This is an archive of the discontinued LLVM Phabricator instance.

Differential D53932

[NFCI][FPEnv] Split constrained intrinsic tests
ClosedPublic

Authored by cameron.mcinally on Oct 31 2018, 7:17 AM.

Details

Reviewers
craig.topper
andrew.w.kaylor
kpn
uweigand
Commits
rG51a91e86e1d6: [NFCI][FPEnv] Split constrained intrinsic tests
rL346137: [NFCI][FPEnv] Split constrained intrinsic tests
Summary

Break out the constrained FMA intrinsic tests into their own file.

Diff Detail

Event Timeline

cameron.mcinally created this revision.Oct 31 2018, 7:17 AM
Herald added a subscriber: llvm-commits. · View Herald TranscriptOct 31 2018, 7:17 AM
kpn added a comment.Oct 31 2018, 7:24 AM

FWIW, I like it.

andrew.w.kaylor accepted this revision.Nov 2 2018, 3:48 PM

lgtm

This revision is now accepted and ready to land.Nov 2 2018, 3:48 PM
craig.topper added inline comments.Nov 2 2018, 3:52 PM
test/CodeGen/X86/vector-constrained-fp-intrinsics-fma.ll
86

Delete the COMMON lines. They look like a leftover.

cameron.mcinally marked an inline comment as done.Nov 5 2018, 7:30 AM
Closed by commit rL346137: [NFCI][FPEnv] Split constrained intrinsic tests (authored by mcinally). · Explain WhyNov 5 2018, 7:30 AM
This revision was automatically updated to reflect the committed changes.

Revision Contents

Diff 171910

test/CodeGen/X86/vector-constrained-fp-intrinsics-fma.ll

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -O3 -mtriple=x86_64-pc-linux -mattr=+fma < %s | FileCheck %s
define <1 x float> @constrained_vector_fma_v1f32() {
; CHECK-LABEL: constrained_vector_fma_v1f32:
; CHECK: # %bb.0: # %entry
; CHECK-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; CHECK-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; CHECK-NEXT: vfmadd213ss {{.*#+}} xmm0 = (xmm1 * xmm0) + mem
; CHECK-NEXT: retq
entry:
%fma = call <1 x float> @llvm.experimental.constrained.fma.v1f32(
<1 x float> <float 0.5>,
<1 x float> <float 2.5>,
<1 x float> <float 4.5>,
metadata !"round.dynamic",
metadata !"fpexcept.strict")
ret <1 x float> %fma
}
define <2 x double> @constrained_vector_fma_v2f64() {
; CHECK-LABEL: constrained_vector_fma_v2f64:
; CHECK: # %bb.0: # %entry
; CHECK-NEXT: vmovapd {{.*#+}} xmm1 = [1.5E+0,5.0E-1]
; CHECK-NEXT: vmovapd {{.*#+}} xmm0 = [3.5E+0,2.5E+0]
; CHECK-NEXT: vfmadd213pd {{.*#+}} xmm0 = (xmm1 * xmm0) + mem
; CHECK-NEXT: retq
entry:
%fma = call <2 x double> @llvm.experimental.constrained.fma.v2f64(
<2 x double> <double 1.5, double 0.5>,
<2 x double> <double 3.5, double 2.5>,
<2 x double> <double 5.5, double 4.5>,
metadata !"round.dynamic",
metadata !"fpexcept.strict")
ret <2 x double> %fma
}
define <3 x float> @constrained_vector_fma_v3f32() {
; CHECK-LABEL: constrained_vector_fma_v3f32:
; CHECK: # %bb.0: # %entry
; CHECK-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; CHECK-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; CHECK-NEXT: vfmadd213ss {{.*#+}} xmm1 = (xmm0 * xmm1) + mem
; CHECK-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; CHECK-NEXT: vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; CHECK-NEXT: vfmadd213ss {{.*#+}} xmm2 = (xmm0 * xmm2) + mem
; CHECK-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; CHECK-NEXT: vmovss {{.*#+}} xmm3 = mem[0],zero,zero,zero
; CHECK-NEXT: vfmadd213ss {{.*#+}} xmm3 = (xmm0 * xmm3) + mem
; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm2[0],xmm3[0],xmm2[2,3]
; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0,1],xmm1[0],xmm0[3]
; CHECK-NEXT: retq
entry:
%fma = call <3 x float> @llvm.experimental.constrained.fma.v3f32(
<3 x float> <float 2.5, float 1.5, float 0.5>,
<3 x float> <float 5.5, float 4.5, float 3.5>,
<3 x float> <float 8.5, float 7.5, float 6.5>,
metadata !"round.dynamic",
metadata !"fpexcept.strict")
ret <3 x float> %fma
}
define <3 x double> @constrained_vector_fma_v3f64() {
; CHECK-LABEL: constrained_vector_fma_v3f64:
; CHECK: # %bb.0: # %entry
; CHECK-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; CHECK-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; CHECK-NEXT: vfmadd213sd {{.*#+}} xmm1 = (xmm0 * xmm1) + mem
; CHECK-NEXT: vmovapd {{.*#+}} xmm0 = [2.5E+0,1.5E+0]
; CHECK-NEXT: vmovapd {{.*#+}} xmm2 = [5.5E+0,4.5E+0]
; CHECK-NEXT: vfmadd213pd {{.*#+}} xmm2 = (xmm0 * xmm2) + mem
; CHECK-NEXT: vinsertf128 $1, %xmm1, %ymm2, %ymm0
; CHECK-NEXT: retq
entry:
%fma = call <3 x double> @llvm.experimental.constrained.fma.v3f64(
<3 x double> <double 2.5, double 1.5, double 0.5>,
<3 x double> <double 5.5, double 4.5, double 3.5>,
<3 x double> <double 8.5, double 7.5, double 6.5>,
metadata !"round.dynamic",
metadata !"fpexcept.strict")
ret <3 x double> %fma
}
define <4 x double> @constrained_vector_fma_v4f64() {
; COMMON-LABEL: constrained_vector_fma_v4f64:
; COMMON: # %bb.0: # %entry
craig.topperUnsubmitted
Done
ReplyInline Actions

Delete the COMMON lines. They look like a leftover.

craig.topper: Delete the COMMON lines. They look like a leftover.
; COMMON-NEXT: vmovapd {{.*#+}} ymm1 = [3.5E+0,2.5E+0,1.5E+0,5.0E-1]
; COMMON-NEXT: vmovapd {{.*#+}} ymm0 = [7.5E+0,6.5E+0,5.5E+0,4.5E+0]
; COMMON-NEXT: vfmadd213pd {{.*#+}} ymm0 = (ymm1 * ymm0) + mem
; COMMON-NEXT: retq
; CHECK-LABEL: constrained_vector_fma_v4f64:
; CHECK: # %bb.0: # %entry
; CHECK-NEXT: vmovapd {{.*#+}} ymm1 = [3.5E+0,2.5E+0,1.5E+0,5.0E-1]
; CHECK-NEXT: vmovapd {{.*#+}} ymm0 = [7.5E+0,6.5E+0,5.5E+0,4.5E+0]
; CHECK-NEXT: vfmadd213pd {{.*#+}} ymm0 = (ymm1 * ymm0) + mem
; CHECK-NEXT: retq
entry:
%fma = call <4 x double> @llvm.experimental.constrained.fma.v4f64(
<4 x double> <double 3.5, double 2.5, double 1.5, double 0.5>,
<4 x double> <double 7.5, double 6.5, double 5.5, double 4.5>,
<4 x double> <double 11.5, double 10.5, double 9.5, double 8.5>,
metadata !"round.dynamic",
metadata !"fpexcept.strict")
ret <4 x double> %fma
}
define <4 x float> @constrained_vector_fma_v4f32() {
; CHECK-LABEL: constrained_vector_fma_v4f32:
; CHECK: # %bb.0: # %entry
; CHECK-NEXT: vmovaps {{.*#+}} xmm1 = [3.5E+0,2.5E+0,1.5E+0,5.0E-1]
; CHECK-NEXT: vmovaps {{.*#+}} xmm0 = [7.5E+0,6.5E+0,5.5E+0,4.5E+0]
; CHECK-NEXT: vfmadd213ps {{.*#+}} xmm0 = (xmm1 * xmm0) + mem
; CHECK-NEXT: retq
entry:
%fma = call <4 x float> @llvm.experimental.constrained.fma.v4f32(
<4 x float> <float 3.5, float 2.5, float 1.5, float 0.5>,
<4 x float> <float 7.5, float 6.5, float 5.5, float 4.5>,
<4 x float> <float 11.5, float 10.5, float 9.5, float 8.5>,
metadata !"round.dynamic",
metadata !"fpexcept.strict")
ret <4 x float> %fma
}
define <8 x float> @constrained_vector_fma_v8f32() {
; CHECK-LABEL: constrained_vector_fma_v8f32:
; CHECK: # %bb.0: # %entry
; CHECK-NEXT: vmovaps {{.*#+}} ymm1 = [3.5E+0,2.5E+0,1.5E+0,5.0E-1,7.5E+0,6.5E+0,5.5E+0,4.5E+0]
; CHECK-NEXT: vmovaps {{.*#+}} ymm0 = [7.5E+0,6.5E+0,5.5E+0,4.5E+0,1.15E+1,1.05E+1,9.5E+0,8.5E+0]
; CHECK-NEXT: vfmadd213ps {{.*#+}} ymm0 = (ymm1 * ymm0) + mem
; CHECK-NEXT: retq
entry:
%fma = call <8 x float> @llvm.experimental.constrained.fma.v8f32(
<8 x float> <float 3.5, float 2.5, float 1.5, float 0.5,
float 7.5, float 6.5, float 5.5, float 4.5>,
<8 x float> <float 7.5, float 6.5, float 5.5, float 4.5,
float 11.5, float 10.5, float 9.5, float 8.5>,
<8 x float> <float 11.5, float 10.5, float 9.5, float 8.5,
float 15.5, float 14.5, float 13.5, float 12.5>,
metadata !"round.dynamic",
metadata !"fpexcept.strict")
ret <8 x float> %fma
}
; Single width declarations
declare <2 x double> @llvm.experimental.constrained.fma.v2f64(<2 x double>, <2 x double>, <2 x double>, metadata, metadata)
declare <4 x float> @llvm.experimental.constrained.fma.v4f32(<4 x float>, <4 x float>, <4 x float>, metadata, metadata)
; Scalar width declarations
declare <1 x float> @llvm.experimental.constrained.fma.v1f32(<1 x float>, <1 x float>, <1 x float>, metadata, metadata)
; Illegal width declarations
declare <3 x float> @llvm.experimental.constrained.fma.v3f32(<3 x float>, <3 x float>, <3 x float>, metadata, metadata)
declare <3 x double> @llvm.experimental.constrained.fma.v3f64(<3 x double>, <3 x double>, <3 x double>, metadata, metadata)
; Double width declarations
declare <4 x double> @llvm.experimental.constrained.fma.v4f64(<4 x double>, <4 x double>, <4 x double>, metadata, metadata)
declare <8 x float> @llvm.experimental.constrained.fma.v8f32(<8 x float>, <8 x float>, <8 x float>, metadata, metadata)

