Skip to content

[SLPVectorizer] Support SLPVectorizer cases of tan across all backends #95517

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 2 commits into from
Jun 27, 2024
Merged

[SLPVectorizer] Support SLPVectorizer cases of tan across all backends #95517

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
3bbc260
[CodeGen] Support SLPVectorizer cases of tan across all backends
farzonl Jun 14, 2024
62d5393
remove target specific changes
farzonl Jun 14, 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
3 changes: 3 additions & 0 deletions llvm/include/llvm/Analysis/TargetLibraryInfo.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -415,10 +415,12 @@ class TargetLibraryInfo {
return false;
switch (F) {
default: break;
// clang-format off
case LibFunc_copysign: case LibFunc_copysignf: case LibFunc_copysignl:
case LibFunc_fabs: case LibFunc_fabsf: case LibFunc_fabsl:
case LibFunc_sin: case LibFunc_sinf: case LibFunc_sinl:
case LibFunc_cos: case LibFunc_cosf: case LibFunc_cosl:
case LibFunc_tan: case LibFunc_tanf: case LibFunc_tanl:
case LibFunc_sqrt: case LibFunc_sqrtf: case LibFunc_sqrtl:
case LibFunc_sqrt_finite: case LibFunc_sqrtf_finite:
case LibFunc_sqrtl_finite:
Expand All @@ -437,6 +439,7 @@ class TargetLibraryInfo {
case LibFunc_memcmp: case LibFunc_bcmp: case LibFunc_strcmp:
case LibFunc_strcpy: case LibFunc_stpcpy: case LibFunc_strlen:
case LibFunc_strnlen: case LibFunc_memchr: case LibFunc_mempcpy:
// clang-format on
return true;
}
return false;
Expand Down
11 changes: 7 additions & 4 deletions llvm/include/llvm/Analysis/TargetTransformInfoImpl.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -156,14 +156,17 @@ class TargetTransformInfoImplBase {
StringRef Name = F->getName();

// These will all likely lower to a single selection DAG node.
// clang-format off
if (Name == "copysign" || Name == "copysignf" || Name == "copysignl" ||
Name == "fabs" || Name == "fabsf" || Name == "fabsl" || Name == "sin" ||
Name == "fabs" || Name == "fabsf" || Name == "fabsl" ||
Name == "fmin" || Name == "fminf" || Name == "fminl" ||
Name == "fmax" || Name == "fmaxf" || Name == "fmaxl" ||
Name == "sinf" || Name == "sinl" || Name == "cos" || Name == "cosf" ||
Name == "cosl" || Name == "sqrt" || Name == "sqrtf" || Name == "sqrtl")
Name == "sin" || Name == "sinf" || Name == "sinl" ||
Name == "cos" || Name == "cosf" || Name == "cosl" ||
Name == "tan" || Name == "tanf" || Name == "tanl" ||
Name == "sqrt" || Name == "sqrtf" || Name == "sqrtl")
return false;

// clang-format on
// These are all likely to be optimized into something smaller.
if (Name == "pow" || Name == "powf" || Name == "powl" || Name == "exp2" ||
Name == "exp2l" || Name == "exp2f" || Name == "floor" ||
Expand Down
4 changes: 4 additions & 0 deletions llvm/lib/Analysis/ValueTracking.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3993,6 +3993,10 @@ Intrinsic::ID llvm::getIntrinsicForCallSite(const CallBase &CB,
case LibFunc_cosf:
case LibFunc_cosl:
return Intrinsic::cos;
case LibFunc_tan:
case LibFunc_tanf:
case LibFunc_tanl:
return Intrinsic::tan;
case LibFunc_exp:
case LibFunc_expf:
case LibFunc_expl:
Expand Down
12 changes: 5 additions & 7 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 @@ -548,13 +548,11 @@ define <4 x float> @tan_4x(ptr %a) {
; NOACCELERATE-NEXT: [[VECEXT_1:%.*]] = extractelement <4 x float> [[TMP0]], i32 1
; NOACCELERATE-NEXT: [[TMP2:%.*]] = tail call fast float @tanf(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 @tanf(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 @tanf(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.tan.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
12 changes: 5 additions & 7 deletions llvm/test/Transforms/SLPVectorizer/AArch64/accelerate-vector-functions.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -548,13 +548,11 @@ define <4 x float> @tan_4x(ptr %a) {
; NOACCELERATE-NEXT: [[VECEXT_1:%.*]] = extractelement <4 x float> [[TMP0]], i32 1
; NOACCELERATE-NEXT: [[TMP2:%.*]] = tail call fast float @tanf(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 @tanf(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 @tanf(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.tan.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
19 changes: 19 additions & 0 deletions llvm/test/Transforms/SLPVectorizer/X86/call.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,6 +6,7 @@ target triple = "x86_64-apple-macosx10.8.0"

declare double @sin(double) nounwind willreturn
declare double @cos(double) nounwind willreturn
declare double @tan(double) nounwind willreturn
declare double @pow(double, double) nounwind willreturn
declare double @exp2(double) nounwind willreturn
declare double @sqrt(double) nounwind willreturn
Expand Down Expand Up @@ -48,6 +49,24 @@ define void @cos_libm(ptr %a, ptr %b) {
ret void
}

define void @tan_libm(ptr %a, ptr %b) {
; CHECK-LABEL: @tan_libm(
; CHECK-NEXT: [[TMP2:%.*]] = load <2 x double>, ptr [[A:%.*]], align 8
; CHECK-NEXT: [[TMP3:%.*]] = call <2 x double> @llvm.tan.v2f64(<2 x double> [[TMP2]])
; CHECK-NEXT: store <2 x double> [[TMP3]], ptr [[B:%.*]], align 8
; CHECK-NEXT: ret void
;
%a0 = load double, ptr %a, align 8
%idx1 = getelementptr inbounds double, ptr %a, i64 1
%a1 = load double, ptr %idx1, align 8
%tan1 = tail call double @tan(double %a0) nounwind readnone
%tan2 = tail call double @tan(double %a1) nounwind readnone
store double %tan1, ptr %b, align 8
%idx2 = getelementptr inbounds double, ptr %b, i64 1
store double %tan2, ptr %idx2, align 8
ret void
}

define void @pow_libm(ptr %a, ptr %b) {
; CHECK-LABEL: @pow_libm(
; CHECK-NEXT: [[TMP2:%.*]] = load <2 x double>, ptr [[A:%.*]], align 8
Expand Down
Loading