+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=corei7-avx -mattr=+avx | FileCheck %s
-; CHECK: vbroadcastsd (%
define <4 x i64> @A(i64* %ptr) nounwind uwtable readnone ssp {
+; CHECK-LABEL: A:
+; CHECK: ## BB#0: ## %entry
+; CHECK-NEXT: vbroadcastsd (%rdi), %ymm0
+; CHECK-NEXT: retq
entry:
- %q = load i64* %ptr, align 8
+ %q = load i64, i64* %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
ret <4 x i64> %vecinit6.i
}
-; CHECK: vbroadcastss (%
define <8 x i32> @B(i32* %ptr) nounwind uwtable readnone ssp {
+; CHECK-LABEL: B:
+; CHECK: ## BB#0: ## %entry
+; CHECK-NEXT: vbroadcastss (%rdi), %ymm0
+; CHECK-NEXT: retq
entry:
- %q = load i32* %ptr, align 4
+ %q = load i32, i32* %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
ret <8 x i32> %vecinit6.i
}
-; CHECK: vbroadcastsd (%
define <4 x double> @C(double* %ptr) nounwind uwtable readnone ssp {
+; CHECK-LABEL: C:
+; CHECK: ## BB#0: ## %entry
+; CHECK-NEXT: vbroadcastsd (%rdi), %ymm0
+; CHECK-NEXT: retq
entry:
- %q = load double* %ptr, align 8
+ %q = load double, double* %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
ret <4 x double> %vecinit6.i
}
-; CHECK: vbroadcastss (%
define <8 x float> @D(float* %ptr) nounwind uwtable readnone ssp {
+; CHECK-LABEL: D:
+; CHECK: ## BB#0: ## %entry
+; CHECK-NEXT: vbroadcastss (%rdi), %ymm0
+; CHECK-NEXT: retq
entry:
- %q = load float* %ptr, align 4
+ %q = load float, float* %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
;;;; 128-bit versions
-; CHECK: vbroadcastss (%
define <4 x float> @e(float* %ptr) nounwind uwtable readnone ssp {
+; CHECK-LABEL: e:
+; CHECK: ## BB#0: ## %entry
+; CHECK-NEXT: vbroadcastss (%rdi), %xmm0
+; CHECK-NEXT: retq
entry:
- %q = load float* %ptr, align 4
+ %q = load float, float* %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
ret <4 x float> %vecinit6.i
}
-
-; CHECK: _e2
-; CHECK-NOT: vbroadcastss
-; CHECK: ret
+; Don't broadcast constants on pre-AVX2 hardware.
define <4 x float> @_e2(float* %ptr) nounwind uwtable readnone ssp {
- %vecinit.i = insertelement <4 x float> undef, float 0xbf80000000000000, i32 0
+; CHECK-LABEL: _e2:
+; CHECK: ## BB#0: ## %entry
+; CHECK-NEXT: vmovaps {{.*#+}} xmm0 = [-7.812500e-03,-7.812500e-03,-7.812500e-03,-7.812500e-03]
+; CHECK-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
}
-; CHECK: vbroadcastss (%
define <4 x i32> @F(i32* %ptr) nounwind uwtable readnone ssp {
+; CHECK-LABEL: F:
+; CHECK: ## BB#0: ## %entry
+; CHECK-NEXT: vbroadcastss (%rdi), %xmm0
+; CHECK-NEXT: retq
entry:
- %q = load i32* %ptr, align 4
+ %q = load i32, i32* %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
; Unsupported vbroadcasts
-; CHECK: _G
-; CHECK-NOT: broadcast (%
-; CHECK: ret
define <2 x i64> @G(i64* %ptr) nounwind uwtable readnone ssp {
+; CHECK-LABEL: G:
+; CHECK: ## BB#0: ## %entry
+; CHECK-NEXT: vmovq {{.*#+}} xmm0 = mem[0],zero
+; CHECK-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
+; CHECK-NEXT: retq
entry:
- %q = load i64* %ptr, align 8
+ %q = load i64, i64* %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
}
-; CHECK: _H
-; CHECK-NOT: broadcast
-; CHECK: ret
define <4 x i32> @H(<4 x i32> %a) {
+; CHECK-LABEL: H:
+; CHECK: ## BB#0: ## %entry
+; CHECK-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[1,1,2,3]
+; CHECK-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
}
-; CHECK: _I
-; CHECK-NOT: broadcast (%
-; CHECK: ret
define <2 x double> @I(double* %ptr) nounwind uwtable readnone ssp {
+; CHECK-LABEL: I:
+; CHECK: ## BB#0: ## %entry
+; CHECK-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0]
+; CHECK-NEXT: retq
entry:
- %q = load double* %ptr, align 4
+ %q = load double, double* %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
}
-; CHECK: _RR
-; CHECK: vbroadcastss (%
-; CHECK: ret
define <4 x float> @_RR(float* %ptr, i32* %k) nounwind uwtable readnone ssp {
+; CHECK-LABEL: _RR:
+; CHECK: ## BB#0: ## %entry
+; CHECK-NEXT: vbroadcastss (%rdi), %xmm0
+; CHECK-NEXT: movl (%rsi), %eax
+; CHECK-NEXT: movl %eax, (%rax)
+; CHECK-NEXT: retq
entry:
- %q = load float* %ptr, align 4
+ %q = load float, float* %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* %k, align 4
+ %j = load i32, i32* %k, align 4
store i32 %j, i32* undef
ret <4 x float> %vecinit6.i
}
+define <4 x float> @_RR2(float* %ptr, i32* %k) nounwind uwtable readnone ssp {
+; CHECK-LABEL: _RR2:
+; CHECK: ## BB#0: ## %entry
+; CHECK-NEXT: vbroadcastss (%rdi), %xmm0
+; CHECK-NEXT: retq
+entry:
+ %q = load float, float* %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(float* %p) {
+; CHECK-LABEL: splat_concat1:
+; CHECK: ## BB#0:
+; CHECK-NEXT: vbroadcastss (%rdi), %ymm0
+; CHECK-NEXT: retq
+ %1 = load float, float* %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(float* %p) {
+; CHECK-LABEL: splat_concat2:
+; CHECK: ## BB#0:
+; CHECK-NEXT: vbroadcastss (%rdi), %ymm0
+; CHECK-NEXT: retq
+ %1 = load float, float* %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(double* %p) {
+; CHECK-LABEL: splat_concat3:
+; CHECK: ## BB#0:
+; CHECK-NEXT: vbroadcastsd (%rdi), %ymm0
+; CHECK-NEXT: retq
+ %1 = load double, double* %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(double* %p) {
+; CHECK-LABEL: splat_concat4:
+; CHECK: ## BB#0:
+; CHECK-NEXT: vbroadcastsd (%rdi), %ymm0
+; CHECK-NEXT: retq
+ %1 = load double, double* %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
+}
projects / oota-llvm.git / blobdiff