[Clang] Add elementwise maximumnum/minimumnum builtin functions #149775
There are no files selected for viewing
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
|
|
@@ -4108,6 +4108,22 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID, | |
| return RValue::get(Result); | ||
| } | ||
|
|
||
| case Builtin::BI__builtin_elementwise_maximumnum: { | ||
| Value *Op0 = EmitScalarExpr(E->getArg(0)); | ||
|
There was a problem hiding this comment. Do we ever ensure we actually HAVE 2 arguments? Both of these refer to the 1st arg, but the prototype accepts 0 or 1. There was a problem hiding this comment.
It is checked at llvm-project/clang/lib/Sema/SemaChecking.cpp Line 15756 in 4d48996
sorry I don't get it. There was a problem hiding this comment. Ah, I stopped chasing down the call-stack 1 too few :) I see it now in I was pointing out that we refer to There was a problem hiding this comment.
I see, thanks @erichkeane |
||
| Value *Op1 = EmitScalarExpr(E->getArg(1)); | ||
| Value *Result = Builder.CreateBinaryIntrinsic( | ||
| Intrinsic::maximumnum, Op0, Op1, nullptr, "elt.maximumnum"); | ||
| return RValue::get(Result); | ||
| } | ||
|
|
||
| case Builtin::BI__builtin_elementwise_minimumnum: { | ||
| Value *Op0 = EmitScalarExpr(E->getArg(0)); | ||
| Value *Op1 = EmitScalarExpr(E->getArg(1)); | ||
| Value *Result = Builder.CreateBinaryIntrinsic( | ||
| Intrinsic::minimumnum, Op0, Op1, nullptr, "elt.minimumnum"); | ||
| return RValue::get(Result); | ||
| } | ||
|
|
||
| case Builtin::BI__builtin_reduce_max: { | ||
| auto GetIntrinsicID = [this](QualType QT) { | ||
| if (auto *VecTy = QT->getAs<VectorType>()) | ||
|
|
||
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,171 @@ | ||
| // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 5 | ||
| // RUN: %clang_cc1 -x c++ -std=c++20 -disable-llvm-passes -O3 -triple x86_64 %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK | ||
|
|
||
| typedef _Float16 half8 __attribute__((ext_vector_type(8))); | ||
| typedef __bf16 bf16x8 __attribute__((ext_vector_type(8))); | ||
| typedef float float4 __attribute__((ext_vector_type(4))); | ||
| typedef double double2 __attribute__((ext_vector_type(2))); | ||
| typedef long double ldouble2 __attribute__((ext_vector_type(2))); | ||
|
|
||
| // CHECK-LABEL: define dso_local noundef <8 x half> @_Z7pfmin16Dv8_DF16_S_( | ||
| // CHECK-SAME: <8 x half> noundef [[A:%.*]], <8 x half> noundef [[B:%.*]]) #[[ATTR0:[0-9]+]] { | ||
| // CHECK-NEXT: [[ENTRY:.*:]] | ||
| // CHECK-NEXT: [[A_ADDR:%.*]] = alloca <8 x half>, align 16 | ||
| // CHECK-NEXT: [[B_ADDR:%.*]] = alloca <8 x half>, align 16 | ||
| // CHECK-NEXT: store <8 x half> [[A]], ptr [[A_ADDR]], align 16, !tbaa [[TBAA2:![0-9]+]] | ||
| // CHECK-NEXT: store <8 x half> [[B]], ptr [[B_ADDR]], align 16, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: [[TMP0:%.*]] = load <8 x half>, ptr [[A_ADDR]], align 16, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: [[TMP1:%.*]] = load <8 x half>, ptr [[B_ADDR]], align 16, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: [[ELT_MINIMUMNUM:%.*]] = call <8 x half> @llvm.minimumnum.v8f16(<8 x half> [[TMP0]], <8 x half> [[TMP1]]) | ||
| // CHECK-NEXT: ret <8 x half> [[ELT_MINIMUMNUM]] | ||
| // | ||
| half8 pfmin16(half8 a, half8 b) { | ||
| return __builtin_elementwise_minimumnum(a, b); | ||
| } | ||
