Intrinsic<[], [], []>;
}
-// Vector load with broadcast
-let TargetPrefix = "x86" in { // All intrinsics start with "llvm.x86.".
- def int_x86_avx_vbroadcastf128_pd_256 :
- GCCBuiltin<"__builtin_ia32_vbroadcastf128_pd256">,
- Intrinsic<[llvm_v4f64_ty], [llvm_ptr_ty], [IntrReadMem, IntrArgMemOnly]>;
- def int_x86_avx_vbroadcastf128_ps_256 :
- GCCBuiltin<"__builtin_ia32_vbroadcastf128_ps256">,
- Intrinsic<[llvm_v8f32_ty], [llvm_ptr_ty], [IntrReadMem, IntrArgMemOnly]>;
-}
-
// SIMD load ops
let TargetPrefix = "x86" in { // All intrinsics start with "llvm.x86.".
def int_x86_avx_ldu_dq_256 : GCCBuiltin<"__builtin_ia32_lddqu256">,
Name.startswith("avx.blend.p") ||
Name == "avx2.pblendw" ||
Name.startswith("avx2.pblendd.") ||
+ Name.startswith("avx.vbroadcastf128") ||
Name == "avx2.vbroadcasti128" ||
Name == "xop.vpcmov" ||
(Name.startswith("xop.vpcom") && F->arg_size() == 2))) {
Value *Trunc0 = Builder.CreateTrunc(CI->getArgOperand(0), Type::getInt32Ty(C));
Rep = Builder.CreateCall(CRC32, {Trunc0, CI->getArgOperand(1)});
Rep = Builder.CreateZExt(Rep, CI->getType(), "");
- } else if (IsX86 && Name.startswith("avx.vbroadcast")) {
+ } else if (IsX86 && Name.startswith("avx.vbroadcast.s")) {
// Replace broadcasts with a series of insertelements.
Type *VecTy = CI->getType();
Type *EltTy = VecTy->getVectorElementType();
bool DoSext = (StringRef::npos != Name.find("pmovsx"));
Rep = DoSext ? Builder.CreateSExt(SV, DstTy)
: Builder.CreateZExt(SV, DstTy);
- } else if (IsX86 && Name == "avx2.vbroadcasti128") {
- // Replace vbroadcasts with a vector shuffle.
- Type *VT = VectorType::get(Type::getInt64Ty(C), 2);
+ } else if (IsX86 && (Name.startswith("avx.vbroadcastf128") ||
+ Name == "avx2.vbroadcasti128")) {
+ // Replace vbroadcastf128/vbroadcasti128 with a vector load+shuffle.
+ Type *EltTy = CI->getType()->getVectorElementType();
+ unsigned NumSrcElts = 128 / EltTy->getPrimitiveSizeInBits();
+ Type *VT = VectorType::get(EltTy, NumSrcElts);
Value *Op = Builder.CreatePointerCast(CI->getArgOperand(0),
PointerType::getUnqual(VT));
Value *Load = Builder.CreateLoad(VT, Op);
- uint32_t Idxs[4] = { 0, 1, 0, 1 };
- Rep = Builder.CreateShuffleVector(Load, UndefValue::get(Load->getType()),
- Idxs);
+ if (NumSrcElts == 2)
+ Rep = Builder.CreateShuffleVector(Load, UndefValue::get(Load->getType()),
+ { 0, 1, 0, 1 });
+ else
+ Rep = Builder.CreateShuffleVector(Load, UndefValue::get(Load->getType()),
+ { 0, 1, 2, 3, 0, 1, 2, 3 });
} else if (IsX86 && (Name.startswith("avx2.pbroadcast") ||
Name.startswith("avx2.vbroadcast") ||
Name.startswith("avx512.pbroadcast") ||
// (insert_subvector (insert_subvector undef, (load addr), 0),
// (load addr + 16), Elts/2)
// --> load32 addr
+ // or a 16-byte broadcast:
+ // (insert_subvector (insert_subvector undef, (load addr), 0),
+ // (load addr), Elts/2)
+ // --> X86SubVBroadcast(load16 addr)
if ((IdxVal == OpVT.getVectorNumElements() / 2) &&
Vec.getOpcode() == ISD::INSERT_SUBVECTOR &&
OpVT.is256BitVector() && SubVecVT.is128BitVector()) {
if (SDValue Ld = EltsFromConsecutiveLoads(OpVT, Ops, dl, DAG, false))
return Ld;
}
+
+ // If lower/upper loads are the same then lower to a VBROADCASTF128.
+ if (SubVec2 == peekThroughBitcasts(SubVec))
+ return DAG.getNode(X86ISD::SUBV_BROADCAST, dl, OpVT, SubVec);
}
}
}
AVX5128IBase, EVEX;
}
+//===----------------------------------------------------------------------===//
+// AVX-512 BROADCAST SUBVECTORS
+//
+
defm VBROADCASTI32X4 : avx512_subvec_broadcast_rm<0x5a, "vbroadcasti32x4",
v16i32_info, v4i32x_info>,
EVEX_V512, EVEX_CD8<32, CD8VT4>;
defm VBROADCASTF32X4Z256 : avx512_subvec_broadcast_rm<0x1a, "vbroadcastf32x4",
v8f32x_info, v4f32x_info>,
EVEX_V256, EVEX_CD8<32, CD8VT4>;
+
+def : Pat<(v16i16 (X86SubVBroadcast (bc_v8i16 (loadv2i64 addr:$src)))),
+ (VBROADCASTI32X4Z256rm addr:$src)>;
+def : Pat<(v32i8 (X86SubVBroadcast (bc_v16i8 (loadv2i64 addr:$src)))),
+ (VBROADCASTI32X4Z256rm addr:$src)>;
}
+
let Predicates = [HasVLX, HasDQI] in {
defm VBROADCASTI64X2Z128 : avx512_subvec_broadcast_rm<0x5a, "vbroadcasti64x2",
v4i64x_info, v2i64x_info>, VEX_W,
v4f64x_info, v2f64x_info>, VEX_W,
EVEX_V256, EVEX_CD8<64, CD8VT2>;
}
+
+let Predicates = [HasVLX, NoDQI] in {
+def : Pat<(v4f64 (X86SubVBroadcast (loadv2f64 addr:$src))),
+ (VBROADCASTF32X4Z256rm addr:$src)>;
+def : Pat<(v4i64 (X86SubVBroadcast (loadv2i64 addr:$src))),
+ (VBROADCASTI32X4Z256rm addr:$src)>;
+}
+
let Predicates = [HasDQI] in {
defm VBROADCASTI64X2 : avx512_subvec_broadcast_rm<0x5a, "vbroadcasti64x2",
v8i64_info, v2i64x_info>, VEX_W,
def VBROADCASTSDYrr : avx2_broadcast_rr<0x19, "vbroadcastsd", VR256,
v4f64, v2f64, WriteFShuffle256>, VEX_L;
+//===----------------------------------------------------------------------===//
+// VBROADCAST*128 - Load from memory and broadcast 128-bit vector to both
+// halves of a 256-bit vector.
