RPCSX/llvm
1
0
Fork 0
mirror of https://github.com/RPCSX/llvm.git synced 2025-04-12 04:56:49 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

[AVX-512] Remove intrinsics for valignd/q and autoupgrade them to native shuffles.

Browse Source 
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@287744 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Craig Topper 2016-11-23 06:54:55 +00:00
parent 92a50aa681
commit fa494a82ff
7 changed files with 143 additions and 172 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

38
include/llvm/IR/IntrinsicsX86.td
View File

@ -5674,44 +5674,6 @@ let TargetPrefix = "x86" in {
[IntrNoMem]>;
}
let TargetPrefix = "x86" in {
def int_x86_avx512_mask_valign_q_512 :
GCCBuiltin<"__builtin_ia32_alignq512_mask">,
Intrinsic<[llvm_v8i64_ty],
[llvm_v8i64_ty, llvm_v8i64_ty, llvm_i32_ty, llvm_v8i64_ty,
llvm_i8_ty], [IntrNoMem]>;
def int_x86_avx512_mask_valign_d_512 :
GCCBuiltin<"__builtin_ia32_alignd512_mask">,
Intrinsic<[llvm_v16i32_ty],
[llvm_v16i32_ty, llvm_v16i32_ty, llvm_i32_ty, llvm_v16i32_ty,
llvm_i16_ty], [IntrNoMem]>;
def int_x86_avx512_mask_valign_q_256 :
GCCBuiltin<"__builtin_ia32_alignq256_mask">,
Intrinsic<[llvm_v4i64_ty],
[llvm_v4i64_ty, llvm_v4i64_ty, llvm_i32_ty, llvm_v4i64_ty,
llvm_i8_ty], [IntrNoMem]>;
def int_x86_avx512_mask_valign_d_256 :
GCCBuiltin<"__builtin_ia32_alignd256_mask">,
Intrinsic<[llvm_v8i32_ty],
[llvm_v8i32_ty, llvm_v8i32_ty, llvm_i32_ty, llvm_v8i32_ty,
llvm_i8_ty], [IntrNoMem]>;
def int_x86_avx512_mask_valign_q_128 :
GCCBuiltin<"__builtin_ia32_alignq128_mask">,
Intrinsic<[llvm_v2i64_ty],
[llvm_v2i64_ty, llvm_v2i64_ty, llvm_i32_ty, llvm_v2i64_ty,
llvm_i8_ty], [IntrNoMem]>;
def int_x86_avx512_mask_valign_d_128 :
GCCBuiltin<"__builtin_ia32_alignd128_mask">,
Intrinsic<[llvm_v4i32_ty],
[llvm_v4i32_ty, llvm_v4i32_ty, llvm_i32_ty, llvm_v4i32_ty,
llvm_i8_ty], [IntrNoMem]>;
}
// Compares
let TargetPrefix = "x86" in {
// 512-bit