test/CodeGen/X86/vector-constrained-fp-intrinsics.ll

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -O3 -mtriple=x86_64-pc-linux < %s | FileCheck --check-prefix=COMMON --check-prefix=NO-FMA --check-prefix=FMACALL64 --check-prefix=FMACALL32 %s ; RUN: llc -O3 -mtriple=x86_64-pc-linux < %s | FileCheck %s
; RUN: llc -O3 -mtriple=x86_64-pc-linux -mattr=+fma < %s | FileCheck -check-prefix=COMMON --check-prefix=HAS-FMA --check-prefix=FMA64 --check-prefix=FMA32 %s
define <1 x float> @constrained_vector_fdiv_v1f32() { define <1 x float> @constrained_vector_fdiv_v1f32() {
; NO-FMA-LABEL: constrained_vector_fdiv_v1f32: ; CHECK-LABEL: constrained_vector_fdiv_v1f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: divss {{.*}}(%rip), %xmm0 ; CHECK-NEXT: divss {{.*}}(%rip), %xmm0
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_fdiv_v1f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vdivss {{.*}}(%rip), %xmm0, %xmm0
; HAS-FMA-NEXT: retq
entry: entry:
%div = call <1 x float> @llvm.experimental.constrained.fdiv.v1f32( %div = call <1 x float> @llvm.experimental.constrained.fdiv.v1f32(
<1 x float> <float 1.000000e+00>, <1 x float> <float 1.000000e+00>,
<1 x float> <float 1.000000e+01>, <1 x float> <float 1.000000e+01>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <1 x float> %div ret <1 x float> %div
} }
define <2 x double> @constrained_vector_fdiv_v2f64() { define <2 x double> @constrained_vector_fdiv_v2f64() {
; NO-FMA-LABEL: constrained_vector_fdiv_v2f64: ; CHECK-LABEL: constrained_vector_fdiv_v2f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: movapd {{.*#+}} xmm0 = [1.0E+0,2.0E+0] ; CHECK-NEXT: movapd {{.*#+}} xmm0 = [1.0E+0,2.0E+0]
; NO-FMA-NEXT: divpd {{.*}}(%rip), %xmm0 ; CHECK-NEXT: divpd {{.*}}(%rip), %xmm0
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_fdiv_v2f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vmovapd {{.*#+}} xmm0 = [1.0E+0,2.0E+0]
; HAS-FMA-NEXT: vdivpd {{.*}}(%rip), %xmm0, %xmm0
; HAS-FMA-NEXT: retq
entry: entry:
%div = call <2 x double> @llvm.experimental.constrained.fdiv.v2f64( %div = call <2 x double> @llvm.experimental.constrained.fdiv.v2f64(
<2 x double> <double 1.000000e+00, double 2.000000e+00>, <2 x double> <double 1.000000e+00, double 2.000000e+00>,
<2 x double> <double 1.000000e+01, double 1.000000e+01>, <2 x double> <double 1.000000e+01, double 1.000000e+01>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <2 x double> %div ret <2 x double> %div
} }
define <3 x float> @constrained_vector_fdiv_v3f32() { define <3 x float> @constrained_vector_fdiv_v3f32() {
; NO-FMA-LABEL: constrained_vector_fdiv_v3f32: ; CHECK-LABEL: constrained_vector_fdiv_v3f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; NO-FMA-NEXT: divss %xmm1, %xmm2 ; CHECK-NEXT: divss %xmm1, %xmm2
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: divss %xmm1, %xmm0 ; CHECK-NEXT: divss %xmm1, %xmm0
; NO-FMA-NEXT: movss {{.*#+}} xmm3 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm3 = mem[0],zero,zero,zero
; NO-FMA-NEXT: divss %xmm1, %xmm3 ; CHECK-NEXT: divss %xmm1, %xmm3
; NO-FMA-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1] ; CHECK-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; NO-FMA-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm2[0] ; CHECK-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm2[0]
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_fdiv_v3f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vdivss %xmm0, %xmm1, %xmm1
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vdivss %xmm0, %xmm2, %xmm2
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm3 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vdivss %xmm0, %xmm3, %xmm0
; HAS-FMA-NEXT: vinsertps {{.*#+}} xmm0 = xmm2[0],xmm0[0],xmm2[2,3]
; HAS-FMA-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0,1],xmm1[0],xmm0[3]
; HAS-FMA-NEXT: retq
entry: entry:
%div = call <3 x float> @llvm.experimental.constrained.fdiv.v3f32( %div = call <3 x float> @llvm.experimental.constrained.fdiv.v3f32(
<3 x float> <float 1.000000e+00, float 2.000000e+00, float 3.000000e+00>, <3 x float> <float 1.000000e+00, float 2.000000e+00, float 3.000000e+00>,
<3 x float> <float 1.000000e+01, float 1.000000e+01, float 1.000000e+01>, <3 x float> <float 1.000000e+01, float 1.000000e+01, float 1.000000e+01>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x float> %div ret <3 x float> %div
} }
define <3 x double> @constrained_vector_fdiv_v3f64() { define <3 x double> @constrained_vector_fdiv_v3f64() {
; NO-FMA-LABEL: constrained_vector_fdiv_v3f64: ; CHECK-LABEL: constrained_vector_fdiv_v3f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: movapd {{.*#+}} xmm0 = [1.0E+0,2.0E+0] ; CHECK-NEXT: movapd {{.*#+}} xmm0 = [1.0E+0,2.0E+0]
; NO-FMA-NEXT: divpd {{.*}}(%rip), %xmm0 ; CHECK-NEXT: divpd {{.*}}(%rip), %xmm0
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: divsd {{.*}}(%rip), %xmm1 ; CHECK-NEXT: divsd {{.*}}(%rip), %xmm1
; NO-FMA-NEXT: movsd %xmm1, -{{[0-9]+}}(%rsp) ; CHECK-NEXT: movsd %xmm1, -{{[0-9]+}}(%rsp)
; NO-FMA-NEXT: movapd %xmm0, %xmm1 ; CHECK-NEXT: movapd %xmm0, %xmm1
; NO-FMA-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1] ; CHECK-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; NO-FMA-NEXT: fldl -{{[0-9]+}}(%rsp) ; CHECK-NEXT: fldl -{{[0-9]+}}(%rsp)
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_fdiv_v3f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vdivsd {{.*}}(%rip), %xmm0, %xmm0
; HAS-FMA-NEXT: vmovapd {{.*#+}} xmm1 = [1.0E+0,2.0E+0]
; HAS-FMA-NEXT: vdivpd {{.*}}(%rip), %xmm1, %xmm1
; HAS-FMA-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; HAS-FMA-NEXT: retq
entry: entry:
%div = call <3 x double> @llvm.experimental.constrained.fdiv.v3f64( %div = call <3 x double> @llvm.experimental.constrained.fdiv.v3f64(
<3 x double> <double 1.000000e+00, double 2.000000e+00, double 3.000000e+00>, <3 x double> <double 1.000000e+00, double 2.000000e+00, double 3.000000e+00>,
<3 x double> <double 1.000000e+01, double 1.000000e+01, double 1.000000e+01>, <3 x double> <double 1.000000e+01, double 1.000000e+01, double 1.000000e+01>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x double> %div ret <3 x double> %div
} }
define <4 x double> @constrained_vector_fdiv_v4f64() { define <4 x double> @constrained_vector_fdiv_v4f64() {
; NO-FMA-LABEL: constrained_vector_fdiv_v4f64: ; CHECK-LABEL: constrained_vector_fdiv_v4f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: movapd {{.*#+}} xmm2 = [1.0E+1,1.0E+1] ; CHECK-NEXT: movapd {{.*#+}} xmm2 = [1.0E+1,1.0E+1]
; NO-FMA-NEXT: movapd {{.*#+}} xmm0 = [1.0E+0,2.0E+0] ; CHECK-NEXT: movapd {{.*#+}} xmm0 = [1.0E+0,2.0E+0]
; NO-FMA-NEXT: divpd %xmm2, %xmm0 ; CHECK-NEXT: divpd %xmm2, %xmm0
; NO-FMA-NEXT: movapd {{.*#+}} xmm1 = [3.0E+0,4.0E+0] ; CHECK-NEXT: movapd {{.*#+}} xmm1 = [3.0E+0,4.0E+0]
; NO-FMA-NEXT: divpd %xmm2, %xmm1 ; CHECK-NEXT: divpd %xmm2, %xmm1
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_fdiv_v4f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vmovapd {{.*#+}} ymm0 = [1.0E+0,2.0E+0,3.0E+0,4.0E+0]
; HAS-FMA-NEXT: vdivpd {{.*}}(%rip), %ymm0, %ymm0
; HAS-FMA-NEXT: retq
entry: entry:
%div = call <4 x double> @llvm.experimental.constrained.fdiv.v4f64( %div = call <4 x double> @llvm.experimental.constrained.fdiv.v4f64(
<4 x double> <double 1.000000e+00, double 2.000000e+00, <4 x double> <double 1.000000e+00, double 2.000000e+00,
double 3.000000e+00, double 4.000000e+00>, double 3.000000e+00, double 4.000000e+00>,
<4 x double> <double 1.000000e+01, double 1.000000e+01, <4 x double> <double 1.000000e+01, double 1.000000e+01,
double 1.000000e+01, double 1.000000e+01>, double 1.000000e+01, double 1.000000e+01>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <4 x double> %div ret <4 x double> %div
} }
define <1 x float> @constrained_vector_frem_v1f32() { define <1 x float> @constrained_vector_frem_v1f32() {
; NO-FMA-LABEL: constrained_vector_frem_v1f32: ; CHECK-LABEL: constrained_vector_frem_v1f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: pushq %rax ; CHECK-NEXT: pushq %rax
; NO-FMA-NEXT: .cfi_def_cfa_offset 16 ; CHECK-NEXT: .cfi_def_cfa_offset 16
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq fmodf ; CHECK-NEXT: callq fmodf
; NO-FMA-NEXT: popq %rax ; CHECK-NEXT: popq %rax
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_frem_v1f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: pushq %rax
; HAS-FMA-NEXT: .cfi_def_cfa_offset 16
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq fmodf
; HAS-FMA-NEXT: popq %rax
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%rem = call <1 x float> @llvm.experimental.constrained.frem.v1f32( %rem = call <1 x float> @llvm.experimental.constrained.frem.v1f32(
<1 x float> <float 1.000000e+00>, <1 x float> <float 1.000000e+00>,
<1 x float> <float 1.000000e+01>, <1 x float> <float 1.000000e+01>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <1 x float> %rem ret <1 x float> %rem
} }
define <2 x double> @constrained_vector_frem_v2f64() { define <2 x double> @constrained_vector_frem_v2f64() {
; NO-FMA-LABEL: constrained_vector_frem_v2f64: ; CHECK-LABEL: constrained_vector_frem_v2f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $24, %rsp ; CHECK-NEXT: subq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 32 ; CHECK-NEXT: .cfi_def_cfa_offset 32
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq fmod ; CHECK-NEXT: callq fmod
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq fmod ; CHECK-NEXT: callq fmod
; NO-FMA-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0] ; CHECK-NEXT: # xmm0 = xmm0[0],mem[0]
; NO-FMA-NEXT: addq $24, %rsp ; CHECK-NEXT: addq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_frem_v2f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $24, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 32
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: callq fmod
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: callq fmod
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: addq $24, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%rem = call <2 x double> @llvm.experimental.constrained.frem.v2f64( %rem = call <2 x double> @llvm.experimental.constrained.frem.v2f64(
<2 x double> <double 1.000000e+00, double 2.000000e+00>, <2 x double> <double 1.000000e+00, double 2.000000e+00>,
<2 x double> <double 1.000000e+01, double 1.000000e+01>, <2 x double> <double 1.000000e+01, double 1.000000e+01>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <2 x double> %rem ret <2 x double> %rem
} }
define <3 x float> @constrained_vector_frem_v3f32() { define <3 x float> @constrained_vector_frem_v3f32() {
; NO-FMA-LABEL: constrained_vector_frem_v3f32: ; CHECK-LABEL: constrained_vector_frem_v3f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $40, %rsp ; CHECK-NEXT: subq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 48 ; CHECK-NEXT: .cfi_def_cfa_offset 48
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq fmodf ; CHECK-NEXT: callq fmodf
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq fmodf ; CHECK-NEXT: callq fmodf
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq fmodf ; CHECK-NEXT: callq fmodf
; NO-FMA-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload ; CHECK-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload
; NO-FMA-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; CHECK-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; NO-FMA-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm1 = xmm1[0],mem[0] ; CHECK-NEXT: # xmm1 = xmm1[0],mem[0]
; NO-FMA-NEXT: movaps %xmm1, %xmm0 ; CHECK-NEXT: movaps %xmm1, %xmm0
; NO-FMA-NEXT: addq $40, %rsp ; CHECK-NEXT: addq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_frem_v3f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 48
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq fmodf
; HAS-FMA-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq fmodf
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq fmodf
; HAS-FMA-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload
; HAS-FMA-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[2,3]
; HAS-FMA-NEXT: vinsertps $32, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0,1],mem[0],xmm0[3]
; HAS-FMA-NEXT: addq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%rem = call <3 x float> @llvm.experimental.constrained.frem.v3f32( %rem = call <3 x float> @llvm.experimental.constrained.frem.v3f32(
<3 x float> <float 1.000000e+00, float 2.000000e+00, float 3.000000e+00>, <3 x float> <float 1.000000e+00, float 2.000000e+00, float 3.000000e+00>,
<3 x float> <float 1.000000e+01, float 1.000000e+01, float 1.000000e+01>, <3 x float> <float 1.000000e+01, float 1.000000e+01, float 1.000000e+01>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x float> %rem ret <3 x float> %rem
} }
define <3 x double> @constrained_vector_frem_v3f64() { define <3 x double> @constrained_vector_frem_v3f64() {
; NO-FMA-LABEL: constrained_vector_frem_v3f64: ; CHECK-LABEL: constrained_vector_frem_v3f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $24, %rsp ; CHECK-NEXT: subq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 32 ; CHECK-NEXT: .cfi_def_cfa_offset 32
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq fmod ; CHECK-NEXT: callq fmod
; NO-FMA-NEXT: movsd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill ; CHECK-NEXT: movsd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq fmod ; CHECK-NEXT: callq fmod
; NO-FMA-NEXT: movsd %xmm0, (%rsp) # 8-byte Spill ; CHECK-NEXT: movsd %xmm0, (%rsp) # 8-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq fmod ; CHECK-NEXT: callq fmod
; NO-FMA-NEXT: movsd %xmm0, {{[0-9]+}}(%rsp) ; CHECK-NEXT: movsd %xmm0, {{[0-9]+}}(%rsp)
; NO-FMA-NEXT: fldl {{[0-9]+}}(%rsp) ; CHECK-NEXT: fldl {{[0-9]+}}(%rsp)
; NO-FMA-NEXT: movsd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 8-byte Reload ; CHECK-NEXT: movsd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 8-byte Reload
; NO-FMA-NEXT: # xmm0 = mem[0],zero ; CHECK-NEXT: # xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd (%rsp), %xmm1 # 8-byte Reload ; CHECK-NEXT: movsd (%rsp), %xmm1 # 8-byte Reload
; NO-FMA-NEXT: # xmm1 = mem[0],zero ; CHECK-NEXT: # xmm1 = mem[0],zero
; NO-FMA-NEXT: addq $24, %rsp ; CHECK-NEXT: addq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_frem_v3f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $56, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 64
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: callq fmod
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: callq fmod
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: vzeroupper
; HAS-FMA-NEXT: callq fmod
; HAS-FMA-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
; HAS-FMA-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; HAS-FMA-NEXT: addq $56, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%rem = call <3 x double> @llvm.experimental.constrained.frem.v3f64( %rem = call <3 x double> @llvm.experimental.constrained.frem.v3f64(
<3 x double> <double 1.000000e+00, double 2.000000e+00, double 3.000000e+00>, <3 x double> <double 1.000000e+00, double 2.000000e+00, double 3.000000e+00>,
<3 x double> <double 1.000000e+01, double 1.000000e+01, double 1.000000e+01>, <3 x double> <double 1.000000e+01, double 1.000000e+01, double 1.000000e+01>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x double> %rem ret <3 x double> %rem
} }
define <4 x double> @constrained_vector_frem_v4f64() { define <4 x double> @constrained_vector_frem_v4f64() {
; NO-FMA-LABEL: constrained_vector_frem_v4f64: ; CHECK-LABEL: constrained_vector_frem_v4f64:
; NO-FMA: # %bb.0: ; CHECK: # %bb.0:
; NO-FMA-NEXT: subq $40, %rsp ; CHECK-NEXT: subq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 48 ; CHECK-NEXT: .cfi_def_cfa_offset 48
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq fmod ; CHECK-NEXT: callq fmod
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq fmod ; CHECK-NEXT: callq fmod
; NO-FMA-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0] ; CHECK-NEXT: # xmm0 = xmm0[0],mem[0]
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq fmod ; CHECK-NEXT: callq fmod
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq fmod ; CHECK-NEXT: callq fmod
; NO-FMA-NEXT: movaps %xmm0, %xmm1 ; CHECK-NEXT: movaps %xmm0, %xmm1
; NO-FMA-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm1 = xmm1[0],mem[0] ; CHECK-NEXT: # xmm1 = xmm1[0],mem[0]
; NO-FMA-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload ; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; NO-FMA-NEXT: addq $40, %rsp ; CHECK-NEXT: addq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_frem_v4f64:
; HAS-FMA: # %bb.0:
; HAS-FMA-NEXT: subq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 48
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: callq fmod
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: callq fmod
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: callq fmod
; HAS-FMA-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: callq fmod
; HAS-FMA-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vinsertf128 $1, (%rsp), %ymm0, %ymm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: addq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
%rem = call <4 x double> @llvm.experimental.constrained.frem.v4f64( %rem = call <4 x double> @llvm.experimental.constrained.frem.v4f64(
<4 x double> <double 1.000000e+00, double 2.000000e+00, <4 x double> <double 1.000000e+00, double 2.000000e+00,
double 3.000000e+00, double 4.000000e+00>, double 3.000000e+00, double 4.000000e+00>,
<4 x double> <double 1.000000e+01, double 1.000000e+01, <4 x double> <double 1.000000e+01, double 1.000000e+01,
double 1.000000e+01, double 1.000000e+01>, double 1.000000e+01, double 1.000000e+01>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <4 x double> %rem ret <4 x double> %rem
} }
define <1 x float> @constrained_vector_fmul_v1f32() { define <1 x float> @constrained_vector_fmul_v1f32() {
; NO-FMA-LABEL: constrained_vector_fmul_v1f32: ; CHECK-LABEL: constrained_vector_fmul_v1f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: mulss {{.*}}(%rip), %xmm0 ; CHECK-NEXT: mulss {{.*}}(%rip), %xmm0
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_fmul_v1f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm0
; HAS-FMA-NEXT: retq
entry: entry:
%mul = call <1 x float> @llvm.experimental.constrained.fmul.v1f32( %mul = call <1 x float> @llvm.experimental.constrained.fmul.v1f32(
<1 x float> <float 0x7FF0000000000000>, <1 x float> <float 0x7FF0000000000000>,
<1 x float> <float 2.000000e+00>, <1 x float> <float 2.000000e+00>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <1 x float> %mul ret <1 x float> %mul
} }
define <2 x double> @constrained_vector_fmul_v2f64() { define <2 x double> @constrained_vector_fmul_v2f64() {
; NO-FMA-LABEL: constrained_vector_fmul_v2f64: ; CHECK-LABEL: constrained_vector_fmul_v2f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: movapd {{.*#+}} xmm0 = [1.7976931348623157E+308,1.7976931348623157E+308] ; CHECK-NEXT: movapd {{.*#+}} xmm0 = [1.7976931348623157E+308,1.7976931348623157E+308]
; NO-FMA-NEXT: mulpd {{.*}}(%rip), %xmm0 ; CHECK-NEXT: mulpd {{.*}}(%rip), %xmm0
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_fmul_v2f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vmovapd {{.*#+}} xmm0 = [1.7976931348623157E+308,1.7976931348623157E+308]
; HAS-FMA-NEXT: vmulpd {{.*}}(%rip), %xmm0, %xmm0
; HAS-FMA-NEXT: retq
entry: entry:
%mul = call <2 x double> @llvm.experimental.constrained.fmul.v2f64( %mul = call <2 x double> @llvm.experimental.constrained.fmul.v2f64(
<2 x double> <double 0x7FEFFFFFFFFFFFFF, double 0x7FEFFFFFFFFFFFFF>, <2 x double> <double 0x7FEFFFFFFFFFFFFF, double 0x7FEFFFFFFFFFFFFF>,
<2 x double> <double 2.000000e+00, double 3.000000e+00>, <2 x double> <double 2.000000e+00, double 3.000000e+00>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <2 x double> %mul ret <2 x double> %mul
} }
define <3 x float> @constrained_vector_fmul_v3f32() { define <3 x float> @constrained_vector_fmul_v3f32() {
; NO-FMA-LABEL: constrained_vector_fmul_v3f32: ; CHECK-LABEL: constrained_vector_fmul_v3f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; NO-FMA-NEXT: mulss %xmm1, %xmm2 ; CHECK-NEXT: mulss %xmm1, %xmm2
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: mulss %xmm1, %xmm0 ; CHECK-NEXT: mulss %xmm1, %xmm0
; NO-FMA-NEXT: mulss {{.*}}(%rip), %xmm1 ; CHECK-NEXT: mulss {{.*}}(%rip), %xmm1
; NO-FMA-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] ; CHECK-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; NO-FMA-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm2[0] ; CHECK-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm2[0]
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_fmul_v3f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm1
; HAS-FMA-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm2
; HAS-FMA-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm0
; HAS-FMA-NEXT: vinsertps {{.*#+}} xmm0 = xmm2[0],xmm0[0],xmm2[2,3]
; HAS-FMA-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0,1],xmm1[0],xmm0[3]
; HAS-FMA-NEXT: retq
entry: entry:
%mul = call <3 x float> @llvm.experimental.constrained.fmul.v3f32( %mul = call <3 x float> @llvm.experimental.constrained.fmul.v3f32(
<3 x float> <float 0x7FF0000000000000, float 0x7FF0000000000000, <3 x float> <float 0x7FF0000000000000, float 0x7FF0000000000000,
float 0x7FF0000000000000>, float 0x7FF0000000000000>,
<3 x float> <float 1.000000e+00, float 1.000000e+01, float 1.000000e+02>, <3 x float> <float 1.000000e+00, float 1.000000e+01, float 1.000000e+02>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x float> %mul ret <3 x float> %mul
} }
define <3 x double> @constrained_vector_fmul_v3f64() { define <3 x double> @constrained_vector_fmul_v3f64() {
; NO-FMA-LABEL: constrained_vector_fmul_v3f64: ; CHECK-LABEL: constrained_vector_fmul_v3f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: movapd {{.*#+}} xmm0 = [1.7976931348623157E+308,1.7976931348623157E+308] ; CHECK-NEXT: movapd {{.*#+}} xmm0 = [1.7976931348623157E+308,1.7976931348623157E+308]
; NO-FMA-NEXT: mulpd {{.*}}(%rip), %xmm0 ; CHECK-NEXT: mulpd {{.*}}(%rip), %xmm0
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: mulsd {{.*}}(%rip), %xmm1 ; CHECK-NEXT: mulsd {{.*}}(%rip), %xmm1
; NO-FMA-NEXT: movsd %xmm1, -{{[0-9]+}}(%rsp) ; CHECK-NEXT: movsd %xmm1, -{{[0-9]+}}(%rsp)
; NO-FMA-NEXT: movapd %xmm0, %xmm1 ; CHECK-NEXT: movapd %xmm0, %xmm1
; NO-FMA-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1] ; CHECK-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; NO-FMA-NEXT: fldl -{{[0-9]+}}(%rsp) ; CHECK-NEXT: fldl -{{[0-9]+}}(%rsp)
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_fmul_v3f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmulsd {{.*}}(%rip), %xmm0, %xmm0
; HAS-FMA-NEXT: vmovapd {{.*#+}} xmm1 = [1.7976931348623157E+308,1.7976931348623157E+308]
; HAS-FMA-NEXT: vmulpd {{.*}}(%rip), %xmm1, %xmm1
; HAS-FMA-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; HAS-FMA-NEXT: retq
entry: entry:
%mul = call <3 x double> @llvm.experimental.constrained.fmul.v3f64( %mul = call <3 x double> @llvm.experimental.constrained.fmul.v3f64(
<3 x double> <double 0x7FEFFFFFFFFFFFFF, double 0x7FEFFFFFFFFFFFFF, <3 x double> <double 0x7FEFFFFFFFFFFFFF, double 0x7FEFFFFFFFFFFFFF,
double 0x7FEFFFFFFFFFFFFF>, double 0x7FEFFFFFFFFFFFFF>,
<3 x double> <double 1.000000e+00, double 1.000000e+01, double 1.000000e+02>, <3 x double> <double 1.000000e+00, double 1.000000e+01, double 1.000000e+02>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x double> %mul ret <3 x double> %mul
} }
define <4 x double> @constrained_vector_fmul_v4f64() { define <4 x double> @constrained_vector_fmul_v4f64() {
; NO-FMA-LABEL: constrained_vector_fmul_v4f64: ; CHECK-LABEL: constrained_vector_fmul_v4f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: movapd {{.*#+}} xmm1 = [1.7976931348623157E+308,1.7976931348623157E+308] ; CHECK-NEXT: movapd {{.*#+}} xmm1 = [1.7976931348623157E+308,1.7976931348623157E+308]
; NO-FMA-NEXT: movapd {{.*#+}} xmm0 = [2.0E+0,3.0E+0] ; CHECK-NEXT: movapd {{.*#+}} xmm0 = [2.0E+0,3.0E+0]
; NO-FMA-NEXT: mulpd %xmm1, %xmm0 ; CHECK-NEXT: mulpd %xmm1, %xmm0
; NO-FMA-NEXT: mulpd {{.*}}(%rip), %xmm1 ; CHECK-NEXT: mulpd {{.*}}(%rip), %xmm1
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_fmul_v4f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vmovapd {{.*#+}} ymm0 = [1.7976931348623157E+308,1.7976931348623157E+308,1.7976931348623157E+308,1.7976931348623157E+308]
; HAS-FMA-NEXT: vmulpd {{.*}}(%rip), %ymm0, %ymm0
; HAS-FMA-NEXT: retq
entry: entry:
%mul = call <4 x double> @llvm.experimental.constrained.fmul.v4f64( %mul = call <4 x double> @llvm.experimental.constrained.fmul.v4f64(
<4 x double> <double 0x7FEFFFFFFFFFFFFF, double 0x7FEFFFFFFFFFFFFF, <4 x double> <double 0x7FEFFFFFFFFFFFFF, double 0x7FEFFFFFFFFFFFFF,
double 0x7FEFFFFFFFFFFFFF, double 0x7FEFFFFFFFFFFFFF>, double 0x7FEFFFFFFFFFFFFF, double 0x7FEFFFFFFFFFFFFF>,
<4 x double> <double 2.000000e+00, double 3.000000e+00, <4 x double> <double 2.000000e+00, double 3.000000e+00,
double 4.000000e+00, double 5.000000e+00>, double 4.000000e+00, double 5.000000e+00>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <4 x double> %mul ret <4 x double> %mul
} }
define <1 x float> @constrained_vector_fadd_v1f32() { define <1 x float> @constrained_vector_fadd_v1f32() {
; NO-FMA-LABEL: constrained_vector_fadd_v1f32: ; CHECK-LABEL: constrained_vector_fadd_v1f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: addss {{.*}}(%rip), %xmm0 ; CHECK-NEXT: addss {{.*}}(%rip), %xmm0
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_fadd_v1f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vaddss {{.*}}(%rip), %xmm0, %xmm0
; HAS-FMA-NEXT: retq
entry: entry:
%add = call <1 x float> @llvm.experimental.constrained.fadd.v1f32( %add = call <1 x float> @llvm.experimental.constrained.fadd.v1f32(
<1 x float> <float 0x7FF0000000000000>, <1 x float> <float 0x7FF0000000000000>,
<1 x float> <float 1.0>, <1 x float> <float 1.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <1 x float> %add ret <1 x float> %add
} }
define <2 x double> @constrained_vector_fadd_v2f64() { define <2 x double> @constrained_vector_fadd_v2f64() {
; NO-FMA-LABEL: constrained_vector_fadd_v2f64: ; CHECK-LABEL: constrained_vector_fadd_v2f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: movapd {{.*#+}} xmm0 = [1.7976931348623157E+308,1.7976931348623157E+308] ; CHECK-NEXT: movapd {{.*#+}} xmm0 = [1.7976931348623157E+308,1.7976931348623157E+308]
; NO-FMA-NEXT: addpd {{.*}}(%rip), %xmm0 ; CHECK-NEXT: addpd {{.*}}(%rip), %xmm0
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_fadd_v2f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vmovapd {{.*#+}} xmm0 = [1.7976931348623157E+308,1.7976931348623157E+308]
; HAS-FMA-NEXT: vaddpd {{.*}}(%rip), %xmm0, %xmm0
; HAS-FMA-NEXT: retq
entry: entry:
%add = call <2 x double> @llvm.experimental.constrained.fadd.v2f64( %add = call <2 x double> @llvm.experimental.constrained.fadd.v2f64(
<2 x double> <double 0x7FEFFFFFFFFFFFFF, double 0x7FEFFFFFFFFFFFFF>, <2 x double> <double 0x7FEFFFFFFFFFFFFF, double 0x7FEFFFFFFFFFFFFF>,