+//
let mayLoad = 1, hasSideEffects = 0, Predicates = [HasAVX2] in
def VBROADCASTI128 : AVX8I<0x5A, MRMSrcMem, (outs VR256:$dst),
(ins i128mem:$src),
"vbroadcasti128\t{$src, $dst|$dst, $src}", []>,
Sched<[WriteLoad]>, VEX, VEX_L;
+let mayLoad = 1, hasSideEffects = 0, Predicates = [HasAVX] in
def VBROADCASTF128 : AVX8I<0x1A, MRMSrcMem, (outs VR256:$dst),
(ins f128mem:$src),
- "vbroadcastf128\t{$src, $dst|$dst, $src}",
- [(set VR256:$dst,
- (int_x86_avx_vbroadcastf128_pd_256 addr:$src))]>,
+ "vbroadcastf128\t{$src, $dst|$dst, $src}", []>,
Sched<[WriteFShuffleLd]>, VEX, VEX_L;
-let Predicates = [HasAVX] in
-def : Pat<(int_x86_avx_vbroadcastf128_ps_256 addr:$src),
+let Predicates = [HasAVX2, NoVLX] in {
+def : Pat<(v4i64 (X86SubVBroadcast (loadv2i64 addr:$src))),
+ (VBROADCASTI128 addr:$src)>;
+def : Pat<(v8i32 (X86SubVBroadcast (bc_v4i32 (loadv2i64 addr:$src)))),
+ (VBROADCASTI128 addr:$src)>;
+def : Pat<(v16i16 (X86SubVBroadcast (bc_v8i16 (loadv2i64 addr:$src)))),
+ (VBROADCASTI128 addr:$src)>;
+def : Pat<(v32i8 (X86SubVBroadcast (bc_v16i8 (loadv2i64 addr:$src)))),
+ (VBROADCASTI128 addr:$src)>;
+}
+
+let Predicates = [HasAVX, NoVLX] in {
+def : Pat<(v4f64 (X86SubVBroadcast (loadv2f64 addr:$src))),
+ (VBROADCASTF128 addr:$src)>;
+def : Pat<(v8f32 (X86SubVBroadcast (loadv4f32 addr:$src))),
(VBROADCASTF128 addr:$src)>;
+}
+let Predicates = [HasAVX1Only] in {
+def : Pat<(v4i64 (X86SubVBroadcast (loadv2i64 addr:$src))),
+ (VBROADCASTF128 addr:$src)>;
+def : Pat<(v8i32 (X86SubVBroadcast (bc_v4i32 (loadv2i64 addr:$src)))),
+ (VBROADCASTF128 addr:$src)>;
+def : Pat<(v16i16 (X86SubVBroadcast (bc_v8i16 (loadv2i64 addr:$src)))),
+ (VBROADCASTF128 addr:$src)>;
+def : Pat<(v32i8 (X86SubVBroadcast (bc_v16i8 (loadv2i64 addr:$src)))),
+ (VBROADCASTF128 addr:$src)>;
+}
//===----------------------------------------------------------------------===//
// VINSERTF128 - Insert packed floating-point values
; X64: # BB#0:
; X64-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
; X64-NEXT: retq
- %arg0 = bitcast <2 x double>* %a0 to i8*
- %res = call <4 x double> @llvm.x86.avx.vbroadcastf128.pd.256(i8* %arg0)
+ %ld = load <2 x double>, <2 x double>* %a0
+ %res = shufflevector <2 x double> %ld, <2 x double> %ld, <4 x i32> <i32 0, i32 1, i32 0, i32 1>
ret <4 x double> %res
}
-declare <4 x double> @llvm.x86.avx.vbroadcastf128.pd.256(i8*) nounwind readonly
define <8 x float> @test_mm256_broadcast_ps(<4 x float>* %a0) nounwind {
; X32-LABEL: test_mm256_broadcast_ps:
; X64: # BB#0:
; X64-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
; X64-NEXT: retq
- %arg0 = bitcast <4 x float>* %a0 to i8*
- %res = call <8 x float> @llvm.x86.avx.vbroadcastf128.ps.256(i8* %arg0)
+ %ld = load <4 x float>, <4 x float>* %a0
+ %res = shufflevector <4 x float> %ld, <4 x float> %ld, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3>
ret <8 x float> %res
}
-declare <8 x float> @llvm.x86.avx.vbroadcastf128.ps.256(i8*) nounwind readonly
define <4 x double> @test_mm256_broadcast_sd(double* %a0) nounwind {
; X32-LABEL: test_mm256_broadcast_sd:
}
+define <4 x double> @test_x86_avx_vbroadcastf128_pd_256(i8* %a0) {
+; CHECK-LABEL: test_x86_avx_vbroadcastf128_pd_256:
+; CHECK: ## BB#0:
+; CHECK-NEXT: movl {{[0-9]+}}(%esp), %eax
+; CHECK-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
+; CHECK-NEXT: retl
+ %res = call <4 x double> @llvm.x86.avx.vbroadcastf128.pd.256(i8* %a0) ; <<4 x double>> [#uses=1]
+ ret <4 x double> %res
+}
+declare <4 x double> @llvm.x86.avx.vbroadcastf128.pd.256(i8*) nounwind readonly
+
+
+define <8 x float> @test_x86_avx_vbroadcastf128_ps_256(i8* %a0) {
+; CHECK-LABEL: test_x86_avx_vbroadcastf128_ps_256:
+; CHECK: ## BB#0:
+; CHECK-NEXT: movl {{[0-9]+}}(%esp), %eax
+; CHECK-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
+; CHECK-NEXT: retl
+ %res = call <8 x float> @llvm.x86.avx.vbroadcastf128.ps.256(i8* %a0) ; <<8 x float>> [#uses=1]
+ ret <8 x float> %res
+}
+declare <8 x float> @llvm.x86.avx.vbroadcastf128.ps.256(i8*) nounwind readonly
+
+
define <4 x double> @test_x86_avx_blend_pd_256(<4 x double> %a0, <4 x double> %a1) {
; CHECK-LABEL: test_x86_avx_blend_pd_256:
; CHECK: ## BB#0:
; CHECK-LABEL: test_x86_sse2_storeu_dq:
; CHECK: ## BB#0:
; CHECK-NEXT: movl {{[0-9]+}}(%esp), %eax
-; CHECK-NEXT: vpaddb LCPI32_0, %xmm0, %xmm0
+; CHECK-NEXT: vpaddb LCPI34_0, %xmm0, %xmm0
; CHECK-NEXT: vmovdqu %xmm0, (%eax)
; CHECK-NEXT: retl
%a2 = add <16 x i8> %a1, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>
declare <8 x float> @llvm.x86.avx.sqrt.ps.256(<8 x float>) nounwind readnone
-define <4 x double> @test_x86_avx_vbroadcastf128_pd_256(i8* %a0) {
-; AVX-LABEL: test_x86_avx_vbroadcastf128_pd_256:
-; AVX: ## BB#0:
-; AVX-NEXT: movl {{[0-9]+}}(%esp), %eax
-; AVX-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
-; AVX-NEXT: retl
-;
-; AVX512VL-LABEL: test_x86_avx_vbroadcastf128_pd_256:
-; AVX512VL: ## BB#0:
-; AVX512VL-NEXT: movl {{[0-9]+}}(%esp), %eax
-; AVX512VL-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
-; AVX512VL-NEXT: retl
- %res = call <4 x double> @llvm.x86.avx.vbroadcastf128.pd.256(i8* %a0) ; <<4 x double>> [#uses=1]
- ret <4 x double> %res
-}
-declare <4 x double> @llvm.x86.avx.vbroadcastf128.pd.256(i8*) nounwind readonly
-
-
-define <8 x float> @test_x86_avx_vbroadcastf128_ps_256(i8* %a0) {
-; AVX-LABEL: test_x86_avx_vbroadcastf128_ps_256:
-; AVX: ## BB#0:
-; AVX-NEXT: movl {{[0-9]+}}(%esp), %eax
-; AVX-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
-; AVX-NEXT: retl
-;
-; AVX512VL-LABEL: test_x86_avx_vbroadcastf128_ps_256:
-; AVX512VL: ## BB#0:
-; AVX512VL-NEXT: movl {{[0-9]+}}(%esp), %eax
-; AVX512VL-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
-; AVX512VL-NEXT: retl
- %res = call <8 x float> @llvm.x86.avx.vbroadcastf128.ps.256(i8* %a0) ; <<8 x float>> [#uses=1]
- ret <8 x float> %res
-}
-declare <8 x float> @llvm.x86.avx.vbroadcastf128.ps.256(i8*) nounwind readonly
-
-
define <4 x double> @test_x86_avx_vperm2f128_pd_256(<4 x double> %a0, <4 x double> %a1) {
; AVX-LABEL: test_x86_avx_vperm2f128_pd_256:
; AVX: ## BB#0:
; AVX-LABEL: movnt_dq:
; AVX: ## BB#0:
; AVX-NEXT: movl {{[0-9]+}}(%esp), %eax
-; AVX-NEXT: vpaddq LCPI256_0, %xmm0, %xmm0
+; AVX-NEXT: vpaddq LCPI254_0, %xmm0, %xmm0
; AVX-NEXT: vmovntdq %ymm0, (%eax)
; AVX-NEXT: vzeroupper
; AVX-NEXT: retl
; AVX512VL-LABEL: movnt_dq:
; AVX512VL: ## BB#0:
; AVX512VL-NEXT: movl {{[0-9]+}}(%esp), %eax
-; AVX512VL-NEXT: vpaddq LCPI256_0, %xmm0, %xmm0
+; AVX512VL-NEXT: vpaddq LCPI254_0, %xmm0, %xmm0
; AVX512VL-NEXT: vmovntdq %ymm0, (%eax)
; AVX512VL-NEXT: retl
%a2 = add <2 x i64> %a1, <i64 1, i64 1>
; X32-LABEL: test_broadcast_2f64_4f64:
; X32: ## BB#0:
; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X32-NEXT: vmovaps (%eax), %xmm0
-; X32-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; X32-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
; X32-NEXT: retl
;
; X64-LABEL: test_broadcast_2f64_4f64:
; X64: ## BB#0:
-; X64-NEXT: vmovaps (%rdi), %xmm0
-; X64-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; X64-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
; X64-NEXT: retq
%1 = load <2 x double>, <2 x double> *%p
%2 = shufflevector <2 x double> %1, <2 x double> undef, <4 x i32> <i32 0, i32 1, i32 0, i32 1>
; X32-LABEL: test_broadcast_2i64_4i64:
; X32: ## BB#0:
; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X32-NEXT: vmovaps (%eax), %xmm0
-; X32-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; X32-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
; X32-NEXT: retl
;
; X64-LABEL: test_broadcast_2i64_4i64:
; X64: ## BB#0:
-; X64-NEXT: vmovaps (%rdi), %xmm0
-; X64-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; X64-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
; X64-NEXT: retq
%1 = load <2 x i64>, <2 x i64> *%p
%2 = shufflevector <2 x i64> %1, <2 x i64> undef, <4 x i32> <i32 0, i32 1, i32 0, i32 1>
; X32-LABEL: test_broadcast_4f32_8f32:
; X32: ## BB#0:
; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X32-NEXT: vmovaps (%eax), %xmm0
-; X32-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; X32-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
; X32-NEXT: retl
;
; X64-LABEL: test_broadcast_4f32_8f32:
; X64: ## BB#0:
-; X64-NEXT: vmovaps (%rdi), %xmm0
-; X64-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; X64-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
; X64-NEXT: retq
%1 = load <4 x float>, <4 x float> *%p