41
lib/IR/AutoUpgrade.cpp
View File

@ -360,6 +360,7 @@ static bool UpgradeIntrinsicFunction1(Function *F, Function *&NewFn) {
Name == "sse42.crc32.64.8" || // Added in 3.4
Name.startswith("avx.vbroadcast.s") || // Added in 3.5
Name.startswith("avx512.mask.palignr.") || // Added in 3.9
Name.startswith("avx512.mask.valign.") || // Added in 4.0
Name.startswith("sse2.psll.dq") || // Added in 3.7
Name.startswith("sse2.psrl.dq") || // Added in 3.7
Name.startswith("avx2.psll.dq") || // Added in 3.7
@ -572,13 +573,23 @@ static Value *EmitX86Select(IRBuilder<> &Builder, Value *Mask,
return Builder.CreateSelect(Mask, Op0, Op1);
}
static Value *UpgradeX86PALIGNRIntrinsics(IRBuilder<> &Builder,
Value *Op0, Value *Op1, Value *Shift,
Value *Passthru, Value *Mask) {
// Handle autoupgrade for masked PALIGNR and VALIGND/Q intrinsics.
// PALIGNR handles large immediates by shifting while VALIGN masks the immediate
// so we need to handle both cases. VALIGN also doesn't have 128-bit lanes.
static Value *UpgradeX86ALIGNIntrinsics(IRBuilder<> &Builder, Value *Op0,
Value *Op1, Value *Shift,
Value *Passthru, Value *Mask,
bool IsVALIGN) {
unsigned ShiftVal = cast<llvm::ConstantInt>(Shift)->getZExtValue();
unsigned NumElts = Op0->getType()->getVectorNumElements();
assert(NumElts % 16 == 0);
assert((IsVALIGN || NumElts % 16 == 0) && "Illegal NumElts for PALIGNR!");
assert((!IsVALIGN || NumElts <= 16) && "NumElts too large for VALIGN!");
assert(isPowerOf2_32(NumElts) && "NumElts not a power of 2!");
// Mask the immediate for VALIGN.
if (IsVALIGN)
ShiftVal &= (NumElts - 1);
// If palignr is shifting the pair of vectors more than the size of two
// lanes, emit zero.
@ -595,10 +606,10 @@ static Value *UpgradeX86PALIGNRIntrinsics(IRBuilder<> &Builder,
uint32_t Indices[64];
// 256-bit palignr operates on 128-bit lanes so we need to handle that
for (unsigned l = 0; l != NumElts; l += 16) {
for (unsigned l = 0; l < NumElts; l += 16) {
for (unsigned i = 0; i != 16; ++i) {
unsigned Idx = ShiftVal + i;
if (Idx >= 16)
if (!IsVALIGN && Idx >= 16) // Disable wrap for VALIGN.
Idx += NumElts - 16; // End of lane, switch operand.
Indices[l + i] = Idx + l;
}
@ -1071,11 +1082,19 @@ void llvm::UpgradeIntrinsicCall(CallInst *CI, Function *NewFn) {
Rep = EmitX86Select(Builder, CI->getArgOperand(2), Rep,
CI->getArgOperand(1));
} else if (IsX86 && Name.startswith("avx512.mask.palignr.")) {
Rep = UpgradeX86PALIGNRIntrinsics(Builder, CI->getArgOperand(0),
CI->getArgOperand(1),
CI->getArgOperand(2),
CI->getArgOperand(3),
CI->getArgOperand(4));
Rep = UpgradeX86ALIGNIntrinsics(Builder, CI->getArgOperand(0),
CI->getArgOperand(1),
CI->getArgOperand(2),
CI->getArgOperand(3),
CI->getArgOperand(4),
false);
} else if (IsX86 && Name.startswith("avx512.mask.valign.")) {
Rep = UpgradeX86ALIGNIntrinsics(Builder, CI->getArgOperand(0),
CI->getArgOperand(1),
CI->getArgOperand(2),
CI->getArgOperand(3),
CI->getArgOperand(4),
true);
} else if (IsX86 && (Name == "sse2.psll.dq" ||
Name == "avx2.psll.dq")) {
// 128/256-bit shift left specified in bits.

12
lib/Target/X86/X86IntrinsicsInfo.h
View File

@ -1171,18 +1171,6 @@ static const IntrinsicData IntrinsicsWithoutChain[] = {
X86_INTRINSIC_DATA(avx512_mask_ucmp_w_128, CMP_MASK_CC, X86ISD::CMPMU, 0),
X86_INTRINSIC_DATA(avx512_mask_ucmp_w_256, CMP_MASK_CC, X86ISD::CMPMU, 0),
X86_INTRINSIC_DATA(avx512_mask_ucmp_w_512, CMP_MASK_CC, X86ISD::CMPMU, 0),
X86_INTRINSIC_DATA(avx512_mask_valign_d_128, INTR_TYPE_3OP_IMM8_MASK,
X86ISD::VALIGN, 0),
X86_INTRINSIC_DATA(avx512_mask_valign_d_256, INTR_TYPE_3OP_IMM8_MASK,
X86ISD::VALIGN, 0),
X86_INTRINSIC_DATA(avx512_mask_valign_d_512, INTR_TYPE_3OP_IMM8_MASK,
X86ISD::VALIGN, 0),
X86_INTRINSIC_DATA(avx512_mask_valign_q_128, INTR_TYPE_3OP_IMM8_MASK,
X86ISD::VALIGN, 0),
X86_INTRINSIC_DATA(avx512_mask_valign_q_256, INTR_TYPE_3OP_IMM8_MASK,
X86ISD::VALIGN, 0),
X86_INTRINSIC_DATA(avx512_mask_valign_q_512, INTR_TYPE_3OP_IMM8_MASK,
X86ISD::VALIGN, 0),
X86_INTRINSIC_DATA(avx512_mask_vcvtph2ps_128, INTR_TYPE_1OP_MASK_RM,
X86ISD::CVTPH2PS, 0),
X86_INTRINSIC_DATA(avx512_mask_vcvtph2ps_256, INTR_TYPE_1OP_MASK_RM,

