Skip to content

[Analysis] isTriviallyVectorizable - add vectorization support for acos/asin/atan and cosh/sinh/tanh intrinsics #106584

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 3 commits into from
Aug 30, 2024
Merged

[Analysis] isTriviallyVectorizable - add vectorization support for acos/asin/atan and cosh/sinh/tanh intrinsics #106584

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
399ec2e
[LoopVectorize][X86] amdlibm-calls.ll - add tests showing failure to …
RKSimon Aug 29, 2024
5560892
[Analysis] isTriviallyVectorizable - add vectorization support for ac…
RKSimon Aug 29, 2024
86e6eb7
[AArch64] Regenerate accelerate-vector-functions.ll tests
RKSimon Aug 30, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 6 additions & 0 deletions llvm/lib/Analysis/VectorUtils.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -66,9 +66,15 @@ bool llvm::isTriviallyVectorizable(Intrinsic::ID ID) {
case Intrinsic::umul_fix:
case Intrinsic::umul_fix_sat:
case Intrinsic::sqrt: // Begin floating-point.
case Intrinsic::asin:
case Intrinsic::acos:
case Intrinsic::atan:
case Intrinsic::sin:
case Intrinsic::cos:
case Intrinsic::tan:
case Intrinsic::sinh:
case Intrinsic::cosh:
case Intrinsic::tanh:
case Intrinsic::exp:
case Intrinsic::exp2:
case Intrinsic::log:
Expand Down
13 changes: 13 additions & 0 deletions llvm/test/Transforms/LoopVectorize/X86/amdlibm-calls.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -414,6 +414,7 @@ for.end:

define void @acos_f32_intrinsic(ptr nocapture %varray) {
; CHECK-LABEL: @acos_f32_intrinsic(
; CHECK-VF2: [[TMP5:%.*]] = call <2 x float> @llvm.acos.v2f32(<2 x float> [[TMP4:%.*]])
; CHECK-VF4: [[TMP5:%.*]] = call <4 x float> @amd_vrs4_acosf(<4 x float> [[TMP4:%.*]])
; CHECK-VF8: [[TMP5:%.*]] = call <8 x float> @amd_vrs8_acosf(<8 x float> [[TMP4:%.*]])
; CHECK-VF16: [[TMP5:%.*]] = call <16 x float> @amd_vrs16_acosf(<16 x float> [[TMP4:%.*]])
Expand Down Expand Up @@ -487,7 +488,10 @@ for.end:

define void @asin_f64_intrinsic(ptr nocapture %varray) {
; CHECK-LABEL: @asin_f64_intrinsic(
; CHECK-VF2: [[TMP5:%.*]] = call <2 x double> @llvm.asin.v2f64(<2 x double> [[TMP4:%.*]])
; CHECK-VF4: [[TMP5:%.*]] = call <4 x double> @llvm.asin.v4f64(<4 x double> [[TMP4:%.*]])
; CHECK-VF8: [[TMP5:%.*]] = call <8 x double> @amd_vrd8_asin(<8 x double> [[TMP4:%.*]])
; CHECK-VF16: [[TMP5:%.*]] = call <16 x double> @llvm.asin.v16f64(<16 x double> [[TMP4:%.*]])
; CHECK: ret void
;
entry:
Expand All @@ -510,6 +514,7 @@ for.end:

define void @asin_f32_intrinsic(ptr nocapture %varray) {
; CHECK-LABEL: @asin_f32_intrinsic(
; CHECK-VF2: [[TMP5:%.*]] = call <2 x float> @llvm.asin.v2f32(<2 x float> [[TMP4:%.*]])
; CHECK-VF4: [[TMP5:%.*]] = call <4 x float> @amd_vrs4_asinf(<4 x float> [[TMP4:%.*]])
; CHECK-VF8: [[TMP5:%.*]] = call <8 x float> @amd_vrs8_asinf(<8 x float> [[TMP4:%.*]])
; CHECK-VF16: [[TMP5:%.*]] = call <16 x float> @amd_vrs16_asinf(<16 x float> [[TMP4:%.*]])
Expand Down Expand Up @@ -588,6 +593,7 @@ define void @atan_f64_intrinsic(ptr nocapture %varray) {
; CHECK-VF2: [[TMP5:%.*]] = call <2 x double> @amd_vrd2_atan(<2 x double> [[TMP4:%.*]])
; CHECK-VF4: [[TMP5:%.*]] = call <4 x double> @amd_vrd4_atan(<4 x double> [[TMP4:%.*]])
; CHECK-VF8: [[TMP5:%.*]] = call <8 x double> @amd_vrd8_atan(<8 x double> [[TMP4:%.*]])
; CHECK-VF16: [[TMP5:%.*]] = call <16 x double> @llvm.atan.v16f64(<16 x double> [[TMP4:%.*]])
; CHECK: ret void
;
entry:
Expand All @@ -610,6 +616,7 @@ for.end:

