NEXT: vaddpd %ymm2, %ymm1, %ymm1 ## encoding: [0x62,0xf1,0xf5,0x28,0x58,0xca]
|
||||
; CHECK-NEXT: vaddpd %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0xfd,0x28,0x58,0xc1]
|
||||
; CHECK-NEXT: retq ## encoding: [0xc3]
|
||||
%res = call <4 x double> @llvm.x86.avx512.mask.vpermil.pd.256(<4 x double> %x0, i32 22, <4 x double> %x2, i8 %x3)
|
||||
%res1 = call <4 x double> @llvm.x86.avx512.mask.vpermil.pd.256(<4 x double> %x0, i32 22, <4 x double> zeroinitializer, i8 %x3)
|
||||
%res2 = call <4 x double> @llvm.x86.avx512.mask.vpermil.pd.256(<4 x double> %x0, i32 22, <4 x double> %x2, i8 -1)
|
||||
%res3 = fadd <4 x double> %res, %res1
|
||||
%res4 = fadd <4 x double> %res2, %res3
|
||||
ret <4 x double> %res4
|
||||
}
|
||||
|
||||
declare <2 x double> @llvm.x86.avx512.mask.vpermil.pd.128(<2 x double>, i32, <2 x double>, i8)
|
||||
|
||||
define <2 x double>@test_int_x86_avx512_mask_vpermil_pd_128(<2 x double> %x0, <2 x double> %x2, i8 %x3) {
|
||||
; CHECK-LABEL: test_int_x86_avx512_mask_vpermil_pd_128:
|
||||
; CHECK: ## BB#0:
|
||||
; CHECK-NEXT: kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
|
||||
; CHECK-NEXT: vpermilpd $1, %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x05,0xc8,0x01]
|
||||
; CHECK-NEXT: ## xmm1 {%k1} = xmm0[1,0]
|
||||
; CHECK-NEXT: vpermilpd $1, %xmm0, %xmm2 {%k1} {z} ## encoding: [0x62,0xf3,0xfd,0x89,0x05,0xd0,0x01]
|
||||
; CHECK-NEXT: ## xmm2 {%k1} {z} = xmm0[1,0]
|
||||
; CHECK-NEXT: vpermilpd $1, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0xfd,0x08,0x05,0xc0,0x01]
|
||||
; CHECK-NEXT: ## xmm0 = xmm0[1,0]
|
||||
; CHECK-NEXT: vaddpd %xmm2, %xmm1, %xmm1 ## encoding: [0x62,0xf1,0xf5,0x08,0x58,0xca]
|
||||
; CHECK-NEXT: vaddpd %xmm0, %xmm1, %xmm0 ## encoding: [0x62,0xf1,0xf5,0x08,0x58,0xc0]
|
||||
; CHECK-NEXT: retq ## encoding: [0xc3]
|
||||
%res = call <2 x double> @llvm.x86.avx512.mask.vpermil.pd.128(<2 x double> %x0, i32 1, <2 x double> %x2, i8 %x3)
|
||||
%res1 = call <2 x double> @llvm.x86.avx512.mask.vpermil.pd.128(<2 x double> %x0, i32 1, <2 x double> zeroinitializer, i8 %x3)
|
||||
%res2 = call <2 x double> @llvm.x86.avx512.mask.vpermil.pd.128(<2 x double> %x0, i32 1, <2 x double> %x2, i8 -1)
|
||||
%res3 = fadd <2 x double> %res, %res1
|
||||
%res4 = fadd <2 x double> %res3, %res2
|
||||
ret <2 x double> %res4
|
||||
}
|
||||
|
||||
declare <8 x float> @llvm.x86.avx512.mask.vpermil.ps.256(<8 x float>, i32, <8 x float>, i8)
|
||||
|
||||
define <8 x float>@test_int_x86_avx512_mask_vpermil_ps_256(<8 x float> %x0, <8 x float> %x2, i8 %x3) {
|
||||
; CHECK-LABEL: test_int_x86_avx512_mask_vpermil_ps_256:
|
||||