35
test/CodeGen/X86/avx512-intrinsics-upgrade.ll
View File

@ -2561,3 +2561,38 @@ define <8 x double>@test_int_x86_avx512_mask_cvt_udq2pd_512(<8 x i32> %x0, <8 x
%res2 = fadd <8 x double> %res, %res1
ret <8 x double> %res2
}
define <8 x i64> @test_valign_q(<8 x i64> %a, <8 x i64> %b) {
; CHECK-LABEL: test_valign_q:
; CHECK: ## BB#0:
; CHECK-NEXT: valignq {{.*#+}} zmm0 = zmm1[2,3,4,5,6,7],zmm0[0,1]
; CHECK-NEXT: retq
%res = call <8 x i64> @llvm.x86.avx512.mask.valign.q.512(<8 x i64> %a, <8 x i64> %b, i32 2, <8 x i64> zeroinitializer, i8 -1)
ret <8 x i64> %res
}
define <8 x i64> @test_mask_valign_q(<8 x i64> %a, <8 x i64> %b, <8 x i64> %src, i8 %mask) {
; CHECK-LABEL: test_mask_valign_q:
; CHECK: ## BB#0:
; CHECK-NEXT: kmovw %edi, %k1
; CHECK-NEXT: valignq {{.*#+}} zmm2 {%k1} = zmm1[2,3,4,5,6,7],zmm0[0,1]
; CHECK-NEXT: vmovdqa64 %zmm2, %zmm0
; CHECK-NEXT: retq
%res = call <8 x i64> @llvm.x86.avx512.mask.valign.q.512(<8 x i64> %a, <8 x i64> %b, i32 2, <8 x i64> %src, i8 %mask)
ret <8 x i64> %res
}
declare <8 x i64> @llvm.x86.avx512.mask.valign.q.512(<8 x i64>, <8 x i64>, i32, <8 x i64>, i8)
define <16 x i32> @test_maskz_valign_d(<16 x i32> %a, <16 x i32> %b, i16 %mask) {
; CHECK-LABEL: test_maskz_valign_d:
; CHECK: ## BB#0:
; CHECK-NEXT: kmovw %edi, %k1
; CHECK-NEXT: valignd {{.*#+}} zmm0 {%k1} {z} = zmm1[5,6,7,8,9,10,11,12,13,14,15],zmm0[0,1,2,3,4]
; CHECK-NEXT: retq
%res = call <16 x i32> @llvm.x86.avx512.mask.valign.d.512(<16 x i32> %a, <16 x i32> %b, i32 5, <16 x i32> zeroinitializer, i16 %mask)
ret <16 x i32> %res
}
declare <16 x i32> @llvm.x86.avx512.mask.valign.d.512(<16 x i32>, <16 x i32>, i32, <16 x i32>, i16)

