| ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py |
| ; RUN: llc < %s -mtriple=i686-apple-darwin -mattr=+avx,+mmx | FileCheck %s --check-prefix=X86 |
| ; RUN: llc < %s -mtriple=x86_64-apple-darwin -mattr=+avx,+mmx | FileCheck %s --check-prefix=X64 |
| |
| define <4 x i64> @A(ptr %ptr) nounwind uwtable readnone ssp { |
| ; X86-LABEL: A: |
| ; X86: ## %bb.0: ## %entry |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: vbroadcastsd (%eax), %ymm0 |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: A: |
| ; X64: ## %bb.0: ## %entry |
| ; X64-NEXT: vbroadcastsd (%rdi), %ymm0 |
| ; X64-NEXT: retq |
| entry: |
| %q = load i64, ptr %ptr, align 8 |
| %vecinit.i = insertelement <4 x i64> undef, i64 %q, i32 0 |
| %vecinit2.i = insertelement <4 x i64> %vecinit.i, i64 %q, i32 1 |
| %vecinit4.i = insertelement <4 x i64> %vecinit2.i, i64 %q, i32 2 |
| %vecinit6.i = insertelement <4 x i64> %vecinit4.i, i64 %q, i32 3 |
| ret <4 x i64> %vecinit6.i |
| } |
| |
| define <4 x i64> @A2(ptr %ptr, ptr %ptr2) nounwind uwtable readnone ssp { |
| ; X86-LABEL: A2: |
| ; X86: ## %bb.0: ## %entry |
| ; X86-NEXT: pushl %esi |
| ; X86-NEXT: .cfi_def_cfa_offset 8 |
| ; X86-NEXT: .cfi_offset %esi, -8 |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx |
| ; X86-NEXT: movl (%ecx), %edx |
| ; X86-NEXT: movl 4(%ecx), %esi |
| ; X86-NEXT: vbroadcastsd (%ecx), %ymm0 |
| ; X86-NEXT: movl %edx, (%eax) |
| ; X86-NEXT: movl %esi, 4(%eax) |
| ; X86-NEXT: popl %esi |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: A2: |
| ; X64: ## %bb.0: ## %entry |
| ; X64-NEXT: movq (%rdi), %rax |
| ; X64-NEXT: movq %rax, (%rsi) |
| ; X64-NEXT: vmovq %rax, %xmm0 |
| ; X64-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,1,0,1] |
| ; X64-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0 |
| ; X64-NEXT: retq |
| entry: |
| %q = load i64, ptr %ptr, align 8 |
| store i64 %q, ptr %ptr2, align 8 ; to create a chain to prevent broadcast |
| %vecinit.i = insertelement <4 x i64> undef, i64 %q, i32 0 |
| %vecinit2.i = insertelement <4 x i64> %vecinit.i, i64 %q, i32 1 |
| %vecinit4.i = insertelement <4 x i64> %vecinit2.i, i64 %q, i32 2 |
| %vecinit6.i = insertelement <4 x i64> %vecinit4.i, i64 %q, i32 3 |
| ret <4 x i64> %vecinit6.i |
| } |
| |
| define <8 x i32> @B(ptr %ptr) nounwind uwtable readnone ssp { |
| ; X86-LABEL: B: |
| ; X86: ## %bb.0: ## %entry |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: vbroadcastss (%eax), %ymm0 |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: B: |
| ; X64: ## %bb.0: ## %entry |
| ; X64-NEXT: vbroadcastss (%rdi), %ymm0 |
| ; X64-NEXT: retq |
| entry: |
| %q = load i32, ptr %ptr, align 4 |
| %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0 |
| %vecinit2.i = insertelement <8 x i32> %vecinit.i, i32 %q, i32 1 |
| %vecinit4.i = insertelement <8 x i32> %vecinit2.i, i32 %q, i32 2 |
| %vecinit6.i = insertelement <8 x i32> %vecinit4.i, i32 %q, i32 3 |
| ret <8 x i32> %vecinit6.i |
| } |
| |
| define <8 x i32> @B2(ptr %ptr) nounwind uwtable readnone ssp { |
| ; X86-LABEL: B2: |
| ; X86: ## %bb.0: ## %entry |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: vbroadcastss (%eax), %ymm0 |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: B2: |
| ; X64: ## %bb.0: ## %entry |
| ; X64-NEXT: vbroadcastss (%rdi), %ymm0 |
| ; X64-NEXT: retq |
| entry: |
| %q = load i32, ptr %ptr, align 4 |
| %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0 |
| %vecinit2.i = insertelement <8 x i32> %vecinit.i, i32 %q, i32 1 |
| %vecinit4.i = insertelement <8 x i32> %vecinit2.i, i32 %q, i32 2 |
| %vecinit6.i = insertelement <8 x i32> %vecinit4.i, i32 %q, i32 3 |
| %vecinit8.i = insertelement <8 x i32> %vecinit6.i, i32 %q, i32 4 |
| %vecinit10.i = insertelement <8 x i32> %vecinit8.i, i32 %q, i32 5 |
| %vecinit12.i = insertelement <8 x i32> %vecinit10.i, i32 %q, i32 6 |
| %vecinit14.i = insertelement <8 x i32> %vecinit12.i, i32 %q, i32 7 |
| ret <8 x i32> %vecinit14.i |
| } |
| |
| define <8 x i32> @B3(ptr %ptr, ptr %ptr2) nounwind uwtable readnone ssp { |
| ; X86-LABEL: B3: |
| ; X86: ## %bb.0: ## %entry |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx |
| ; X86-NEXT: movl (%ecx), %ecx |
| ; X86-NEXT: movl %ecx, (%eax) |
| ; X86-NEXT: vmovd %ecx, %xmm0 |
| ; X86-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,0,0,0] |
| ; X86-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0 |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: B3: |
| ; X64: ## %bb.0: ## %entry |
| ; X64-NEXT: movl (%rdi), %eax |
| ; X64-NEXT: movl %eax, (%rsi) |
| ; X64-NEXT: vmovd %eax, %xmm0 |
| ; X64-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,0,0,0] |
| ; X64-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0 |
| ; X64-NEXT: retq |
| entry: |
| %q = load i32, ptr %ptr, align 4 |
| store i32 %q, ptr %ptr2, align 4 ; to create a chain to prevent broadcast |