<2 x double> <double 1.000000e+00, double 1.000000e-01>, <2 x double> <double 1.000000e+00, double 1.000000e-01>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <2 x double> %add ret <2 x double> %add
} }
define <3 x float> @constrained_vector_fadd_v3f32() { define <3 x float> @constrained_vector_fadd_v3f32() {
; NO-FMA-LABEL: constrained_vector_fadd_v3f32: ; CHECK-LABEL: constrained_vector_fadd_v3f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: xorps %xmm1, %xmm1 ; CHECK-NEXT: xorps %xmm1, %xmm1
; NO-FMA-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; NO-FMA-NEXT: addss %xmm2, %xmm1 ; CHECK-NEXT: addss %xmm2, %xmm1
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: addss %xmm2, %xmm0 ; CHECK-NEXT: addss %xmm2, %xmm0
; NO-FMA-NEXT: addss {{.*}}(%rip), %xmm2 ; CHECK-NEXT: addss {{.*}}(%rip), %xmm2
; NO-FMA-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1] ; CHECK-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
; NO-FMA-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0] ; CHECK-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_fadd_v3f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vxorps %xmm0, %xmm0, %xmm0
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vaddss %xmm0, %xmm1, %xmm0
; HAS-FMA-NEXT: vaddss {{.*}}(%rip), %xmm1, %xmm2
; HAS-FMA-NEXT: vaddss {{.*}}(%rip), %xmm1, %xmm1
; HAS-FMA-NEXT: vinsertps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[2,3]
; HAS-FMA-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0,1],xmm0[0],xmm1[3]
; HAS-FMA-NEXT: retq
entry: entry:
%add = call <3 x float> @llvm.experimental.constrained.fadd.v3f32( %add = call <3 x float> @llvm.experimental.constrained.fadd.v3f32(
<3 x float> <float 0xFFFFFFFFE0000000, float 0xFFFFFFFFE0000000, <3 x float> <float 0xFFFFFFFFE0000000, float 0xFFFFFFFFE0000000,
float 0xFFFFFFFFE0000000>, float 0xFFFFFFFFE0000000>,
<3 x float> <float 2.0, float 1.0, float 0.0>, <3 x float> <float 2.0, float 1.0, float 0.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x float> %add ret <3 x float> %add
} }
define <3 x double> @constrained_vector_fadd_v3f64() { define <3 x double> @constrained_vector_fadd_v3f64() {
; NO-FMA-LABEL: constrained_vector_fadd_v3f64: ; CHECK-LABEL: constrained_vector_fadd_v3f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: movapd {{.*#+}} xmm0 = [1.7976931348623157E+308,1.7976931348623157E+308] ; CHECK-NEXT: movapd {{.*#+}} xmm0 = [1.7976931348623157E+308,1.7976931348623157E+308]
; NO-FMA-NEXT: addpd {{.*}}(%rip), %xmm0 ; CHECK-NEXT: addpd {{.*}}(%rip), %xmm0
; NO-FMA-NEXT: xorpd %xmm1, %xmm1 ; CHECK-NEXT: xorpd %xmm1, %xmm1
; NO-FMA-NEXT: addsd {{.*}}(%rip), %xmm1 ; CHECK-NEXT: addsd {{.*}}(%rip), %xmm1
; NO-FMA-NEXT: movsd %xmm1, -{{[0-9]+}}(%rsp) ; CHECK-NEXT: movsd %xmm1, -{{[0-9]+}}(%rsp)
; NO-FMA-NEXT: movapd %xmm0, %xmm1 ; CHECK-NEXT: movapd %xmm0, %xmm1
; NO-FMA-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1] ; CHECK-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; NO-FMA-NEXT: fldl -{{[0-9]+}}(%rsp) ; CHECK-NEXT: fldl -{{[0-9]+}}(%rsp)
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_fadd_v3f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vxorpd %xmm0, %xmm0, %xmm0
; HAS-FMA-NEXT: vaddsd {{.*}}(%rip), %xmm0, %xmm0
; HAS-FMA-NEXT: vmovapd {{.*#+}} xmm1 = [1.7976931348623157E+308,1.7976931348623157E+308]
; HAS-FMA-NEXT: vaddpd {{.*}}(%rip), %xmm1, %xmm1
; HAS-FMA-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; HAS-FMA-NEXT: retq
entry: entry:
%add = call <3 x double> @llvm.experimental.constrained.fadd.v3f64( %add = call <3 x double> @llvm.experimental.constrained.fadd.v3f64(
<3 x double> <double 0x7FEFFFFFFFFFFFFF, double 0x7FEFFFFFFFFFFFFF, <3 x double> <double 0x7FEFFFFFFFFFFFFF, double 0x7FEFFFFFFFFFFFFF,
double 0x7FEFFFFFFFFFFFFF>, double 0x7FEFFFFFFFFFFFFF>,
<3 x double> <double 2.0, double 1.0, double 0.0>, <3 x double> <double 2.0, double 1.0, double 0.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x double> %add ret <3 x double> %add
} }
define <4 x double> @constrained_vector_fadd_v4f64() { define <4 x double> @constrained_vector_fadd_v4f64() {
; NO-FMA-LABEL: constrained_vector_fadd_v4f64: ; CHECK-LABEL: constrained_vector_fadd_v4f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: movapd {{.*#+}} xmm1 = [1.7976931348623157E+308,1.7976931348623157E+308] ; CHECK-NEXT: movapd {{.*#+}} xmm1 = [1.7976931348623157E+308,1.7976931348623157E+308]
; NO-FMA-NEXT: movapd {{.*#+}} xmm0 = [1.0E+0,1.0000000000000001E-1] ; CHECK-NEXT: movapd {{.*#+}} xmm0 = [1.0E+0,1.0000000000000001E-1]
; NO-FMA-NEXT: addpd %xmm1, %xmm0 ; CHECK-NEXT: addpd %xmm1, %xmm0
; NO-FMA-NEXT: addpd {{.*}}(%rip), %xmm1 ; CHECK-NEXT: addpd {{.*}}(%rip), %xmm1
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_fadd_v4f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vmovapd {{.*#+}} ymm0 = [1.7976931348623157E+308,1.7976931348623157E+308,1.7976931348623157E+308,1.7976931348623157E+308]
; HAS-FMA-NEXT: vaddpd {{.*}}(%rip), %ymm0, %ymm0
; HAS-FMA-NEXT: retq
entry: entry:
%add = call <4 x double> @llvm.experimental.constrained.fadd.v4f64( %add = call <4 x double> @llvm.experimental.constrained.fadd.v4f64(
<4 x double> <double 0x7FEFFFFFFFFFFFFF, double 0x7FEFFFFFFFFFFFFF, <4 x double> <double 0x7FEFFFFFFFFFFFFF, double 0x7FEFFFFFFFFFFFFF,
double 0x7FEFFFFFFFFFFFFF, double 0x7FEFFFFFFFFFFFFF>, double 0x7FEFFFFFFFFFFFFF, double 0x7FEFFFFFFFFFFFFF>,
<4 x double> <double 1.000000e+00, double 1.000000e-01, <4 x double> <double 1.000000e+00, double 1.000000e-01,
double 2.000000e+00, double 2.000000e-01>, double 2.000000e+00, double 2.000000e-01>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <4 x double> %add ret <4 x double> %add
} }
define <1 x float> @constrained_vector_fsub_v1f32() { define <1 x float> @constrained_vector_fsub_v1f32() {
; NO-FMA-LABEL: constrained_vector_fsub_v1f32: ; CHECK-LABEL: constrained_vector_fsub_v1f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: subss {{.*}}(%rip), %xmm0 ; CHECK-NEXT: subss {{.*}}(%rip), %xmm0
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_fsub_v1f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vsubss {{.*}}(%rip), %xmm0, %xmm0
; HAS-FMA-NEXT: retq
entry: entry:
%sub = call <1 x float> @llvm.experimental.constrained.fsub.v1f32( %sub = call <1 x float> @llvm.experimental.constrained.fsub.v1f32(
<1 x float> <float 0x7FF0000000000000>, <1 x float> <float 0x7FF0000000000000>,
<1 x float> <float 1.000000e+00>, <1 x float> <float 1.000000e+00>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <1 x float> %sub ret <1 x float> %sub
} }
define <2 x double> @constrained_vector_fsub_v2f64() { define <2 x double> @constrained_vector_fsub_v2f64() {
; NO-FMA-LABEL: constrained_vector_fsub_v2f64: ; CHECK-LABEL: constrained_vector_fsub_v2f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: movapd {{.*#+}} xmm0 = [-1.7976931348623157E+308,-1.7976931348623157E+308] ; CHECK-NEXT: movapd {{.*#+}} xmm0 = [-1.7976931348623157E+308,-1.7976931348623157E+308]
; NO-FMA-NEXT: subpd {{.*}}(%rip), %xmm0 ; CHECK-NEXT: subpd {{.*}}(%rip), %xmm0
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_fsub_v2f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vmovapd {{.*#+}} xmm0 = [-1.7976931348623157E+308,-1.7976931348623157E+308]
; HAS-FMA-NEXT: vsubpd {{.*}}(%rip), %xmm0, %xmm0
; HAS-FMA-NEXT: retq
entry: entry:
%sub = call <2 x double> @llvm.experimental.constrained.fsub.v2f64( %sub = call <2 x double> @llvm.experimental.constrained.fsub.v2f64(
<2 x double> <double 0xFFEFFFFFFFFFFFFF, double 0xFFEFFFFFFFFFFFFF>, <2 x double> <double 0xFFEFFFFFFFFFFFFF, double 0xFFEFFFFFFFFFFFFF>,
<2 x double> <double 1.000000e+00, double 1.000000e-01>, <2 x double> <double 1.000000e+00, double 1.000000e-01>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <2 x double> %sub ret <2 x double> %sub
} }
define <3 x float> @constrained_vector_fsub_v3f32() { define <3 x float> @constrained_vector_fsub_v3f32() {
; NO-FMA-LABEL: constrained_vector_fsub_v3f32: ; CHECK-LABEL: constrained_vector_fsub_v3f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: xorps %xmm0, %xmm0 ; CHECK-NEXT: xorps %xmm0, %xmm0
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movaps %xmm1, %xmm2 ; CHECK-NEXT: movaps %xmm1, %xmm2
; NO-FMA-NEXT: subss %xmm0, %xmm2 ; CHECK-NEXT: subss %xmm0, %xmm2
; NO-FMA-NEXT: movaps %xmm1, %xmm0 ; CHECK-NEXT: movaps %xmm1, %xmm0
; NO-FMA-NEXT: subss {{.*}}(%rip), %xmm0 ; CHECK-NEXT: subss {{.*}}(%rip), %xmm0
; NO-FMA-NEXT: subss {{.*}}(%rip), %xmm1 ; CHECK-NEXT: subss {{.*}}(%rip), %xmm1
; NO-FMA-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] ; CHECK-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; NO-FMA-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm2[0] ; CHECK-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm2[0]
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_fsub_v3f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vxorps %xmm0, %xmm0, %xmm0
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vsubss %xmm0, %xmm1, %xmm0
; HAS-FMA-NEXT: vsubss {{.*}}(%rip), %xmm1, %xmm2
; HAS-FMA-NEXT: vsubss {{.*}}(%rip), %xmm1, %xmm1
; HAS-FMA-NEXT: vinsertps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[2,3]
; HAS-FMA-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0,1],xmm0[0],xmm1[3]
; HAS-FMA-NEXT: retq
entry: entry:
%sub = call <3 x float> @llvm.experimental.constrained.fsub.v3f32( %sub = call <3 x float> @llvm.experimental.constrained.fsub.v3f32(
<3 x float> <float 0xFFFFFFFFE0000000, float 0xFFFFFFFFE0000000, <3 x float> <float 0xFFFFFFFFE0000000, float 0xFFFFFFFFE0000000,
float 0xFFFFFFFFE0000000>, float 0xFFFFFFFFE0000000>,
<3 x float> <float 2.0, float 1.0, float 0.0>, <3 x float> <float 2.0, float 1.0, float 0.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x float> %sub ret <3 x float> %sub
} }
define <3 x double> @constrained_vector_fsub_v3f64() { define <3 x double> @constrained_vector_fsub_v3f64() {
; NO-FMA-LABEL: constrained_vector_fsub_v3f64: ; CHECK-LABEL: constrained_vector_fsub_v3f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: xorpd %xmm0, %xmm0 ; CHECK-NEXT: xorpd %xmm0, %xmm0
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: subsd %xmm0, %xmm1 ; CHECK-NEXT: subsd %xmm0, %xmm1
; NO-FMA-NEXT: movapd {{.*#+}} xmm0 = [-1.7976931348623157E+308,-1.7976931348623157E+308] ; CHECK-NEXT: movapd {{.*#+}} xmm0 = [-1.7976931348623157E+308,-1.7976931348623157E+308]
; NO-FMA-NEXT: subpd {{.*}}(%rip), %xmm0 ; CHECK-NEXT: subpd {{.*}}(%rip), %xmm0
; NO-FMA-NEXT: movsd %xmm1, -{{[0-9]+}}(%rsp) ; CHECK-NEXT: movsd %xmm1, -{{[0-9]+}}(%rsp)
; NO-FMA-NEXT: movapd %xmm0, %xmm1 ; CHECK-NEXT: movapd %xmm0, %xmm1
; NO-FMA-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1] ; CHECK-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; NO-FMA-NEXT: fldl -{{[0-9]+}}(%rsp) ; CHECK-NEXT: fldl -{{[0-9]+}}(%rsp)
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_fsub_v3f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vxorpd %xmm0, %xmm0, %xmm0
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: vsubsd %xmm0, %xmm1, %xmm0
; HAS-FMA-NEXT: vmovapd {{.*#+}} xmm1 = [-1.7976931348623157E+308,-1.7976931348623157E+308]
; HAS-FMA-NEXT: vsubpd {{.*}}(%rip), %xmm1, %xmm1
; HAS-FMA-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; HAS-FMA-NEXT: retq
entry: entry:
%sub = call <3 x double> @llvm.experimental.constrained.fsub.v3f64( %sub = call <3 x double> @llvm.experimental.constrained.fsub.v3f64(
<3 x double> <double 0xFFEFFFFFFFFFFFFF, double 0xFFEFFFFFFFFFFFFF, <3 x double> <double 0xFFEFFFFFFFFFFFFF, double 0xFFEFFFFFFFFFFFFF,
double 0xFFEFFFFFFFFFFFFF>, double 0xFFEFFFFFFFFFFFFF>,
<3 x double> <double 2.0, double 1.0, double 0.0>, <3 x double> <double 2.0, double 1.0, double 0.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x double> %sub ret <3 x double> %sub
} }
define <4 x double> @constrained_vector_fsub_v4f64() { define <4 x double> @constrained_vector_fsub_v4f64() {
; NO-FMA-LABEL: constrained_vector_fsub_v4f64: ; CHECK-LABEL: constrained_vector_fsub_v4f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: movapd {{.*#+}} xmm1 = [-1.7976931348623157E+308,-1.7976931348623157E+308] ; CHECK-NEXT: movapd {{.*#+}} xmm1 = [-1.7976931348623157E+308,-1.7976931348623157E+308]
; NO-FMA-NEXT: movapd %xmm1, %xmm0 ; CHECK-NEXT: movapd %xmm1, %xmm0
; NO-FMA-NEXT: subpd {{.*}}(%rip), %xmm0 ; CHECK-NEXT: subpd {{.*}}(%rip), %xmm0
; NO-FMA-NEXT: subpd {{.*}}(%rip), %xmm1 ; CHECK-NEXT: subpd {{.*}}(%rip), %xmm1
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_fsub_v4f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vmovapd {{.*#+}} ymm0 = [-1.7976931348623157E+308,-1.7976931348623157E+308,-1.7976931348623157E+308,-1.7976931348623157E+308]
; HAS-FMA-NEXT: vsubpd {{.*}}(%rip), %ymm0, %ymm0
; HAS-FMA-NEXT: retq
entry: entry:
%sub = call <4 x double> @llvm.experimental.constrained.fsub.v4f64( %sub = call <4 x double> @llvm.experimental.constrained.fsub.v4f64(
<4 x double> <double 0xFFEFFFFFFFFFFFFF, double 0xFFEFFFFFFFFFFFFF, <4 x double> <double 0xFFEFFFFFFFFFFFFF, double 0xFFEFFFFFFFFFFFFF,
double 0xFFEFFFFFFFFFFFFF, double 0xFFEFFFFFFFFFFFFF>, double 0xFFEFFFFFFFFFFFFF, double 0xFFEFFFFFFFFFFFFF>,
<4 x double> <double 1.000000e+00, double 1.000000e-01, <4 x double> <double 1.000000e+00, double 1.000000e-01,
double 2.000000e+00, double 2.000000e-01>, double 2.000000e+00, double 2.000000e-01>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <4 x double> %sub ret <4 x double> %sub
} }
define <1 x float> @constrained_vector_fma_v1f32() {
; NO-FMA-LABEL: constrained_vector_fma_v1f32:
; NO-FMA: # %bb.0: # %entry
; NO-FMA-NEXT: pushq %rax
; NO-FMA-NEXT: .cfi_def_cfa_offset 16
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq fmaf
; NO-FMA-NEXT: popq %rax
; NO-FMA-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_fma_v1f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vfmadd213ss {{.*#+}} xmm0 = (xmm1 * xmm0) + mem
; HAS-FMA-NEXT: retq
entry:
%fma = call <1 x float> @llvm.experimental.constrained.fma.v1f32(
<1 x float> <float 0.5>,
<1 x float> <float 2.5>,
<1 x float> <float 4.5>,
metadata !"round.dynamic",
metadata !"fpexcept.strict")
ret <1 x float> %fma
}
define <2 x double> @constrained_vector_fma_v2f64() {
; NO-FMA-LABEL: constrained_vector_fma_v2f64:
; NO-FMA: # %bb.0: # %entry
; NO-FMA-NEXT: subq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 32
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm2 = mem[0],zero
; NO-FMA-NEXT: callq fma
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm2 = mem[0],zero
; NO-FMA-NEXT: callq fma
; NO-FMA-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; NO-FMA-NEXT: addq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_fma_v2f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vmovapd {{.*#+}} xmm1 = [1.5E+0,5.0E-1]
; HAS-FMA-NEXT: vmovapd {{.*#+}} xmm0 = [3.5E+0,2.5E+0]
; HAS-FMA-NEXT: vfmadd213pd {{.*#+}} xmm0 = (xmm1 * xmm0) + mem
; HAS-FMA-NEXT: retq
entry:
%fma = call <2 x double> @llvm.experimental.constrained.fma.v2f64(
<2 x double> <double 1.5, double 0.5>,
<2 x double> <double 3.5, double 2.5>,
<2 x double> <double 5.5, double 4.5>,
metadata !"round.dynamic",
metadata !"fpexcept.strict")
ret <2 x double> %fma
}
define <3 x float> @constrained_vector_fma_v3f32() {
; NO-FMA-LABEL: constrained_vector_fma_v3f32:
; NO-FMA: # %bb.0: # %entry
; NO-FMA-NEXT: subq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 48
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq fmaf
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq fmaf
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq fmaf
; NO-FMA-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload
; NO-FMA-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; NO-FMA-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm1 = xmm1[0],mem[0]
; NO-FMA-NEXT: movaps %xmm1, %xmm0
; NO-FMA-NEXT: addq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_fma_v3f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vfmadd213ss {{.*#+}} xmm1 = (xmm0 * xmm1) + mem
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vfmadd213ss {{.*#+}} xmm2 = (xmm0 * xmm2) + mem
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm3 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vfmadd213ss {{.*#+}} xmm3 = (xmm0 * xmm3) + mem
; HAS-FMA-NEXT: vinsertps {{.*#+}} xmm0 = xmm2[0],xmm3[0],xmm2[2,3]
; HAS-FMA-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0,1],xmm1[0],xmm0[3]
; HAS-FMA-NEXT: retq
entry:
%fma = call <3 x float> @llvm.experimental.constrained.fma.v3f32(
<3 x float> <float 2.5, float 1.5, float 0.5>,
<3 x float> <float 5.5, float 4.5, float 3.5>,
<3 x float> <float 8.5, float 7.5, float 6.5>,
metadata !"round.dynamic",
metadata !"fpexcept.strict")
ret <3 x float> %fma
}
define <3 x double> @constrained_vector_fma_v3f64() {
; NO-FMA-LABEL: constrained_vector_fma_v3f64:
; NO-FMA: # %bb.0: # %entry
; NO-FMA-NEXT: subq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 32
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm2 = mem[0],zero
; NO-FMA-NEXT: callq fma
; NO-FMA-NEXT: movsd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm2 = mem[0],zero
; NO-FMA-NEXT: callq fma
; NO-FMA-NEXT: movsd %xmm0, (%rsp) # 8-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm2 = mem[0],zero
; NO-FMA-NEXT: callq fma
; NO-FMA-NEXT: movsd %xmm0, {{[0-9]+}}(%rsp)
; NO-FMA-NEXT: fldl {{[0-9]+}}(%rsp)
; NO-FMA-NEXT: movsd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 8-byte Reload
; NO-FMA-NEXT: # xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd (%rsp), %xmm1 # 8-byte Reload
; NO-FMA-NEXT: # xmm1 = mem[0],zero
; NO-FMA-NEXT: addq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_fma_v3f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: vfmadd213sd {{.*#+}} xmm1 = (xmm0 * xmm1) + mem
; HAS-FMA-NEXT: vmovapd {{.*#+}} xmm0 = [2.5E+0,1.5E+0]
; HAS-FMA-NEXT: vmovapd {{.*#+}} xmm2 = [5.5E+0,4.5E+0]
; HAS-FMA-NEXT: vfmadd213pd {{.*#+}} xmm2 = (xmm0 * xmm2) + mem
; HAS-FMA-NEXT: vinsertf128 $1, %xmm1, %ymm2, %ymm0
; HAS-FMA-NEXT: retq
entry:
%fma = call <3 x double> @llvm.experimental.constrained.fma.v3f64(
<3 x double> <double 2.5, double 1.5, double 0.5>,
<3 x double> <double 5.5, double 4.5, double 3.5>,
<3 x double> <double 8.5, double 7.5, double 6.5>,
metadata !"round.dynamic",
metadata !"fpexcept.strict")
ret <3 x double> %fma
}
define <4 x double> @constrained_vector_fma_v4f64() {
; NO-FMA-LABEL: constrained_vector_fma_v4f64:
; NO-FMA: # %bb.0: # %entry
; NO-FMA-NEXT: subq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 48
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm2 = mem[0],zero
; NO-FMA-NEXT: callq fma
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm2 = mem[0],zero
; NO-FMA-NEXT: callq fma
; NO-FMA-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm2 = mem[0],zero
; NO-FMA-NEXT: callq fma
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm2 = mem[0],zero
; NO-FMA-NEXT: callq fma
; NO-FMA-NEXT: movaps %xmm0, %xmm1
; NO-FMA-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm1 = xmm1[0],mem[0]
; NO-FMA-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; NO-FMA-NEXT: addq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_fma_v4f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vmovapd {{.*#+}} ymm1 = [3.5E+0,2.5E+0,1.5E+0,5.0E-1]
; HAS-FMA-NEXT: vmovapd {{.*#+}} ymm0 = [7.5E+0,6.5E+0,5.5E+0,4.5E+0]
; HAS-FMA-NEXT: vfmadd213pd {{.*#+}} ymm0 = (ymm1 * ymm0) + mem
; HAS-FMA-NEXT: retq
entry:
%fma = call <4 x double> @llvm.experimental.constrained.fma.v4f64(
<4 x double> <double 3.5, double 2.5, double 1.5, double 0.5>,
<4 x double> <double 7.5, double 6.5, double 5.5, double 4.5>,
<4 x double> <double 11.5, double 10.5, double 9.5, double 8.5>,
metadata !"round.dynamic",
metadata !"fpexcept.strict")
ret <4 x double> %fma
}
define <4 x float> @constrained_vector_fma_v4f32() {
; NO-FMA-LABEL: constrained_vector_fma_v4f32:
; NO-FMA: # %bb.0: # %entry
; NO-FMA-NEXT: subq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 48
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq fmaf
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq fmaf
; NO-FMA-NEXT: unpcklps (%rsp), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq fmaf
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq fmaf
; NO-FMA-NEXT: unpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; NO-FMA-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; NO-FMA-NEXT: addq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_fma_v4f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vmovaps {{.*#+}} xmm1 = [3.5E+0,2.5E+0,1.5E+0,5.0E-1]
; HAS-FMA-NEXT: vmovaps {{.*#+}} xmm0 = [7.5E+0,6.5E+0,5.5E+0,4.5E+0]
; HAS-FMA-NEXT: vfmadd213ps {{.*#+}} xmm0 = (xmm1 * xmm0) + mem
; HAS-FMA-NEXT: retq
entry:
%fma = call <4 x float> @llvm.experimental.constrained.fma.v4f32(
<4 x float> <float 3.5, float 2.5, float 1.5, float 0.5>,
<4 x float> <float 7.5, float 6.5, float 5.5, float 4.5>,
<4 x float> <float 11.5, float 10.5, float 9.5, float 8.5>,
metadata !"round.dynamic",
metadata !"fpexcept.strict")
ret <4 x float> %fma
}
define <8 x float> @constrained_vector_fma_v8f32() {
; NO-FMA-LABEL: constrained_vector_fma_v8f32:
; NO-FMA: # %bb.0: # %entry
; NO-FMA-NEXT: subq $56, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 64
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq fmaf
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq fmaf
; NO-FMA-NEXT: unpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq fmaf
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq fmaf
; NO-FMA-NEXT: unpcklps (%rsp), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; NO-FMA-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq fmaf
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq fmaf
; NO-FMA-NEXT: unpcklps (%rsp), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq fmaf
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq fmaf
; NO-FMA-NEXT: movaps %xmm0, %xmm1
; NO-FMA-NEXT: unpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
; NO-FMA-NEXT: unpcklpd (%rsp), %xmm1 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm1 = xmm1[0],mem[0]
; NO-FMA-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; NO-FMA-NEXT: addq $56, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_fma_v8f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vmovaps {{.*#+}} ymm1 = [3.5E+0,2.5E+0,1.5E+0,5.0E-1,7.5E+0,6.5E+0,5.5E+0,4.5E+0]
; HAS-FMA-NEXT: vmovaps {{.*#+}} ymm0 = [7.5E+0,6.5E+0,5.5E+0,4.5E+0,1.15E+1,1.05E+1,9.5E+0,8.5E+0]
; HAS-FMA-NEXT: vfmadd213ps {{.*#+}} ymm0 = (ymm1 * ymm0) + mem
; HAS-FMA-NEXT: retq
entry:
%fma = call <8 x float> @llvm.experimental.constrained.fma.v8f32(
<8 x float> <float 3.5, float 2.5, float 1.5, float 0.5,
float 7.5, float 6.5, float 5.5, float 4.5>,
<8 x float> <float 7.5, float 6.5, float 5.5, float 4.5,
float 11.5, float 10.5, float 9.5, float 8.5>,
<8 x float> <float 11.5, float 10.5, float 9.5, float 8.5,
float 15.5, float 14.5, float 13.5, float 12.5>,
metadata !"round.dynamic",
metadata !"fpexcept.strict")
ret <8 x float> %fma
}
define <1 x float> @constrained_vector_sqrt_v1f32() { define <1 x float> @constrained_vector_sqrt_v1f32() {
; NO-FMA-LABEL: constrained_vector_sqrt_v1f32: ; CHECK-LABEL: constrained_vector_sqrt_v1f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: sqrtss %xmm0, %xmm0 ; CHECK-NEXT: sqrtss %xmm0, %xmm0
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_sqrt_v1f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vsqrtss %xmm0, %xmm0, %xmm0
; HAS-FMA-NEXT: retq
entry: entry:
%sqrt = call <1 x float> @llvm.experimental.constrained.sqrt.v1f32( %sqrt = call <1 x float> @llvm.experimental.constrained.sqrt.v1f32(
<1 x float> <float 42.0>, <1 x float> <float 42.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <1 x float> %sqrt ret <1 x float> %sqrt
} }
define <2 x double> @constrained_vector_sqrt_v2f64() { define <2 x double> @constrained_vector_sqrt_v2f64() {
; NO-FMA-LABEL: constrained_vector_sqrt_v2f64: ; CHECK-LABEL: constrained_vector_sqrt_v2f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: sqrtpd {{.*}}(%rip), %xmm0 ; CHECK-NEXT: sqrtpd {{.*}}(%rip), %xmm0
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_sqrt_v2f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vsqrtpd {{.*}}(%rip), %xmm0
; HAS-FMA-NEXT: retq
entry: entry:
%sqrt = call <2 x double> @llvm.experimental.constrained.sqrt.v2f64( %sqrt = call <2 x double> @llvm.experimental.constrained.sqrt.v2f64(
<2 x double> <double 42.0, double 42.1>, <2 x double> <double 42.0, double 42.1>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <2 x double> %sqrt ret <2 x double> %sqrt
} }
define <3 x float> @constrained_vector_sqrt_v3f32() { define <3 x float> @constrained_vector_sqrt_v3f32() {
; NO-FMA-LABEL: constrained_vector_sqrt_v3f32: ; CHECK-LABEL: constrained_vector_sqrt_v3f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: sqrtss %xmm0, %xmm1 ; CHECK-NEXT: sqrtss %xmm0, %xmm1
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: sqrtss %xmm0, %xmm0 ; CHECK-NEXT: sqrtss %xmm0, %xmm0
; NO-FMA-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; NO-FMA-NEXT: sqrtss %xmm2, %xmm2 ; CHECK-NEXT: sqrtss %xmm2, %xmm2
; NO-FMA-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1] ; CHECK-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
; NO-FMA-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0] ; CHECK-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_sqrt_v3f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vsqrtss %xmm0, %xmm0, %xmm0
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vsqrtss %xmm1, %xmm1, %xmm1
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vsqrtss %xmm2, %xmm2, %xmm2
; HAS-FMA-NEXT: vinsertps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[2,3]
; HAS-FMA-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0,1],xmm0[0],xmm1[3]
; HAS-FMA-NEXT: retq
entry: entry:
%sqrt = call <3 x float> @llvm.experimental.constrained.sqrt.v3f32( %sqrt = call <3 x float> @llvm.experimental.constrained.sqrt.v3f32(
<3 x float> <float 42.0, float 43.0, float 44.0>, <3 x float> <float 42.0, float 43.0, float 44.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x float> %sqrt ret <3 x float> %sqrt
} }
define <3 x double> @constrained_vector_sqrt_v3f64() { define <3 x double> @constrained_vector_sqrt_v3f64() {
; NO-FMA-LABEL: constrained_vector_sqrt_v3f64: ; CHECK-LABEL: constrained_vector_sqrt_v3f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: sqrtsd %xmm0, %xmm1 ; CHECK-NEXT: sqrtsd %xmm0, %xmm1
; NO-FMA-NEXT: sqrtpd {{.*}}(%rip), %xmm0 ; CHECK-NEXT: sqrtpd {{.*}}(%rip), %xmm0
; NO-FMA-NEXT: movsd %xmm1, -{{[0-9]+}}(%rsp) ; CHECK-NEXT: movsd %xmm1, -{{[0-9]+}}(%rsp)
; NO-FMA-NEXT: movapd %xmm0, %xmm1 ; CHECK-NEXT: movapd %xmm0, %xmm1
; NO-FMA-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1] ; CHECK-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; NO-FMA-NEXT: fldl -{{[0-9]+}}(%rsp) ; CHECK-NEXT: fldl -{{[0-9]+}}(%rsp)
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_sqrt_v3f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vsqrtsd %xmm0, %xmm0, %xmm0
; HAS-FMA-NEXT: vsqrtpd {{.*}}(%rip), %xmm1
; HAS-FMA-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; HAS-FMA-NEXT: retq
entry: entry:
%sqrt = call <3 x double> @llvm.experimental.constrained.sqrt.v3f64( %sqrt = call <3 x double> @llvm.experimental.constrained.sqrt.v3f64(