define void @atan_f32_intrinsic(ptr nocapture %varray) {
; CHECK-LABEL: @atan_f32_intrinsic(
; CHECK-VF2: [[TMP5:%.*]] = call <2 x float> @llvm.atan.v2f32(<2 x float> [[TMP4:%.*]])
; CHECK-VF4: [[TMP5:%.*]] = call <4 x float> @amd_vrs4_atanf(<4 x float> [[TMP4:%.*]])
; CHECK-VF8: [[TMP5:%.*]] = call <8 x float> @amd_vrs8_atanf(<8 x float> [[TMP4:%.*]])
; CHECK-VF16: [[TMP5:%.*]] = call <16 x float> @amd_vrs16_atanf(<16 x float> [[TMP4:%.*]])
Expand Down Expand Up @@ -683,6 +690,9 @@ for.end:
define void @cosh_f64_intrinsic(ptr nocapture %varray) {
; CHECK-LABEL: @cosh_f64_intrinsic(
; CHECK-VF2: [[TMP5:%.*]] = call <2 x double> @amd_vrd2_cosh(<2 x double> [[TMP4:%.*]])
; CHECK-VF4: [[TMP5:%.*]] = call <4 x double> @llvm.cosh.v4f64(<4 x double> [[TMP4:%.*]])
; CHECK-VF8: [[TMP5:%.*]] = call <8 x double> @llvm.cosh.v8f64(<8 x double> [[TMP4:%.*]])
; CHECK-VF16: [[TMP5:%.*]] = call <16 x double> @llvm.cosh.v16f64(<16 x double> [[TMP4:%.*]])
; CHECK: ret void
;
entry:
Expand All @@ -705,8 +715,10 @@ for.end:

define void @cosh_f32_intrinsic(ptr nocapture %varray) {
; CHECK-LABEL: @cosh_f32_intrinsic(
; CHECK-VF2: [[TMP5:%.*]] = call <2 x float> @llvm.cosh.v2f32(<2 x float> [[TMP4:%.*]])
; CHECK-VF4: [[TMP5:%.*]] = call <4 x float> @amd_vrs4_coshf(<4 x float> [[TMP4:%.*]])
; CHECK-VF8: [[TMP5:%.*]] = call <8 x float> @amd_vrs8_coshf(<8 x float> [[TMP4:%.*]])
; CHECK-VF16: [[TMP5:%.*]] = call <16 x float> @llvm.cosh.v16f32(<16 x float> [[TMP4:%.*]])
; CHECK: ret void
;
entry:
Expand Down Expand Up @@ -754,6 +766,7 @@ for.end:

define void @tanh_f32_intrinsic(ptr nocapture %varray) {
; CHECK-LABEL: @tanh_f32_intrinsic(
; CHECK-VF2: [[TMP5:%.*]] = call <2 x float> @llvm.tanh.v2f32(<2 x float> [[TMP4:%.*]])
; CHECK-VF4: [[TMP5:%.*]] = call <4 x float> @amd_vrs4_tanhf(<4 x float> [[TMP4:%.*]])
; CHECK-VF8: [[TMP5:%.*]] = call <8 x float> @amd_vrs8_tanhf(<8 x float> [[TMP4:%.*]])
; CHECK-VF16: [[TMP5:%.*]] = call <16 x float> @amd_vrs16_tanhf(<16 x float> [[TMP4:%.*]])
Expand Down
72 changes: 30 additions & 42 deletions llvm/test/Transforms/SLPVectorizer/AArch64/accelerate-vector-functions-inseltpoison.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -627,13 +627,11 @@ define <4 x float> @int_asin_4x(ptr %a) {
; NOACCELERATE-NEXT: [[VECEXT_1:%.*]] = extractelement <4 x float> [[TMP0]], i32 1
; NOACCELERATE-NEXT: [[TMP2:%.*]] = tail call fast float @llvm.asin.f32(float [[VECEXT_1]])
; NOACCELERATE-NEXT: [[VECINS_1:%.*]] = insertelement <4 x float> [[VECINS]], float [[TMP2]], i32 1
; NOACCELERATE-NEXT: [[VECEXT_2:%.*]] = extractelement <4 x float> [[TMP0]], i32 2
; NOACCELERATE-NEXT: [[TMP3:%.*]] = tail call fast float @llvm.asin.f32(float [[VECEXT_2]])
; NOACCELERATE-NEXT: [[VECINS_2:%.*]] = insertelement <4 x float> [[VECINS_1]], float [[TMP3]], i32 2
; NOACCELERATE-NEXT: [[VECEXT_3:%.*]] = extractelement <4 x float> [[TMP0]], i32 3
; NOACCELERATE-NEXT: [[TMP4:%.*]] = tail call fast float @llvm.asin.f32(float [[VECEXT_3]])
; NOACCELERATE-NEXT: [[VECINS_3:%.*]] = insertelement <4 x float> [[VECINS_2]], float [[TMP4]], i32 3
; NOACCELERATE-NEXT: ret <4 x float> [[VECINS_3]]
; NOACCELERATE-NEXT: [[TMP3:%.*]] = shufflevector <4 x float> [[TMP0]], <4 x float> poison, <2 x i32> <i32 2, i32 3>
; NOACCELERATE-NEXT: [[TMP4:%.*]] = call fast <2 x float> @llvm.asin.v2f32(<2 x float> [[TMP3]])
; NOACCELERATE-NEXT: [[TMP5:%.*]] = shufflevector <2 x float> [[TMP4]], <2 x float> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
; NOACCELERATE-NEXT: [[VECINS_31:%.*]] = shufflevector <4 x float> [[VECINS_1]], <4 x float> [[TMP5]], <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; NOACCELERATE-NEXT: ret <4 x float> [[VECINS_31]]
;
entry:
%0 = load <4 x float>, ptr %a, align 16
Expand Down Expand Up @@ -708,13 +706,11 @@ define <4 x float> @int_acos_4x(ptr %a) {
; NOACCELERATE-NEXT: [[VECEXT_1:%.*]] = extractelement <4 x float> [[TMP0]], i32 1
; NOACCELERATE-NEXT: [[TMP2:%.*]] = tail call fast float @llvm.acos.f32(float [[VECEXT_1]])
; NOACCELERATE-NEXT: [[VECINS_1:%.*]] = insertelement <4 x float> [[VECINS]], float [[TMP2]], i32 1
; NOACCELERATE-NEXT: [[VECEXT_2:%.*]] = extractelement <4 x float> [[TMP0]], i32 2
; NOACCELERATE-NEXT: [[TMP3:%.*]] = tail call fast float @llvm.acos.f32(float [[VECEXT_2]])
; NOACCELERATE-NEXT: [[VECINS_2:%.*]] = insertelement <4 x float> [[VECINS_1]], float [[TMP3]], i32 2
; NOACCELERATE-NEXT: [[VECEXT_3:%.*]] = extractelement <4 x float> [[TMP0]], i32 3
; NOACCELERATE-NEXT: [[TMP4:%.*]] = tail call fast float @llvm.acos.f32(float [[VECEXT_3]])
; NOACCELERATE-NEXT: [[VECINS_3:%.*]] = insertelement <4 x float> [[VECINS_2]], float [[TMP4]], i32 3
; NOACCELERATE-NEXT: ret <4 x float> [[VECINS_3]]
; NOACCELERATE-NEXT: [[TMP3:%.*]] = shufflevector <4 x float> [[TMP0]], <4 x float> poison, <2 x i32> <i32 2, i32 3>
; NOACCELERATE-NEXT: [[TMP4:%.*]] = call fast <2 x float> @llvm.acos.v2f32(<2 x float> [[TMP3]])