%2 = shufflevector <4 x float> %1, <4 x float> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3>
; X32-LABEL: test_broadcast_4i32_8i32:
; X32: ## BB#0:
; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X32-NEXT: vmovaps (%eax), %xmm0
-; X32-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; X32-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
; X32-NEXT: retl
;
; X64-LABEL: test_broadcast_4i32_8i32:
; X64: ## BB#0:
-; X64-NEXT: vmovaps (%rdi), %xmm0
-; X64-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; X64-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
; X64-NEXT: retq
%1 = load <4 x i32>, <4 x i32> *%p
%2 = shufflevector <4 x i32> %1, <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3>
; X32-LABEL: test_broadcast_8i16_16i16:
; X32: ## BB#0:
; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X32-NEXT: vmovaps (%eax), %xmm0
-; X32-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; X32-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
; X32-NEXT: retl
;
; X64-LABEL: test_broadcast_8i16_16i16:
; X64: ## BB#0:
-; X64-NEXT: vmovaps (%rdi), %xmm0
-; X64-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; X64-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
; X64-NEXT: retq
%1 = load <8 x i16>, <8 x i16> *%p
%2 = shufflevector <8 x i16> %1, <8 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
; X32-LABEL: test_broadcast_16i8_32i8:
; X32: ## BB#0:
; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X32-NEXT: vmovaps (%eax), %xmm0
-; X32-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; X32-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
; X32-NEXT: retl
;
; X64-LABEL: test_broadcast_16i8_32i8:
; X64: ## BB#0:
-; X64-NEXT: vmovaps (%rdi), %xmm0
-; X64-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; X64-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
; X64-NEXT: retq
%1 = load <16 x i8>, <16 x i8> *%p
%2 = shufflevector <16 x i8> %1, <16 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
; X32-LABEL: test_mm256_broadcastsi128_si256_mem:
; X32: # BB#0:
; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X32-NEXT: vmovaps (%eax), %xmm0
-; X32-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; X32-NEXT: vbroadcasti128 {{.*#+}} ymm0 = mem[0,1,0,1]
; X32-NEXT: retl
;
; X64-LABEL: test_mm256_broadcastsi128_si256_mem:
; X64: # BB#0:
-; X64-NEXT: vmovaps (%rdi), %xmm0
-; X64-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; X64-NEXT: vbroadcasti128 {{.*#+}} ymm0 = mem[0,1,0,1]
; X64-NEXT: retq
%a0 = load <2 x i64>, <2 x i64>* %p0
%res = shufflevector <2 x i64> %a0, <2 x i64> undef, <4 x i32> <i32 0, i32 1, i32 0, i32 1>
; X32-LABEL: test_broadcast_2f64_4f64:
; X32: ## BB#0:
; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X32-NEXT: vmovapd (%eax), %xmm0
-; X32-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; X32-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
; X32-NEXT: vaddpd LCPI0_0, %ymm0, %ymm0
; X32-NEXT: retl
;
; X64-LABEL: test_broadcast_2f64_4f64:
; X64: ## BB#0:
-; X64-NEXT: vmovapd (%rdi), %xmm0
-; X64-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; X64-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
; X64-NEXT: vaddpd {{.*}}(%rip), %ymm0, %ymm0
; X64-NEXT: retq
%1 = load <2 x double>, <2 x double> *%p
; X32-LABEL: test_broadcast_2i64_4i64:
; X32: ## BB#0:
; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X32-NEXT: vmovdqa (%eax), %xmm0
-; X32-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; X32-NEXT: vbroadcasti128 {{.*#+}} ymm0 = mem[0,1,0,1]
; X32-NEXT: vpaddq LCPI1_0, %ymm0, %ymm0
; X32-NEXT: retl
;
; X64-LABEL: test_broadcast_2i64_4i64:
; X64: ## BB#0:
-; X64-NEXT: vmovdqa (%rdi), %xmm0
-; X64-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; X64-NEXT: vbroadcasti128 {{.*#+}} ymm0 = mem[0,1,0,1]
; X64-NEXT: vpaddq {{.*}}(%rip), %ymm0, %ymm0
; X64-NEXT: retq
%1 = load <2 x i64>, <2 x i64> *%p
; X32-LABEL: test_broadcast_4f32_8f32:
; X32: ## BB#0:
; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X32-NEXT: vmovaps (%eax), %xmm0
-; X32-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; X32-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
; X32-NEXT: vaddps LCPI2_0, %ymm0, %ymm0