<3 x double> <double 42.0, double 42.1, double 42.2>, <3 x double> <double 42.0, double 42.1, double 42.2>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x double> %sqrt ret <3 x double> %sqrt
} }
define <4 x double> @constrained_vector_sqrt_v4f64() { define <4 x double> @constrained_vector_sqrt_v4f64() {
; NO-FMA-LABEL: constrained_vector_sqrt_v4f64: ; CHECK-LABEL: constrained_vector_sqrt_v4f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: sqrtpd {{.*}}(%rip), %xmm0 ; CHECK-NEXT: sqrtpd {{.*}}(%rip), %xmm0
; NO-FMA-NEXT: sqrtpd {{.*}}(%rip), %xmm1 ; CHECK-NEXT: sqrtpd {{.*}}(%rip), %xmm1
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_sqrt_v4f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vsqrtpd {{.*}}(%rip), %ymm0
; HAS-FMA-NEXT: retq
entry: entry:
%sqrt = call <4 x double> @llvm.experimental.constrained.sqrt.v4f64( %sqrt = call <4 x double> @llvm.experimental.constrained.sqrt.v4f64(
<4 x double> <double 42.0, double 42.1, <4 x double> <double 42.0, double 42.1,
double 42.2, double 42.3>, double 42.2, double 42.3>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <4 x double> %sqrt ret <4 x double> %sqrt
} }
define <1 x float> @constrained_vector_pow_v1f32() { define <1 x float> @constrained_vector_pow_v1f32() {
; NO-FMA-LABEL: constrained_vector_pow_v1f32: ; CHECK-LABEL: constrained_vector_pow_v1f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: pushq %rax ; CHECK-NEXT: pushq %rax
; NO-FMA-NEXT: .cfi_def_cfa_offset 16 ; CHECK-NEXT: .cfi_def_cfa_offset 16
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq powf ; CHECK-NEXT: callq powf
; NO-FMA-NEXT: popq %rax ; CHECK-NEXT: popq %rax
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_pow_v1f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: pushq %rax
; HAS-FMA-NEXT: .cfi_def_cfa_offset 16
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq powf
; HAS-FMA-NEXT: popq %rax
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%pow = call <1 x float> @llvm.experimental.constrained.pow.v1f32( %pow = call <1 x float> @llvm.experimental.constrained.pow.v1f32(
<1 x float> <float 42.0>, <1 x float> <float 42.0>,
<1 x float> <float 3.0>, <1 x float> <float 3.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <1 x float> %pow ret <1 x float> %pow
} }
define <2 x double> @constrained_vector_pow_v2f64() { define <2 x double> @constrained_vector_pow_v2f64() {
; NO-FMA-LABEL: constrained_vector_pow_v2f64: ; CHECK-LABEL: constrained_vector_pow_v2f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $24, %rsp ; CHECK-NEXT: subq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 32 ; CHECK-NEXT: .cfi_def_cfa_offset 32
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq pow ; CHECK-NEXT: callq pow
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq pow ; CHECK-NEXT: callq pow
; NO-FMA-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0] ; CHECK-NEXT: # xmm0 = xmm0[0],mem[0]
; NO-FMA-NEXT: addq $24, %rsp ; CHECK-NEXT: addq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_pow_v2f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $24, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 32
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: callq pow
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: callq pow
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: addq $24, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%pow = call <2 x double> @llvm.experimental.constrained.pow.v2f64( %pow = call <2 x double> @llvm.experimental.constrained.pow.v2f64(
<2 x double> <double 42.1, double 42.2>, <2 x double> <double 42.1, double 42.2>,
<2 x double> <double 3.0, double 3.0>, <2 x double> <double 3.0, double 3.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <2 x double> %pow ret <2 x double> %pow
} }
define <3 x float> @constrained_vector_pow_v3f32() { define <3 x float> @constrained_vector_pow_v3f32() {
; NO-FMA-LABEL: constrained_vector_pow_v3f32: ; CHECK-LABEL: constrained_vector_pow_v3f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $40, %rsp ; CHECK-NEXT: subq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 48 ; CHECK-NEXT: .cfi_def_cfa_offset 48
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq powf ; CHECK-NEXT: callq powf
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq powf ; CHECK-NEXT: callq powf
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq powf ; CHECK-NEXT: callq powf
; NO-FMA-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload ; CHECK-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload
; NO-FMA-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; CHECK-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; NO-FMA-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm1 = xmm1[0],mem[0] ; CHECK-NEXT: # xmm1 = xmm1[0],mem[0]
; NO-FMA-NEXT: movaps %xmm1, %xmm0 ; CHECK-NEXT: movaps %xmm1, %xmm0
; NO-FMA-NEXT: addq $40, %rsp ; CHECK-NEXT: addq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_pow_v3f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 48
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq powf
; HAS-FMA-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq powf
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq powf
; HAS-FMA-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload
; HAS-FMA-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[2,3]
; HAS-FMA-NEXT: vinsertps $32, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0,1],mem[0],xmm0[3]
; HAS-FMA-NEXT: addq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%pow = call <3 x float> @llvm.experimental.constrained.pow.v3f32( %pow = call <3 x float> @llvm.experimental.constrained.pow.v3f32(
<3 x float> <float 42.0, float 43.0, float 44.0>, <3 x float> <float 42.0, float 43.0, float 44.0>,
<3 x float> <float 3.0, float 3.0, float 3.0>, <3 x float> <float 3.0, float 3.0, float 3.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x float> %pow ret <3 x float> %pow
} }
define <3 x double> @constrained_vector_pow_v3f64() { define <3 x double> @constrained_vector_pow_v3f64() {
; NO-FMA-LABEL: constrained_vector_pow_v3f64: ; CHECK-LABEL: constrained_vector_pow_v3f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $24, %rsp ; CHECK-NEXT: subq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 32 ; CHECK-NEXT: .cfi_def_cfa_offset 32
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq pow ; CHECK-NEXT: callq pow
; NO-FMA-NEXT: movsd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill ; CHECK-NEXT: movsd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq pow ; CHECK-NEXT: callq pow
; NO-FMA-NEXT: movsd %xmm0, (%rsp) # 8-byte Spill ; CHECK-NEXT: movsd %xmm0, (%rsp) # 8-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq pow ; CHECK-NEXT: callq pow
; NO-FMA-NEXT: movsd %xmm0, {{[0-9]+}}(%rsp) ; CHECK-NEXT: movsd %xmm0, {{[0-9]+}}(%rsp)
; NO-FMA-NEXT: fldl {{[0-9]+}}(%rsp) ; CHECK-NEXT: fldl {{[0-9]+}}(%rsp)
; NO-FMA-NEXT: movsd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 8-byte Reload ; CHECK-NEXT: movsd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 8-byte Reload
; NO-FMA-NEXT: # xmm0 = mem[0],zero ; CHECK-NEXT: # xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd (%rsp), %xmm1 # 8-byte Reload ; CHECK-NEXT: movsd (%rsp), %xmm1 # 8-byte Reload
; NO-FMA-NEXT: # xmm1 = mem[0],zero ; CHECK-NEXT: # xmm1 = mem[0],zero
; NO-FMA-NEXT: addq $24, %rsp ; CHECK-NEXT: addq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_pow_v3f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $56, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 64
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: callq pow
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: callq pow
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: vzeroupper
; HAS-FMA-NEXT: callq pow
; HAS-FMA-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
; HAS-FMA-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; HAS-FMA-NEXT: addq $56, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%pow = call <3 x double> @llvm.experimental.constrained.pow.v3f64( %pow = call <3 x double> @llvm.experimental.constrained.pow.v3f64(
<3 x double> <double 42.0, double 42.1, double 42.2>, <3 x double> <double 42.0, double 42.1, double 42.2>,
<3 x double> <double 3.0, double 3.0, double 3.0>, <3 x double> <double 3.0, double 3.0, double 3.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x double> %pow ret <3 x double> %pow
} }
define <4 x double> @constrained_vector_pow_v4f64() { define <4 x double> @constrained_vector_pow_v4f64() {
; NO-FMA-LABEL: constrained_vector_pow_v4f64: ; CHECK-LABEL: constrained_vector_pow_v4f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $40, %rsp ; CHECK-NEXT: subq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 48 ; CHECK-NEXT: .cfi_def_cfa_offset 48
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq pow ; CHECK-NEXT: callq pow
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq pow ; CHECK-NEXT: callq pow
; NO-FMA-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0] ; CHECK-NEXT: # xmm0 = xmm0[0],mem[0]
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq pow ; CHECK-NEXT: callq pow
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq pow ; CHECK-NEXT: callq pow
; NO-FMA-NEXT: movaps %xmm0, %xmm1 ; CHECK-NEXT: movaps %xmm0, %xmm1
; NO-FMA-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm1 = xmm1[0],mem[0] ; CHECK-NEXT: # xmm1 = xmm1[0],mem[0]
; NO-FMA-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload ; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; NO-FMA-NEXT: addq $40, %rsp ; CHECK-NEXT: addq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_pow_v4f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 48
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: callq pow
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: callq pow
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: callq pow
; HAS-FMA-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: callq pow
; HAS-FMA-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vinsertf128 $1, (%rsp), %ymm0, %ymm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: addq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%pow = call <4 x double> @llvm.experimental.constrained.pow.v4f64( %pow = call <4 x double> @llvm.experimental.constrained.pow.v4f64(
<4 x double> <double 42.1, double 42.2, <4 x double> <double 42.1, double 42.2,
double 42.3, double 42.4>, double 42.3, double 42.4>,
<4 x double> <double 3.0, double 3.0, <4 x double> <double 3.0, double 3.0,
double 3.0, double 3.0>, double 3.0, double 3.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <4 x double> %pow ret <4 x double> %pow
} }
define <1 x float> @constrained_vector_powi_v1f32() { define <1 x float> @constrained_vector_powi_v1f32() {
; NO-FMA-LABEL: constrained_vector_powi_v1f32: ; CHECK-LABEL: constrained_vector_powi_v1f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: pushq %rax ; CHECK-NEXT: pushq %rax
; NO-FMA-NEXT: .cfi_def_cfa_offset 16 ; CHECK-NEXT: .cfi_def_cfa_offset 16
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movl $3, %edi ; CHECK-NEXT: movl $3, %edi
; NO-FMA-NEXT: callq __powisf2 ; CHECK-NEXT: callq __powisf2
; NO-FMA-NEXT: popq %rax ; CHECK-NEXT: popq %rax
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_powi_v1f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: pushq %rax
; HAS-FMA-NEXT: .cfi_def_cfa_offset 16
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: movl $3, %edi
; HAS-FMA-NEXT: callq __powisf2
; HAS-FMA-NEXT: popq %rax
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%powi = call <1 x float> @llvm.experimental.constrained.powi.v1f32( %powi = call <1 x float> @llvm.experimental.constrained.powi.v1f32(
<1 x float> <float 42.0>, <1 x float> <float 42.0>,
i32 3, i32 3,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <1 x float> %powi ret <1 x float> %powi
} }
define <2 x double> @constrained_vector_powi_v2f64() { define <2 x double> @constrained_vector_powi_v2f64() {
; NO-FMA-LABEL: constrained_vector_powi_v2f64: ; CHECK-LABEL: constrained_vector_powi_v2f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $24, %rsp ; CHECK-NEXT: subq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 32 ; CHECK-NEXT: .cfi_def_cfa_offset 32
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movl $3, %edi ; CHECK-NEXT: movl $3, %edi
; NO-FMA-NEXT: callq __powidf2 ; CHECK-NEXT: callq __powidf2
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movl $3, %edi ; CHECK-NEXT: movl $3, %edi
; NO-FMA-NEXT: callq __powidf2 ; CHECK-NEXT: callq __powidf2
; NO-FMA-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0] ; CHECK-NEXT: # xmm0 = xmm0[0],mem[0]
; NO-FMA-NEXT: addq $24, %rsp ; CHECK-NEXT: addq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_powi_v2f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $24, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 32
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: movl $3, %edi
; HAS-FMA-NEXT: callq __powidf2
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: movl $3, %edi
; HAS-FMA-NEXT: callq __powidf2
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: addq $24, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%powi = call <2 x double> @llvm.experimental.constrained.powi.v2f64( %powi = call <2 x double> @llvm.experimental.constrained.powi.v2f64(
<2 x double> <double 42.1, double 42.2>, <2 x double> <double 42.1, double 42.2>,
i32 3, i32 3,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <2 x double> %powi ret <2 x double> %powi
} }
define <3 x float> @constrained_vector_powi_v3f32() { define <3 x float> @constrained_vector_powi_v3f32() {
; NO-FMA-LABEL: constrained_vector_powi_v3f32: ; CHECK-LABEL: constrained_vector_powi_v3f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $40, %rsp ; CHECK-NEXT: subq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 48 ; CHECK-NEXT: .cfi_def_cfa_offset 48
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movl $3, %edi ; CHECK-NEXT: movl $3, %edi
; NO-FMA-NEXT: callq __powisf2 ; CHECK-NEXT: callq __powisf2
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movl $3, %edi ; CHECK-NEXT: movl $3, %edi
; NO-FMA-NEXT: callq __powisf2 ; CHECK-NEXT: callq __powisf2
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movl $3, %edi ; CHECK-NEXT: movl $3, %edi
; NO-FMA-NEXT: callq __powisf2 ; CHECK-NEXT: callq __powisf2
; NO-FMA-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload ; CHECK-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload
; NO-FMA-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; CHECK-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; NO-FMA-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm1 = xmm1[0],mem[0] ; CHECK-NEXT: # xmm1 = xmm1[0],mem[0]
; NO-FMA-NEXT: movaps %xmm1, %xmm0 ; CHECK-NEXT: movaps %xmm1, %xmm0
; NO-FMA-NEXT: addq $40, %rsp ; CHECK-NEXT: addq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_powi_v3f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 48
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: movl $3, %edi
; HAS-FMA-NEXT: callq __powisf2
; HAS-FMA-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: movl $3, %edi
; HAS-FMA-NEXT: callq __powisf2
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: movl $3, %edi
; HAS-FMA-NEXT: callq __powisf2
; HAS-FMA-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload
; HAS-FMA-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[2,3]
; HAS-FMA-NEXT: vinsertps $32, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0,1],mem[0],xmm0[3]
; HAS-FMA-NEXT: addq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%powi = call <3 x float> @llvm.experimental.constrained.powi.v3f32( %powi = call <3 x float> @llvm.experimental.constrained.powi.v3f32(
<3 x float> <float 42.0, float 43.0, float 44.0>, <3 x float> <float 42.0, float 43.0, float 44.0>,
i32 3, i32 3,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x float> %powi ret <3 x float> %powi
} }
define <3 x double> @constrained_vector_powi_v3f64() { define <3 x double> @constrained_vector_powi_v3f64() {
; NO-FMA-LABEL: constrained_vector_powi_v3f64: ; CHECK-LABEL: constrained_vector_powi_v3f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $24, %rsp ; CHECK-NEXT: subq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 32 ; CHECK-NEXT: .cfi_def_cfa_offset 32
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movl $3, %edi ; CHECK-NEXT: movl $3, %edi
; NO-FMA-NEXT: callq __powidf2 ; CHECK-NEXT: callq __powidf2
; NO-FMA-NEXT: movsd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill ; CHECK-NEXT: movsd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movl $3, %edi ; CHECK-NEXT: movl $3, %edi
; NO-FMA-NEXT: callq __powidf2 ; CHECK-NEXT: callq __powidf2
; NO-FMA-NEXT: movsd %xmm0, (%rsp) # 8-byte Spill ; CHECK-NEXT: movsd %xmm0, (%rsp) # 8-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movl $3, %edi ; CHECK-NEXT: movl $3, %edi
; NO-FMA-NEXT: callq __powidf2 ; CHECK-NEXT: callq __powidf2
; NO-FMA-NEXT: movsd %xmm0, {{[0-9]+}}(%rsp) ; CHECK-NEXT: movsd %xmm0, {{[0-9]+}}(%rsp)
; NO-FMA-NEXT: fldl {{[0-9]+}}(%rsp) ; CHECK-NEXT: fldl {{[0-9]+}}(%rsp)
; NO-FMA-NEXT: movsd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 8-byte Reload ; CHECK-NEXT: movsd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 8-byte Reload
; NO-FMA-NEXT: # xmm0 = mem[0],zero ; CHECK-NEXT: # xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd (%rsp), %xmm1 # 8-byte Reload ; CHECK-NEXT: movsd (%rsp), %xmm1 # 8-byte Reload
; NO-FMA-NEXT: # xmm1 = mem[0],zero ; CHECK-NEXT: # xmm1 = mem[0],zero
; NO-FMA-NEXT: addq $24, %rsp ; CHECK-NEXT: addq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_powi_v3f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $56, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 64
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: movl $3, %edi
; HAS-FMA-NEXT: callq __powidf2
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: movl $3, %edi
; HAS-FMA-NEXT: callq __powidf2
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: movl $3, %edi
; HAS-FMA-NEXT: vzeroupper
; HAS-FMA-NEXT: callq __powidf2
; HAS-FMA-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
; HAS-FMA-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; HAS-FMA-NEXT: addq $56, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%powi = call <3 x double> @llvm.experimental.constrained.powi.v3f64( %powi = call <3 x double> @llvm.experimental.constrained.powi.v3f64(
<3 x double> <double 42.0, double 42.1, double 42.2>, <3 x double> <double 42.0, double 42.1, double 42.2>,
i32 3, i32 3,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x double> %powi ret <3 x double> %powi
} }
define <4 x double> @constrained_vector_powi_v4f64() { define <4 x double> @constrained_vector_powi_v4f64() {
; NO-FMA-LABEL: constrained_vector_powi_v4f64: ; CHECK-LABEL: constrained_vector_powi_v4f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $40, %rsp ; CHECK-NEXT: subq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 48 ; CHECK-NEXT: .cfi_def_cfa_offset 48
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movl $3, %edi ; CHECK-NEXT: movl $3, %edi
; NO-FMA-NEXT: callq __powidf2 ; CHECK-NEXT: callq __powidf2
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movl $3, %edi ; CHECK-NEXT: movl $3, %edi
; NO-FMA-NEXT: callq __powidf2 ; CHECK-NEXT: callq __powidf2
; NO-FMA-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0] ; CHECK-NEXT: # xmm0 = xmm0[0],mem[0]
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movl $3, %edi ; CHECK-NEXT: movl $3, %edi
; NO-FMA-NEXT: callq __powidf2 ; CHECK-NEXT: callq __powidf2
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movl $3, %edi ; CHECK-NEXT: movl $3, %edi
; NO-FMA-NEXT: callq __powidf2 ; CHECK-NEXT: callq __powidf2
; NO-FMA-NEXT: movaps %xmm0, %xmm1 ; CHECK-NEXT: movaps %xmm0, %xmm1
; NO-FMA-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm1 = xmm1[0],mem[0] ; CHECK-NEXT: # xmm1 = xmm1[0],mem[0]
; NO-FMA-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload ; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; NO-FMA-NEXT: addq $40, %rsp ; CHECK-NEXT: addq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_powi_v4f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 48
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: movl $3, %edi
; HAS-FMA-NEXT: callq __powidf2
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: movl $3, %edi
; HAS-FMA-NEXT: callq __powidf2
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: movl $3, %edi
; HAS-FMA-NEXT: callq __powidf2
; HAS-FMA-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: movl $3, %edi
; HAS-FMA-NEXT: callq __powidf2
; HAS-FMA-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vinsertf128 $1, (%rsp), %ymm0, %ymm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: addq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%powi = call <4 x double> @llvm.experimental.constrained.powi.v4f64( %powi = call <4 x double> @llvm.experimental.constrained.powi.v4f64(
<4 x double> <double 42.1, double 42.2, <4 x double> <double 42.1, double 42.2,
double 42.3, double 42.4>, double 42.3, double 42.4>,
i32 3, i32 3,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <4 x double> %powi ret <4 x double> %powi
} }
define <1 x float> @constrained_vector_sin_v1f32() { define <1 x float> @constrained_vector_sin_v1f32() {
; NO-FMA-LABEL: constrained_vector_sin_v1f32: ; CHECK-LABEL: constrained_vector_sin_v1f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: pushq %rax ; CHECK-NEXT: pushq %rax
; NO-FMA-NEXT: .cfi_def_cfa_offset 16 ; CHECK-NEXT: .cfi_def_cfa_offset 16
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq sinf ; CHECK-NEXT: callq sinf
; NO-FMA-NEXT: popq %rax ; CHECK-NEXT: popq %rax
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_sin_v1f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: pushq %rax
; HAS-FMA-NEXT: .cfi_def_cfa_offset 16
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq sinf
; HAS-FMA-NEXT: popq %rax
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%sin = call <1 x float> @llvm.experimental.constrained.sin.v1f32( %sin = call <1 x float> @llvm.experimental.constrained.sin.v1f32(
<1 x float> <float 42.0>, <1 x float> <float 42.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <1 x float> %sin ret <1 x float> %sin
} }
define <2 x double> @constrained_vector_sin_v2f64() { define <2 x double> @constrained_vector_sin_v2f64() {
; NO-FMA-LABEL: constrained_vector_sin_v2f64: ; CHECK-LABEL: constrained_vector_sin_v2f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $24, %rsp ; CHECK-NEXT: subq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 32 ; CHECK-NEXT: .cfi_def_cfa_offset 32
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq sin ; CHECK-NEXT: callq sin
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq sin ; CHECK-NEXT: callq sin
; NO-FMA-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0] ; CHECK-NEXT: # xmm0 = xmm0[0],mem[0]
; NO-FMA-NEXT: addq $24, %rsp ; CHECK-NEXT: addq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_sin_v2f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $24, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 32
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq sin
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq sin
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: addq $24, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%sin = call <2 x double> @llvm.experimental.constrained.sin.v2f64( %sin = call <2 x double> @llvm.experimental.constrained.sin.v2f64(
<2 x double> <double 42.0, double 42.1>, <2 x double> <double 42.0, double 42.1>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <2 x double> %sin ret <2 x double> %sin
} }
define <3 x float> @constrained_vector_sin_v3f32() { define <3 x float> @constrained_vector_sin_v3f32() {
; NO-FMA-LABEL: constrained_vector_sin_v3f32: ; CHECK-LABEL: constrained_vector_sin_v3f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $40, %rsp ; CHECK-NEXT: subq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 48 ; CHECK-NEXT: .cfi_def_cfa_offset 48
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq sinf ; CHECK-NEXT: callq sinf
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq sinf ; CHECK-NEXT: callq sinf
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq sinf ; CHECK-NEXT: callq sinf
; NO-FMA-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload ; CHECK-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload
; NO-FMA-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; CHECK-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; NO-FMA-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm1 = xmm1[0],mem[0] ; CHECK-NEXT: # xmm1 = xmm1[0],mem[0]
; NO-FMA-NEXT: movaps %xmm1, %xmm0 ; CHECK-NEXT: movaps %xmm1, %xmm0
; NO-FMA-NEXT: addq $40, %rsp ; CHECK-NEXT: addq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_sin_v3f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 48
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq sinf
; HAS-FMA-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq sinf
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq sinf
; HAS-FMA-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload
; HAS-FMA-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[2,3]
; HAS-FMA-NEXT: vinsertps $32, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0,1],mem[0],xmm0[3]
; HAS-FMA-NEXT: addq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%sin = call <3 x float> @llvm.experimental.constrained.sin.v3f32( %sin = call <3 x float> @llvm.experimental.constrained.sin.v3f32(
<3 x float> <float 42.0, float 43.0, float 44.0>, <3 x float> <float 42.0, float 43.0, float 44.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x float> %sin ret <3 x float> %sin
} }
define <3 x double> @constrained_vector_sin_v3f64() { define <3 x double> @constrained_vector_sin_v3f64() {
; NO-FMA-LABEL: constrained_vector_sin_v3f64: ; CHECK-LABEL: constrained_vector_sin_v3f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $24, %rsp ; CHECK-NEXT: subq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 32 ; CHECK-NEXT: .cfi_def_cfa_offset 32
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq sin ; CHECK-NEXT: callq sin
; NO-FMA-NEXT: movsd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill ; CHECK-NEXT: movsd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq sin ; CHECK-NEXT: callq sin
; NO-FMA-NEXT: movsd %xmm0, (%rsp) # 8-byte Spill ; CHECK-NEXT: movsd %xmm0, (%rsp) # 8-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq sin ; CHECK-NEXT: callq sin
; NO-FMA-NEXT: movsd %xmm0, {{[0-9]+}}(%rsp) ; CHECK-NEXT: movsd %xmm0, {{[0-9]+}}(%rsp)
; NO-FMA-NEXT: fldl {{[0-9]+}}(%rsp) ; CHECK-NEXT: fldl {{[0-9]+}}(%rsp)
; NO-FMA-NEXT: movsd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 8-byte Reload ; CHECK-NEXT: movsd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 8-byte Reload
; NO-FMA-NEXT: # xmm0 = mem[0],zero ; CHECK-NEXT: # xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd (%rsp), %xmm1 # 8-byte Reload ; CHECK-NEXT: movsd (%rsp), %xmm1 # 8-byte Reload
; NO-FMA-NEXT: # xmm1 = mem[0],zero ; CHECK-NEXT: # xmm1 = mem[0],zero