| %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0 |
| %vecinit2.i = insertelement <8 x i32> %vecinit.i, i32 %q, i32 1 |
| %vecinit4.i = insertelement <8 x i32> %vecinit2.i, i32 %q, i32 2 |
| %vecinit6.i = insertelement <8 x i32> %vecinit4.i, i32 %q, i32 3 |
| %vecinit8.i = insertelement <8 x i32> %vecinit6.i, i32 %q, i32 4 |
| %vecinit10.i = insertelement <8 x i32> %vecinit8.i, i32 %q, i32 5 |
| %vecinit12.i = insertelement <8 x i32> %vecinit10.i, i32 %q, i32 6 |
| %vecinit14.i = insertelement <8 x i32> %vecinit12.i, i32 %q, i32 7 |
| ret <8 x i32> %vecinit14.i |
| } |
| |
| define <4 x double> @C(ptr %ptr) nounwind uwtable readnone ssp { |
| ; X86-LABEL: C: |
| ; X86: ## %bb.0: ## %entry |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: vbroadcastsd (%eax), %ymm0 |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: C: |
| ; X64: ## %bb.0: ## %entry |
| ; X64-NEXT: vbroadcastsd (%rdi), %ymm0 |
| ; X64-NEXT: retq |
| entry: |
| %q = load double, ptr %ptr, align 8 |
| %vecinit.i = insertelement <4 x double> undef, double %q, i32 0 |
| %vecinit2.i = insertelement <4 x double> %vecinit.i, double %q, i32 1 |
| %vecinit4.i = insertelement <4 x double> %vecinit2.i, double %q, i32 2 |
| %vecinit6.i = insertelement <4 x double> %vecinit4.i, double %q, i32 3 |
| ret <4 x double> %vecinit6.i |
| } |
| |
| define <4 x double> @C2(ptr %ptr, ptr %ptr2) nounwind uwtable readnone ssp { |
| ; X86-LABEL: C2: |
| ; X86: ## %bb.0: ## %entry |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx |
| ; X86-NEXT: vbroadcastsd (%ecx), %ymm0 |
| ; X86-NEXT: vmovlps %xmm0, (%eax) |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: C2: |
| ; X64: ## %bb.0: ## %entry |
| ; X64-NEXT: vbroadcastsd (%rdi), %ymm0 |
| ; X64-NEXT: vmovlps %xmm0, (%rsi) |
| ; X64-NEXT: retq |
| entry: |
| %q = load double, ptr %ptr, align 8 |
| store double %q, ptr %ptr2, align 8 ; to create a chain to prevent broadcast |
| %vecinit.i = insertelement <4 x double> undef, double %q, i32 0 |
| %vecinit2.i = insertelement <4 x double> %vecinit.i, double %q, i32 1 |
| %vecinit4.i = insertelement <4 x double> %vecinit2.i, double %q, i32 2 |
| %vecinit6.i = insertelement <4 x double> %vecinit4.i, double %q, i32 3 |
| ret <4 x double> %vecinit6.i |
| } |
| |
| define <8 x float> @D(ptr %ptr) nounwind uwtable readnone ssp { |
| ; X86-LABEL: D: |
| ; X86: ## %bb.0: ## %entry |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: vbroadcastss (%eax), %ymm0 |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: D: |
| ; X64: ## %bb.0: ## %entry |
| ; X64-NEXT: vbroadcastss (%rdi), %ymm0 |
| ; X64-NEXT: retq |
| entry: |
| %q = load float, ptr %ptr, align 4 |
| %vecinit.i = insertelement <8 x float> undef, float %q, i32 0 |
| %vecinit2.i = insertelement <8 x float> %vecinit.i, float %q, i32 1 |
| %vecinit4.i = insertelement <8 x float> %vecinit2.i, float %q, i32 2 |
| %vecinit6.i = insertelement <8 x float> %vecinit4.i, float %q, i32 3 |
| ret <8 x float> %vecinit6.i |
| } |
| |
| define <8 x float> @D2(ptr %ptr) nounwind uwtable readnone ssp { |
| ; X86-LABEL: D2: |
| ; X86: ## %bb.0: ## %entry |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: vbroadcastss (%eax), %ymm0 |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: D2: |
| ; X64: ## %bb.0: ## %entry |
| ; X64-NEXT: vbroadcastss (%rdi), %ymm0 |
| ; X64-NEXT: retq |
| entry: |
| %q = load float, ptr %ptr, align 4 |
| %vecinit.i = insertelement <8 x float> undef, float %q, i32 0 |
| %vecinit2.i = insertelement <8 x float> %vecinit.i, float %q, i32 1 |
| %vecinit4.i = insertelement <8 x float> %vecinit2.i, float %q, i32 2 |
| %vecinit6.i = insertelement <8 x float> %vecinit4.i, float %q, i32 3 |
| %vecinit8.i = insertelement <8 x float> %vecinit6.i, float %q, i32 4 |
| %vecinit10.i = insertelement <8 x float> %vecinit8.i, float %q, i32 5 |
| %vecinit12.i = insertelement <8 x float> %vecinit10.i, float %q, i32 6 |
| %vecinit14.i = insertelement <8 x float> %vecinit12.i, float %q, i32 7 |
| ret <8 x float> %vecinit14.i |
| } |
| |
| define <8 x float> @D3(ptr %ptr, ptr %ptr2) nounwind uwtable readnone ssp { |
| ; X86-LABEL: D3: |
| ; X86: ## %bb.0: ## %entry |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx |
| ; X86-NEXT: vbroadcastss (%ecx), %ymm0 |
| ; X86-NEXT: vmovss %xmm0, (%eax) |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: D3: |
| ; X64: ## %bb.0: ## %entry |
| ; X64-NEXT: vbroadcastss (%rdi), %ymm0 |
| ; X64-NEXT: vmovss %xmm0, (%rsi) |
| ; X64-NEXT: retq |
| entry: |
| %q = load float, ptr %ptr, align 4 |
| store float %q, ptr %ptr2, align 4 ; to create a chain to prevent broadcast |
| %vecinit.i = insertelement <8 x float> undef, float %q, i32 0 |
| %vecinit2.i = insertelement <8 x float> %vecinit.i, float %q, i32 1 |
| %vecinit4.i = insertelement <8 x float> %vecinit2.i, float %q, i32 2 |
| %vecinit6.i = insertelement <8 x float> %vecinit4.i, float %q, i32 3 |