34
test/CodeGen/X86/avx512-intrinsics.ll
View File

@ -843,40 +843,6 @@ define i16 @test_vptestmd(<16 x i32> %a0, <16 x i32> %a1, i16 %m) {
}
declare i16 @llvm.x86.avx512.ptestm.d.512(<16 x i32>, <16 x i32>, i16)
define <8 x i64> @test_valign_q(<8 x i64> %a, <8 x i64> %b) {
; CHECK-LABEL: test_valign_q:
; CHECK: ## BB#0:
; CHECK-NEXT: valignq {{.*#+}} zmm0 = zmm1[2,3,4,5,6,7],zmm0[0,1]
; CHECK-NEXT: retq
%res = call <8 x i64> @llvm.x86.avx512.mask.valign.q.512(<8 x i64> %a, <8 x i64> %b, i32 2, <8 x i64> zeroinitializer, i8 -1)
ret <8 x i64> %res
}
define <8 x i64> @test_mask_valign_q(<8 x i64> %a, <8 x i64> %b, <8 x i64> %src, i8 %mask) {
; CHECK-LABEL: test_mask_valign_q:
; CHECK: ## BB#0:
; CHECK-NEXT: kmovw %edi, %k1
; CHECK-NEXT: valignq {{.*#+}} zmm2 {%k1} = zmm1[2,3,4,5,6,7],zmm0[0,1]
; CHECK-NEXT: vmovdqa64 %zmm2, %zmm0
; CHECK-NEXT: retq
%res = call <8 x i64> @llvm.x86.avx512.mask.valign.q.512(<8 x i64> %a, <8 x i64> %b, i32 2, <8 x i64> %src, i8 %mask)
ret <8 x i64> %res
}
declare <8 x i64> @llvm.x86.avx512.mask.valign.q.512(<8 x i64>, <8 x i64>, i32, <8 x i64>, i8)
define <16 x i32> @test_maskz_valign_d(<16 x i32> %a, <16 x i32> %b, i16 %mask) {
; CHECK-LABEL: test_maskz_valign_d:
; CHECK: ## BB#0:
; CHECK-NEXT: kmovw %edi, %k1
; CHECK-NEXT: valignd {{.*#+}} zmm0 {%k1} {z} = zmm1[5,6,7,8,9,10,11,12,13,14,15],zmm0[0,1,2,3,4]
; CHECK-NEXT: retq
%res = call <16 x i32> @llvm.x86.avx512.mask.valign.d.512(<16 x i32> %a, <16 x i32> %b, i32 5, <16 x i32> zeroinitializer, i16 %mask)
ret <16 x i32> %res
}
declare <16 x i32> @llvm.x86.avx512.mask.valign.d.512(<16 x i32>, <16 x i32>, i32, <16 x i32>, i16)
define void @test_mask_store_ss(i8* %ptr, <4 x float> %data, i8 %mask) {
; CHECK-LABEL: test_mask_store_ss:
; CHECK: ## BB#0:

78
test/CodeGen/X86/avx512vl-intrinsics-upgrade.ll
View File

@ -4615,3 +4615,81 @@ define <4 x double>@test_int_x86_avx512_mask_cvt_udq2pd_256(<4 x i32> %x0, <4 x
%res2 = fadd <4 x double> %res, %res1
ret <4 x double> %res2
}
declare <4 x i32> @llvm.x86.avx512.mask.valign.d.128(<4 x i32>, <4 x i32>, i32, <4 x i32>, i8)
define <4 x i32>@test_int_x86_avx512_mask_valign_d_128(<4 x i32> %x0, <4 x i32> %x1,<4 x i32> %x3, i8 %x4) {
; CHECK-LABEL: test_int_x86_avx512_mask_valign_d_128:
; CHECK: ## BB#0:
; CHECK-NEXT: valignd $2, %xmm1, %xmm0, %xmm3 ## encoding: [0x62,0xf3,0x7d,0x08,0x03,0xd9,0x02]
; CHECK-NEXT: ## xmm3 = xmm1[2,3],xmm0[0,1]
; CHECK-NEXT: kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT: valignd $2, %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x03,0xd1,0x02]
; CHECK-NEXT: ## xmm2 {%k1} = xmm1[2,3],xmm0[0,1]
; CHECK-NEXT: valignd $2, %xmm1, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf3,0x7d,0x89,0x03,0xc1,0x02]
; CHECK-NEXT: ## xmm0 {%k1} {z} = xmm1[2,3],xmm0[0,1]
; CHECK-NEXT: vpaddd %xmm3, %xmm2, %xmm1 ## encoding: [0x62,0xf1,0x6d,0x08,0xfe,0xcb]
; CHECK-NEXT: vpaddd %xmm0, %xmm1, %xmm0 ## encoding: [0x62,0xf1,0x75,0x08,0xfe,0xc0]
; CHECK-NEXT: retq ## encoding: [0xc3]
%res = call <4 x i32> @llvm.x86.avx512.mask.valign.d.128(<4 x i32> %x0, <4 x i32> %x1, i32 2, <4 x i32> %x3, i8 %x4)
%res1 = call <4 x i32> @llvm.x86.avx512.mask.valign.d.128(<4 x i32> %x0, <4 x i32> %x1, i32 2, <4 x i32> %x3, i8 -1)
%res2 = call <4 x i32> @llvm.x86.avx512.mask.valign.d.128(<4 x i32> %x0, <4 x i32> %x1, i32 2, <4 x i32> zeroinitializer,i8 %x4)
%res3 = add <4 x i32> %res, %res1
%res4 = add <4 x i32> %res3, %res2
ret <4 x i32> %res4
}
declare <8 x i32> @llvm.x86.avx512.mask.valign.d.256(<8 x i32>, <8 x i32>, i32, <8 x i32>, i8)
define <8 x i32>@test_int_x86_avx512_mask_valign_d_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x3, i8 %x4) {
; CHECK-LABEL: test_int_x86_avx512_mask_valign_d_256:
; CHECK: ## BB#0:
; CHECK-NEXT: valignd $6, %ymm1, %ymm0, %ymm3 ## encoding: [0x62,0xf3,0x7d,0x28,0x03,0xd9,0x06]
; CHECK-NEXT: ## ymm3 = ymm1[6,7],ymm0[0,1,2,3,4,5]
; CHECK-NEXT: kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT: valignd $6, %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x03,0xd1,0x06]
; CHECK-NEXT: ## ymm2 {%k1} = ymm1[6,7],ymm0[0,1,2,3,4,5]
; CHECK-NEXT: vpaddd %ymm3, %ymm2, %ymm0 ## encoding: [0x62,0xf1,0x6d,0x28,0xfe,0xc3]