; NO-FMA-NEXT: addq $24, %rsp ; CHECK-NEXT: addq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_sin_v3f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $56, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 64
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq sin
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq sin
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vzeroupper
; HAS-FMA-NEXT: callq sin
; HAS-FMA-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
; HAS-FMA-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; HAS-FMA-NEXT: addq $56, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%sin = call <3 x double> @llvm.experimental.constrained.sin.v3f64( %sin = call <3 x double> @llvm.experimental.constrained.sin.v3f64(
<3 x double> <double 42.0, double 42.1, double 42.2>, <3 x double> <double 42.0, double 42.1, double 42.2>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x double> %sin ret <3 x double> %sin
} }
define <4 x double> @constrained_vector_sin_v4f64() { define <4 x double> @constrained_vector_sin_v4f64() {
; NO-FMA-LABEL: constrained_vector_sin_v4f64: ; CHECK-LABEL: constrained_vector_sin_v4f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $40, %rsp ; CHECK-NEXT: subq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 48 ; CHECK-NEXT: .cfi_def_cfa_offset 48
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq sin ; CHECK-NEXT: callq sin
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq sin ; CHECK-NEXT: callq sin
; NO-FMA-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0] ; CHECK-NEXT: # xmm0 = xmm0[0],mem[0]
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq sin ; CHECK-NEXT: callq sin
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq sin ; CHECK-NEXT: callq sin
; NO-FMA-NEXT: movaps %xmm0, %xmm1 ; CHECK-NEXT: movaps %xmm0, %xmm1
; NO-FMA-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm1 = xmm1[0],mem[0] ; CHECK-NEXT: # xmm1 = xmm1[0],mem[0]
; NO-FMA-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload ; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; NO-FMA-NEXT: addq $40, %rsp ; CHECK-NEXT: addq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_sin_v4f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 48
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq sin
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq sin
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq sin
; HAS-FMA-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq sin
; HAS-FMA-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vinsertf128 $1, (%rsp), %ymm0, %ymm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: addq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%sin = call <4 x double> @llvm.experimental.constrained.sin.v4f64( %sin = call <4 x double> @llvm.experimental.constrained.sin.v4f64(
<4 x double> <double 42.0, double 42.1, <4 x double> <double 42.0, double 42.1,
double 42.2, double 42.3>, double 42.2, double 42.3>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <4 x double> %sin ret <4 x double> %sin
} }
define <1 x float> @constrained_vector_cos_v1f32() { define <1 x float> @constrained_vector_cos_v1f32() {
; NO-FMA-LABEL: constrained_vector_cos_v1f32: ; CHECK-LABEL: constrained_vector_cos_v1f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: pushq %rax ; CHECK-NEXT: pushq %rax
; NO-FMA-NEXT: .cfi_def_cfa_offset 16 ; CHECK-NEXT: .cfi_def_cfa_offset 16
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq cosf ; CHECK-NEXT: callq cosf
; NO-FMA-NEXT: popq %rax ; CHECK-NEXT: popq %rax
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_cos_v1f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: pushq %rax
; HAS-FMA-NEXT: .cfi_def_cfa_offset 16
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq cosf
; HAS-FMA-NEXT: popq %rax
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%cos = call <1 x float> @llvm.experimental.constrained.cos.v1f32( %cos = call <1 x float> @llvm.experimental.constrained.cos.v1f32(
<1 x float> <float 42.0>, <1 x float> <float 42.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <1 x float> %cos ret <1 x float> %cos
} }
define <2 x double> @constrained_vector_cos_v2f64() { define <2 x double> @constrained_vector_cos_v2f64() {
; NO-FMA-LABEL: constrained_vector_cos_v2f64: ; CHECK-LABEL: constrained_vector_cos_v2f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $24, %rsp ; CHECK-NEXT: subq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 32 ; CHECK-NEXT: .cfi_def_cfa_offset 32
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq cos ; CHECK-NEXT: callq cos
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq cos ; CHECK-NEXT: callq cos
; NO-FMA-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0] ; CHECK-NEXT: # xmm0 = xmm0[0],mem[0]
; NO-FMA-NEXT: addq $24, %rsp ; CHECK-NEXT: addq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_cos_v2f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $24, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 32
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq cos
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq cos
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: addq $24, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%cos = call <2 x double> @llvm.experimental.constrained.cos.v2f64( %cos = call <2 x double> @llvm.experimental.constrained.cos.v2f64(
<2 x double> <double 42.0, double 42.1>, <2 x double> <double 42.0, double 42.1>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <2 x double> %cos ret <2 x double> %cos
} }
define <3 x float> @constrained_vector_cos_v3f32() { define <3 x float> @constrained_vector_cos_v3f32() {
; NO-FMA-LABEL: constrained_vector_cos_v3f32: ; CHECK-LABEL: constrained_vector_cos_v3f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $40, %rsp ; CHECK-NEXT: subq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 48 ; CHECK-NEXT: .cfi_def_cfa_offset 48
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq cosf ; CHECK-NEXT: callq cosf
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq cosf ; CHECK-NEXT: callq cosf
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq cosf ; CHECK-NEXT: callq cosf
; NO-FMA-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload ; CHECK-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload
; NO-FMA-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; CHECK-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; NO-FMA-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm1 = xmm1[0],mem[0] ; CHECK-NEXT: # xmm1 = xmm1[0],mem[0]
; NO-FMA-NEXT: movaps %xmm1, %xmm0 ; CHECK-NEXT: movaps %xmm1, %xmm0
; NO-FMA-NEXT: addq $40, %rsp ; CHECK-NEXT: addq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_cos_v3f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 48
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq cosf
; HAS-FMA-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq cosf
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq cosf
; HAS-FMA-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload
; HAS-FMA-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[2,3]
; HAS-FMA-NEXT: vinsertps $32, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0,1],mem[0],xmm0[3]
; HAS-FMA-NEXT: addq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%cos = call <3 x float> @llvm.experimental.constrained.cos.v3f32( %cos = call <3 x float> @llvm.experimental.constrained.cos.v3f32(
<3 x float> <float 42.0, float 43.0, float 44.0>, <3 x float> <float 42.0, float 43.0, float 44.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x float> %cos ret <3 x float> %cos
} }
define <3 x double> @constrained_vector_cos_v3f64() { define <3 x double> @constrained_vector_cos_v3f64() {
; NO-FMA-LABEL: constrained_vector_cos_v3f64: ; CHECK-LABEL: constrained_vector_cos_v3f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $24, %rsp ; CHECK-NEXT: subq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 32 ; CHECK-NEXT: .cfi_def_cfa_offset 32
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq cos ; CHECK-NEXT: callq cos
; NO-FMA-NEXT: movsd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill ; CHECK-NEXT: movsd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq cos ; CHECK-NEXT: callq cos
; NO-FMA-NEXT: movsd %xmm0, (%rsp) # 8-byte Spill ; CHECK-NEXT: movsd %xmm0, (%rsp) # 8-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq cos ; CHECK-NEXT: callq cos
; NO-FMA-NEXT: movsd %xmm0, {{[0-9]+}}(%rsp) ; CHECK-NEXT: movsd %xmm0, {{[0-9]+}}(%rsp)
; NO-FMA-NEXT: fldl {{[0-9]+}}(%rsp) ; CHECK-NEXT: fldl {{[0-9]+}}(%rsp)
; NO-FMA-NEXT: movsd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 8-byte Reload ; CHECK-NEXT: movsd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 8-byte Reload
; NO-FMA-NEXT: # xmm0 = mem[0],zero ; CHECK-NEXT: # xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd (%rsp), %xmm1 # 8-byte Reload ; CHECK-NEXT: movsd (%rsp), %xmm1 # 8-byte Reload
; NO-FMA-NEXT: # xmm1 = mem[0],zero ; CHECK-NEXT: # xmm1 = mem[0],zero
; NO-FMA-NEXT: addq $24, %rsp ; CHECK-NEXT: addq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_cos_v3f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $56, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 64
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq cos
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq cos
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vzeroupper
; HAS-FMA-NEXT: callq cos
; HAS-FMA-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
; HAS-FMA-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; HAS-FMA-NEXT: addq $56, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%cos = call <3 x double> @llvm.experimental.constrained.cos.v3f64( %cos = call <3 x double> @llvm.experimental.constrained.cos.v3f64(
<3 x double> <double 42.0, double 42.1, double 42.2>, <3 x double> <double 42.0, double 42.1, double 42.2>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x double> %cos ret <3 x double> %cos
} }
define <4 x double> @constrained_vector_cos_v4f64() { define <4 x double> @constrained_vector_cos_v4f64() {
; NO-FMA-LABEL: constrained_vector_cos_v4f64: ; CHECK-LABEL: constrained_vector_cos_v4f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $40, %rsp ; CHECK-NEXT: subq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 48 ; CHECK-NEXT: .cfi_def_cfa_offset 48
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq cos ; CHECK-NEXT: callq cos
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq cos ; CHECK-NEXT: callq cos
; NO-FMA-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0] ; CHECK-NEXT: # xmm0 = xmm0[0],mem[0]
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq cos ; CHECK-NEXT: callq cos
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq cos ; CHECK-NEXT: callq cos
; NO-FMA-NEXT: movaps %xmm0, %xmm1 ; CHECK-NEXT: movaps %xmm0, %xmm1
; NO-FMA-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm1 = xmm1[0],mem[0] ; CHECK-NEXT: # xmm1 = xmm1[0],mem[0]
; NO-FMA-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload ; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; NO-FMA-NEXT: addq $40, %rsp ; CHECK-NEXT: addq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_cos_v4f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 48
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq cos
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq cos
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq cos
; HAS-FMA-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq cos
; HAS-FMA-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vinsertf128 $1, (%rsp), %ymm0, %ymm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: addq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%cos = call <4 x double> @llvm.experimental.constrained.cos.v4f64( %cos = call <4 x double> @llvm.experimental.constrained.cos.v4f64(
<4 x double> <double 42.0, double 42.1, <4 x double> <double 42.0, double 42.1,
double 42.2, double 42.3>, double 42.2, double 42.3>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <4 x double> %cos ret <4 x double> %cos
} }
define <1 x float> @constrained_vector_exp_v1f32() { define <1 x float> @constrained_vector_exp_v1f32() {
; NO-FMA-LABEL: constrained_vector_exp_v1f32: ; CHECK-LABEL: constrained_vector_exp_v1f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: pushq %rax ; CHECK-NEXT: pushq %rax
; NO-FMA-NEXT: .cfi_def_cfa_offset 16 ; CHECK-NEXT: .cfi_def_cfa_offset 16
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq expf ; CHECK-NEXT: callq expf
; NO-FMA-NEXT: popq %rax ; CHECK-NEXT: popq %rax
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_exp_v1f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: pushq %rax
; HAS-FMA-NEXT: .cfi_def_cfa_offset 16
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq expf
; HAS-FMA-NEXT: popq %rax
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%exp = call <1 x float> @llvm.experimental.constrained.exp.v1f32( %exp = call <1 x float> @llvm.experimental.constrained.exp.v1f32(
<1 x float> <float 42.0>, <1 x float> <float 42.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <1 x float> %exp ret <1 x float> %exp
} }
define <2 x double> @constrained_vector_exp_v2f64() { define <2 x double> @constrained_vector_exp_v2f64() {
; NO-FMA-LABEL: constrained_vector_exp_v2f64: ; CHECK-LABEL: constrained_vector_exp_v2f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $24, %rsp ; CHECK-NEXT: subq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 32 ; CHECK-NEXT: .cfi_def_cfa_offset 32
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq exp ; CHECK-NEXT: callq exp
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq exp ; CHECK-NEXT: callq exp
; NO-FMA-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0] ; CHECK-NEXT: # xmm0 = xmm0[0],mem[0]
; NO-FMA-NEXT: addq $24, %rsp ; CHECK-NEXT: addq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_exp_v2f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $24, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 32
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq exp
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq exp
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: addq $24, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%exp = call <2 x double> @llvm.experimental.constrained.exp.v2f64( %exp = call <2 x double> @llvm.experimental.constrained.exp.v2f64(
<2 x double> <double 42.0, double 42.1>, <2 x double> <double 42.0, double 42.1>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <2 x double> %exp ret <2 x double> %exp
} }
define <3 x float> @constrained_vector_exp_v3f32() { define <3 x float> @constrained_vector_exp_v3f32() {
; NO-FMA-LABEL: constrained_vector_exp_v3f32: ; CHECK-LABEL: constrained_vector_exp_v3f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $40, %rsp ; CHECK-NEXT: subq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 48 ; CHECK-NEXT: .cfi_def_cfa_offset 48
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq expf ; CHECK-NEXT: callq expf
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq expf ; CHECK-NEXT: callq expf
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq expf ; CHECK-NEXT: callq expf
; NO-FMA-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload ; CHECK-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload
; NO-FMA-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; CHECK-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; NO-FMA-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm1 = xmm1[0],mem[0] ; CHECK-NEXT: # xmm1 = xmm1[0],mem[0]
; NO-FMA-NEXT: movaps %xmm1, %xmm0 ; CHECK-NEXT: movaps %xmm1, %xmm0
; NO-FMA-NEXT: addq $40, %rsp ; CHECK-NEXT: addq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_exp_v3f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 48
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq expf
; HAS-FMA-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq expf
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq expf
; HAS-FMA-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload
; HAS-FMA-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[2,3]
; HAS-FMA-NEXT: vinsertps $32, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0,1],mem[0],xmm0[3]
; HAS-FMA-NEXT: addq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%exp = call <3 x float> @llvm.experimental.constrained.exp.v3f32( %exp = call <3 x float> @llvm.experimental.constrained.exp.v3f32(
<3 x float> <float 42.0, float 43.0, float 44.0>, <3 x float> <float 42.0, float 43.0, float 44.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x float> %exp ret <3 x float> %exp
} }
define <3 x double> @constrained_vector_exp_v3f64() { define <3 x double> @constrained_vector_exp_v3f64() {
; NO-FMA-LABEL: constrained_vector_exp_v3f64: ; CHECK-LABEL: constrained_vector_exp_v3f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $24, %rsp ; CHECK-NEXT: subq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 32 ; CHECK-NEXT: .cfi_def_cfa_offset 32
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq exp ; CHECK-NEXT: callq exp
; NO-FMA-NEXT: movsd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill ; CHECK-NEXT: movsd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq exp ; CHECK-NEXT: callq exp
; NO-FMA-NEXT: movsd %xmm0, (%rsp) # 8-byte Spill ; CHECK-NEXT: movsd %xmm0, (%rsp) # 8-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq exp ; CHECK-NEXT: callq exp
; NO-FMA-NEXT: movsd %xmm0, {{[0-9]+}}(%rsp) ; CHECK-NEXT: movsd %xmm0, {{[0-9]+}}(%rsp)
; NO-FMA-NEXT: fldl {{[0-9]+}}(%rsp) ; CHECK-NEXT: fldl {{[0-9]+}}(%rsp)
; NO-FMA-NEXT: movsd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 8-byte Reload ; CHECK-NEXT: movsd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 8-byte Reload
; NO-FMA-NEXT: # xmm0 = mem[0],zero ; CHECK-NEXT: # xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd (%rsp), %xmm1 # 8-byte Reload ; CHECK-NEXT: movsd (%rsp), %xmm1 # 8-byte Reload
; NO-FMA-NEXT: # xmm1 = mem[0],zero ; CHECK-NEXT: # xmm1 = mem[0],zero
; NO-FMA-NEXT: addq $24, %rsp ; CHECK-NEXT: addq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_exp_v3f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $56, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 64
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq exp
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq exp
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vzeroupper
; HAS-FMA-NEXT: callq exp
; HAS-FMA-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
; HAS-FMA-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; HAS-FMA-NEXT: addq $56, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%exp = call <3 x double> @llvm.experimental.constrained.exp.v3f64( %exp = call <3 x double> @llvm.experimental.constrained.exp.v3f64(
<3 x double> <double 42.0, double 42.1, double 42.2>, <3 x double> <double 42.0, double 42.1, double 42.2>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x double> %exp ret <3 x double> %exp
} }
define <4 x double> @constrained_vector_exp_v4f64() { define <4 x double> @constrained_vector_exp_v4f64() {
; NO-FMA-LABEL: constrained_vector_exp_v4f64: ; CHECK-LABEL: constrained_vector_exp_v4f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $40, %rsp ; CHECK-NEXT: subq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 48 ; CHECK-NEXT: .cfi_def_cfa_offset 48
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq exp ; CHECK-NEXT: callq exp
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq exp ; CHECK-NEXT: callq exp
; NO-FMA-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0] ; CHECK-NEXT: # xmm0 = xmm0[0],mem[0]
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq exp ; CHECK-NEXT: callq exp
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq exp ; CHECK-NEXT: callq exp
; NO-FMA-NEXT: movaps %xmm0, %xmm1 ; CHECK-NEXT: movaps %xmm0, %xmm1
; NO-FMA-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm1 = xmm1[0],mem[0] ; CHECK-NEXT: # xmm1 = xmm1[0],mem[0]
; NO-FMA-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload ; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; NO-FMA-NEXT: addq $40, %rsp ; CHECK-NEXT: addq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_exp_v4f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 48
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq exp
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq exp
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq exp
; HAS-FMA-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq exp
; HAS-FMA-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vinsertf128 $1, (%rsp), %ymm0, %ymm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: addq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%exp = call <4 x double> @llvm.experimental.constrained.exp.v4f64( %exp = call <4 x double> @llvm.experimental.constrained.exp.v4f64(
<4 x double> <double 42.0, double 42.1, <4 x double> <double 42.0, double 42.1,
double 42.2, double 42.3>, double 42.2, double 42.3>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <4 x double> %exp ret <4 x double> %exp
} }
define <1 x float> @constrained_vector_exp2_v1f32() { define <1 x float> @constrained_vector_exp2_v1f32() {
; NO-FMA-LABEL: constrained_vector_exp2_v1f32: ; CHECK-LABEL: constrained_vector_exp2_v1f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: pushq %rax ; CHECK-NEXT: pushq %rax
; NO-FMA-NEXT: .cfi_def_cfa_offset 16 ; CHECK-NEXT: .cfi_def_cfa_offset 16
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq exp2f ; CHECK-NEXT: callq exp2f
; NO-FMA-NEXT: popq %rax ; CHECK-NEXT: popq %rax
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_exp2_v1f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: pushq %rax
; HAS-FMA-NEXT: .cfi_def_cfa_offset 16
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq exp2f
; HAS-FMA-NEXT: popq %rax
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%exp2 = call <1 x float> @llvm.experimental.constrained.exp2.v1f32( %exp2 = call <1 x float> @llvm.experimental.constrained.exp2.v1f32(
<1 x float> <float 42.0>, <1 x float> <float 42.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <1 x float> %exp2 ret <1 x float> %exp2
} }
define <2 x double> @constrained_vector_exp2_v2f64() { define <2 x double> @constrained_vector_exp2_v2f64() {
; NO-FMA-LABEL: constrained_vector_exp2_v2f64: ; CHECK-LABEL: constrained_vector_exp2_v2f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $24, %rsp ; CHECK-NEXT: subq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 32 ; CHECK-NEXT: .cfi_def_cfa_offset 32
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq exp2 ; CHECK-NEXT: callq exp2
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq exp2 ; CHECK-NEXT: callq exp2
; NO-FMA-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0] ; CHECK-NEXT: # xmm0 = xmm0[0],mem[0]
; NO-FMA-NEXT: addq $24, %rsp ; CHECK-NEXT: addq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_exp2_v2f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $24, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 32
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq exp2
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq exp2
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: addq $24, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%exp2 = call <2 x double> @llvm.experimental.constrained.exp2.v2f64( %exp2 = call <2 x double> @llvm.experimental.constrained.exp2.v2f64(
<2 x double> <double 42.1, double 42.0>, <2 x double> <double 42.1, double 42.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <2 x double> %exp2 ret <2 x double> %exp2
} }
define <3 x float> @constrained_vector_exp2_v3f32() { define <3 x float> @constrained_vector_exp2_v3f32() {
; NO-FMA-LABEL: constrained_vector_exp2_v3f32: ; CHECK-LABEL: constrained_vector_exp2_v3f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $40, %rsp ; CHECK-NEXT: subq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 48 ; CHECK-NEXT: .cfi_def_cfa_offset 48
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq exp2f ; CHECK-NEXT: callq exp2f
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq exp2f ; CHECK-NEXT: callq exp2f
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq exp2f ; CHECK-NEXT: callq exp2f
; NO-FMA-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload ; CHECK-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload
; NO-FMA-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; CHECK-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; NO-FMA-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm1 = xmm1[0],mem[0] ; CHECK-NEXT: # xmm1 = xmm1[0],mem[0]
; NO-FMA-NEXT: movaps %xmm1, %xmm0 ; CHECK-NEXT: movaps %xmm1, %xmm0
; NO-FMA-NEXT: addq $40, %rsp ; CHECK-NEXT: addq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_exp2_v3f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 48
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq exp2f
; HAS-FMA-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq exp2f
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq exp2f
; HAS-FMA-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload
; HAS-FMA-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[2,3]
; HAS-FMA-NEXT: vinsertps $32, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0,1],mem[0],xmm0[3]
; HAS-FMA-NEXT: addq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%exp2 = call <3 x float> @llvm.experimental.constrained.exp2.v3f32( %exp2 = call <3 x float> @llvm.experimental.constrained.exp2.v3f32(
<3 x float> <float 42.0, float 43.0, float 44.0>, <3 x float> <float 42.0, float 43.0, float 44.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x float> %exp2 ret <3 x float> %exp2
} }
define <3 x double> @constrained_vector_exp2_v3f64() { define <3 x double> @constrained_vector_exp2_v3f64() {
; NO-FMA-LABEL: constrained_vector_exp2_v3f64: ; CHECK-LABEL: constrained_vector_exp2_v3f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $24, %rsp ; CHECK-NEXT: subq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 32 ; CHECK-NEXT: .cfi_def_cfa_offset 32
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq exp2 ; CHECK-NEXT: callq exp2
; NO-FMA-NEXT: movsd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill ; CHECK-NEXT: movsd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq exp2 ; CHECK-NEXT: callq exp2
; NO-FMA-NEXT: movsd %xmm0, (%rsp) # 8-byte Spill ; CHECK-NEXT: movsd %xmm0, (%rsp) # 8-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq exp2 ; CHECK-NEXT: callq exp2
; NO-FMA-NEXT: movsd %xmm0, {{[0-9]+}}(%rsp) ; CHECK-NEXT: movsd %xmm0, {{[0-9]+}}(%rsp)
; NO-FMA-NEXT: fldl {{[0-9]+}}(%rsp) ; CHECK-NEXT: fldl {{[0-9]+}}(%rsp)
; NO-FMA-NEXT: movsd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 8-byte Reload ; CHECK-NEXT: movsd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 8-byte Reload
; NO-FMA-NEXT: # xmm0 = mem[0],zero ; CHECK-NEXT: # xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd (%rsp), %xmm1 # 8-byte Reload ; CHECK-NEXT: movsd (%rsp), %xmm1 # 8-byte Reload
; NO-FMA-NEXT: # xmm1 = mem[0],zero ; CHECK-NEXT: # xmm1 = mem[0],zero
; NO-FMA-NEXT: addq $24, %rsp ; CHECK-NEXT: addq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_exp2_v3f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $56, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 64
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq exp2