| %vecinit8.i = insertelement <8 x float> %vecinit6.i, float %q, i32 4 |
| %vecinit10.i = insertelement <8 x float> %vecinit8.i, float %q, i32 5 |
| %vecinit12.i = insertelement <8 x float> %vecinit10.i, float %q, i32 6 |
| %vecinit14.i = insertelement <8 x float> %vecinit12.i, float %q, i32 7 |
| ret <8 x float> %vecinit14.i |
| } |
| |
| ;;;; 128-bit versions |
| |
| define <4 x float> @e(ptr %ptr) nounwind uwtable readnone ssp { |
| ; X86-LABEL: e: |
| ; X86: ## %bb.0: ## %entry |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: vbroadcastss (%eax), %xmm0 |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: e: |
| ; X64: ## %bb.0: ## %entry |
| ; X64-NEXT: vbroadcastss (%rdi), %xmm0 |
| ; X64-NEXT: retq |
| entry: |
| %q = load float, ptr %ptr, align 4 |
| %vecinit.i = insertelement <4 x float> undef, float %q, i32 0 |
| %vecinit2.i = insertelement <4 x float> %vecinit.i, float %q, i32 1 |
| %vecinit4.i = insertelement <4 x float> %vecinit2.i, float %q, i32 2 |
| %vecinit6.i = insertelement <4 x float> %vecinit4.i, float %q, i32 3 |
| ret <4 x float> %vecinit6.i |
| } |
| |
| define <4 x float> @e2(ptr %ptr, ptr %ptr2) nounwind uwtable readnone ssp { |
| ; X86-LABEL: e2: |
| ; X86: ## %bb.0: ## %entry |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx |
| ; X86-NEXT: vbroadcastss (%ecx), %xmm0 |
| ; X86-NEXT: vmovss %xmm0, (%eax) |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: e2: |
| ; X64: ## %bb.0: ## %entry |
| ; X64-NEXT: vbroadcastss (%rdi), %xmm0 |
| ; X64-NEXT: vmovss %xmm0, (%rsi) |
| ; X64-NEXT: retq |
| entry: |
| %q = load float, ptr %ptr, align 4 |
| store float %q, ptr %ptr2, align 4 ; to create a chain to prevent broadcast |
| %vecinit.i = insertelement <4 x float> undef, float %q, i32 0 |
| %vecinit2.i = insertelement <4 x float> %vecinit.i, float %q, i32 1 |
| %vecinit4.i = insertelement <4 x float> %vecinit2.i, float %q, i32 2 |
| %vecinit6.i = insertelement <4 x float> %vecinit4.i, float %q, i32 3 |
| ret <4 x float> %vecinit6.i |
| } |
| |
| ; Don't broadcast constants on pre-AVX2 hardware. |
| define <4 x float> @_e2(ptr %ptr) nounwind uwtable readnone ssp { |
| ; X86-LABEL: _e2: |
| ; X86: ## %bb.0: ## %entry |
| ; X86-NEXT: vbroadcastss {{.*#+}} xmm0 = [-7.8125E-3,-7.8125E-3,-7.8125E-3,-7.8125E-3] |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: _e2: |
| ; X64: ## %bb.0: ## %entry |
| ; X64-NEXT: vbroadcastss {{.*#+}} xmm0 = [-7.8125E-3,-7.8125E-3,-7.8125E-3,-7.8125E-3] |
| ; X64-NEXT: retq |
| entry: |
| %vecinit.i = insertelement <4 x float> undef, float 0xbf80000000000000, i32 0 |
| %vecinit2.i = insertelement <4 x float> %vecinit.i, float 0xbf80000000000000, i32 1 |
| %vecinit4.i = insertelement <4 x float> %vecinit2.i, float 0xbf80000000000000, i32 2 |
| %vecinit6.i = insertelement <4 x float> %vecinit4.i, float 0xbf80000000000000, i32 3 |
| ret <4 x float> %vecinit6.i |
| } |
| |
| |
| define <4 x i32> @F(ptr %ptr) nounwind uwtable readnone ssp { |
| ; X86-LABEL: F: |
| ; X86: ## %bb.0: ## %entry |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: vbroadcastss (%eax), %xmm0 |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: F: |
| ; X64: ## %bb.0: ## %entry |
| ; X64-NEXT: vbroadcastss (%rdi), %xmm0 |
| ; X64-NEXT: retq |
| entry: |
| %q = load i32, ptr %ptr, align 4 |
| %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0 |
| %vecinit2.i = insertelement <4 x i32> %vecinit.i, i32 %q, i32 1 |
| %vecinit4.i = insertelement <4 x i32> %vecinit2.i, i32 %q, i32 2 |
| %vecinit6.i = insertelement <4 x i32> %vecinit4.i, i32 %q, i32 3 |
| ret <4 x i32> %vecinit6.i |
| } |
| |
| define <4 x i32> @F2(ptr %ptr, ptr %ptr2) nounwind uwtable readnone ssp { |
| ; X86-LABEL: F2: |
| ; X86: ## %bb.0: ## %entry |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx |
| ; X86-NEXT: movl (%ecx), %ecx |
| ; X86-NEXT: movl %ecx, (%eax) |
| ; X86-NEXT: vmovd %ecx, %xmm0 |
| ; X86-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,0,0,0] |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: F2: |
| ; X64: ## %bb.0: ## %entry |
| ; X64-NEXT: movl (%rdi), %eax |
| ; X64-NEXT: movl %eax, (%rsi) |
| ; X64-NEXT: vmovd %eax, %xmm0 |
| ; X64-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,0,0,0] |
| ; X64-NEXT: retq |
| entry: |
| %q = load i32, ptr %ptr, align 4 |
| store i32 %q, ptr %ptr2, align 4 ; to create a chain to prevent broadcast |
| %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0 |
| %vecinit2.i = insertelement <4 x i32> %vecinit.i, i32 %q, i32 1 |
| %vecinit4.i = insertelement <4 x i32> %vecinit2.i, i32 %q, i32 2 |
| %vecinit6.i = insertelement <4 x i32> %vecinit4.i, i32 %q, i32 3 |
| ret <4 x i32> %vecinit6.i |
| } |
| |
| ; FIXME: Pointer adjusted broadcasts |
| |
| define <4 x i32> @load_splat_4i32_4i32_1111(ptr %ptr) nounwind uwtable readnone ssp { |
| ; X86-LABEL: load_splat_4i32_4i32_1111: |