; CHECK: ## BB#0:
|
||||
; CHECK-NEXT: kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
|
||||
; CHECK-NEXT: vpermilps $22, %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x04,0xc8,0x16]
|
||||
; CHECK-NEXT: ## ymm1 {%k1} = ymm0[2,1,1,0,6,5,5,4]
|
||||
; CHECK-NEXT: vpermilps $22, %ymm0, %ymm2 {%k1} {z} ## encoding: [0x62,0xf3,0x7d,0xa9,0x04,0xd0,0x16]
|
||||
; CHECK-NEXT: ## ymm2 {%k1} {z} = ymm0[2,1,1,0,6,5,5,4]
|
||||
; CHECK-NEXT: vpermilps $22, %ymm0, %ymm0 ## encoding: [0x62,0xf3,0x7d,0x28,0x04,0xc0,0x16]
|
||||
; CHECK-NEXT: ## ymm0 = ymm0[2,1,1,0,6,5,5,4]
|
||||
; CHECK-NEXT: vaddps %ymm2, %ymm1, %ymm1 ## encoding: [0x62,0xf1,0x74,0x28,0x58,0xca]
|
||||
; CHECK-NEXT: vaddps %ymm0, %ymm1, %ymm0 ## encoding: [0x62,0xf1,0x74,0x28,0x58,0xc0]
|
||||
; CHECK-NEXT: retq ## encoding: [0xc3]
|
||||
%res = call <8 x float> @llvm.x86.avx512.mask.vpermil.ps.256(<8 x float> %x0, i32 22, <8 x float> %x2, i8 %x3)
|
||||
%res1 = call <8 x float> @llvm.x86.avx512.mask.vpermil.ps.256(<8 x float> %x0, i32 22, <8 x float> zeroinitializer, i8 %x3)
|
||||
%res2 = call <8 x float> @llvm.x86.avx512.mask.vpermil.ps.256(<8 x float> %x0, i32 22, <8 x float> %x2, i8 -1)
|
||||
%res3 = fadd <8 x float> %res, %res1
|
||||
%res4 = fadd <8 x float> %res3, %res2
|
||||
ret <8 x float> %res4
|
||||
}
|
||||
|
||||
declare <4 x float> @llvm.x86.avx512.mask.vpermil.ps.128(<4 x float>, i32, <4 x float>, i8)
|
||||
|
||||
define <4 x float>@test_int_x86_avx512_mask_vpermil_ps_128(<4 x float> %x0, <4 x float> %x2, i8 %x3) {
|
||||
; CHECK-LABEL: test_int_x86_avx512_mask_vpermil_ps_128:
|
||||
; CHECK: ## BB#0:
|
||||
; CHECK-NEXT: kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
|
||||
; CHECK-NEXT: vpermilps $22, %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x04,0xc8,0x16]
|
||||
; CHECK-NEXT: ## xmm1 {%k1} = xmm0[2,1,1,0]
|
||||
; CHECK-NEXT: vpermilps $22, %xmm0, %xmm2 {%k1} {z} ## encoding: [0x62,0xf3,0x7d,0x89,0x04,0xd0,0x16]
|
||||
; CHECK-NEXT: ## xmm2 {%k1} {z} = xmm0[2,1,1,0]
|
||||
; CHECK-NEXT: vpermilps $22, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x04,0xc0,0x16]
|
||||
; CHECK-NEXT: ## xmm0 = xmm0[2,1,1,0]
|
||||
; CHECK-NEXT: vaddps %xmm2, %xmm1, %xmm1 ## encoding: [0x62,0xf1,0x74,0x08,0x58,0xca]
|
||||
; CHECK-NEXT: vaddps %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7c,0x08,0x58,0xc1]
|
||||
; CHECK-NEXT: retq ## encoding: [0xc3]
|
||||
%res = call <4 x float> @llvm.x86.avx512.mask.vpermil.ps.128(<4 x float> %x0, i32 22, <4 x float> %x2, i8 %x3)
|
||||
%res1 = call <4 x float> @llvm.x86.avx512.mask.vpermil.ps.128(<4 x float> %x0, i32 22, <4 x float> zeroinitializer, i8 %x3)