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq exp2
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vzeroupper
; HAS-FMA-NEXT: callq exp2
; HAS-FMA-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
; HAS-FMA-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; HAS-FMA-NEXT: addq $56, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%exp2 = call <3 x double> @llvm.experimental.constrained.exp2.v3f64( %exp2 = call <3 x double> @llvm.experimental.constrained.exp2.v3f64(
<3 x double> <double 42.0, double 42.1, double 42.2>, <3 x double> <double 42.0, double 42.1, double 42.2>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x double> %exp2 ret <3 x double> %exp2
} }
define <4 x double> @constrained_vector_exp2_v4f64() { define <4 x double> @constrained_vector_exp2_v4f64() {
; NO-FMA-LABEL: constrained_vector_exp2_v4f64: ; CHECK-LABEL: constrained_vector_exp2_v4f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $40, %rsp ; CHECK-NEXT: subq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 48 ; CHECK-NEXT: .cfi_def_cfa_offset 48
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq exp2 ; CHECK-NEXT: callq exp2
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq exp2 ; CHECK-NEXT: callq exp2
; NO-FMA-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0] ; CHECK-NEXT: # xmm0 = xmm0[0],mem[0]
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq exp2 ; CHECK-NEXT: callq exp2
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq exp2 ; CHECK-NEXT: callq exp2
; NO-FMA-NEXT: movaps %xmm0, %xmm1 ; CHECK-NEXT: movaps %xmm0, %xmm1
; NO-FMA-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm1 = xmm1[0],mem[0] ; CHECK-NEXT: # xmm1 = xmm1[0],mem[0]
; NO-FMA-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload ; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; NO-FMA-NEXT: addq $40, %rsp ; CHECK-NEXT: addq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_exp2_v4f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 48
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq exp2
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq exp2
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq exp2
; HAS-FMA-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq exp2
; HAS-FMA-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vinsertf128 $1, (%rsp), %ymm0, %ymm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: addq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%exp2 = call <4 x double> @llvm.experimental.constrained.exp2.v4f64( %exp2 = call <4 x double> @llvm.experimental.constrained.exp2.v4f64(
<4 x double> <double 42.1, double 42.2, <4 x double> <double 42.1, double 42.2,
double 42.3, double 42.4>, double 42.3, double 42.4>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <4 x double> %exp2 ret <4 x double> %exp2
} }
define <1 x float> @constrained_vector_log_v1f32() { define <1 x float> @constrained_vector_log_v1f32() {
; NO-FMA-LABEL: constrained_vector_log_v1f32: ; CHECK-LABEL: constrained_vector_log_v1f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: pushq %rax ; CHECK-NEXT: pushq %rax
; NO-FMA-NEXT: .cfi_def_cfa_offset 16 ; CHECK-NEXT: .cfi_def_cfa_offset 16
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq logf ; CHECK-NEXT: callq logf
; NO-FMA-NEXT: popq %rax ; CHECK-NEXT: popq %rax
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_log_v1f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: pushq %rax
; HAS-FMA-NEXT: .cfi_def_cfa_offset 16
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq logf
; HAS-FMA-NEXT: popq %rax
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%log = call <1 x float> @llvm.experimental.constrained.log.v1f32( %log = call <1 x float> @llvm.experimental.constrained.log.v1f32(
<1 x float> <float 42.0>, <1 x float> <float 42.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <1 x float> %log ret <1 x float> %log
} }
define <2 x double> @constrained_vector_log_v2f64() { define <2 x double> @constrained_vector_log_v2f64() {
; NO-FMA-LABEL: constrained_vector_log_v2f64: ; CHECK-LABEL: constrained_vector_log_v2f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $24, %rsp ; CHECK-NEXT: subq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 32 ; CHECK-NEXT: .cfi_def_cfa_offset 32
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq log ; CHECK-NEXT: callq log
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq log ; CHECK-NEXT: callq log
; NO-FMA-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0] ; CHECK-NEXT: # xmm0 = xmm0[0],mem[0]
; NO-FMA-NEXT: addq $24, %rsp ; CHECK-NEXT: addq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_log_v2f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $24, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 32
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq log
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq log
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: addq $24, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%log = call <2 x double> @llvm.experimental.constrained.log.v2f64( %log = call <2 x double> @llvm.experimental.constrained.log.v2f64(
<2 x double> <double 42.0, double 42.1>, <2 x double> <double 42.0, double 42.1>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <2 x double> %log ret <2 x double> %log
} }
define <3 x float> @constrained_vector_log_v3f32() { define <3 x float> @constrained_vector_log_v3f32() {
; NO-FMA-LABEL: constrained_vector_log_v3f32: ; CHECK-LABEL: constrained_vector_log_v3f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $40, %rsp ; CHECK-NEXT: subq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 48 ; CHECK-NEXT: .cfi_def_cfa_offset 48
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq logf ; CHECK-NEXT: callq logf
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq logf ; CHECK-NEXT: callq logf
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq logf ; CHECK-NEXT: callq logf
; NO-FMA-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload ; CHECK-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload
; NO-FMA-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; CHECK-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; NO-FMA-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm1 = xmm1[0],mem[0] ; CHECK-NEXT: # xmm1 = xmm1[0],mem[0]
; NO-FMA-NEXT: movaps %xmm1, %xmm0 ; CHECK-NEXT: movaps %xmm1, %xmm0
; NO-FMA-NEXT: addq $40, %rsp ; CHECK-NEXT: addq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_log_v3f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 48
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq logf
; HAS-FMA-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq logf
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq logf
; HAS-FMA-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload
; HAS-FMA-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[2,3]
; HAS-FMA-NEXT: vinsertps $32, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0,1],mem[0],xmm0[3]
; HAS-FMA-NEXT: addq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%log = call <3 x float> @llvm.experimental.constrained.log.v3f32( %log = call <3 x float> @llvm.experimental.constrained.log.v3f32(
<3 x float> <float 42.0, float 43.0, float 44.0>, <3 x float> <float 42.0, float 43.0, float 44.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x float> %log ret <3 x float> %log
} }
define <3 x double> @constrained_vector_log_v3f64() { define <3 x double> @constrained_vector_log_v3f64() {
; NO-FMA-LABEL: constrained_vector_log_v3f64: ; CHECK-LABEL: constrained_vector_log_v3f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $24, %rsp ; CHECK-NEXT: subq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 32 ; CHECK-NEXT: .cfi_def_cfa_offset 32
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq log ; CHECK-NEXT: callq log
; NO-FMA-NEXT: movsd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill ; CHECK-NEXT: movsd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq log ; CHECK-NEXT: callq log
; NO-FMA-NEXT: movsd %xmm0, (%rsp) # 8-byte Spill ; CHECK-NEXT: movsd %xmm0, (%rsp) # 8-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq log ; CHECK-NEXT: callq log
; NO-FMA-NEXT: movsd %xmm0, {{[0-9]+}}(%rsp) ; CHECK-NEXT: movsd %xmm0, {{[0-9]+}}(%rsp)
; NO-FMA-NEXT: fldl {{[0-9]+}}(%rsp) ; CHECK-NEXT: fldl {{[0-9]+}}(%rsp)
; NO-FMA-NEXT: movsd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 8-byte Reload ; CHECK-NEXT: movsd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 8-byte Reload
; NO-FMA-NEXT: # xmm0 = mem[0],zero ; CHECK-NEXT: # xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd (%rsp), %xmm1 # 8-byte Reload ; CHECK-NEXT: movsd (%rsp), %xmm1 # 8-byte Reload
; NO-FMA-NEXT: # xmm1 = mem[0],zero ; CHECK-NEXT: # xmm1 = mem[0],zero
; NO-FMA-NEXT: addq $24, %rsp ; CHECK-NEXT: addq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_log_v3f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $56, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 64
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq log
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq log
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vzeroupper
; HAS-FMA-NEXT: callq log
; HAS-FMA-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
; HAS-FMA-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; HAS-FMA-NEXT: addq $56, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%log = call <3 x double> @llvm.experimental.constrained.log.v3f64( %log = call <3 x double> @llvm.experimental.constrained.log.v3f64(
<3 x double> <double 42.0, double 42.1, double 42.2>, <3 x double> <double 42.0, double 42.1, double 42.2>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x double> %log ret <3 x double> %log
} }
define <4 x double> @constrained_vector_log_v4f64() { define <4 x double> @constrained_vector_log_v4f64() {
; NO-FMA-LABEL: constrained_vector_log_v4f64: ; CHECK-LABEL: constrained_vector_log_v4f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $40, %rsp ; CHECK-NEXT: subq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 48 ; CHECK-NEXT: .cfi_def_cfa_offset 48
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq log ; CHECK-NEXT: callq log
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq log ; CHECK-NEXT: callq log
; NO-FMA-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0] ; CHECK-NEXT: # xmm0 = xmm0[0],mem[0]
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq log ; CHECK-NEXT: callq log
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq log ; CHECK-NEXT: callq log
; NO-FMA-NEXT: movaps %xmm0, %xmm1 ; CHECK-NEXT: movaps %xmm0, %xmm1
; NO-FMA-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm1 = xmm1[0],mem[0] ; CHECK-NEXT: # xmm1 = xmm1[0],mem[0]
; NO-FMA-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload ; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; NO-FMA-NEXT: addq $40, %rsp ; CHECK-NEXT: addq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_log_v4f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 48
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq log
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq log
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq log
; HAS-FMA-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq log
; HAS-FMA-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vinsertf128 $1, (%rsp), %ymm0, %ymm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: addq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%log = call <4 x double> @llvm.experimental.constrained.log.v4f64( %log = call <4 x double> @llvm.experimental.constrained.log.v4f64(
<4 x double> <double 42.0, double 42.1, <4 x double> <double 42.0, double 42.1,
double 42.2, double 42.3>, double 42.2, double 42.3>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <4 x double> %log ret <4 x double> %log
} }
define <1 x float> @constrained_vector_log10_v1f32() { define <1 x float> @constrained_vector_log10_v1f32() {
; NO-FMA-LABEL: constrained_vector_log10_v1f32: ; CHECK-LABEL: constrained_vector_log10_v1f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: pushq %rax ; CHECK-NEXT: pushq %rax
; NO-FMA-NEXT: .cfi_def_cfa_offset 16 ; CHECK-NEXT: .cfi_def_cfa_offset 16
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq log10f ; CHECK-NEXT: callq log10f
; NO-FMA-NEXT: popq %rax ; CHECK-NEXT: popq %rax
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_log10_v1f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: pushq %rax
; HAS-FMA-NEXT: .cfi_def_cfa_offset 16
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq log10f
; HAS-FMA-NEXT: popq %rax
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%log10 = call <1 x float> @llvm.experimental.constrained.log10.v1f32( %log10 = call <1 x float> @llvm.experimental.constrained.log10.v1f32(
<1 x float> <float 42.0>, <1 x float> <float 42.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <1 x float> %log10 ret <1 x float> %log10
} }
define <2 x double> @constrained_vector_log10_v2f64() { define <2 x double> @constrained_vector_log10_v2f64() {
; NO-FMA-LABEL: constrained_vector_log10_v2f64: ; CHECK-LABEL: constrained_vector_log10_v2f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $24, %rsp ; CHECK-NEXT: subq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 32 ; CHECK-NEXT: .cfi_def_cfa_offset 32
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq log10 ; CHECK-NEXT: callq log10
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq log10 ; CHECK-NEXT: callq log10
; NO-FMA-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0] ; CHECK-NEXT: # xmm0 = xmm0[0],mem[0]
; NO-FMA-NEXT: addq $24, %rsp ; CHECK-NEXT: addq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_log10_v2f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $24, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 32
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq log10
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq log10
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: addq $24, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%log10 = call <2 x double> @llvm.experimental.constrained.log10.v2f64( %log10 = call <2 x double> @llvm.experimental.constrained.log10.v2f64(
<2 x double> <double 42.0, double 42.1>, <2 x double> <double 42.0, double 42.1>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <2 x double> %log10 ret <2 x double> %log10
} }
define <3 x float> @constrained_vector_log10_v3f32() { define <3 x float> @constrained_vector_log10_v3f32() {
; NO-FMA-LABEL: constrained_vector_log10_v3f32: ; CHECK-LABEL: constrained_vector_log10_v3f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $40, %rsp ; CHECK-NEXT: subq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 48 ; CHECK-NEXT: .cfi_def_cfa_offset 48
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq log10f ; CHECK-NEXT: callq log10f
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq log10f ; CHECK-NEXT: callq log10f
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq log10f ; CHECK-NEXT: callq log10f
; NO-FMA-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload ; CHECK-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload
; NO-FMA-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; CHECK-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; NO-FMA-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm1 = xmm1[0],mem[0] ; CHECK-NEXT: # xmm1 = xmm1[0],mem[0]
; NO-FMA-NEXT: movaps %xmm1, %xmm0 ; CHECK-NEXT: movaps %xmm1, %xmm0
; NO-FMA-NEXT: addq $40, %rsp ; CHECK-NEXT: addq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_log10_v3f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 48
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq log10f
; HAS-FMA-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq log10f
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq log10f
; HAS-FMA-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload
; HAS-FMA-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[2,3]
; HAS-FMA-NEXT: vinsertps $32, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0,1],mem[0],xmm0[3]
; HAS-FMA-NEXT: addq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%log10 = call <3 x float> @llvm.experimental.constrained.log10.v3f32( %log10 = call <3 x float> @llvm.experimental.constrained.log10.v3f32(
<3 x float> <float 42.0, float 43.0, float 44.0>, <3 x float> <float 42.0, float 43.0, float 44.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x float> %log10 ret <3 x float> %log10
} }
define <3 x double> @constrained_vector_log10_v3f64() { define <3 x double> @constrained_vector_log10_v3f64() {
; NO-FMA-LABEL: constrained_vector_log10_v3f64: ; CHECK-LABEL: constrained_vector_log10_v3f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $24, %rsp ; CHECK-NEXT: subq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 32 ; CHECK-NEXT: .cfi_def_cfa_offset 32
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq log10 ; CHECK-NEXT: callq log10
; NO-FMA-NEXT: movsd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill ; CHECK-NEXT: movsd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq log10 ; CHECK-NEXT: callq log10
; NO-FMA-NEXT: movsd %xmm0, (%rsp) # 8-byte Spill ; CHECK-NEXT: movsd %xmm0, (%rsp) # 8-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq log10 ; CHECK-NEXT: callq log10
; NO-FMA-NEXT: movsd %xmm0, {{[0-9]+}}(%rsp) ; CHECK-NEXT: movsd %xmm0, {{[0-9]+}}(%rsp)
; NO-FMA-NEXT: fldl {{[0-9]+}}(%rsp) ; CHECK-NEXT: fldl {{[0-9]+}}(%rsp)
; NO-FMA-NEXT: movsd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 8-byte Reload ; CHECK-NEXT: movsd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 8-byte Reload
; NO-FMA-NEXT: # xmm0 = mem[0],zero ; CHECK-NEXT: # xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd (%rsp), %xmm1 # 8-byte Reload ; CHECK-NEXT: movsd (%rsp), %xmm1 # 8-byte Reload
; NO-FMA-NEXT: # xmm1 = mem[0],zero ; CHECK-NEXT: # xmm1 = mem[0],zero
; NO-FMA-NEXT: addq $24, %rsp ; CHECK-NEXT: addq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_log10_v3f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $56, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 64
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq log10
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq log10
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vzeroupper
; HAS-FMA-NEXT: callq log10
; HAS-FMA-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
; HAS-FMA-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; HAS-FMA-NEXT: addq $56, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%log10 = call <3 x double> @llvm.experimental.constrained.log10.v3f64( %log10 = call <3 x double> @llvm.experimental.constrained.log10.v3f64(
<3 x double> <double 42.0, double 42.1, double 42.2>, <3 x double> <double 42.0, double 42.1, double 42.2>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x double> %log10 ret <3 x double> %log10
} }
define <4 x double> @constrained_vector_log10_v4f64() { define <4 x double> @constrained_vector_log10_v4f64() {
; NO-FMA-LABEL: constrained_vector_log10_v4f64: ; CHECK-LABEL: constrained_vector_log10_v4f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $40, %rsp ; CHECK-NEXT: subq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 48 ; CHECK-NEXT: .cfi_def_cfa_offset 48
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq log10 ; CHECK-NEXT: callq log10
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq log10 ; CHECK-NEXT: callq log10
; NO-FMA-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0] ; CHECK-NEXT: # xmm0 = xmm0[0],mem[0]
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq log10 ; CHECK-NEXT: callq log10
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq log10 ; CHECK-NEXT: callq log10
; NO-FMA-NEXT: movaps %xmm0, %xmm1 ; CHECK-NEXT: movaps %xmm0, %xmm1
; NO-FMA-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm1 = xmm1[0],mem[0] ; CHECK-NEXT: # xmm1 = xmm1[0],mem[0]
; NO-FMA-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload ; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; NO-FMA-NEXT: addq $40, %rsp ; CHECK-NEXT: addq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_log10_v4f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 48
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq log10
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq log10
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq log10
; HAS-FMA-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq log10
; HAS-FMA-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vinsertf128 $1, (%rsp), %ymm0, %ymm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: addq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%log10 = call <4 x double> @llvm.experimental.constrained.log10.v4f64( %log10 = call <4 x double> @llvm.experimental.constrained.log10.v4f64(
<4 x double> <double 42.0, double 42.1, <4 x double> <double 42.0, double 42.1,
double 42.2, double 42.3>, double 42.2, double 42.3>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <4 x double> %log10 ret <4 x double> %log10
} }
define <1 x float> @constrained_vector_log2_v1f32() { define <1 x float> @constrained_vector_log2_v1f32() {
; NO-FMA-LABEL: constrained_vector_log2_v1f32: ; CHECK-LABEL: constrained_vector_log2_v1f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: pushq %rax ; CHECK-NEXT: pushq %rax
; NO-FMA-NEXT: .cfi_def_cfa_offset 16 ; CHECK-NEXT: .cfi_def_cfa_offset 16
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq log2f ; CHECK-NEXT: callq log2f
; NO-FMA-NEXT: popq %rax ; CHECK-NEXT: popq %rax
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_log2_v1f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: pushq %rax
; HAS-FMA-NEXT: .cfi_def_cfa_offset 16
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq log2f
; HAS-FMA-NEXT: popq %rax
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%log2 = call <1 x float> @llvm.experimental.constrained.log2.v1f32( %log2 = call <1 x float> @llvm.experimental.constrained.log2.v1f32(
<1 x float> <float 42.0>, <1 x float> <float 42.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <1 x float> %log2 ret <1 x float> %log2
} }
define <2 x double> @constrained_vector_log2_v2f64() { define <2 x double> @constrained_vector_log2_v2f64() {
; NO-FMA-LABEL: constrained_vector_log2_v2f64: ; CHECK-LABEL: constrained_vector_log2_v2f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $24, %rsp ; CHECK-NEXT: subq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 32 ; CHECK-NEXT: .cfi_def_cfa_offset 32
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq log2 ; CHECK-NEXT: callq log2
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq log2 ; CHECK-NEXT: callq log2
; NO-FMA-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0] ; CHECK-NEXT: # xmm0 = xmm0[0],mem[0]
; NO-FMA-NEXT: addq $24, %rsp ; CHECK-NEXT: addq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_log2_v2f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $24, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 32
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq log2
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq log2
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: addq $24, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%log2 = call <2 x double> @llvm.experimental.constrained.log2.v2f64( %log2 = call <2 x double> @llvm.experimental.constrained.log2.v2f64(
<2 x double> <double 42.0, double 42.1>, <2 x double> <double 42.0, double 42.1>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <2 x double> %log2 ret <2 x double> %log2
} }
define <3 x float> @constrained_vector_log2_v3f32() { define <3 x float> @constrained_vector_log2_v3f32() {
; NO-FMA-LABEL: constrained_vector_log2_v3f32: ; CHECK-LABEL: constrained_vector_log2_v3f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $40, %rsp ; CHECK-NEXT: subq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 48 ; CHECK-NEXT: .cfi_def_cfa_offset 48
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq log2f ; CHECK-NEXT: callq log2f
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq log2f ; CHECK-NEXT: callq log2f
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq log2f ; CHECK-NEXT: callq log2f
; NO-FMA-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload ; CHECK-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload
; NO-FMA-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; CHECK-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; NO-FMA-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm1 = xmm1[0],mem[0] ; CHECK-NEXT: # xmm1 = xmm1[0],mem[0]
; NO-FMA-NEXT: movaps %xmm1, %xmm0 ; CHECK-NEXT: movaps %xmm1, %xmm0
; NO-FMA-NEXT: addq $40, %rsp ; CHECK-NEXT: addq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_log2_v3f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 48
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq log2f
; HAS-FMA-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq log2f
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq log2f
; HAS-FMA-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload
; HAS-FMA-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[2,3]
; HAS-FMA-NEXT: vinsertps $32, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0,1],mem[0],xmm0[3]
; HAS-FMA-NEXT: addq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%log2 = call <3 x float> @llvm.experimental.constrained.log2.v3f32( %log2 = call <3 x float> @llvm.experimental.constrained.log2.v3f32(
<3 x float> <float 42.0, float 43.0, float 44.0>, <3 x float> <float 42.0, float 43.0, float 44.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x float> %log2 ret <3 x float> %log2
} }
define <3 x double> @constrained_vector_log2_v3f64() { define <3 x double> @constrained_vector_log2_v3f64() {
; NO-FMA-LABEL: constrained_vector_log2_v3f64: ; CHECK-LABEL: constrained_vector_log2_v3f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $24, %rsp ; CHECK-NEXT: subq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 32 ; CHECK-NEXT: .cfi_def_cfa_offset 32
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq log2 ; CHECK-NEXT: callq log2
; NO-FMA-NEXT: movsd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill ; CHECK-NEXT: movsd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq log2 ; CHECK-NEXT: callq log2
; NO-FMA-NEXT: movsd %xmm0, (%rsp) # 8-byte Spill ; CHECK-NEXT: movsd %xmm0, (%rsp) # 8-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq log2 ; CHECK-NEXT: callq log2
; NO-FMA-NEXT: movsd %xmm0, {{[0-9]+}}(%rsp) ; CHECK-NEXT: movsd %xmm0, {{[0-9]+}}(%rsp)
; NO-FMA-NEXT: fldl {{[0-9]+}}(%rsp) ; CHECK-NEXT: fldl {{[0-9]+}}(%rsp)
; NO-FMA-NEXT: movsd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 8-byte Reload ; CHECK-NEXT: movsd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 8-byte Reload
; NO-FMA-NEXT: # xmm0 = mem[0],zero ; CHECK-NEXT: # xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd (%rsp), %xmm1 # 8-byte Reload ; CHECK-NEXT: movsd (%rsp), %xmm1 # 8-byte Reload
; NO-FMA-NEXT: # xmm1 = mem[0],zero ; CHECK-NEXT: # xmm1 = mem[0],zero
; NO-FMA-NEXT: addq $24, %rsp ; CHECK-NEXT: addq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_log2_v3f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $56, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 64
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq log2