| // CHECK-LABEL: define dso_local noundef <8 x bfloat> @_Z8pfmin16bDv8_DF16bS_( | ||
| // CHECK-SAME: <8 x bfloat> noundef [[A:%.*]], <8 x bfloat> noundef [[B:%.*]]) #[[ATTR0]] { | ||
| // CHECK-NEXT: [[ENTRY:.*:]] | ||
| // CHECK-NEXT: [[A_ADDR:%.*]] = alloca <8 x bfloat>, align 16 | ||
| // CHECK-NEXT: [[B_ADDR:%.*]] = alloca <8 x bfloat>, align 16 | ||
| // CHECK-NEXT: store <8 x bfloat> [[A]], ptr [[A_ADDR]], align 16, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: store <8 x bfloat> [[B]], ptr [[B_ADDR]], align 16, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: [[TMP0:%.*]] = load <8 x bfloat>, ptr [[A_ADDR]], align 16, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: [[TMP1:%.*]] = load <8 x bfloat>, ptr [[B_ADDR]], align 16, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: [[ELT_MINIMUMNUM:%.*]] = call <8 x bfloat> @llvm.minimumnum.v8bf16(<8 x bfloat> [[TMP0]], <8 x bfloat> [[TMP1]]) | ||
| // CHECK-NEXT: ret <8 x bfloat> [[ELT_MINIMUMNUM]] | ||
| // | ||
| bf16x8 pfmin16b(bf16x8 a, bf16x8 b) { | ||
| return __builtin_elementwise_minimumnum(a, b); | ||
| } | ||
| // CHECK-LABEL: define dso_local noundef <4 x float> @_Z7pfmin32Dv4_fS_( | ||
| // CHECK-SAME: <4 x float> noundef [[A:%.*]], <4 x float> noundef [[B:%.*]]) #[[ATTR0]] { | ||
| // CHECK-NEXT: [[ENTRY:.*:]] | ||
| // CHECK-NEXT: [[A_ADDR:%.*]] = alloca <4 x float>, align 16 | ||
| // CHECK-NEXT: [[B_ADDR:%.*]] = alloca <4 x float>, align 16 | ||
| // CHECK-NEXT: store <4 x float> [[A]], ptr [[A_ADDR]], align 16, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: store <4 x float> [[B]], ptr [[B_ADDR]], align 16, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: [[TMP0:%.*]] = load <4 x float>, ptr [[A_ADDR]], align 16, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: [[TMP1:%.*]] = load <4 x float>, ptr [[B_ADDR]], align 16, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: [[ELT_MINIMUMNUM:%.*]] = call <4 x float> @llvm.minimumnum.v4f32(<4 x float> [[TMP0]], <4 x float> [[TMP1]]) | ||
| // CHECK-NEXT: ret <4 x float> [[ELT_MINIMUMNUM]] | ||
| // | ||
| float4 pfmin32(float4 a, float4 b) { | ||
| return __builtin_elementwise_minimumnum(a, b); | ||
| } | ||
| // CHECK-LABEL: define dso_local noundef <2 x double> @_Z7pfmin64Dv2_dS_( | ||
| // CHECK-SAME: <2 x double> noundef [[A:%.*]], <2 x double> noundef [[B:%.*]]) #[[ATTR0]] { | ||
| // CHECK-NEXT: [[ENTRY:.*:]] | ||
| // CHECK-NEXT: [[A_ADDR:%.*]] = alloca <2 x double>, align 16 | ||
| // CHECK-NEXT: [[B_ADDR:%.*]] = alloca <2 x double>, align 16 | ||
| // CHECK-NEXT: store <2 x double> [[A]], ptr [[A_ADDR]], align 16, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: store <2 x double> [[B]], ptr [[B_ADDR]], align 16, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: [[TMP0:%.*]] = load <2 x double>, ptr [[A_ADDR]], align 16, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: [[TMP1:%.*]] = load <2 x double>, ptr [[B_ADDR]], align 16, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: [[ELT_MINIMUMNUM:%.*]] = call <2 x double> @llvm.minimumnum.v2f64(<2 x double> [[TMP0]], <2 x double> [[TMP1]]) | ||
| // CHECK-NEXT: ret <2 x double> [[ELT_MINIMUMNUM]] | ||
| // | ||
| double2 pfmin64(double2 a, double2 b) { | ||
| return __builtin_elementwise_minimumnum(a, b); | ||
| } | ||
| // CHECK-LABEL: define dso_local noundef <2 x x86_fp80> @_Z7pfmin80Dv2_eS_( | ||