|
||||
%res2 = call <4 x float> @llvm.x86.avx512.mask.vpermil.ps.128(<4 x float> %x0, i32 22, <4 x float> %x2, i8 -1)
|
||||
%res3 = fadd <4 x float> %res, %res1
|
||||
%res4 = fadd <4 x float> %res2, %res3
|
||||
ret <4 x float> %res4
|
||||
}
|
||||
|
||||
declare <4 x double> @llvm.x86.avx512.mask.vpermilvar.pd.256(<4 x double>, <4 x i64>, <4 x double>, i8)
|
||||
|
||||
define <4 x double>@test_int_x86_avx512_mask_vpermilvar_pd_256(<4 x double> %x0, <4 x i64> %x1, <4 x double> %x2, i8 %x3) {
|
||||
|
|
@ -7218,9 +7126,9 @@ define <8 x i32>@test_int_x86_avx512_mask_psrav8_si_const() {
|
|||
; CHECK: ## BB#0:
|
||||
; CHECK-NEXT: vmovdqa32 {{.*#+}} ymm0 = [2,9,4294967284,23,4294967270,37,4294967256,51]
|
||||
; CHECK-NEXT: ## encoding: [0x62,0xf1,0x7d,0x28,0x6f,0x05,A,A,A,A]
|
||||
; CHECK-NEXT: ## fixup A - offset: 6, value: LCPI472_0-4, kind: reloc_riprel_4byte
|
||||
; CHECK-NEXT: ## fixup A - offset: 6, value: LCPI468_0-4, kind: reloc_riprel_4byte
|
||||
; CHECK-NEXT: vpsravd {{.*}}(%rip), %ymm0, %ymm0 ## encoding: [0x62,0xf2,0x7d,0x28,0x46,0x05,A,A,A,A]
|
||||
; CHECK-NEXT: ## fixup A - offset: 6, value: LCPI472_1-4, kind: reloc_riprel_4byte
|
||||
; CHECK-NEXT: ## fixup A - offset: 6, value: LCPI468_1-4, kind: reloc_riprel_4byte
|
||||
; CHECK-NEXT: retq ## encoding: [0xc3]
|
||||
%res = call <8 x i32> @llvm.x86.avx512.mask.psrav8.si(<8 x i32> <i32 2, i32 9, i32 -12, i32 23, i32 -26, i32 37, i32 -40, i32 51>, <8 x i32> <i32 1, i32 18, i32 35, i32 52, i32 69, i32 15, i32 32, i32 49>, <8 x i32> zeroinitializer, i8 -1)
|
||||
ret <8 x i32> %res
|
||||
|
|
@ -7251,9 +7159,9 @@ define <2 x i64>@test_int_x86_avx512_mask_psrav_q_128_const(i8 %x3) {
|
|||
; CHECK: ## BB#0:
|
||||
; CHECK-NEXT: vmovdqa64 {{.*#+}} xmm0 = [2,18446744073709551607]
|
||||
; CHECK-NEXT: ## encoding: [0x62,0xf1,0xfd,0x08,0x6f,0x05,A,A,A,A]
|
||||
; CHECK-NEXT: ## fixup A - offset: 6, value: LCPI474_0-4, kind: reloc_riprel_4byte
|
||||
; CHECK-NEXT: ## fixup A - offset: 6, value: LCPI470_0-4, kind: reloc_riprel_4byte
|
||||
; CHECK-NEXT: vpsravq {{.*}}(%rip), %xmm0, %xmm0 ## encoding: [0x62,0xf2,0xfd,0x08,0x46,0x05,A,A,A,A]
|
||||
; CHECK-NEXT: ## fixup A - offset: 6, value: LCPI474_1-4, kind: reloc_riprel_4byte
|
||||
; CHECK-NEXT: ## fixup A - offset: 6, value: LCPI470_1-4, kind: reloc_riprel_4byte
|
||||
; CHECK-NEXT: retq ## encoding: [0xc3]
|
||||
%res = call <2 x i64> @llvm.x86.avx512.mask.psrav.q.128(<2 x i64> <i64 2, i64 -9>, <2 x i64> <i64 1, i64 90>, <2 x i64> zeroinitializer, i8 -1)
|
||||
ret <2 x i64> %res
|
||||
|
|
|
|||
Loading…
Reference in New Issue