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq log2
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vzeroupper
; HAS-FMA-NEXT: callq log2
; HAS-FMA-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
; HAS-FMA-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; HAS-FMA-NEXT: addq $56, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%log2 = call <3 x double> @llvm.experimental.constrained.log2.v3f64( %log2 = call <3 x double> @llvm.experimental.constrained.log2.v3f64(
<3 x double> <double 42.0, double 42.1, double 42.2>, <3 x double> <double 42.0, double 42.1, double 42.2>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x double> %log2 ret <3 x double> %log2
} }
define <4 x double> @constrained_vector_log2_v4f64() { define <4 x double> @constrained_vector_log2_v4f64() {
; NO-FMA-LABEL: constrained_vector_log2_v4f64: ; CHECK-LABEL: constrained_vector_log2_v4f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $40, %rsp ; CHECK-NEXT: subq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 48 ; CHECK-NEXT: .cfi_def_cfa_offset 48
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq log2 ; CHECK-NEXT: callq log2
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq log2 ; CHECK-NEXT: callq log2
; NO-FMA-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0] ; CHECK-NEXT: # xmm0 = xmm0[0],mem[0]
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq log2 ; CHECK-NEXT: callq log2
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq log2 ; CHECK-NEXT: callq log2
; NO-FMA-NEXT: movaps %xmm0, %xmm1 ; CHECK-NEXT: movaps %xmm0, %xmm1
; NO-FMA-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm1 = xmm1[0],mem[0] ; CHECK-NEXT: # xmm1 = xmm1[0],mem[0]
; NO-FMA-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload ; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; NO-FMA-NEXT: addq $40, %rsp ; CHECK-NEXT: addq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_log2_v4f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 48
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq log2
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq log2
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq log2
; HAS-FMA-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: callq log2
; HAS-FMA-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vinsertf128 $1, (%rsp), %ymm0, %ymm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: addq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%log2 = call <4 x double> @llvm.experimental.constrained.log2.v4f64( %log2 = call <4 x double> @llvm.experimental.constrained.log2.v4f64(
<4 x double> <double 42.0, double 42.1, <4 x double> <double 42.0, double 42.1,
double 42.2, double 42.3>, double 42.2, double 42.3>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <4 x double> %log2 ret <4 x double> %log2
} }
define <1 x float> @constrained_vector_rint_v1f32() { define <1 x float> @constrained_vector_rint_v1f32() {
; NO-FMA-LABEL: constrained_vector_rint_v1f32: ; CHECK-LABEL: constrained_vector_rint_v1f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: pushq %rax ; CHECK-NEXT: pushq %rax
; NO-FMA-NEXT: .cfi_def_cfa_offset 16 ; CHECK-NEXT: .cfi_def_cfa_offset 16
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq rintf ; CHECK-NEXT: callq rintf
; NO-FMA-NEXT: popq %rax ; CHECK-NEXT: popq %rax
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_rint_v1f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vroundss $4, %xmm0, %xmm0, %xmm0
; HAS-FMA-NEXT: retq
entry: entry:
%rint = call <1 x float> @llvm.experimental.constrained.rint.v1f32( %rint = call <1 x float> @llvm.experimental.constrained.rint.v1f32(
<1 x float> <float 42.0>, <1 x float> <float 42.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <1 x float> %rint ret <1 x float> %rint
} }
define <2 x double> @constrained_vector_rint_v2f64() { define <2 x double> @constrained_vector_rint_v2f64() {
; NO-FMA-LABEL: constrained_vector_rint_v2f64: ; CHECK-LABEL: constrained_vector_rint_v2f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $24, %rsp ; CHECK-NEXT: subq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 32 ; CHECK-NEXT: .cfi_def_cfa_offset 32
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq rint ; CHECK-NEXT: callq rint
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq rint ; CHECK-NEXT: callq rint
; NO-FMA-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0] ; CHECK-NEXT: # xmm0 = xmm0[0],mem[0]
; NO-FMA-NEXT: addq $24, %rsp ; CHECK-NEXT: addq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_rint_v2f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vroundpd $4, {{.*}}(%rip), %xmm0
; HAS-FMA-NEXT: retq
entry: entry:
%rint = call <2 x double> @llvm.experimental.constrained.rint.v2f64( %rint = call <2 x double> @llvm.experimental.constrained.rint.v2f64(
<2 x double> <double 42.1, double 42.0>, <2 x double> <double 42.1, double 42.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <2 x double> %rint ret <2 x double> %rint
} }
define <3 x float> @constrained_vector_rint_v3f32() { define <3 x float> @constrained_vector_rint_v3f32() {
; NO-FMA-LABEL: constrained_vector_rint_v3f32: ; CHECK-LABEL: constrained_vector_rint_v3f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $40, %rsp ; CHECK-NEXT: subq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 48 ; CHECK-NEXT: .cfi_def_cfa_offset 48
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq rintf ; CHECK-NEXT: callq rintf
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq rintf ; CHECK-NEXT: callq rintf
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq rintf ; CHECK-NEXT: callq rintf
; NO-FMA-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload ; CHECK-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload
; NO-FMA-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; CHECK-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; NO-FMA-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm1 = xmm1[0],mem[0] ; CHECK-NEXT: # xmm1 = xmm1[0],mem[0]
; NO-FMA-NEXT: movaps %xmm1, %xmm0 ; CHECK-NEXT: movaps %xmm1, %xmm0
; NO-FMA-NEXT: addq $40, %rsp ; CHECK-NEXT: addq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_rint_v3f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vroundss $4, %xmm0, %xmm0, %xmm0
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vroundss $4, %xmm1, %xmm1, %xmm1
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vroundss $4, %xmm2, %xmm2, %xmm2
; HAS-FMA-NEXT: vinsertps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[2,3]
; HAS-FMA-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0,1],xmm0[0],xmm1[3]
; HAS-FMA-NEXT: retq
entry: entry:
%rint = call <3 x float> @llvm.experimental.constrained.rint.v3f32( %rint = call <3 x float> @llvm.experimental.constrained.rint.v3f32(
<3 x float> <float 42.0, float 43.0, float 44.0>, <3 x float> <float 42.0, float 43.0, float 44.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x float> %rint ret <3 x float> %rint
} }
define <3 x double> @constrained_vector_rint_v3f64() { define <3 x double> @constrained_vector_rint_v3f64() {
; NO-FMA-LABEL: constrained_vector_rint_v3f64: ; CHECK-LABEL: constrained_vector_rint_v3f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $24, %rsp ; CHECK-NEXT: subq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 32 ; CHECK-NEXT: .cfi_def_cfa_offset 32
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq rint ; CHECK-NEXT: callq rint
; NO-FMA-NEXT: movsd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill ; CHECK-NEXT: movsd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq rint ; CHECK-NEXT: callq rint
; NO-FMA-NEXT: movsd %xmm0, (%rsp) # 8-byte Spill ; CHECK-NEXT: movsd %xmm0, (%rsp) # 8-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq rint ; CHECK-NEXT: callq rint
; NO-FMA-NEXT: movsd %xmm0, {{[0-9]+}}(%rsp) ; CHECK-NEXT: movsd %xmm0, {{[0-9]+}}(%rsp)
; NO-FMA-NEXT: fldl {{[0-9]+}}(%rsp) ; CHECK-NEXT: fldl {{[0-9]+}}(%rsp)
; NO-FMA-NEXT: movsd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 8-byte Reload ; CHECK-NEXT: movsd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 8-byte Reload
; NO-FMA-NEXT: # xmm0 = mem[0],zero ; CHECK-NEXT: # xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd (%rsp), %xmm1 # 8-byte Reload ; CHECK-NEXT: movsd (%rsp), %xmm1 # 8-byte Reload
; NO-FMA-NEXT: # xmm1 = mem[0],zero ; CHECK-NEXT: # xmm1 = mem[0],zero
; NO-FMA-NEXT: addq $24, %rsp ; CHECK-NEXT: addq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_rint_v3f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vroundsd $4, %xmm0, %xmm0, %xmm0
; HAS-FMA-NEXT: vroundpd $4, {{.*}}(%rip), %xmm1
; HAS-FMA-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; HAS-FMA-NEXT: retq
entry: entry:
%rint = call <3 x double> @llvm.experimental.constrained.rint.v3f64( %rint = call <3 x double> @llvm.experimental.constrained.rint.v3f64(
<3 x double> <double 42.0, double 42.1, double 42.2>, <3 x double> <double 42.0, double 42.1, double 42.2>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x double> %rint ret <3 x double> %rint
} }
define <4 x double> @constrained_vector_rint_v4f64() { define <4 x double> @constrained_vector_rint_v4f64() {
; NO-FMA-LABEL: constrained_vector_rint_v4f64: ; CHECK-LABEL: constrained_vector_rint_v4f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $40, %rsp ; CHECK-NEXT: subq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 48 ; CHECK-NEXT: .cfi_def_cfa_offset 48
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq rint ; CHECK-NEXT: callq rint
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq rint ; CHECK-NEXT: callq rint
; NO-FMA-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0] ; CHECK-NEXT: # xmm0 = xmm0[0],mem[0]
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq rint ; CHECK-NEXT: callq rint
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq rint ; CHECK-NEXT: callq rint
; NO-FMA-NEXT: movaps %xmm0, %xmm1 ; CHECK-NEXT: movaps %xmm0, %xmm1
; NO-FMA-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm1 = xmm1[0],mem[0] ; CHECK-NEXT: # xmm1 = xmm1[0],mem[0]
; NO-FMA-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload ; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; NO-FMA-NEXT: addq $40, %rsp ; CHECK-NEXT: addq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_rint_v4f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vroundpd $4, {{.*}}(%rip), %ymm0
; HAS-FMA-NEXT: retq
entry: entry:
%rint = call <4 x double> @llvm.experimental.constrained.rint.v4f64( %rint = call <4 x double> @llvm.experimental.constrained.rint.v4f64(
<4 x double> <double 42.1, double 42.2, <4 x double> <double 42.1, double 42.2,
double 42.3, double 42.4>, double 42.3, double 42.4>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <4 x double> %rint ret <4 x double> %rint
} }
define <1 x float> @constrained_vector_nearbyint_v1f32() { define <1 x float> @constrained_vector_nearbyint_v1f32() {
; NO-FMA-LABEL: constrained_vector_nearbyint_v1f32: ; CHECK-LABEL: constrained_vector_nearbyint_v1f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: pushq %rax ; CHECK-NEXT: pushq %rax
; NO-FMA-NEXT: .cfi_def_cfa_offset 16 ; CHECK-NEXT: .cfi_def_cfa_offset 16
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq nearbyintf ; CHECK-NEXT: callq nearbyintf
; NO-FMA-NEXT: popq %rax ; CHECK-NEXT: popq %rax
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_nearbyint_v1f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vroundss $12, %xmm0, %xmm0, %xmm0
; HAS-FMA-NEXT: retq
entry: entry:
%nearby = call <1 x float> @llvm.experimental.constrained.nearbyint.v1f32( %nearby = call <1 x float> @llvm.experimental.constrained.nearbyint.v1f32(
<1 x float> <float 42.0>, <1 x float> <float 42.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <1 x float> %nearby ret <1 x float> %nearby
} }
define <2 x double> @constrained_vector_nearbyint_v2f64() { define <2 x double> @constrained_vector_nearbyint_v2f64() {
; NO-FMA-LABEL: constrained_vector_nearbyint_v2f64: ; CHECK-LABEL: constrained_vector_nearbyint_v2f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $24, %rsp ; CHECK-NEXT: subq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 32 ; CHECK-NEXT: .cfi_def_cfa_offset 32
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq nearbyint ; CHECK-NEXT: callq nearbyint
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq nearbyint ; CHECK-NEXT: callq nearbyint
; NO-FMA-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0] ; CHECK-NEXT: # xmm0 = xmm0[0],mem[0]
; NO-FMA-NEXT: addq $24, %rsp ; CHECK-NEXT: addq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_nearbyint_v2f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vroundpd $12, {{.*}}(%rip), %xmm0
; HAS-FMA-NEXT: retq
entry: entry:
%nearby = call <2 x double> @llvm.experimental.constrained.nearbyint.v2f64( %nearby = call <2 x double> @llvm.experimental.constrained.nearbyint.v2f64(
<2 x double> <double 42.1, double 42.0>, <2 x double> <double 42.1, double 42.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <2 x double> %nearby ret <2 x double> %nearby
} }
define <3 x float> @constrained_vector_nearbyint_v3f32() { define <3 x float> @constrained_vector_nearbyint_v3f32() {
; NO-FMA-LABEL: constrained_vector_nearbyint_v3f32: ; CHECK-LABEL: constrained_vector_nearbyint_v3f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $40, %rsp ; CHECK-NEXT: subq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 48 ; CHECK-NEXT: .cfi_def_cfa_offset 48
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq nearbyintf ; CHECK-NEXT: callq nearbyintf
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq nearbyintf ; CHECK-NEXT: callq nearbyintf
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq nearbyintf ; CHECK-NEXT: callq nearbyintf
; NO-FMA-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload ; CHECK-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload
; NO-FMA-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; CHECK-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; NO-FMA-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm1 = xmm1[0],mem[0] ; CHECK-NEXT: # xmm1 = xmm1[0],mem[0]
; NO-FMA-NEXT: movaps %xmm1, %xmm0 ; CHECK-NEXT: movaps %xmm1, %xmm0
; NO-FMA-NEXT: addq $40, %rsp ; CHECK-NEXT: addq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_nearbyint_v3f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vroundss $12, %xmm0, %xmm0, %xmm0
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vroundss $12, %xmm1, %xmm1, %xmm1
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vroundss $12, %xmm2, %xmm2, %xmm2
; HAS-FMA-NEXT: vinsertps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[2,3]
; HAS-FMA-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0,1],xmm0[0],xmm1[3]
; HAS-FMA-NEXT: retq
entry: entry:
%nearby = call <3 x float> @llvm.experimental.constrained.nearbyint.v3f32( %nearby = call <3 x float> @llvm.experimental.constrained.nearbyint.v3f32(
<3 x float> <float 42.0, float 43.0, float 44.0>, <3 x float> <float 42.0, float 43.0, float 44.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x float> %nearby ret <3 x float> %nearby
} }
define <3 x double> @constrained_vector_nearby_v3f64() { define <3 x double> @constrained_vector_nearby_v3f64() {
; NO-FMA-LABEL: constrained_vector_nearby_v3f64: ; CHECK-LABEL: constrained_vector_nearby_v3f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $24, %rsp ; CHECK-NEXT: subq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 32 ; CHECK-NEXT: .cfi_def_cfa_offset 32
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq nearbyint ; CHECK-NEXT: callq nearbyint
; NO-FMA-NEXT: movsd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill ; CHECK-NEXT: movsd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq nearbyint ; CHECK-NEXT: callq nearbyint
; NO-FMA-NEXT: movsd %xmm0, (%rsp) # 8-byte Spill ; CHECK-NEXT: movsd %xmm0, (%rsp) # 8-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq nearbyint ; CHECK-NEXT: callq nearbyint
; NO-FMA-NEXT: movsd %xmm0, {{[0-9]+}}(%rsp) ; CHECK-NEXT: movsd %xmm0, {{[0-9]+}}(%rsp)
; NO-FMA-NEXT: fldl {{[0-9]+}}(%rsp) ; CHECK-NEXT: fldl {{[0-9]+}}(%rsp)
; NO-FMA-NEXT: movsd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 8-byte Reload ; CHECK-NEXT: movsd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 8-byte Reload
; NO-FMA-NEXT: # xmm0 = mem[0],zero ; CHECK-NEXT: # xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd (%rsp), %xmm1 # 8-byte Reload ; CHECK-NEXT: movsd (%rsp), %xmm1 # 8-byte Reload
; NO-FMA-NEXT: # xmm1 = mem[0],zero ; CHECK-NEXT: # xmm1 = mem[0],zero
; NO-FMA-NEXT: addq $24, %rsp ; CHECK-NEXT: addq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_nearby_v3f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vroundsd $12, %xmm0, %xmm0, %xmm0
; HAS-FMA-NEXT: vroundpd $12, {{.*}}(%rip), %xmm1
; HAS-FMA-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; HAS-FMA-NEXT: retq
entry: entry:
%nearby = call <3 x double> @llvm.experimental.constrained.nearbyint.v3f64( %nearby = call <3 x double> @llvm.experimental.constrained.nearbyint.v3f64(
<3 x double> <double 42.0, double 42.1, double 42.2>, <3 x double> <double 42.0, double 42.1, double 42.2>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x double> %nearby ret <3 x double> %nearby
} }
define <4 x double> @constrained_vector_nearbyint_v4f64() { define <4 x double> @constrained_vector_nearbyint_v4f64() {
; NO-FMA-LABEL: constrained_vector_nearbyint_v4f64: ; CHECK-LABEL: constrained_vector_nearbyint_v4f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $40, %rsp ; CHECK-NEXT: subq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 48 ; CHECK-NEXT: .cfi_def_cfa_offset 48
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq nearbyint ; CHECK-NEXT: callq nearbyint
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq nearbyint ; CHECK-NEXT: callq nearbyint
; NO-FMA-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0] ; CHECK-NEXT: # xmm0 = xmm0[0],mem[0]
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq nearbyint ; CHECK-NEXT: callq nearbyint
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: callq nearbyint ; CHECK-NEXT: callq nearbyint
; NO-FMA-NEXT: movaps %xmm0, %xmm1 ; CHECK-NEXT: movaps %xmm0, %xmm1
; NO-FMA-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm1 = xmm1[0],mem[0] ; CHECK-NEXT: # xmm1 = xmm1[0],mem[0]
; NO-FMA-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload ; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; NO-FMA-NEXT: addq $40, %rsp ; CHECK-NEXT: addq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_nearbyint_v4f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: vroundpd $12, {{.*}}(%rip), %ymm0
; HAS-FMA-NEXT: retq
entry: entry:
%nearby = call <4 x double> @llvm.experimental.constrained.nearbyint.v4f64( %nearby = call <4 x double> @llvm.experimental.constrained.nearbyint.v4f64(
<4 x double> <double 42.1, double 42.2, <4 x double> <double 42.1, double 42.2,
double 42.3, double 42.4>, double 42.3, double 42.4>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <4 x double> %nearby ret <4 x double> %nearby
} }
define <1 x float> @constrained_vector_maxnum_v1f32() { define <1 x float> @constrained_vector_maxnum_v1f32() {
; NO-FMA-LABEL: constrained_vector_maxnum_v1f32: ; CHECK-LABEL: constrained_vector_maxnum_v1f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: pushq %rax ; CHECK-NEXT: pushq %rax
; NO-FMA-NEXT: .cfi_def_cfa_offset 16 ; CHECK-NEXT: .cfi_def_cfa_offset 16
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq fmaxf ; CHECK-NEXT: callq fmaxf
; NO-FMA-NEXT: popq %rax ; CHECK-NEXT: popq %rax
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_maxnum_v1f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: pushq %rax
; HAS-FMA-NEXT: .cfi_def_cfa_offset 16
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq fmaxf
; HAS-FMA-NEXT: popq %rax
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%max = call <1 x float> @llvm.experimental.constrained.maxnum.v1f32( %max = call <1 x float> @llvm.experimental.constrained.maxnum.v1f32(
<1 x float> <float 42.0>, <1 x float> <float 41.0>, <1 x float> <float 42.0>, <1 x float> <float 41.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <1 x float> %max ret <1 x float> %max
} }
define <2 x double> @constrained_vector_maxnum_v2f64() { define <2 x double> @constrained_vector_maxnum_v2f64() {
; NO-FMA-LABEL: constrained_vector_maxnum_v2f64: ; CHECK-LABEL: constrained_vector_maxnum_v2f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $24, %rsp ; CHECK-NEXT: subq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 32 ; CHECK-NEXT: .cfi_def_cfa_offset 32
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq fmax ; CHECK-NEXT: callq fmax
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq fmax ; CHECK-NEXT: callq fmax
; NO-FMA-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0] ; CHECK-NEXT: # xmm0 = xmm0[0],mem[0]
; NO-FMA-NEXT: addq $24, %rsp ; CHECK-NEXT: addq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_maxnum_v2f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $24, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 32
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: callq fmax
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: callq fmax
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: addq $24, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%max = call <2 x double> @llvm.experimental.constrained.maxnum.v2f64( %max = call <2 x double> @llvm.experimental.constrained.maxnum.v2f64(
<2 x double> <double 43.0, double 42.0>, <2 x double> <double 43.0, double 42.0>,
<2 x double> <double 41.0, double 40.0>, <2 x double> <double 41.0, double 40.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <2 x double> %max ret <2 x double> %max
} }
define <3 x float> @constrained_vector_maxnum_v3f32() { define <3 x float> @constrained_vector_maxnum_v3f32() {
; NO-FMA-LABEL: constrained_vector_maxnum_v3f32: ; CHECK-LABEL: constrained_vector_maxnum_v3f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $40, %rsp ; CHECK-NEXT: subq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 48 ; CHECK-NEXT: .cfi_def_cfa_offset 48
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq fmaxf ; CHECK-NEXT: callq fmaxf
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq fmaxf ; CHECK-NEXT: callq fmaxf
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq fmaxf ; CHECK-NEXT: callq fmaxf
; NO-FMA-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload ; CHECK-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload
; NO-FMA-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; CHECK-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; NO-FMA-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm1 = xmm1[0],mem[0] ; CHECK-NEXT: # xmm1 = xmm1[0],mem[0]
; NO-FMA-NEXT: movaps %xmm1, %xmm0 ; CHECK-NEXT: movaps %xmm1, %xmm0
; NO-FMA-NEXT: addq $40, %rsp ; CHECK-NEXT: addq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_maxnum_v3f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 48
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq fmaxf
; HAS-FMA-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq fmaxf
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq fmaxf
; HAS-FMA-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload
; HAS-FMA-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[2,3]
; HAS-FMA-NEXT: vinsertps $32, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0,1],mem[0],xmm0[3]
; HAS-FMA-NEXT: addq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%max = call <3 x float> @llvm.experimental.constrained.maxnum.v3f32( %max = call <3 x float> @llvm.experimental.constrained.maxnum.v3f32(
<3 x float> <float 43.0, float 44.0, float 45.0>, <3 x float> <float 43.0, float 44.0, float 45.0>,
<3 x float> <float 41.0, float 42.0, float 43.0>, <3 x float> <float 41.0, float 42.0, float 43.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x float> %max ret <3 x float> %max
} }
define <3 x double> @constrained_vector_max_v3f64() { define <3 x double> @constrained_vector_max_v3f64() {
; NO-FMA-LABEL: constrained_vector_max_v3f64: ; CHECK-LABEL: constrained_vector_max_v3f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $24, %rsp ; CHECK-NEXT: subq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 32 ; CHECK-NEXT: .cfi_def_cfa_offset 32
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq fmax ; CHECK-NEXT: callq fmax
; NO-FMA-NEXT: movsd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill ; CHECK-NEXT: movsd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq fmax ; CHECK-NEXT: callq fmax
; NO-FMA-NEXT: movsd %xmm0, (%rsp) # 8-byte Spill ; CHECK-NEXT: movsd %xmm0, (%rsp) # 8-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq fmax ; CHECK-NEXT: callq fmax
; NO-FMA-NEXT: movsd %xmm0, {{[0-9]+}}(%rsp) ; CHECK-NEXT: movsd %xmm0, {{[0-9]+}}(%rsp)
; NO-FMA-NEXT: fldl {{[0-9]+}}(%rsp) ; CHECK-NEXT: fldl {{[0-9]+}}(%rsp)
; NO-FMA-NEXT: movsd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 8-byte Reload ; CHECK-NEXT: movsd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 8-byte Reload
; NO-FMA-NEXT: # xmm0 = mem[0],zero ; CHECK-NEXT: # xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd (%rsp), %xmm1 # 8-byte Reload ; CHECK-NEXT: movsd (%rsp), %xmm1 # 8-byte Reload
; NO-FMA-NEXT: # xmm1 = mem[0],zero ; CHECK-NEXT: # xmm1 = mem[0],zero
; NO-FMA-NEXT: addq $24, %rsp ; CHECK-NEXT: addq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_max_v3f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $56, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 64
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: callq fmax
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: callq fmax
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: vzeroupper
; HAS-FMA-NEXT: callq fmax
; HAS-FMA-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
; HAS-FMA-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; HAS-FMA-NEXT: addq $56, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%max = call <3 x double> @llvm.experimental.constrained.maxnum.v3f64( %max = call <3 x double> @llvm.experimental.constrained.maxnum.v3f64(
<3 x double> <double 43.0, double 44.0, double 45.0>, <3 x double> <double 43.0, double 44.0, double 45.0>,
<3 x double> <double 40.0, double 41.0, double 42.0>, <3 x double> <double 40.0, double 41.0, double 42.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x double> %max ret <3 x double> %max
} }
define <4 x double> @constrained_vector_maxnum_v4f64() { define <4 x double> @constrained_vector_maxnum_v4f64() {
; NO-FMA-LABEL: constrained_vector_maxnum_v4f64: ; CHECK-LABEL: constrained_vector_maxnum_v4f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $40, %rsp ; CHECK-NEXT: subq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 48 ; CHECK-NEXT: .cfi_def_cfa_offset 48
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq fmax ; CHECK-NEXT: callq fmax
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq fmax ; CHECK-NEXT: callq fmax
; NO-FMA-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0] ; CHECK-NEXT: # xmm0 = xmm0[0],mem[0]
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq fmax ; CHECK-NEXT: callq fmax
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq fmax ; CHECK-NEXT: callq fmax