| ; X86: ## %bb.0: ## %entry |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: vpermilps {{.*#+}} xmm0 = mem[1,1,1,1] |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: load_splat_4i32_4i32_1111: |
| ; X64: ## %bb.0: ## %entry |
| ; X64-NEXT: vpermilps {{.*#+}} xmm0 = mem[1,1,1,1] |
| ; X64-NEXT: retq |
| entry: |
| %ld = load <4 x i32>, ptr %ptr |
| %ret = shufflevector <4 x i32> %ld, <4 x i32> undef, <4 x i32> <i32 1, i32 1, i32 1, i32 1> |
| ret <4 x i32> %ret |
| } |
| |
| define <8 x i32> @load_splat_8i32_4i32_33333333(ptr %ptr) nounwind uwtable readnone ssp { |
| ; X86-LABEL: load_splat_8i32_4i32_33333333: |
| ; X86: ## %bb.0: ## %entry |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: vbroadcastss 12(%eax), %ymm0 |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: load_splat_8i32_4i32_33333333: |
| ; X64: ## %bb.0: ## %entry |
| ; X64-NEXT: vbroadcastss 12(%rdi), %ymm0 |
| ; X64-NEXT: retq |
| entry: |
| %ld = load <4 x i32>, ptr %ptr |
| %ret = shufflevector <4 x i32> %ld, <4 x i32> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3> |
| ret <8 x i32> %ret |
| } |
| |
| define <8 x i32> @load_splat_8i32_8i32_55555555(ptr %ptr) nounwind uwtable readnone ssp { |
| ; X86-LABEL: load_splat_8i32_8i32_55555555: |
| ; X86: ## %bb.0: ## %entry |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: vbroadcastss 20(%eax), %ymm0 |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: load_splat_8i32_8i32_55555555: |
| ; X64: ## %bb.0: ## %entry |
| ; X64-NEXT: vbroadcastss 20(%rdi), %ymm0 |
| ; X64-NEXT: retq |
| entry: |
| %ld = load <8 x i32>, ptr %ptr |
| %ret = shufflevector <8 x i32> %ld, <8 x i32> undef, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5> |
| ret <8 x i32> %ret |
| } |
| |
| define <4 x float> @load_splat_4f32_4f32_1111(ptr %ptr) nounwind uwtable readnone ssp { |
| ; X86-LABEL: load_splat_4f32_4f32_1111: |
| ; X86: ## %bb.0: ## %entry |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: vbroadcastss 4(%eax), %xmm0 |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: load_splat_4f32_4f32_1111: |
| ; X64: ## %bb.0: ## %entry |
| ; X64-NEXT: vbroadcastss 4(%rdi), %xmm0 |
| ; X64-NEXT: retq |
| entry: |
| %ld = load <4 x float>, ptr %ptr |
| %ret = shufflevector <4 x float> %ld, <4 x float> undef, <4 x i32> <i32 1, i32 1, i32 1, i32 1> |
| ret <4 x float> %ret |
| } |
| |
| define <8 x float> @load_splat_8f32_4f32_33333333(ptr %ptr) nounwind uwtable readnone ssp { |
| ; X86-LABEL: load_splat_8f32_4f32_33333333: |
| ; X86: ## %bb.0: ## %entry |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: vbroadcastss 12(%eax), %ymm0 |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: load_splat_8f32_4f32_33333333: |
| ; X64: ## %bb.0: ## %entry |
| ; X64-NEXT: vbroadcastss 12(%rdi), %ymm0 |
| ; X64-NEXT: retq |
| entry: |
| %ld = load <4 x float>, ptr %ptr |
| %ret = shufflevector <4 x float> %ld, <4 x float> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3> |
| ret <8 x float> %ret |
| } |
| |
| define <8 x float> @load_splat_8f32_8f32_55555555(ptr %ptr) nounwind uwtable readnone ssp { |
| ; X86-LABEL: load_splat_8f32_8f32_55555555: |
| ; X86: ## %bb.0: ## %entry |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: vbroadcastss 20(%eax), %ymm0 |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: load_splat_8f32_8f32_55555555: |
| ; X64: ## %bb.0: ## %entry |
| ; X64-NEXT: vbroadcastss 20(%rdi), %ymm0 |
| ; X64-NEXT: retq |
| entry: |
| %ld = load <8 x float>, ptr %ptr |
| %ret = shufflevector <8 x float> %ld, <8 x float> undef, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5> |
| ret <8 x float> %ret |
| } |
| |
| define <2 x i64> @load_splat_2i64_2i64_1111(ptr %ptr) nounwind uwtable readnone ssp { |
| ; X86-LABEL: load_splat_2i64_2i64_1111: |
| ; X86: ## %bb.0: ## %entry |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: vpermilps {{.*#+}} xmm0 = mem[2,3,2,3] |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: load_splat_2i64_2i64_1111: |
| ; X64: ## %bb.0: ## %entry |
| ; X64-NEXT: vpermilps {{.*#+}} xmm0 = mem[2,3,2,3] |
| ; X64-NEXT: retq |
| entry: |
| %ld = load <2 x i64>, ptr %ptr |
| %ret = shufflevector <2 x i64> %ld, <2 x i64> undef, <2 x i32> <i32 1, i32 1> |
| ret <2 x i64> %ret |
| } |
| |
| define <4 x i64> @load_splat_4i64_2i64_1111(ptr %ptr) nounwind uwtable readnone ssp { |
| ; X86-LABEL: load_splat_4i64_2i64_1111: |
| ; X86: ## %bb.0: ## %entry |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: vbroadcastsd 8(%eax), %ymm0 |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: load_splat_4i64_2i64_1111: |
| ; X64: ## %bb.0: ## %entry |
| ; X64-NEXT: vbroadcastsd 8(%rdi), %ymm0 |
| ; X64-NEXT: retq |
| entry: |
| %ld = load <2 x i64>, ptr %ptr |
| %ret = shufflevector <2 x i64> %ld, <2 x i64> undef, <4 x i32> <i32 1, i32 1, i32 1, i32 1> |
| ret <4 x i64> %ret |
| } |
| |
| define <4 x i64> @load_splat_4i64_4i64_2222(ptr %ptr) nounwind uwtable readnone ssp { |