; X32-NEXT: retl
;
; X64-LABEL: test_broadcast_4f32_8f32:
; X64: ## BB#0:
-; X64-NEXT: vmovaps (%rdi), %xmm0
-; X64-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; X64-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
; X64-NEXT: vaddps {{.*}}(%rip), %ymm0, %ymm0
; X64-NEXT: retq
%1 = load <4 x float>, <4 x float> *%p
; X32-LABEL: test_broadcast_4i32_8i32:
; X32: ## BB#0:
; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X32-NEXT: vmovdqa (%eax), %xmm0
-; X32-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; X32-NEXT: vbroadcasti128 {{.*#+}} ymm0 = mem[0,1,0,1]
; X32-NEXT: vpaddd LCPI3_0, %ymm0, %ymm0
; X32-NEXT: retl
;
; X64-LABEL: test_broadcast_4i32_8i32:
; X64: ## BB#0:
-; X64-NEXT: vmovdqa (%rdi), %xmm0
-; X64-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; X64-NEXT: vbroadcasti128 {{.*#+}} ymm0 = mem[0,1,0,1]
; X64-NEXT: vpaddd {{.*}}(%rip), %ymm0, %ymm0
; X64-NEXT: retq
%1 = load <4 x i32>, <4 x i32> *%p
; X32-LABEL: test_broadcast_8i16_16i16:
; X32: ## BB#0:
; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X32-NEXT: vmovdqa (%eax), %xmm0
-; X32-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; X32-NEXT: vbroadcasti128 {{.*#+}} ymm0 = mem[0,1,0,1]
; X32-NEXT: vpaddw LCPI4_0, %ymm0, %ymm0
; X32-NEXT: retl
;
; X64-LABEL: test_broadcast_8i16_16i16:
; X64: ## BB#0:
-; X64-NEXT: vmovdqa (%rdi), %xmm0
-; X64-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; X64-NEXT: vbroadcasti128 {{.*#+}} ymm0 = mem[0,1,0,1]
; X64-NEXT: vpaddw {{.*}}(%rip), %ymm0, %ymm0
; X64-NEXT: retq
%1 = load <8 x i16>, <8 x i16> *%p
; X32-LABEL: test_broadcast_16i8_32i8:
; X32: ## BB#0:
; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X32-NEXT: vmovdqa (%eax), %xmm0
-; X32-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; X32-NEXT: vbroadcasti128 {{.*#+}} ymm0 = mem[0,1,0,1]
; X32-NEXT: vpaddb LCPI5_0, %ymm0, %ymm0
; X32-NEXT: retl
;
; X64-LABEL: test_broadcast_16i8_32i8:
; X64: ## BB#0:
-; X64-NEXT: vmovdqa (%rdi), %xmm0
-; X64-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; X64-NEXT: vbroadcasti128 {{.*#+}} ymm0 = mem[0,1,0,1]
; X64-NEXT: vpaddb {{.*}}(%rip), %ymm0, %ymm0
; X64-NEXT: retq
%1 = load <16 x i8>, <16 x i8> *%p
define <4 x double> @test_broadcast_2f64_4f64(<2 x double> *%p) nounwind {
; X64-AVX512VL-LABEL: test_broadcast_2f64_4f64:
; X64-AVX512VL: ## BB#0:
-; X64-AVX512VL-NEXT: vmovapd (%rdi), %xmm0
-; X64-AVX512VL-NEXT: vinsertf32x4 $1, %xmm0, %ymm0, %ymm0
+; X64-AVX512VL-NEXT: vbroadcastf32x4 (%rdi), %ymm0
; X64-AVX512VL-NEXT: vaddpd {{.*}}(%rip), %ymm0, %ymm0
; X64-AVX512VL-NEXT: retq
;
; X64-AVX512BWVL-LABEL: test_broadcast_2f64_4f64:
; X64-AVX512BWVL: ## BB#0:
-; X64-AVX512BWVL-NEXT: vmovapd (%rdi), %xmm0
-; X64-AVX512BWVL-NEXT: vinsertf32x4 $1, %xmm0, %ymm0, %ymm0
+; X64-AVX512BWVL-NEXT: vbroadcastf32x4 (%rdi), %ymm0
; X64-AVX512BWVL-NEXT: vaddpd {{.*}}(%rip), %ymm0, %ymm0
; X64-AVX512BWVL-NEXT: retq
;
; X64-AVX512DQVL-LABEL: test_broadcast_2f64_4f64:
; X64-AVX512DQVL: ## BB#0:
-; X64-AVX512DQVL-NEXT: vmovapd (%rdi), %xmm0
-; X64-AVX512DQVL-NEXT: vinsertf64x2 $1, %xmm0, %ymm0, %ymm0
+; X64-AVX512DQVL-NEXT: vbroadcastf64x2 (%rdi), %ymm0
; X64-AVX512DQVL-NEXT: vaddpd {{.*}}(%rip), %ymm0, %ymm0
; X64-AVX512DQVL-NEXT: retq
%1 = load <2 x double>, <2 x double> *%p
define <4 x i64> @test_broadcast_2i64_4i64(<2 x i64> *%p) nounwind {
; X64-AVX512VL-LABEL: test_broadcast_2i64_4i64:
; X64-AVX512VL: ## BB#0:
-; X64-AVX512VL-NEXT: vmovdqa64 (%rdi), %xmm0
-; X64-AVX512VL-NEXT: vinserti32x4 $1, %xmm0, %ymm0, %ymm0
+; X64-AVX512VL-NEXT: vbroadcasti32x4 (%rdi), %ymm0
; X64-AVX512VL-NEXT: vpaddq {{.*}}(%rip), %ymm0, %ymm0
; X64-AVX512VL-NEXT: retq
;
; X64-AVX512BWVL-LABEL: test_broadcast_2i64_4i64:
; X64-AVX512BWVL: ## BB#0:
-; X64-AVX512BWVL-NEXT: vmovdqa64 (%rdi), %xmm0
-; X64-AVX512BWVL-NEXT: vinserti32x4 $1, %xmm0, %ymm0, %ymm0
+; X64-AVX512BWVL-NEXT: vbroadcasti32x4 (%rdi), %ymm0
; X64-AVX512BWVL-NEXT: vpaddq {{.*}}(%rip), %ymm0, %ymm0
; X64-AVX512BWVL-NEXT: retq
;
; X64-AVX512DQVL-LABEL: test_broadcast_2i64_4i64:
; X64-AVX512DQVL: ## BB#0:
-; X64-AVX512DQVL-NEXT: vmovdqa64 (%rdi), %xmm0
-; X64-AVX512DQVL-NEXT: vinserti64x2 $1, %xmm0, %ymm0, %ymm0
+; X64-AVX512DQVL-NEXT: vbroadcasti64x2 (%rdi), %ymm0