; NOACCELERATE-NEXT: [[TMP5:%.*]] = shufflevector <2 x float> [[TMP4]], <2 x float> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
; NOACCELERATE-NEXT: [[VECINS_31:%.*]] = shufflevector <4 x float> [[VECINS_1]], <4 x float> [[TMP5]], <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; NOACCELERATE-NEXT: ret <4 x float> [[VECINS_31]]
;
entry:
%0 = load <4 x float>, ptr %a, align 16
Expand Down Expand Up @@ -789,13 +785,11 @@ define <4 x float> @int_atan_4x(ptr %a) {
; NOACCELERATE-NEXT: [[VECEXT_1:%.*]] = extractelement <4 x float> [[TMP0]], i32 1
; NOACCELERATE-NEXT: [[TMP2:%.*]] = tail call fast float @llvm.atan.f32(float [[VECEXT_1]])
; NOACCELERATE-NEXT: [[VECINS_1:%.*]] = insertelement <4 x float> [[VECINS]], float [[TMP2]], i32 1
; NOACCELERATE-NEXT: [[VECEXT_2:%.*]] = extractelement <4 x float> [[TMP0]], i32 2
; NOACCELERATE-NEXT: [[TMP3:%.*]] = tail call fast float @llvm.atan.f32(float [[VECEXT_2]])
; NOACCELERATE-NEXT: [[VECINS_2:%.*]] = insertelement <4 x float> [[VECINS_1]], float [[TMP3]], i32 2
; NOACCELERATE-NEXT: [[VECEXT_3:%.*]] = extractelement <4 x float> [[TMP0]], i32 3
; NOACCELERATE-NEXT: [[TMP4:%.*]] = tail call fast float @llvm.atan.f32(float [[VECEXT_3]])
; NOACCELERATE-NEXT: [[VECINS_3:%.*]] = insertelement <4 x float> [[VECINS_2]], float [[TMP4]], i32 3
; NOACCELERATE-NEXT: ret <4 x float> [[VECINS_3]]
; NOACCELERATE-NEXT: [[TMP3:%.*]] = shufflevector <4 x float> [[TMP0]], <4 x float> poison, <2 x i32> <i32 2, i32 3>
; NOACCELERATE-NEXT: [[TMP4:%.*]] = call fast <2 x float> @llvm.atan.v2f32(<2 x float> [[TMP3]])
; NOACCELERATE-NEXT: [[TMP5:%.*]] = shufflevector <2 x float> [[TMP4]], <2 x float> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
; NOACCELERATE-NEXT: [[VECINS_31:%.*]] = shufflevector <4 x float> [[VECINS_1]], <4 x float> [[TMP5]], <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; NOACCELERATE-NEXT: ret <4 x float> [[VECINS_31]]
;
entry:
%0 = load <4 x float>, ptr %a, align 16
Expand Down Expand Up @@ -870,13 +864,11 @@ define <4 x float> @int_sinh_4x(ptr %a) {
; NOACCELERATE-NEXT: [[VECEXT_1:%.*]] = extractelement <4 x float> [[TMP0]], i32 1
; NOACCELERATE-NEXT: [[TMP2:%.*]] = tail call fast float @llvm.sinh.f32(float [[VECEXT_1]])
; NOACCELERATE-NEXT: [[VECINS_1:%.*]] = insertelement <4 x float> [[VECINS]], float [[TMP2]], i32 1
; NOACCELERATE-NEXT: [[VECEXT_2:%.*]] = extractelement <4 x float> [[TMP0]], i32 2
; NOACCELERATE-NEXT: [[TMP3:%.*]] = tail call fast float @llvm.sinh.f32(float [[VECEXT_2]])
; NOACCELERATE-NEXT: [[VECINS_2:%.*]] = insertelement <4 x float> [[VECINS_1]], float [[TMP3]], i32 2
; NOACCELERATE-NEXT: [[VECEXT_3:%.*]] = extractelement <4 x float> [[TMP0]], i32 3
; NOACCELERATE-NEXT: [[TMP4:%.*]] = tail call fast float @llvm.sinh.f32(float [[VECEXT_3]])
; NOACCELERATE-NEXT: [[VECINS_3:%.*]] = insertelement <4 x float> [[VECINS_2]], float [[TMP4]], i32 3
; NOACCELERATE-NEXT: ret <4 x float> [[VECINS_3]]
; NOACCELERATE-NEXT: [[TMP3:%.*]] = shufflevector <4 x float> [[TMP0]], <4 x float> poison, <2 x i32> <i32 2, i32 3>
; NOACCELERATE-NEXT: [[TMP4:%.*]] = call fast <2 x float> @llvm.sinh.v2f32(<2 x float> [[TMP3]])
; NOACCELERATE-NEXT: [[TMP5:%.*]] = shufflevector <2 x float> [[TMP4]], <2 x float> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