; NO-FMA-NEXT: movaps %xmm0, %xmm1 ; CHECK-NEXT: movaps %xmm0, %xmm1
; NO-FMA-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm1 = xmm1[0],mem[0] ; CHECK-NEXT: # xmm1 = xmm1[0],mem[0]
; NO-FMA-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload ; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; NO-FMA-NEXT: addq $40, %rsp ; CHECK-NEXT: addq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_maxnum_v4f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 48
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: callq fmax
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: callq fmax
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: callq fmax
; HAS-FMA-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: callq fmax
; HAS-FMA-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vinsertf128 $1, (%rsp), %ymm0, %ymm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: addq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%max = call <4 x double> @llvm.experimental.constrained.maxnum.v4f64( %max = call <4 x double> @llvm.experimental.constrained.maxnum.v4f64(
<4 x double> <double 44.0, double 45.0, <4 x double> <double 44.0, double 45.0,
double 46.0, double 47.0>, double 46.0, double 47.0>,
<4 x double> <double 40.0, double 41.0, <4 x double> <double 40.0, double 41.0,
double 42.0, double 43.0>, double 42.0, double 43.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <4 x double> %max ret <4 x double> %max
} }
define <1 x float> @constrained_vector_minnum_v1f32() { define <1 x float> @constrained_vector_minnum_v1f32() {
; NO-FMA-LABEL: constrained_vector_minnum_v1f32: ; CHECK-LABEL: constrained_vector_minnum_v1f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: pushq %rax ; CHECK-NEXT: pushq %rax
; NO-FMA-NEXT: .cfi_def_cfa_offset 16 ; CHECK-NEXT: .cfi_def_cfa_offset 16
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq fminf ; CHECK-NEXT: callq fminf
; NO-FMA-NEXT: popq %rax ; CHECK-NEXT: popq %rax
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_minnum_v1f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: pushq %rax
; HAS-FMA-NEXT: .cfi_def_cfa_offset 16
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq fminf
; HAS-FMA-NEXT: popq %rax
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%min = call <1 x float> @llvm.experimental.constrained.minnum.v1f32( %min = call <1 x float> @llvm.experimental.constrained.minnum.v1f32(
<1 x float> <float 42.0>, <1 x float> <float 41.0>, <1 x float> <float 42.0>, <1 x float> <float 41.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <1 x float> %min ret <1 x float> %min
} }
define <2 x double> @constrained_vector_minnum_v2f64() { define <2 x double> @constrained_vector_minnum_v2f64() {
; NO-FMA-LABEL: constrained_vector_minnum_v2f64: ; CHECK-LABEL: constrained_vector_minnum_v2f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $24, %rsp ; CHECK-NEXT: subq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 32 ; CHECK-NEXT: .cfi_def_cfa_offset 32
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq fmin ; CHECK-NEXT: callq fmin
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq fmin ; CHECK-NEXT: callq fmin
; NO-FMA-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0] ; CHECK-NEXT: # xmm0 = xmm0[0],mem[0]
; NO-FMA-NEXT: addq $24, %rsp ; CHECK-NEXT: addq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_minnum_v2f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $24, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 32
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: callq fmin
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: callq fmin
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: addq $24, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%min = call <2 x double> @llvm.experimental.constrained.minnum.v2f64( %min = call <2 x double> @llvm.experimental.constrained.minnum.v2f64(
<2 x double> <double 43.0, double 42.0>, <2 x double> <double 43.0, double 42.0>,
<2 x double> <double 41.0, double 40.0>, <2 x double> <double 41.0, double 40.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <2 x double> %min ret <2 x double> %min
} }
define <3 x float> @constrained_vector_minnum_v3f32() { define <3 x float> @constrained_vector_minnum_v3f32() {
; NO-FMA-LABEL: constrained_vector_minnum_v3f32: ; CHECK-LABEL: constrained_vector_minnum_v3f32:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $40, %rsp ; CHECK-NEXT: subq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 48 ; CHECK-NEXT: .cfi_def_cfa_offset 48
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq fminf ; CHECK-NEXT: callq fminf
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq fminf ; CHECK-NEXT: callq fminf
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; NO-FMA-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero ; CHECK-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; NO-FMA-NEXT: callq fminf ; CHECK-NEXT: callq fminf
; NO-FMA-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload ; CHECK-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload
; NO-FMA-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; CHECK-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; NO-FMA-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm1 = xmm1[0],mem[0] ; CHECK-NEXT: # xmm1 = xmm1[0],mem[0]
; NO-FMA-NEXT: movaps %xmm1, %xmm0 ; CHECK-NEXT: movaps %xmm1, %xmm0
; NO-FMA-NEXT: addq $40, %rsp ; CHECK-NEXT: addq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_minnum_v3f32:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 48
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq fminf
; HAS-FMA-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq fminf
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; HAS-FMA-NEXT: callq fminf
; HAS-FMA-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload
; HAS-FMA-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[2,3]
; HAS-FMA-NEXT: vinsertps $32, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0,1],mem[0],xmm0[3]
; HAS-FMA-NEXT: addq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%min = call <3 x float> @llvm.experimental.constrained.minnum.v3f32( %min = call <3 x float> @llvm.experimental.constrained.minnum.v3f32(
<3 x float> <float 43.0, float 44.0, float 45.0>, <3 x float> <float 43.0, float 44.0, float 45.0>,
<3 x float> <float 41.0, float 42.0, float 43.0>, <3 x float> <float 41.0, float 42.0, float 43.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x float> %min ret <3 x float> %min
} }
define <3 x double> @constrained_vector_min_v3f64() {entry: define <3 x double> @constrained_vector_min_v3f64() {
; NO-FMA-LABEL: constrained_vector_min_v3f64: ; CHECK-LABEL: constrained_vector_min_v3f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $24, %rsp ; CHECK-NEXT: subq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 32 ; CHECK-NEXT: .cfi_def_cfa_offset 32
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq fmin ; CHECK-NEXT: callq fmin
; NO-FMA-NEXT: movsd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill ; CHECK-NEXT: movsd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq fmin ; CHECK-NEXT: callq fmin
; NO-FMA-NEXT: movsd %xmm0, (%rsp) # 8-byte Spill ; CHECK-NEXT: movsd %xmm0, (%rsp) # 8-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq fmin ; CHECK-NEXT: callq fmin
; NO-FMA-NEXT: movsd %xmm0, {{[0-9]+}}(%rsp) ; CHECK-NEXT: movsd %xmm0, {{[0-9]+}}(%rsp)
; NO-FMA-NEXT: fldl {{[0-9]+}}(%rsp) ; CHECK-NEXT: fldl {{[0-9]+}}(%rsp)
; NO-FMA-NEXT: movsd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 8-byte Reload ; CHECK-NEXT: movsd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 8-byte Reload
; NO-FMA-NEXT: # xmm0 = mem[0],zero ; CHECK-NEXT: # xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd (%rsp), %xmm1 # 8-byte Reload ; CHECK-NEXT: movsd (%rsp), %xmm1 # 8-byte Reload
; NO-FMA-NEXT: # xmm1 = mem[0],zero ; CHECK-NEXT: # xmm1 = mem[0],zero
; NO-FMA-NEXT: addq $24, %rsp ; CHECK-NEXT: addq $24, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
; entry:
; HAS-FMA-LABEL: constrained_vector_min_v3f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $56, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 64
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: callq fmin
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: callq fmin
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: vzeroupper
; HAS-FMA-NEXT: callq fmin
; HAS-FMA-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
; HAS-FMA-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; HAS-FMA-NEXT: addq $56, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
%min = call <3 x double> @llvm.experimental.constrained.minnum.v3f64( %min = call <3 x double> @llvm.experimental.constrained.minnum.v3f64(
<3 x double> <double 43.0, double 44.0, double 45.0>, <3 x double> <double 43.0, double 44.0, double 45.0>,
<3 x double> <double 40.0, double 41.0, double 42.0>, <3 x double> <double 40.0, double 41.0, double 42.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <3 x double> %min ret <3 x double> %min
} }
define <4 x double> @constrained_vector_minnum_v4f64() { define <4 x double> @constrained_vector_minnum_v4f64() {
; NO-FMA-LABEL: constrained_vector_minnum_v4f64: ; CHECK-LABEL: constrained_vector_minnum_v4f64:
; NO-FMA: # %bb.0: # %entry ; CHECK: # %bb.0: # %entry
; NO-FMA-NEXT: subq $40, %rsp ; CHECK-NEXT: subq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 48 ; CHECK-NEXT: .cfi_def_cfa_offset 48
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq fmin ; CHECK-NEXT: callq fmin
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq fmin ; CHECK-NEXT: callq fmin
; NO-FMA-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd (%rsp), %xmm0 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm0 = xmm0[0],mem[0] ; CHECK-NEXT: # xmm0 = xmm0[0],mem[0]
; NO-FMA-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq fmin ; CHECK-NEXT: callq fmin
; NO-FMA-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; NO-FMA-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; NO-FMA-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero ; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; NO-FMA-NEXT: callq fmin ; CHECK-NEXT: callq fmin
; NO-FMA-NEXT: movaps %xmm0, %xmm1 ; CHECK-NEXT: movaps %xmm0, %xmm1
; NO-FMA-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; CHECK-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; NO-FMA-NEXT: # xmm1 = xmm1[0],mem[0] ; CHECK-NEXT: # xmm1 = xmm1[0],mem[0]
; NO-FMA-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload ; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; NO-FMA-NEXT: addq $40, %rsp ; CHECK-NEXT: addq $40, %rsp
; NO-FMA-NEXT: .cfi_def_cfa_offset 8 ; CHECK-NEXT: .cfi_def_cfa_offset 8
; NO-FMA-NEXT: retq ; CHECK-NEXT: retq
;
; HAS-FMA-LABEL: constrained_vector_minnum_v4f64:
; HAS-FMA: # %bb.0: # %entry
; HAS-FMA-NEXT: subq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 48
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: callq fmin
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: callq fmin
; HAS-FMA-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: callq fmin
; HAS-FMA-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; HAS-FMA-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; HAS-FMA-NEXT: callq fmin
; HAS-FMA-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: # xmm0 = xmm0[0],mem[0]
; HAS-FMA-NEXT: vinsertf128 $1, (%rsp), %ymm0, %ymm0 # 16-byte Folded Reload
; HAS-FMA-NEXT: addq $40, %rsp
; HAS-FMA-NEXT: .cfi_def_cfa_offset 8
; HAS-FMA-NEXT: retq
entry: entry:
%min = call <4 x double> @llvm.experimental.constrained.minnum.v4f64( %min = call <4 x double> @llvm.experimental.constrained.minnum.v4f64(
<4 x double> <double 44.0, double 45.0, <4 x double> <double 44.0, double 45.0,
double 46.0, double 47.0>, double 46.0, double 47.0>,
<4 x double> <double 40.0, double 41.0, <4 x double> <double 40.0, double 41.0,
double 42.0, double 43.0>, double 42.0, double 43.0>,
metadata !"round.dynamic", metadata !"round.dynamic",
metadata !"fpexcept.strict") metadata !"fpexcept.strict")
ret <4 x double> %min ret <4 x double> %min
} }
; Single width declarations ; Single width declarations
declare <2 x double> @llvm.experimental.constrained.fadd.v2f64(<2 x double>, <2 x double>, metadata, metadata) declare <2 x double> @llvm.experimental.constrained.fadd.v2f64(<2 x double>, <2 x double>, metadata, metadata)
declare <2 x double> @llvm.experimental.constrained.fsub.v2f64(<2 x double>, <2 x double>, metadata, metadata) declare <2 x double> @llvm.experimental.constrained.fsub.v2f64(<2 x double>, <2 x double>, metadata, metadata)
declare <2 x double> @llvm.experimental.constrained.fmul.v2f64(<2 x double>, <2 x double>, metadata, metadata) declare <2 x double> @llvm.experimental.constrained.fmul.v2f64(<2 x double>, <2 x double>, metadata, metadata)
declare <2 x double> @llvm.experimental.constrained.fdiv.v2f64(<2 x double>, <2 x double>, metadata, metadata) declare <2 x double> @llvm.experimental.constrained.fdiv.v2f64(<2 x double>, <2 x double>, metadata, metadata)
declare <2 x double> @llvm.experimental.constrained.frem.v2f64(<2 x double>, <2 x double>, metadata, metadata) declare <2 x double> @llvm.experimental.constrained.frem.v2f64(<2 x double>, <2 x double>, metadata, metadata)
declare <2 x double> @llvm.experimental.constrained.fma.v2f64(<2 x double>, <2 x double>, <2 x double>, metadata, metadata)
declare <4 x float> @llvm.experimental.constrained.fma.v4f32(<4 x float>, <4 x float>, <4 x float>, metadata, metadata)
declare <2 x double> @llvm.experimental.constrained.sqrt.v2f64(<2 x double>, metadata, metadata) declare <2 x double> @llvm.experimental.constrained.sqrt.v2f64(<2 x double>, metadata, metadata)
declare <2 x double> @llvm.experimental.constrained.pow.v2f64(<2 x double>, <2 x double>, metadata, metadata) declare <2 x double> @llvm.experimental.constrained.pow.v2f64(<2 x double>, <2 x double>, metadata, metadata)
declare <2 x double> @llvm.experimental.constrained.powi.v2f64(<2 x double>, i32, metadata, metadata) declare <2 x double> @llvm.experimental.constrained.powi.v2f64(<2 x double>, i32, metadata, metadata)
declare <2 x double> @llvm.experimental.constrained.sin.v2f64(<2 x double>, metadata, metadata) declare <2 x double> @llvm.experimental.constrained.sin.v2f64(<2 x double>, metadata, metadata)
declare <2 x double> @llvm.experimental.constrained.cos.v2f64(<2 x double>, metadata, metadata) declare <2 x double> @llvm.experimental.constrained.cos.v2f64(<2 x double>, metadata, metadata)
declare <2 x double> @llvm.experimental.constrained.exp.v2f64(<2 x double>, metadata, metadata) declare <2 x double> @llvm.experimental.constrained.exp.v2f64(<2 x double>, metadata, metadata)
declare <2 x double> @llvm.experimental.constrained.exp2.v2f64(<2 x double>, metadata, metadata) declare <2 x double> @llvm.experimental.constrained.exp2.v2f64(<2 x double>, metadata, metadata)
declare <2 x double> @llvm.experimental.constrained.log.v2f64(<2 x double>, metadata, metadata) declare <2 x double> @llvm.experimental.constrained.log.v2f64(<2 x double>, metadata, metadata)
declare <2 x double> @llvm.experimental.constrained.log10.v2f64(<2 x double>, metadata, metadata) declare <2 x double> @llvm.experimental.constrained.log10.v2f64(<2 x double>, metadata, metadata)
declare <2 x double> @llvm.experimental.constrained.log2.v2f64(<2 x double>, metadata, metadata) declare <2 x double> @llvm.experimental.constrained.log2.v2f64(<2 x double>, metadata, metadata)
declare <2 x double> @llvm.experimental.constrained.rint.v2f64(<2 x double>, metadata, metadata) declare <2 x double> @llvm.experimental.constrained.rint.v2f64(<2 x double>, metadata, metadata)
declare <2 x double> @llvm.experimental.constrained.nearbyint.v2f64(<2 x double>, metadata, metadata) declare <2 x double> @llvm.experimental.constrained.nearbyint.v2f64(<2 x double>, metadata, metadata)
declare <2 x double> @llvm.experimental.constrained.maxnum.v2f64(<2 x double>, <2 x double>, metadata, metadata) declare <2 x double> @llvm.experimental.constrained.maxnum.v2f64(<2 x double>, <2 x double>, metadata, metadata)
declare <2 x double> @llvm.experimental.constrained.minnum.v2f64(<2 x double>, <2 x double>, metadata, metadata) declare <2 x double> @llvm.experimental.constrained.minnum.v2f64(<2 x double>, <2 x double>, metadata, metadata)
; Scalar width declarations ; Scalar width declarations
declare <1 x float> @llvm.experimental.constrained.fadd.v1f32(<1 x float>, <1 x float>, metadata, metadata) declare <1 x float> @llvm.experimental.constrained.fadd.v1f32(<1 x float>, <1 x float>, metadata, metadata)
declare <1 x float> @llvm.experimental.constrained.fsub.v1f32(<1 x float>, <1 x float>, metadata, metadata) declare <1 x float> @llvm.experimental.constrained.fsub.v1f32(<1 x float>, <1 x float>, metadata, metadata)
declare <1 x float> @llvm.experimental.constrained.fmul.v1f32(<1 x float>, <1 x float>, metadata, metadata) declare <1 x float> @llvm.experimental.constrained.fmul.v1f32(<1 x float>, <1 x float>, metadata, metadata)
declare <1 x float> @llvm.experimental.constrained.fdiv.v1f32(<1 x float>, <1 x float>, metadata, metadata) declare <1 x float> @llvm.experimental.constrained.fdiv.v1f32(<1 x float>, <1 x float>, metadata, metadata)
declare <1 x float> @llvm.experimental.constrained.frem.v1f32(<1 x float>, <1 x float>, metadata, metadata) declare <1 x float> @llvm.experimental.constrained.frem.v1f32(<1 x float>, <1 x float>, metadata, metadata)
declare <1 x float> @llvm.experimental.constrained.fma.v1f32(<1 x float>, <1 x float>, <1 x float>, metadata, metadata)
declare <1 x float> @llvm.experimental.constrained.sqrt.v1f32(<1 x float>, metadata, metadata) declare <1 x float> @llvm.experimental.constrained.sqrt.v1f32(<1 x float>, metadata, metadata)
declare <1 x float> @llvm.experimental.constrained.pow.v1f32(<1 x float>, <1 x float>, metadata, metadata) declare <1 x float> @llvm.experimental.constrained.pow.v1f32(<1 x float>, <1 x float>, metadata, metadata)
declare <1 x float> @llvm.experimental.constrained.powi.v1f32(<1 x float>, i32, metadata, metadata) declare <1 x float> @llvm.experimental.constrained.powi.v1f32(<1 x float>, i32, metadata, metadata)
declare <1 x float> @llvm.experimental.constrained.sin.v1f32(<1 x float>, metadata, metadata) declare <1 x float> @llvm.experimental.constrained.sin.v1f32(<1 x float>, metadata, metadata)
declare <1 x float> @llvm.experimental.constrained.cos.v1f32(<1 x float>, metadata, metadata) declare <1 x float> @llvm.experimental.constrained.cos.v1f32(<1 x float>, metadata, metadata)
declare <1 x float> @llvm.experimental.constrained.exp.v1f32(<1 x float>, metadata, metadata) declare <1 x float> @llvm.experimental.constrained.exp.v1f32(<1 x float>, metadata, metadata)
declare <1 x float> @llvm.experimental.constrained.exp2.v1f32(<1 x float>, metadata, metadata) declare <1 x float> @llvm.experimental.constrained.exp2.v1f32(<1 x float>, metadata, metadata)
declare <1 x float> @llvm.experimental.constrained.log.v1f32(<1 x float>, metadata, metadata) declare <1 x float> @llvm.experimental.constrained.log.v1f32(<1 x float>, metadata, metadata)
Show All 10 Lines
declare <3 x float> @llvm.experimental.constrained.fsub.v3f32(<3 x float>, <3 x float>, metadata, metadata) declare <3 x float> @llvm.experimental.constrained.fsub.v3f32(<3 x float>, <3 x float>, metadata, metadata)
declare <3 x double> @llvm.experimental.constrained.fsub.v3f64(<3 x double>, <3 x double>, metadata, metadata) declare <3 x double> @llvm.experimental.constrained.fsub.v3f64(<3 x double>, <3 x double>, metadata, metadata)
declare <3 x float> @llvm.experimental.constrained.fmul.v3f32(<3 x float>, <3 x float>, metadata, metadata) declare <3 x float> @llvm.experimental.constrained.fmul.v3f32(<3 x float>, <3 x float>, metadata, metadata)
declare <3 x double> @llvm.experimental.constrained.fmul.v3f64(<3 x double>, <3 x double>, metadata, metadata) declare <3 x double> @llvm.experimental.constrained.fmul.v3f64(<3 x double>, <3 x double>, metadata, metadata)
declare <3 x float> @llvm.experimental.constrained.fdiv.v3f32(<3 x float>, <3 x float>, metadata, metadata) declare <3 x float> @llvm.experimental.constrained.fdiv.v3f32(<3 x float>, <3 x float>, metadata, metadata)
declare <3 x double> @llvm.experimental.constrained.fdiv.v3f64(<3 x double>, <3 x double>, metadata, metadata) declare <3 x double> @llvm.experimental.constrained.fdiv.v3f64(<3 x double>, <3 x double>, metadata, metadata)
declare <3 x float> @llvm.experimental.constrained.frem.v3f32(<3 x float>, <3 x float>, metadata, metadata) declare <3 x float> @llvm.experimental.constrained.frem.v3f32(<3 x float>, <3 x float>, metadata, metadata)
declare <3 x double> @llvm.experimental.constrained.frem.v3f64(<3 x double>, <3 x double>, metadata, metadata) declare <3 x double> @llvm.experimental.constrained.frem.v3f64(<3 x double>, <3 x double>, metadata, metadata)
declare <3 x float> @llvm.experimental.constrained.fma.v3f32(<3 x float>, <3 x float>, <3 x float>, metadata, metadata)
declare <3 x double> @llvm.experimental.constrained.fma.v3f64(<3 x double>, <3 x double>, <3 x double>, metadata, metadata)
declare <3 x float> @llvm.experimental.constrained.sqrt.v3f32(<3 x float>, metadata, metadata) declare <3 x float> @llvm.experimental.constrained.sqrt.v3f32(<3 x float>, metadata, metadata)
declare <3 x double> @llvm.experimental.constrained.sqrt.v3f64(<3 x double>, metadata, metadata) declare <3 x double> @llvm.experimental.constrained.sqrt.v3f64(<3 x double>, metadata, metadata)
declare <3 x float> @llvm.experimental.constrained.pow.v3f32(<3 x float>, <3 x float>, metadata, metadata) declare <3 x float> @llvm.experimental.constrained.pow.v3f32(<3 x float>, <3 x float>, metadata, metadata)
declare <3 x double> @llvm.experimental.constrained.pow.v3f64(<3 x double>, <3 x double>, metadata, metadata) declare <3 x double> @llvm.experimental.constrained.pow.v3f64(<3 x double>, <3 x double>, metadata, metadata)
declare <3 x float> @llvm.experimental.constrained.powi.v3f32(<3 x float>, i32, metadata, metadata) declare <3 x float> @llvm.experimental.constrained.powi.v3f32(<3 x float>, i32, metadata, metadata)
declare <3 x double> @llvm.experimental.constrained.powi.v3f64(<3 x double>, i32, metadata, metadata) declare <3 x double> @llvm.experimental.constrained.powi.v3f64(<3 x double>, i32, metadata, metadata)
declare <3 x float> @llvm.experimental.constrained.sin.v3f32(<3 x float>, metadata, metadata) declare <3 x float> @llvm.experimental.constrained.sin.v3f32(<3 x float>, metadata, metadata)
declare <3 x double> @llvm.experimental.constrained.sin.v3f64(<3 x double>, metadata, metadata) declare <3 x double> @llvm.experimental.constrained.sin.v3f64(<3 x double>, metadata, metadata)
Show All 19 Lines
declare <3 x double> @llvm.experimental.constrained.minnum.v3f64(<3 x double>, <3 x double>, metadata, metadata) declare <3 x double> @llvm.experimental.constrained.minnum.v3f64(<3 x double>, <3 x double>, metadata, metadata)
; Double width declarations ; Double width declarations
declare <4 x double> @llvm.experimental.constrained.fadd.v4f64(<4 x double>, <4 x double>, metadata, metadata) declare <4 x double> @llvm.experimental.constrained.fadd.v4f64(<4 x double>, <4 x double>, metadata, metadata)
declare <4 x double> @llvm.experimental.constrained.fsub.v4f64(<4 x double>, <4 x double>, metadata, metadata) declare <4 x double> @llvm.experimental.constrained.fsub.v4f64(<4 x double>, <4 x double>, metadata, metadata)
declare <4 x double> @llvm.experimental.constrained.fmul.v4f64(<4 x double>, <4 x double>, metadata, metadata) declare <4 x double> @llvm.experimental.constrained.fmul.v4f64(<4 x double>, <4 x double>, metadata, metadata)
declare <4 x double> @llvm.experimental.constrained.fdiv.v4f64(<4 x double>, <4 x double>, metadata, metadata) declare <4 x double> @llvm.experimental.constrained.fdiv.v4f64(<4 x double>, <4 x double>, metadata, metadata)
declare <4 x double> @llvm.experimental.constrained.frem.v4f64(<4 x double>, <4 x double>, metadata, metadata) declare <4 x double> @llvm.experimental.constrained.frem.v4f64(<4 x double>, <4 x double>, metadata, metadata)
declare <4 x double> @llvm.experimental.constrained.fma.v4f64(<4 x double>, <4 x double>, <4 x double>, metadata, metadata)
declare <8 x float> @llvm.experimental.constrained.fma.v8f32(<8 x float>, <8 x float>, <8 x float>, metadata, metadata)
declare <4 x double> @llvm.experimental.constrained.sqrt.v4f64(<4 x double>, metadata, metadata) declare <4 x double> @llvm.experimental.constrained.sqrt.v4f64(<4 x double>, metadata, metadata)
declare <4 x double> @llvm.experimental.constrained.pow.v4f64(<4 x double>, <4 x double>, metadata, metadata) declare <4 x double> @llvm.experimental.constrained.pow.v4f64(<4 x double>, <4 x double>, metadata, metadata)
declare <4 x double> @llvm.experimental.constrained.powi.v4f64(<4 x double>, i32, metadata, metadata) declare <4 x double> @llvm.experimental.constrained.powi.v4f64(<4 x double>, i32, metadata, metadata)
declare <4 x double> @llvm.experimental.constrained.sin.v4f64(<4 x double>, metadata, metadata) declare <4 x double> @llvm.experimental.constrained.sin.v4f64(<4 x double>, metadata, metadata)
declare <4 x double> @llvm.experimental.constrained.cos.v4f64(<4 x double>, metadata, metadata) declare <4 x double> @llvm.experimental.constrained.cos.v4f64(<4 x double>, metadata, metadata)
declare <4 x double> @llvm.experimental.constrained.exp.v4f64(<4 x double>, metadata, metadata) declare <4 x double> @llvm.experimental.constrained.exp.v4f64(<4 x double>, metadata, metadata)
declare <4 x double> @llvm.experimental.constrained.exp2.v4f64(<4 x double>, metadata, metadata) declare <4 x double> @llvm.experimental.constrained.exp2.v4f64(<4 x double>, metadata, metadata)
declare <4 x double> @llvm.experimental.constrained.log.v4f64(<4 x double>, metadata, metadata) declare <4 x double> @llvm.experimental.constrained.log.v4f64(<4 x double>, metadata, metadata)
declare <4 x double> @llvm.experimental.constrained.log10.v4f64(<4 x double>, metadata, metadata) declare <4 x double> @llvm.experimental.constrained.log10.v4f64(<4 x double>, metadata, metadata)
declare <4 x double> @llvm.experimental.constrained.log2.v4f64(<4 x double>, metadata, metadata) declare <4 x double> @llvm.experimental.constrained.log2.v4f64(<4 x double>, metadata, metadata)
declare <4 x double> @llvm.experimental.constrained.rint.v4f64(<4 x double>, metadata, metadata) declare <4 x double> @llvm.experimental.constrained.rint.v4f64(<4 x double>, metadata, metadata)
declare <4 x double> @llvm.experimental.constrained.nearbyint.v4f64(<4 x double>, metadata, metadata) declare <4 x double> @llvm.experimental.constrained.nearbyint.v4f64(<4 x double>, metadata, metadata)
declare <4 x double> @llvm.experimental.constrained.maxnum.v4f64(<4 x double>, <4 x double>, metadata, metadata) declare <4 x double> @llvm.experimental.constrained.maxnum.v4f64(<4 x double>, <4 x double>, metadata, metadata)
declare <4 x double> @llvm.experimental.constrained.minnum.v4f64(<4 x double>, <4 x double>, metadata, metadata) declare <4 x double> @llvm.experimental.constrained.minnum.v4f64(<4 x double>, <4 x double>, metadata, metadata)
test/CodeGen/X86/vector-constrained-fp-intrinsics-fma.ll