; X64-AVX512DQVL-NEXT: vpaddq {{.*}}(%rip), %ymm0, %ymm0
; X64-AVX512DQVL-NEXT: retq
%1 = load <2 x i64>, <2 x i64> *%p
define <8 x float> @test_broadcast_4f32_8f32(<4 x float> *%p) nounwind {
; X64-AVX512-LABEL: test_broadcast_4f32_8f32:
; X64-AVX512: ## BB#0:
-; X64-AVX512-NEXT: vmovaps (%rdi), %xmm0
-; X64-AVX512-NEXT: vinsertf32x4 $1, %xmm0, %ymm0, %ymm0
+; X64-AVX512-NEXT: vbroadcastf32x4 (%rdi), %ymm0
; X64-AVX512-NEXT: vaddps {{.*}}(%rip), %ymm0, %ymm0
; X64-AVX512-NEXT: retq
%1 = load <4 x float>, <4 x float> *%p
define <8 x i32> @test_broadcast_4i32_8i32(<4 x i32> *%p) nounwind {
; X64-AVX512-LABEL: test_broadcast_4i32_8i32:
; X64-AVX512: ## BB#0:
-; X64-AVX512-NEXT: vmovdqa32 (%rdi), %xmm0
-; X64-AVX512-NEXT: vinserti32x4 $1, %xmm0, %ymm0, %ymm0
+; X64-AVX512-NEXT: vbroadcasti32x4 (%rdi), %ymm0
; X64-AVX512-NEXT: vpaddd {{.*}}(%rip), %ymm0, %ymm0
; X64-AVX512-NEXT: retq
%1 = load <4 x i32>, <4 x i32> *%p
}
define <16 x i16> @test_broadcast_8i16_16i16(<8 x i16> *%p) nounwind {
-; X64-AVX512VL-LABEL: test_broadcast_8i16_16i16:
-; X64-AVX512VL: ## BB#0:
-; X64-AVX512VL-NEXT: vmovdqa64 (%rdi), %xmm0
-; X64-AVX512VL-NEXT: vinserti32x4 $1, %xmm0, %ymm0, %ymm0
-; X64-AVX512VL-NEXT: vpaddw {{.*}}(%rip), %ymm0, %ymm0
-; X64-AVX512VL-NEXT: retq
-;
-; X64-AVX512BWVL-LABEL: test_broadcast_8i16_16i16:
-; X64-AVX512BWVL: ## BB#0:
-; X64-AVX512BWVL-NEXT: vmovdqu16 (%rdi), %xmm0
-; X64-AVX512BWVL-NEXT: vinserti32x4 $1, %xmm0, %ymm0, %ymm0
-; X64-AVX512BWVL-NEXT: vpaddw {{.*}}(%rip), %ymm0, %ymm0
-; X64-AVX512BWVL-NEXT: retq
-;
-; X64-AVX512DQVL-LABEL: test_broadcast_8i16_16i16:
-; X64-AVX512DQVL: ## BB#0:
-; X64-AVX512DQVL-NEXT: vmovdqa64 (%rdi), %xmm0
-; X64-AVX512DQVL-NEXT: vinserti32x4 $1, %xmm0, %ymm0, %ymm0
-; X64-AVX512DQVL-NEXT: vpaddw {{.*}}(%rip), %ymm0, %ymm0
-; X64-AVX512DQVL-NEXT: retq
+; X64-AVX512-LABEL: test_broadcast_8i16_16i16:
+; X64-AVX512: ## BB#0:
+; X64-AVX512-NEXT: vbroadcasti32x4 (%rdi), %ymm0
+; X64-AVX512-NEXT: vpaddw {{.*}}(%rip), %ymm0, %ymm0
+; X64-AVX512-NEXT: retq
%1 = load <8 x i16>, <8 x i16> *%p
%2 = shufflevector <8 x i16> %1, <8 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
%3 = add <16 x i16> %2, <i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16>
}
define <32 x i8> @test_broadcast_16i8_32i8(<16 x i8> *%p) nounwind {
-; X64-AVX512VL-LABEL: test_broadcast_16i8_32i8:
-; X64-AVX512VL: ## BB#0:
-; X64-AVX512VL-NEXT: vmovdqa64 (%rdi), %xmm0
-; X64-AVX512VL-NEXT: vinserti32x4 $1, %xmm0, %ymm0, %ymm0
-; X64-AVX512VL-NEXT: vpaddb {{.*}}(%rip), %ymm0, %ymm0
-; X64-AVX512VL-NEXT: retq
-;
-; X64-AVX512BWVL-LABEL: test_broadcast_16i8_32i8:
-; X64-AVX512BWVL: ## BB#0:
-; X64-AVX512BWVL-NEXT: vmovdqu8 (%rdi), %xmm0
-; X64-AVX512BWVL-NEXT: vinserti32x4 $1, %xmm0, %ymm0, %ymm0
-; X64-AVX512BWVL-NEXT: vpaddb {{.*}}(%rip), %ymm0, %ymm0
-; X64-AVX512BWVL-NEXT: retq
-;
-; X64-AVX512DQVL-LABEL: test_broadcast_16i8_32i8:
-; X64-AVX512DQVL: ## BB#0:
-; X64-AVX512DQVL-NEXT: vmovdqa64 (%rdi), %xmm0
-; X64-AVX512DQVL-NEXT: vinserti32x4 $1, %xmm0, %ymm0, %ymm0
-; X64-AVX512DQVL-NEXT: vpaddb {{.*}}(%rip), %ymm0, %ymm0
-; X64-AVX512DQVL-NEXT: retq
+; X64-AVX512-LABEL: test_broadcast_16i8_32i8:
+; X64-AVX512: ## BB#0:
+; X64-AVX512-NEXT: vbroadcasti32x4 (%rdi), %ymm0
+; X64-AVX512-NEXT: vpaddb {{.*}}(%rip), %ymm0, %ymm0
+; X64-AVX512-NEXT: retq
%1 = load <16 x i8>, <16 x i8> *%p
%2 = shufflevector <16 x i8> %1, <16 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
%3 = add <32 x i8> %2, <i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 20, i8 21, i8 22, i8 23, i8 24, i8 25, i8 26, i8 27, i8 28, i8 29, i8 30, i8 31, i8 32>
define <8 x double> @test_broadcast_2f64_8f64(<2 x double> *%p) nounwind {
; X64-AVX512VL-LABEL: test_broadcast_2f64_8f64:
; X64-AVX512VL: ## BB#0:
-; X64-AVX512VL-NEXT: vmovapd (%rdi), %xmm0
-; X64-AVX512VL-NEXT: vinsertf32x4 $1, %xmm0, %ymm0, %ymm0
+; X64-AVX512VL-NEXT: vbroadcastf32x4 (%rdi), %ymm0
; X64-AVX512VL-NEXT: vinsertf64x4 $1, %ymm0, %zmm0, %zmm0
; X64-AVX512VL-NEXT: vaddpd {{.*}}(%rip), %zmm0, %zmm0
; X64-AVX512VL-NEXT: retq
;
; X64-AVX512BWVL-LABEL: test_broadcast_2f64_8f64:
; X64-AVX512BWVL: ## BB#0:
-; X64-AVX512BWVL-NEXT: vmovapd (%rdi), %xmm0