| ; X86-LABEL: load_splat_4i64_4i64_2222: |
| ; X86: ## %bb.0: ## %entry |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: vbroadcastsd 16(%eax), %ymm0 |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: load_splat_4i64_4i64_2222: |
| ; X64: ## %bb.0: ## %entry |
| ; X64-NEXT: vbroadcastsd 16(%rdi), %ymm0 |
| ; X64-NEXT: retq |
| entry: |
| %ld = load <4 x i64>, ptr %ptr |
| %ret = shufflevector <4 x i64> %ld, <4 x i64> undef, <4 x i32> <i32 2, i32 2, i32 2, i32 2> |
| ret <4 x i64> %ret |
| } |
| |
| define <2 x double> @load_splat_2f64_2f64_1111(ptr %ptr) nounwind uwtable readnone ssp { |
| ; X86-LABEL: load_splat_2f64_2f64_1111: |
| ; X86: ## %bb.0: ## %entry |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0] |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: load_splat_2f64_2f64_1111: |
| ; X64: ## %bb.0: ## %entry |
| ; X64-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0] |
| ; X64-NEXT: retq |
| entry: |
| %ld = load <2 x double>, ptr %ptr |
| %ret = shufflevector <2 x double> %ld, <2 x double> undef, <2 x i32> <i32 1, i32 1> |
| ret <2 x double> %ret |
| } |
| |
| define <4 x double> @load_splat_4f64_2f64_1111(ptr %ptr) nounwind uwtable readnone ssp { |
| ; X86-LABEL: load_splat_4f64_2f64_1111: |
| ; X86: ## %bb.0: ## %entry |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: vbroadcastsd 8(%eax), %ymm0 |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: load_splat_4f64_2f64_1111: |
| ; X64: ## %bb.0: ## %entry |
| ; X64-NEXT: vbroadcastsd 8(%rdi), %ymm0 |
| ; X64-NEXT: retq |
| entry: |
| %ld = load <2 x double>, ptr %ptr |
| %ret = shufflevector <2 x double> %ld, <2 x double> undef, <4 x i32> <i32 1, i32 1, i32 1, i32 1> |
| ret <4 x double> %ret |
| } |
| |
| define <4 x double> @load_splat_4f64_4f64_2222(ptr %ptr) nounwind uwtable readnone ssp { |
| ; X86-LABEL: load_splat_4f64_4f64_2222: |
| ; X86: ## %bb.0: ## %entry |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: vbroadcastsd 16(%eax), %ymm0 |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: load_splat_4f64_4f64_2222: |
| ; X64: ## %bb.0: ## %entry |
| ; X64-NEXT: vbroadcastsd 16(%rdi), %ymm0 |
| ; X64-NEXT: retq |
| entry: |
| %ld = load <4 x double>, ptr %ptr |
| %ret = shufflevector <4 x double> %ld, <4 x double> undef, <4 x i32> <i32 2, i32 2, i32 2, i32 2> |
| ret <4 x double> %ret |
| } |
| |
| ; Unsupported vbroadcasts |
| |
| define <2 x i64> @G(ptr %ptr) nounwind uwtable readnone ssp { |
| ; X86-LABEL: G: |
| ; X86: ## %bb.0: ## %entry |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0] |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: G: |
| ; X64: ## %bb.0: ## %entry |
| ; X64-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0] |
| ; X64-NEXT: retq |
| entry: |
| %q = load i64, ptr %ptr, align 8 |
| %vecinit.i = insertelement <2 x i64> undef, i64 %q, i32 0 |
| %vecinit2.i = insertelement <2 x i64> %vecinit.i, i64 %q, i32 1 |
| ret <2 x i64> %vecinit2.i |
| } |
| |
| define <2 x i64> @G2(ptr %ptr, ptr %ptr2) nounwind uwtable readnone ssp { |
| ; X86-LABEL: G2: |
| ; X86: ## %bb.0: ## %entry |
| ; X86-NEXT: pushl %esi |
| ; X86-NEXT: .cfi_def_cfa_offset 8 |
| ; X86-NEXT: .cfi_offset %esi, -8 |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx |
| ; X86-NEXT: movl (%ecx), %edx |
| ; X86-NEXT: movl 4(%ecx), %esi |
| ; X86-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0] |
| ; X86-NEXT: movl %edx, (%eax) |
| ; X86-NEXT: movl %esi, 4(%eax) |
| ; X86-NEXT: popl %esi |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: G2: |
| ; X64: ## %bb.0: ## %entry |
| ; X64-NEXT: movq (%rdi), %rax |
| ; X64-NEXT: movq %rax, (%rsi) |
| ; X64-NEXT: vmovq %rax, %xmm0 |
| ; X64-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,1,0,1] |
| ; X64-NEXT: retq |
| entry: |
| %q = load i64, ptr %ptr, align 8 |
| store i64 %q, ptr %ptr2, align 8 ; to create a chain to prevent broadcast |
| %vecinit.i = insertelement <2 x i64> undef, i64 %q, i32 0 |
| %vecinit2.i = insertelement <2 x i64> %vecinit.i, i64 %q, i32 1 |
| ret <2 x i64> %vecinit2.i |
| } |
| |
| define <4 x i32> @H(<4 x i32> %a) { |
| ; X86-LABEL: H: |
| ; X86: ## %bb.0: ## %entry |
| ; X86-NEXT: vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1] |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: H: |
| ; X64: ## %bb.0: ## %entry |
| ; X64-NEXT: vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1] |
| ; X64-NEXT: retq |
| entry: |
| %x = shufflevector <4 x i32> %a, <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef> |
| ret <4 x i32> %x |
| } |
| |
| define <2 x double> @I(ptr %ptr) nounwind uwtable readnone ssp { |
| ; X86-LABEL: I: |
| ; X86: ## %bb.0: ## %entry |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0] |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: I: |
| ; X64: ## %bb.0: ## %entry |
| ; X64-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0] |