; CHECK-NEXT: retq ## encoding: [0xc3]
%res = call <8 x i32> @llvm.x86.avx512.mask.valign.d.256(<8 x i32> %x0, <8 x i32> %x1, i32 6, <8 x i32> %x3, i8 %x4)
%res1 = call <8 x i32> @llvm.x86.avx512.mask.valign.d.256(<8 x i32> %x0, <8 x i32> %x1, i32 6, <8 x i32> %x3, i8 -1)
%res2 = add <8 x i32> %res, %res1
ret <8 x i32> %res2
}
declare <2 x i64> @llvm.x86.avx512.mask.valign.q.128(<2 x i64>, <2 x i64>, i32, <2 x i64>, i8)
define <2 x i64>@test_int_x86_avx512_mask_valign_q_128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x3, i8 %x4) {
; CHECK-LABEL: test_int_x86_avx512_mask_valign_q_128:
; CHECK: ## BB#0:
; CHECK-NEXT: valignq $1, %xmm1, %xmm0, %xmm3 ## encoding: [0x62,0xf3,0xfd,0x08,0x03,0xd9,0x01]
; CHECK-NEXT: ## xmm3 = xmm1[1],xmm0[0]
; CHECK-NEXT: kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT: valignq $1, %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x03,0xd1,0x01]
; CHECK-NEXT: ## xmm2 {%k1} = xmm1[1],xmm0[0]
; CHECK-NEXT: vpaddq %xmm3, %xmm2, %xmm0 ## encoding: [0x62,0xf1,0xed,0x08,0xd4,0xc3]
; CHECK-NEXT: retq ## encoding: [0xc3]
%res = call <2 x i64> @llvm.x86.avx512.mask.valign.q.128(<2 x i64> %x0, <2 x i64> %x1, i32 1, <2 x i64> %x3, i8 %x4)
%res1 = call <2 x i64> @llvm.x86.avx512.mask.valign.q.128(<2 x i64> %x0, <2 x i64> %x1, i32 1, <2 x i64> %x3, i8 -1)
%res2 = add <2 x i64> %res, %res1
ret <2 x i64> %res2
}
declare <4 x i64> @llvm.x86.avx512.mask.valign.q.256(<4 x i64>, <4 x i64>, i32, <4 x i64>, i8)
define <4 x i64>@test_int_x86_avx512_mask_valign_q_256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x3, i8 %x4) {
; CHECK-LABEL: test_int_x86_avx512_mask_valign_q_256:
; CHECK: ## BB#0:
; CHECK-NEXT: valignq $3, %ymm1, %ymm0, %ymm3 ## encoding: [0x62,0xf3,0xfd,0x28,0x03,0xd9,0x03]
; CHECK-NEXT: ## ymm3 = ymm1[3],ymm0[0,1,2]
; CHECK-NEXT: kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT: valignq $3, %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x03,0xd1,0x03]
; CHECK-NEXT: ## ymm2 {%k1} = ymm1[3],ymm0[0,1,2]
; CHECK-NEXT: vpaddq %ymm3, %ymm2, %ymm0 ## encoding: [0x62,0xf1,0xed,0x28,0xd4,0xc3]
; CHECK-NEXT: retq ## encoding: [0xc3]
%res = call <4 x i64> @llvm.x86.avx512.mask.valign.q.256(<4 x i64> %x0, <4 x i64> %x1, i32 3, <4 x i64> %x3, i8 %x4)
%res1 = call <4 x i64> @llvm.x86.avx512.mask.valign.q.256(<4 x i64> %x0, <4 x i64> %x1, i32 3, <4 x i64> %x3, i8 -1)
%res2 = add <4 x i64> %res, %res1
ret <4 x i64> %res2
}