-; X64-AVX512BWVL-NEXT: vinsertf32x4 $1, %xmm0, %ymm0, %ymm0
+; X64-AVX512BWVL-NEXT: vbroadcastf32x4 (%rdi), %ymm0
; X64-AVX512BWVL-NEXT: vinsertf64x4 $1, %ymm0, %zmm0, %zmm0
; X64-AVX512BWVL-NEXT: vaddpd {{.*}}(%rip), %zmm0, %zmm0
; X64-AVX512BWVL-NEXT: retq
;
; X64-AVX512DQVL-LABEL: test_broadcast_2f64_8f64:
; X64-AVX512DQVL: ## BB#0:
-; X64-AVX512DQVL-NEXT: vmovapd (%rdi), %xmm0
-; X64-AVX512DQVL-NEXT: vinsertf64x2 $1, %xmm0, %ymm0, %ymm0
+; X64-AVX512DQVL-NEXT: vbroadcastf64x2 (%rdi), %ymm0
; X64-AVX512DQVL-NEXT: vinsertf64x4 $1, %ymm0, %zmm0, %zmm0
; X64-AVX512DQVL-NEXT: vaddpd {{.*}}(%rip), %zmm0, %zmm0
; X64-AVX512DQVL-NEXT: retq
define <8 x i64> @test_broadcast_2i64_8i64(<2 x i64> *%p) nounwind {
; X64-AVX512VL-LABEL: test_broadcast_2i64_8i64:
; X64-AVX512VL: ## BB#0:
-; X64-AVX512VL-NEXT: vmovdqa64 (%rdi), %xmm0
-; X64-AVX512VL-NEXT: vinserti32x4 $1, %xmm0, %ymm0, %ymm0
+; X64-AVX512VL-NEXT: vbroadcasti32x4 (%rdi), %ymm0
; X64-AVX512VL-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm0
; X64-AVX512VL-NEXT: vpaddq {{.*}}(%rip), %zmm0, %zmm0
; X64-AVX512VL-NEXT: retq
;
; X64-AVX512BWVL-LABEL: test_broadcast_2i64_8i64:
; X64-AVX512BWVL: ## BB#0:
-; X64-AVX512BWVL-NEXT: vmovdqa64 (%rdi), %xmm0
-; X64-AVX512BWVL-NEXT: vinserti32x4 $1, %xmm0, %ymm0, %ymm0
+; X64-AVX512BWVL-NEXT: vbroadcasti32x4 (%rdi), %ymm0
; X64-AVX512BWVL-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm0
; X64-AVX512BWVL-NEXT: vpaddq {{.*}}(%rip), %zmm0, %zmm0
; X64-AVX512BWVL-NEXT: retq
;
; X64-AVX512DQVL-LABEL: test_broadcast_2i64_8i64:
; X64-AVX512DQVL: ## BB#0:
-; X64-AVX512DQVL-NEXT: vmovdqa64 (%rdi), %xmm0
-; X64-AVX512DQVL-NEXT: vinserti64x2 $1, %xmm0, %ymm0, %ymm0
+; X64-AVX512DQVL-NEXT: vbroadcasti64x2 (%rdi), %ymm0
; X64-AVX512DQVL-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm0
; X64-AVX512DQVL-NEXT: vpaddq {{.*}}(%rip), %zmm0, %zmm0
; X64-AVX512DQVL-NEXT: retq
define <16 x float> @test_broadcast_4f32_16f32(<4 x float> *%p) nounwind {
; X64-AVX512VL-LABEL: test_broadcast_4f32_16f32:
; X64-AVX512VL: ## BB#0:
-; X64-AVX512VL-NEXT: vmovaps (%rdi), %xmm0
-; X64-AVX512VL-NEXT: vinsertf32x4 $1, %xmm0, %ymm0, %ymm0
+; X64-AVX512VL-NEXT: vbroadcastf32x4 (%rdi), %ymm0
; X64-AVX512VL-NEXT: vinsertf64x4 $1, %ymm0, %zmm0, %zmm0
; X64-AVX512VL-NEXT: vaddps {{.*}}(%rip), %zmm0, %zmm0
; X64-AVX512VL-NEXT: retq
;
; X64-AVX512BWVL-LABEL: test_broadcast_4f32_16f32:
; X64-AVX512BWVL: ## BB#0:
-; X64-AVX512BWVL-NEXT: vmovaps (%rdi), %xmm0
-; X64-AVX512BWVL-NEXT: vinsertf32x4 $1, %xmm0, %ymm0, %ymm0
+; X64-AVX512BWVL-NEXT: vbroadcastf32x4 (%rdi), %ymm0
; X64-AVX512BWVL-NEXT: vinsertf64x4 $1, %ymm0, %zmm0, %zmm0
; X64-AVX512BWVL-NEXT: vaddps {{.*}}(%rip), %zmm0, %zmm0
; X64-AVX512BWVL-NEXT: retq
;
; X64-AVX512DQVL-LABEL: test_broadcast_4f32_16f32:
; X64-AVX512DQVL: ## BB#0:
-; X64-AVX512DQVL-NEXT: vmovaps (%rdi), %xmm0
-; X64-AVX512DQVL-NEXT: vinsertf32x4 $1, %xmm0, %ymm0, %ymm0
+; X64-AVX512DQVL-NEXT: vbroadcastf32x4 (%rdi), %ymm0
; X64-AVX512DQVL-NEXT: vinsertf32x8 $1, %ymm0, %zmm0, %zmm0
; X64-AVX512DQVL-NEXT: vaddps {{.*}}(%rip), %zmm0, %zmm0
; X64-AVX512DQVL-NEXT: retq
define <16 x i32> @test_broadcast_4i32_16i32(<4 x i32> *%p) nounwind {
; X64-AVX512VL-LABEL: test_broadcast_4i32_16i32:
; X64-AVX512VL: ## BB#0:
-; X64-AVX512VL-NEXT: vmovdqa32 (%rdi), %xmm0
-; X64-AVX512VL-NEXT: vinserti32x4 $1, %xmm0, %ymm0, %ymm0
+; X64-AVX512VL-NEXT: vbroadcasti32x4 (%rdi), %ymm0
; X64-AVX512VL-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm0
; X64-AVX512VL-NEXT: vpaddd {{.*}}(%rip), %zmm0, %zmm0
; X64-AVX512VL-NEXT: retq
;
; X64-AVX512BWVL-LABEL: test_broadcast_4i32_16i32:
; X64-AVX512BWVL: ## BB#0:
-; X64-AVX512BWVL-NEXT: vmovdqa32 (%rdi), %xmm0
-; X64-AVX512BWVL-NEXT: vinserti32x4 $1, %xmm0, %ymm0, %ymm0
+; X64-AVX512BWVL-NEXT: vbroadcasti32x4 (%rdi), %ymm0
; X64-AVX512BWVL-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm0
; X64-AVX512BWVL-NEXT: vpaddd {{.*}}(%rip), %zmm0, %zmm0
; X64-AVX512BWVL-NEXT: retq
;
; X64-AVX512DQVL-LABEL: test_broadcast_4i32_16i32:
; X64-AVX512DQVL: ## BB#0:
-; X64-AVX512DQVL-NEXT: vmovdqa32 (%rdi), %xmm0
-; X64-AVX512DQVL-NEXT: vinserti32x4 $1, %xmm0, %ymm0, %ymm0
+; X64-AVX512DQVL-NEXT: vbroadcasti32x4 (%rdi), %ymm0