| ; X64-NEXT: retq |
| entry: |
| %q = load double, ptr %ptr, align 4 |
| %vecinit.i = insertelement <2 x double> undef, double %q, i32 0 |
| %vecinit2.i = insertelement <2 x double> %vecinit.i, double %q, i32 1 |
| ret <2 x double> %vecinit2.i |
| } |
| |
| define <2 x double> @I2(ptr %ptr, ptr %ptr2) nounwind uwtable readnone ssp { |
| ; X86-LABEL: I2: |
| ; X86: ## %bb.0: ## %entry |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx |
| ; X86-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0] |
| ; X86-NEXT: vmovlps %xmm0, (%eax) |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: I2: |
| ; X64: ## %bb.0: ## %entry |
| ; X64-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0] |
| ; X64-NEXT: vmovlps %xmm0, (%rsi) |
| ; X64-NEXT: retq |
| entry: |
| %q = load double, ptr %ptr, align 4 |
| store double %q, ptr %ptr2, align 4 ; to create a chain to prevent broadcast |
| %vecinit.i = insertelement <2 x double> undef, double %q, i32 0 |
| %vecinit2.i = insertelement <2 x double> %vecinit.i, double %q, i32 1 |
| ret <2 x double> %vecinit2.i |
| } |
| |
| define <4 x float> @_RR(ptr %ptr, ptr %k) nounwind uwtable readnone ssp { |
| ; X86-LABEL: _RR: |
| ; X86: ## %bb.0: ## %entry |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx |
| ; X86-NEXT: vbroadcastss (%ecx), %xmm0 |
| ; X86-NEXT: movl (%eax), %eax |
| ; X86-NEXT: movl %eax, (%eax) |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: _RR: |
| ; X64: ## %bb.0: ## %entry |
| ; X64-NEXT: vbroadcastss (%rdi), %xmm0 |
| ; X64-NEXT: movl (%rsi), %eax |
| ; X64-NEXT: movl %eax, (%rax) |
| ; X64-NEXT: retq |
| entry: |
| %q = load float, ptr %ptr, align 4 |
| %vecinit.i = insertelement <4 x float> undef, float %q, i32 0 |
| %vecinit2.i = insertelement <4 x float> %vecinit.i, float %q, i32 1 |
| %vecinit4.i = insertelement <4 x float> %vecinit2.i, float %q, i32 2 |
| %vecinit6.i = insertelement <4 x float> %vecinit4.i, float %q, i32 3 |
| ; force a chain |
| %j = load i32, ptr %k, align 4 |
| store i32 %j, ptr undef |
| ret <4 x float> %vecinit6.i |
| } |
| |
| define <4 x float> @_RR2(ptr %ptr, ptr %k) nounwind uwtable readnone ssp { |
| ; X86-LABEL: _RR2: |
| ; X86: ## %bb.0: ## %entry |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: vbroadcastss (%eax), %xmm0 |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: _RR2: |
| ; X64: ## %bb.0: ## %entry |
| ; X64-NEXT: vbroadcastss (%rdi), %xmm0 |
| ; X64-NEXT: retq |
| entry: |
| %q = load float, ptr %ptr, align 4 |
| %v = insertelement <4 x float> undef, float %q, i32 0 |
| %t = shufflevector <4 x float> %v, <4 x float> undef, <4 x i32> zeroinitializer |
| ret <4 x float> %t |
| } |
| |
| ; These tests check that a vbroadcast instruction is used when we have a splat |
| ; formed from a concat_vectors (via the shufflevector) of two BUILD_VECTORs |
| ; (via the insertelements). |
| |
| define <8 x float> @splat_concat1(ptr %p) { |
| ; X86-LABEL: splat_concat1: |
| ; X86: ## %bb.0: |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: vbroadcastss (%eax), %ymm0 |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: splat_concat1: |
| ; X64: ## %bb.0: |
| ; X64-NEXT: vbroadcastss (%rdi), %ymm0 |
| ; X64-NEXT: retq |
| %1 = load float, ptr %p, align 4 |
| %2 = insertelement <4 x float> undef, float %1, i32 0 |
| %3 = insertelement <4 x float> %2, float %1, i32 1 |
| %4 = insertelement <4 x float> %3, float %1, i32 2 |
| %5 = insertelement <4 x float> %4, float %1, i32 3 |
| %6 = shufflevector <4 x float> %5, <4 x float> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3> |
| ret <8 x float> %6 |
| } |
| |
| define <8 x float> @splat_concat2(ptr %p) { |
| ; X86-LABEL: splat_concat2: |
| ; X86: ## %bb.0: |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: vbroadcastss (%eax), %ymm0 |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: splat_concat2: |
| ; X64: ## %bb.0: |
| ; X64-NEXT: vbroadcastss (%rdi), %ymm0 |
| ; X64-NEXT: retq |
| %1 = load float, ptr %p, align 4 |
| %2 = insertelement <4 x float> undef, float %1, i32 0 |
| %3 = insertelement <4 x float> %2, float %1, i32 1 |
| %4 = insertelement <4 x float> %3, float %1, i32 2 |
| %5 = insertelement <4 x float> %4, float %1, i32 3 |
| %6 = insertelement <4 x float> undef, float %1, i32 0 |
| %7 = insertelement <4 x float> %6, float %1, i32 1 |
| %8 = insertelement <4 x float> %7, float %1, i32 2 |
| %9 = insertelement <4 x float> %8, float %1, i32 3 |
| %10 = shufflevector <4 x float> %5, <4 x float> %9, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7> |
| ret <8 x float> %10 |
| } |
| |
| define <4 x double> @splat_concat3(ptr %p) { |
| ; X86-LABEL: splat_concat3: |
| ; X86: ## %bb.0: |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: vbroadcastsd (%eax), %ymm0 |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: splat_concat3: |
| ; X64: ## %bb.0: |