77
test/CodeGen/X86/avx512vl-intrinsics.ll
View File

@ -3619,83 +3619,6 @@ define <8 x float>@test_int_x86_avx512_mask_getmant_ps_256(<8 x float> %x0, <8 x
ret <8 x float> %res2
}
declare <4 x i32> @llvm.x86.avx512.mask.valign.d.128(<4 x i32>, <4 x i32>, i32, <4 x i32>, i8)
define <4 x i32>@test_int_x86_avx512_mask_valign_d_128(<4 x i32> %x0, <4 x i32> %x1,<4 x i32> %x3, i8 %x4) {
; CHECK-LABEL: test_int_x86_avx512_mask_valign_d_128:
; CHECK: ## BB#0:
; CHECK-NEXT: kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT: valignd $22, %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x03,0xd1,0x16]
; CHECK-NEXT: ## xmm2 {%k1} = xmm1[2,3],xmm0[0,1]
; CHECK-NEXT: valignd $22, %xmm1, %xmm0, %xmm3 {%k1} {z} ## encoding: [0x62,0xf3,0x7d,0x89,0x03,0xd9,0x16]
; CHECK-NEXT: ## xmm3 {%k1} {z} = xmm1[2,3],xmm0[0,1]
; CHECK-NEXT: valignd $22, %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x03,0xc1,0x16]
; CHECK-NEXT: ## xmm0 = xmm1[2,3],xmm0[0,1]
; CHECK-NEXT: vpaddd %xmm0, %xmm2, %xmm0 ## encoding: [0x62,0xf1,0x6d,0x08,0xfe,0xc0]
; CHECK-NEXT: vpaddd %xmm3, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xfe,0xc3]
; CHECK-NEXT: retq ## encoding: [0xc3]
%res = call <4 x i32> @llvm.x86.avx512.mask.valign.d.128(<4 x i32> %x0, <4 x i32> %x1, i32 22, <4 x i32> %x3, i8 %x4)
%res1 = call <4 x i32> @llvm.x86.avx512.mask.valign.d.128(<4 x i32> %x0, <4 x i32> %x1, i32 22, <4 x i32> %x3, i8 -1)
%res2 = call <4 x i32> @llvm.x86.avx512.mask.valign.d.128(<4 x i32> %x0, <4 x i32> %x1, i32 22, <4 x i32> zeroinitializer,i8 %x4)
%res3 = add <4 x i32> %res, %res1
%res4 = add <4 x i32> %res3, %res2
ret <4 x i32> %res4
}
declare <8 x i32> @llvm.x86.avx512.mask.valign.d.256(<8 x i32>, <8 x i32>, i32, <8 x i32>, i8)
define <8 x i32>@test_int_x86_avx512_mask_valign_d_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x3, i8 %x4) {
; CHECK-LABEL: test_int_x86_avx512_mask_valign_d_256:
; CHECK: ## BB#0:
; CHECK-NEXT: kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT: valignd $22, %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x03,0xd1,0x16]
; CHECK-NEXT: ## ymm2 {%k1} = ymm1[6,7],ymm0[0,1,2,3,4,5]
; CHECK-NEXT: valignd $22, %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf3,0x7d,0x28,0x03,0xc1,0x16]
; CHECK-NEXT: ## ymm0 = ymm1[6,7],ymm0[0,1,2,3,4,5]
; CHECK-NEXT: vpaddd %ymm0, %ymm2, %ymm0 ## encoding: [0x62,0xf1,0x6d,0x28,0xfe,0xc0]
; CHECK-NEXT: retq ## encoding: [0xc3]
%res = call <8 x i32> @llvm.x86.avx512.mask.valign.d.256(<8 x i32> %x0, <8 x i32> %x1, i32 22, <8 x i32> %x3, i8 %x4)
%res1 = call <8 x i32> @llvm.x86.avx512.mask.valign.d.256(<8 x i32> %x0, <8 x i32> %x1, i32 22, <8 x i32> %x3, i8 -1)
%res2 = add <8 x i32> %res, %res1
ret <8 x i32> %res2
}
declare <2 x i64> @llvm.x86.avx512.mask.valign.q.128(<2 x i64>, <2 x i64>, i32, <2 x i64>, i8)
define <2 x i64>@test_int_x86_avx512_mask_valign_q_128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x3, i8 %x4) {
; CHECK-LABEL: test_int_x86_avx512_mask_valign_q_128:
; CHECK: ## BB#0:
; CHECK-NEXT: kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT: valignq $22, %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x03,0xd1,0x16]
; CHECK-NEXT: ## xmm2 {%k1} = xmm1[0,1]
; CHECK-NEXT: valignq $22, %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0xfd,0x08,0x03,0xc1,0x16]
; CHECK-NEXT: ## xmm0 = xmm1[0,1]
; CHECK-NEXT: vpaddq %xmm0, %xmm2, %xmm0 ## encoding: [0x62,0xf1,0xed,0x08,0xd4,0xc0]
; CHECK-NEXT: retq ## encoding: [0xc3]
%res = call <2 x i64> @llvm.x86.avx512.mask.valign.q.128(<2 x i64> %x0, <2 x i64> %x1, i32 22, <2 x i64> %x3, i8 %x4)
%res1 = call <2 x i64> @llvm.x86.avx512.mask.valign.q.128(<2 x i64> %x0, <2 x i64> %x1, i32 22, <2 x i64> %x3, i8 -1)
%res2 = add <2 x i64> %res, %res1
ret <2 x i64> %res2
}
declare <4 x i64> @llvm.x86.avx512.mask.valign.q.256(<4 x i64>, <4 x i64>, i32, <4 x i64>, i8)
define <4 x i64>@test_int_x86_avx512_mask_valign_q_256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x3, i8 %x4) {
; CHECK-LABEL: test_int_x86_avx512_mask_valign_q_256:
; CHECK: ## BB#0:
; CHECK-NEXT: kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT: valignq $22, %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x03,0xd1,0x16]
; CHECK-NEXT: ## ymm2 {%k1} = ymm1[2,3],ymm0[0,1]
; CHECK-NEXT: valignq $22, %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf3,0xfd,0x28,0x03,0xc1,0x16]
; CHECK-NEXT: ## ymm0 = ymm1[2,3],ymm0[0,1]
; CHECK-NEXT: vpaddq %ymm0, %ymm2, %ymm0 ## encoding: [0x62,0xf1,0xed,0x28,0xd4,0xc0]
; CHECK-NEXT: retq ## encoding: [0xc3]
%res = call <4 x i64> @llvm.x86.avx512.mask.valign.q.256(<4 x i64> %x0, <4 x i64> %x1, i32 22, <4 x i64> %x3, i8 %x4)
%res1 = call <4 x i64> @llvm.x86.avx512.mask.valign.q.256(<4 x i64> %x0, <4 x i64> %x1, i32 22, <4 x i64> %x3, i8 -1)
%res2 = add <4 x i64> %res, %res1
ret <4 x i64> %res2
}
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) {