; X64-AVX512DQVL-NEXT: vinserti32x8 $1, %ymm0, %zmm0, %zmm0
; X64-AVX512DQVL-NEXT: vpaddd {{.*}}(%rip), %zmm0, %zmm0
; X64-AVX512DQVL-NEXT: retq
define <32 x i16> @test_broadcast_8i16_32i16(<8 x i16> *%p) nounwind {
; X64-AVX512VL-LABEL: test_broadcast_8i16_32i16:
; X64-AVX512VL: ## BB#0:
-; X64-AVX512VL-NEXT: vmovdqa64 (%rdi), %xmm0
-; X64-AVX512VL-NEXT: vinserti32x4 $1, %xmm0, %ymm0, %ymm1
+; X64-AVX512VL-NEXT: vbroadcasti32x4 (%rdi), %ymm1
; X64-AVX512VL-NEXT: vpaddw {{.*}}(%rip), %ymm1, %ymm0
; X64-AVX512VL-NEXT: vpaddw {{.*}}(%rip), %ymm1, %ymm1
; X64-AVX512VL-NEXT: retq
;
; X64-AVX512BWVL-LABEL: test_broadcast_8i16_32i16:
; X64-AVX512BWVL: ## BB#0:
-; X64-AVX512BWVL-NEXT: vmovdqu16 (%rdi), %xmm0
-; X64-AVX512BWVL-NEXT: vinserti32x4 $1, %xmm0, %ymm0, %ymm0
+; X64-AVX512BWVL-NEXT: vbroadcasti32x4 (%rdi), %ymm0
; X64-AVX512BWVL-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm0
; X64-AVX512BWVL-NEXT: vpaddw {{.*}}(%rip), %zmm0, %zmm0
; X64-AVX512BWVL-NEXT: retq
;
; X64-AVX512DQVL-LABEL: test_broadcast_8i16_32i16:
; X64-AVX512DQVL: ## BB#0:
-; X64-AVX512DQVL-NEXT: vmovdqa64 (%rdi), %xmm0
-; X64-AVX512DQVL-NEXT: vinserti32x4 $1, %xmm0, %ymm0, %ymm1
+; X64-AVX512DQVL-NEXT: vbroadcasti32x4 (%rdi), %ymm1
; X64-AVX512DQVL-NEXT: vpaddw {{.*}}(%rip), %ymm1, %ymm0
; X64-AVX512DQVL-NEXT: vpaddw {{.*}}(%rip), %ymm1, %ymm1
; X64-AVX512DQVL-NEXT: retq
define <64 x i8> @test_broadcast_16i8_64i8(<16 x i8> *%p) nounwind {
; X64-AVX512VL-LABEL: test_broadcast_16i8_64i8:
; X64-AVX512VL: ## BB#0:
-; X64-AVX512VL-NEXT: vmovdqa64 (%rdi), %xmm0
-; X64-AVX512VL-NEXT: vinserti32x4 $1, %xmm0, %ymm0, %ymm1
+; X64-AVX512VL-NEXT: vbroadcasti32x4 (%rdi), %ymm1
; X64-AVX512VL-NEXT: vpaddb {{.*}}(%rip), %ymm1, %ymm0
; X64-AVX512VL-NEXT: vpaddb {{.*}}(%rip), %ymm1, %ymm1
; X64-AVX512VL-NEXT: retq
;
; X64-AVX512BWVL-LABEL: test_broadcast_16i8_64i8:
; X64-AVX512BWVL: ## BB#0:
-; X64-AVX512BWVL-NEXT: vmovdqu8 (%rdi), %xmm0
-; X64-AVX512BWVL-NEXT: vinserti32x4 $1, %xmm0, %ymm0, %ymm0
+; X64-AVX512BWVL-NEXT: vbroadcasti32x4 (%rdi), %ymm0
; X64-AVX512BWVL-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm0
; X64-AVX512BWVL-NEXT: vpaddb {{.*}}(%rip), %zmm0, %zmm0
; X64-AVX512BWVL-NEXT: retq
;
; X64-AVX512DQVL-LABEL: test_broadcast_16i8_64i8:
; X64-AVX512DQVL: ## BB#0:
-; X64-AVX512DQVL-NEXT: vmovdqa64 (%rdi), %xmm0
-; X64-AVX512DQVL-NEXT: vinserti32x4 $1, %xmm0, %ymm0, %ymm1
+; X64-AVX512DQVL-NEXT: vbroadcasti32x4 (%rdi), %ymm1
; X64-AVX512DQVL-NEXT: vpaddb {{.*}}(%rip), %ymm1, %ymm0
; X64-AVX512DQVL-NEXT: vpaddb {{.*}}(%rip), %ymm1, %ymm1
; X64-AVX512DQVL-NEXT: retq
define <4 x i64> @splat128_mem_v4i64_from_v2i64(<2 x i64>* %ptr) {
; AVX1-LABEL: splat128_mem_v4i64_from_v2i64:
; AVX1: # BB#0:
-; AVX1-NEXT: vmovaps (%rdi), %xmm0
-; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
; AVX1-NEXT: retq
;
; AVX2-LABEL: splat128_mem_v4i64_from_v2i64:
; AVX2: # BB#0:
-; AVX2-NEXT: vmovaps (%rdi), %xmm0
-; AVX2-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-NEXT: vbroadcasti128 {{.*#+}} ymm0 = mem[0,1,0,1]
; AVX2-NEXT: retq
;
; AVX512VL-LABEL: splat128_mem_v4i64_from_v2i64:
; AVX512VL: # BB#0:
-; AVX512VL-NEXT: vmovdqa64 (%rdi), %xmm0
-; AVX512VL-NEXT: vinserti32x4 $1, %xmm0, %ymm0, %ymm0
+; AVX512VL-NEXT: vbroadcasti32x4 (%rdi), %ymm0
; AVX512VL-NEXT: retq
%v = load <2 x i64>, <2 x i64>* %ptr
%shuffle = shufflevector <2 x i64> %v, <2 x i64> undef, <4 x i32> <i32 0, i32 1, i32 0, i32 1>
define <4 x double> @splat128_mem_v4f64_from_v2f64(<2 x double>* %ptr) {
; AVX1-LABEL: splat128_mem_v4f64_from_v2f64:
; AVX1: # BB#0:
-; AVX1-NEXT: vmovaps (%rdi), %xmm0
-; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
; AVX1-NEXT: retq
;
; AVX2-LABEL: splat128_mem_v4f64_from_v2f64:
; AVX2: # BB#0:
-; AVX2-NEXT: vmovaps (%rdi), %xmm0
-; AVX2-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
; AVX2-NEXT: retq
;
; AVX512VL-LABEL: splat128_mem_v4f64_from_v2f64:
; AVX512VL: # BB#0:
-; AVX512VL-NEXT: vmovapd (%rdi), %xmm0
-; AVX512VL-NEXT: vinsertf32x4 $1, %xmm0, %ymm0, %ymm0
+; AVX512VL-NEXT: vbroadcastf32x4 (%rdi), %ymm0
; AVX512VL-NEXT: retq
%v = load <2 x double>, <2 x double>* %ptr
%shuffle = shufflevector <2 x double> %v, <2 x double> undef, <4 x i32> <i32 0, i32 1, i32 0, i32 1>