| ; X64-NEXT: vbroadcastsd (%rdi), %ymm0 |
| ; X64-NEXT: retq |
| %1 = load double, ptr %p, align 8 |
| %2 = insertelement <2 x double> undef, double %1, i32 0 |
| %3 = insertelement <2 x double> %2, double %1, i32 1 |
| %4 = shufflevector <2 x double> %3, <2 x double> undef, <4 x i32> <i32 0, i32 1, i32 0, i32 1> |
| ret <4 x double> %4 |
| } |
| |
| define <4 x double> @splat_concat4(ptr %p) { |
| ; X86-LABEL: splat_concat4: |
| ; X86: ## %bb.0: |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: vbroadcastsd (%eax), %ymm0 |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: splat_concat4: |
| ; X64: ## %bb.0: |
| ; X64-NEXT: vbroadcastsd (%rdi), %ymm0 |
| ; X64-NEXT: retq |
| %1 = load double, ptr %p, align 8 |
| %2 = insertelement <2 x double> undef, double %1, i32 0 |
| %3 = insertelement <2 x double> %2, double %1, i32 1 |
| %4 = insertelement <2 x double> undef, double %1, i32 0 |
| %5 = insertelement <2 x double> %2, double %1, i32 1 |
| %6 = shufflevector <2 x double> %3, <2 x double> %5, <4 x i32> <i32 0, i32 1, i32 2, i32 3> |
| ret <4 x double> %6 |
| } |
| |
| ; PR34041 |
| define <4 x double> @broadcast_shuffle_1000(ptr %p) { |
| ; X86-LABEL: broadcast_shuffle_1000: |
| ; X86: ## %bb.0: |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: vbroadcastsd (%eax), %ymm0 |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: broadcast_shuffle_1000: |
| ; X64: ## %bb.0: |
| ; X64-NEXT: vbroadcastsd (%rdi), %ymm0 |
| ; X64-NEXT: retq |
| %1 = load double, ptr %p |
| %2 = insertelement <2 x double> undef, double %1, i32 0 |
| %3 = shufflevector <2 x double> %2, <2 x double> undef, <4 x i32> <i32 1, i32 0, i32 0, i32 0> |
| ret <4 x double> %3 |
| } |
| |
| define <4 x double> @broadcast_shuffle1032(ptr %p) { |
| ; X86-LABEL: broadcast_shuffle1032: |
| ; X86: ## %bb.0: |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: vbroadcastsd (%eax), %ymm0 |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: broadcast_shuffle1032: |
| ; X64: ## %bb.0: |
| ; X64-NEXT: vbroadcastsd (%rdi), %ymm0 |
| ; X64-NEXT: retq |
| %1 = load double, ptr %p |
| %2 = insertelement <2 x double> undef, double %1, i32 1 |
| %3 = insertelement <2 x double> undef, double %1, i32 0 |
| %4 = shufflevector <2 x double> %2, <2 x double> %3, <4 x i32> <i32 1, i32 0, i32 3, i32 2> |
| ret <4 x double> %4 |
| } |
| |
| define void @broadcast_v16i32(ptr %a, ptr %b) { |
| ; X86-LABEL: broadcast_v16i32: |
| ; X86: ## %bb.0: |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx |
| ; X86-NEXT: vbroadcastss (%ecx), %ymm0 |
| ; X86-NEXT: vmovups %ymm0, 32(%eax) |
| ; X86-NEXT: vmovups %ymm0, (%eax) |
| ; X86-NEXT: vzeroupper |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: broadcast_v16i32: |
| ; X64: ## %bb.0: |
| ; X64-NEXT: vbroadcastss (%rdi), %ymm0 |
| ; X64-NEXT: vmovups %ymm0, 32(%rsi) |
| ; X64-NEXT: vmovups %ymm0, (%rsi) |
| ; X64-NEXT: vzeroupper |
| ; X64-NEXT: retq |
| %1 = load i32, ptr %a, align 4 |
| %2 = insertelement <8 x i32> undef, i32 %1, i32 0 |
| %3 = shufflevector <8 x i32> %2, <8 x i32> undef, <8 x i32> zeroinitializer |
| %4 = shufflevector <8 x i32> undef, <8 x i32> %3, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15> |
| store <16 x i32> %4, ptr %b, align 4 |
| ret void |
| } |
| |
| ; |
| ; Broadcast scale factor for xyz vector - slp will have vectorized xy. |
| ; |
| define double @broadcast_scale_xyz(ptr nocapture readonly, ptr nocapture readonly) nounwind { |
| ; X86-LABEL: broadcast_scale_xyz: |
| ; X86: ## %bb.0: |
| ; X86-NEXT: subl $12, %esp |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax |
| ; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx |
| ; X86-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0] |
| ; X86-NEXT: vmulpd (%eax), %xmm0, %xmm1 |
| ; X86-NEXT: vmulsd 16(%eax), %xmm0, %xmm0 |
| ; X86-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0] |
| ; X86-NEXT: vaddsd %xmm2, %xmm1, %xmm1 |
| ; X86-NEXT: vaddsd %xmm1, %xmm0, %xmm0 |
| ; X86-NEXT: vmovsd %xmm0, (%esp) |
| ; X86-NEXT: fldl (%esp) |
| ; X86-NEXT: addl $12, %esp |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: broadcast_scale_xyz: |
| ; X64: ## %bb.0: |
| ; X64-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0] |
| ; X64-NEXT: vmulpd (%rsi), %xmm0, %xmm1 |
| ; X64-NEXT: vmulsd 16(%rsi), %xmm0, %xmm0 |
| ; X64-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0] |
| ; X64-NEXT: vaddsd %xmm2, %xmm1, %xmm1 |
| ; X64-NEXT: vaddsd %xmm1, %xmm0, %xmm0 |
| ; X64-NEXT: retq |
| %3 = load <2 x double>, ptr %1, align 8 |
| %4 = getelementptr inbounds double, ptr %1, i64 2 |
| %5 = load double, ptr %4, align 8 |
| %6 = load double, ptr %0, align 8 |
| %7 = insertelement <2 x double> undef, double %6, i32 0 |
| %8 = shufflevector <2 x double> %7, <2 x double> undef, <2 x i32> zeroinitializer |
| %9 = fmul <2 x double> %3, %8 |
| %10 = fmul double %5, %6 |
| %11 = extractelement <2 x double> %9, i32 0 |