; NOACCELERATE-NEXT: [[VECINS_31:%.*]] = shufflevector <4 x float> [[VECINS_1]], <4 x float> [[TMP5]], <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; NOACCELERATE-NEXT: ret <4 x float> [[VECINS_31]]
;
entry:
%0 = load <4 x float>, ptr %a, align 16
Expand Down Expand Up @@ -951,13 +943,11 @@ define <4 x float> @int_cosh_4x(ptr %a) {
; NOACCELERATE-NEXT: [[VECEXT_1:%.*]] = extractelement <4 x float> [[TMP0]], i32 1
; NOACCELERATE-NEXT: [[TMP2:%.*]] = tail call fast float @llvm.cosh.f32(float [[VECEXT_1]])
; NOACCELERATE-NEXT: [[VECINS_1:%.*]] = insertelement <4 x float> [[VECINS]], float [[TMP2]], i32 1
; NOACCELERATE-NEXT: [[VECEXT_2:%.*]] = extractelement <4 x float> [[TMP0]], i32 2
; NOACCELERATE-NEXT: [[TMP3:%.*]] = tail call fast float @llvm.cosh.f32(float [[VECEXT_2]])
; NOACCELERATE-NEXT: [[VECINS_2:%.*]] = insertelement <4 x float> [[VECINS_1]], float [[TMP3]], i32 2
; NOACCELERATE-NEXT: [[VECEXT_3:%.*]] = extractelement <4 x float> [[TMP0]], i32 3
; NOACCELERATE-NEXT: [[TMP4:%.*]] = tail call fast float @llvm.cosh.f32(float [[VECEXT_3]])
; NOACCELERATE-NEXT: [[VECINS_3:%.*]] = insertelement <4 x float> [[VECINS_2]], float [[TMP4]], i32 3
; NOACCELERATE-NEXT: ret <4 x float> [[VECINS_3]]
; NOACCELERATE-NEXT: [[TMP3:%.*]] = shufflevector <4 x float> [[TMP0]], <4 x float> poison, <2 x i32> <i32 2, i32 3>
; NOACCELERATE-NEXT: [[TMP4:%.*]] = call fast <2 x float> @llvm.cosh.v2f32(<2 x float> [[TMP3]])
; NOACCELERATE-NEXT: [[TMP5:%.*]] = shufflevector <2 x float> [[TMP4]], <2 x float> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
; NOACCELERATE-NEXT: [[VECINS_31:%.*]] = shufflevector <4 x float> [[VECINS_1]], <4 x float> [[TMP5]], <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; NOACCELERATE-NEXT: ret <4 x float> [[VECINS_31]]
;
entry:
%0 = load <4 x float>, ptr %a, align 16
Expand Down Expand Up @@ -1032,13 +1022,11 @@ define <4 x float> @int_tanh_4x(ptr %a) {
; NOACCELERATE-NEXT: [[VECEXT_1:%.*]] = extractelement <4 x float> [[TMP0]], i32 1
; NOACCELERATE-NEXT: [[TMP2:%.*]] = tail call fast float @llvm.tanh.f32(float [[VECEXT_1]])
; NOACCELERATE-NEXT: [[VECINS_1:%.*]] = insertelement <4 x float> [[VECINS]], float [[TMP2]], i32 1
; NOACCELERATE-NEXT: [[VECEXT_2:%.*]] = extractelement <4 x float> [[TMP0]], i32 2
; NOACCELERATE-NEXT: [[TMP3:%.*]] = tail call fast float @llvm.tanh.f32(float [[VECEXT_2]])
; NOACCELERATE-NEXT: [[VECINS_2:%.*]] = insertelement <4 x float> [[VECINS_1]], float [[TMP3]], i32 2
; NOACCELERATE-NEXT: [[VECEXT_3:%.*]] = extractelement <4 x float> [[TMP0]], i32 3
; NOACCELERATE-NEXT: [[TMP4:%.*]] = tail call fast float @llvm.tanh.f32(float [[VECEXT_3]])
; NOACCELERATE-NEXT: [[VECINS_3:%.*]] = insertelement <4 x float> [[VECINS_2]], float [[TMP4]], i32 3
; NOACCELERATE-NEXT: ret <4 x float> [[VECINS_3]]
; NOACCELERATE-NEXT: [[TMP3:%.*]] = shufflevector <4 x float> [[TMP0]], <4 x float> poison, <2 x i32> <i32 2, i32 3>
; NOACCELERATE-NEXT: [[TMP4:%.*]] = call fast <2 x float> @llvm.tanh.v2f32(<2 x float> [[TMP3]])
; NOACCELERATE-NEXT: [[TMP5:%.*]] = shufflevector <2 x float> [[TMP4]], <2 x float> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
; NOACCELERATE-NEXT: [[VECINS_31:%.*]] = shufflevector <4 x float> [[VECINS_1]], <4 x float> [[TMP5]], <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; NOACCELERATE-NEXT: ret <4 x float> [[VECINS_31]]
;
entry:
%0 = load <4 x float>, ptr %a, align 16
Expand Down
Loading
Loading