| %12 = extractelement <2 x double> %9, i32 1 |
| %13 = fadd double %11, %12 |
| %14 = fadd double %10, %13 |
| ret double %14 |
| } |
| |
| ; |
| ; Broadcast v2f32 non-uniform constant via vmovddup |
| ; |
| define void @fmul_by_v2f32_broadcast() nounwind { |
| ; X86-LABEL: fmul_by_v2f32_broadcast: |
| ; X86: ## %bb.0: |
| ; X86-NEXT: vmovddup {{.*#+}} xmm0 = [3.1E+1,0.0E+0,3.1E+1,0.0E+0] |
| ; X86-NEXT: ## xmm0 = mem[0,0] |
| ; X86-NEXT: ## implicit-def: $xmm1 |
| ; X86-NEXT: .p2align 4, 0x90 |
| ; X86-NEXT: LBB42_1: ## =>This Inner Loop Header: Depth=1 |
| ; X86-NEXT: vmovsd {{.*#+}} xmm2 = mem[0],zero |
| ; X86-NEXT: vmulps %xmm0, %xmm2, %xmm2 |
| ; X86-NEXT: vmovlps %xmm2, (%eax) |
| ; X86-NEXT: vmulps %xmm0, %xmm1, %xmm1 |
| ; X86-NEXT: vmovlps %xmm1, (%eax) |
| ; X86-NEXT: jmp LBB42_1 |
| ; |
| ; X64-LABEL: fmul_by_v2f32_broadcast: |
| ; X64: ## %bb.0: |
| ; X64-NEXT: vmovddup {{.*#+}} xmm0 = [3.1E+1,0.0E+0,3.1E+1,0.0E+0] |
| ; X64-NEXT: ## xmm0 = mem[0,0] |
| ; X64-NEXT: ## implicit-def: $xmm1 |
| ; X64-NEXT: .p2align 4, 0x90 |
| ; X64-NEXT: LBB42_1: ## =>This Inner Loop Header: Depth=1 |
| ; X64-NEXT: vmovsd {{.*#+}} xmm2 = mem[0],zero |
| ; X64-NEXT: vmulps %xmm0, %xmm2, %xmm2 |
| ; X64-NEXT: vmovlps %xmm2, (%rax) |
| ; X64-NEXT: vmulps %xmm0, %xmm1, %xmm1 |
| ; X64-NEXT: vmovlps %xmm1, (%rax) |
| ; X64-NEXT: jmp LBB42_1 |
| br label %1 |
| 1: |
| %2 = phi <2 x float> [ undef, %0 ], [ %5, %1 ] |
| %3 = load <2 x float>, ptr poison, align 8 |
| %4 = fmul <2 x float> %3, <float 3.100000e+01, float 0.000000e+00> |
| store <2 x float> %4, ptr poison, align 8 |
| %5 = fmul <2 x float> %2, <float 3.100000e+01, float 0.000000e+00> |
| store <2 x float> %5, ptr poison, align 8 |
| br label %1 |
| } |
| |
| ; |
| ; When VBROADCAST replaces an existing load, ensure it still respects lifetime dependencies. |
| ; |
| define float @broadcast_lifetime() nounwind { |
| ; X86-LABEL: broadcast_lifetime: |
| ; X86: ## %bb.0: |
| ; X86-NEXT: pushl %esi |
| ; X86-NEXT: subl $40, %esp |
| ; X86-NEXT: leal {{[0-9]+}}(%esp), %esi |
| ; X86-NEXT: movl %esi, (%esp) |
| ; X86-NEXT: calll _gfunc |
| ; X86-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero |
| ; X86-NEXT: vmovss %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) ## 4-byte Spill |
| ; X86-NEXT: movl %esi, (%esp) |
| ; X86-NEXT: calll _gfunc |
| ; X86-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero |
| ; X86-NEXT: vsubss {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 ## 4-byte Folded Reload |
| ; X86-NEXT: vmovss %xmm0, {{[0-9]+}}(%esp) |
| ; X86-NEXT: flds {{[0-9]+}}(%esp) |
| ; X86-NEXT: addl $40, %esp |
| ; X86-NEXT: popl %esi |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: broadcast_lifetime: |
| ; X64: ## %bb.0: |
| ; X64-NEXT: subq $40, %rsp |
| ; X64-NEXT: leaq {{[0-9]+}}(%rsp), %rdi |
| ; X64-NEXT: callq _gfunc |
| ; X64-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero |
| ; X64-NEXT: vmovss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) ## 4-byte Spill |
| ; X64-NEXT: leaq {{[0-9]+}}(%rsp), %rdi |
| ; X64-NEXT: callq _gfunc |
| ; X64-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero |
| ; X64-NEXT: vsubss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 ## 4-byte Folded Reload |
| ; X64-NEXT: addq $40, %rsp |
| ; X64-NEXT: retq |
| %1 = alloca <4 x float>, align 16 |
| %2 = alloca <4 x float>, align 16 |
| |
| call void @llvm.lifetime.start.p0(i64 16, ptr %1) |
| call void @gfunc(ptr %1) |
| %3 = load <4 x float>, ptr %1, align 16 |
| call void @llvm.lifetime.end.p0(i64 16, ptr %1) |
| |
| call void @llvm.lifetime.start.p0(i64 16, ptr %2) |
| call void @gfunc(ptr %2) |
| %4 = load <4 x float>, ptr %2, align 16 |
| call void @llvm.lifetime.end.p0(i64 16, ptr %2) |
| |
| %5 = extractelement <4 x float> %3, i32 1 |
| %6 = extractelement <4 x float> %4, i32 1 |
| %7 = fsub float %6, %5 |
| ret float %7 |
| } |
| |
| define <8 x i16> @broadcast_x86_mmx(x86_mmx %tmp) nounwind { |
| ; X86-LABEL: broadcast_x86_mmx: |
| ; X86: ## %bb.0: ## %bb |
| ; X86-NEXT: subl $12, %esp |
| ; X86-NEXT: movq %mm0, (%esp) |
| ; X86-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0] |
| ; X86-NEXT: addl $12, %esp |
| ; X86-NEXT: retl |
| ; |
| ; X64-LABEL: broadcast_x86_mmx: |
| ; X64: ## %bb.0: ## %bb |
| ; X64-NEXT: movdq2q %xmm0, %mm0 |
| ; X64-NEXT: movq2dq %mm0, %xmm0 |
| ; X64-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,1,0,1] |
| ; X64-NEXT: retq |
| bb: |
| %tmp1 = bitcast x86_mmx %tmp to i64 |
| %tmp2 = insertelement <2 x i64> undef, i64 %tmp1, i32 0 |
| %tmp3 = bitcast <2 x i64> %tmp2 to <8 x i16> |
| %tmp4 = shufflevector <8 x i16> %tmp3, <8 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3> |
| ret <8 x i16> %tmp4 |
| } |
| |
| declare void @gfunc(ptr) |
| declare void @llvm.lifetime.start.p0(i64, ptr) |
| declare void @llvm.lifetime.end.p0(i64, ptr) |