| // CHECK-SAME: ptr noundef byval(<2 x x86_fp80>) align 32 [[TMP0:%.*]], ptr noundef byval(<2 x x86_fp80>) align 32 [[TMP1:%.*]]) #[[ATTR2:[0-9]+]] { | ||
| // CHECK-NEXT: [[ENTRY:.*:]] | ||
| // CHECK-NEXT: [[A_ADDR:%.*]] = alloca <2 x x86_fp80>, align 32 | ||
| // CHECK-NEXT: [[B_ADDR:%.*]] = alloca <2 x x86_fp80>, align 32 | ||
| // CHECK-NEXT: [[A:%.*]] = load <2 x x86_fp80>, ptr [[TMP0]], align 32, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: [[B:%.*]] = load <2 x x86_fp80>, ptr [[TMP1]], align 32, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: store <2 x x86_fp80> [[A]], ptr [[A_ADDR]], align 32, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: store <2 x x86_fp80> [[B]], ptr [[B_ADDR]], align 32, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: [[TMP2:%.*]] = load <2 x x86_fp80>, ptr [[A_ADDR]], align 32, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: [[TMP3:%.*]] = load <2 x x86_fp80>, ptr [[B_ADDR]], align 32, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: [[ELT_MINIMUMNUM:%.*]] = call <2 x x86_fp80> @llvm.minimumnum.v2f80(<2 x x86_fp80> [[TMP2]], <2 x x86_fp80> [[TMP3]]) | ||
| // CHECK-NEXT: ret <2 x x86_fp80> [[ELT_MINIMUMNUM]] | ||
| // | ||
| ldouble2 pfmin80(ldouble2 a, ldouble2 b) { | ||
| return __builtin_elementwise_minimumnum(a, b); | ||
| } | ||
|
|
||
| // CHECK-LABEL: define dso_local noundef <8 x half> @_Z7pfmax16Dv8_DF16_S_( | ||
| // CHECK-SAME: <8 x half> noundef [[A:%.*]], <8 x half> noundef [[B:%.*]]) #[[ATTR0]] { | ||
| // CHECK-NEXT: [[ENTRY:.*:]] | ||
| // CHECK-NEXT: [[A_ADDR:%.*]] = alloca <8 x half>, align 16 | ||
| // CHECK-NEXT: [[B_ADDR:%.*]] = alloca <8 x half>, align 16 | ||
| // CHECK-NEXT: store <8 x half> [[A]], ptr [[A_ADDR]], align 16, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: store <8 x half> [[B]], ptr [[B_ADDR]], align 16, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: [[TMP0:%.*]] = load <8 x half>, ptr [[A_ADDR]], align 16, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: [[TMP1:%.*]] = load <8 x half>, ptr [[B_ADDR]], align 16, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: [[ELT_MAXIMUMNUM:%.*]] = call <8 x half> @llvm.maximumnum.v8f16(<8 x half> [[TMP0]], <8 x half> [[TMP1]]) | ||
| // CHECK-NEXT: ret <8 x half> [[ELT_MAXIMUMNUM]] | ||
| // | ||
| half8 pfmax16(half8 a, half8 b) { | ||
| return __builtin_elementwise_maximumnum(a, b); | ||
| } | ||
| // CHECK-LABEL: define dso_local noundef <8 x bfloat> @_Z8pfmax16bDv8_DF16bS_( | ||
| // CHECK-SAME: <8 x bfloat> noundef [[A:%.*]], <8 x bfloat> noundef [[B:%.*]]) #[[ATTR0]] { | ||
| // CHECK-NEXT: [[ENTRY:.*:]] | ||
| // CHECK-NEXT: [[A_ADDR:%.*]] = alloca <8 x bfloat>, align 16 | ||
| // CHECK-NEXT: [[B_ADDR:%.*]] = alloca <8 x bfloat>, align 16 | ||
| // CHECK-NEXT: store <8 x bfloat> [[A]], ptr [[A_ADDR]], align 16, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: store <8 x bfloat> [[B]], ptr [[B_ADDR]], align 16, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: [[TMP0:%.*]] = load <8 x bfloat>, ptr [[A_ADDR]], align 16, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: [[TMP1:%.*]] = load <8 x bfloat>, ptr [[B_ADDR]], align 16, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: [[ELT_MAXIMUMNUM:%.*]] = call <8 x bfloat> @llvm.maximumnum.v8bf16(<8 x bfloat> [[TMP0]], <8 x bfloat> [[TMP1]]) | ||
| // CHECK-NEXT: ret <8 x bfloat> [[ELT_MAXIMUMNUM]] | ||
| // | ||
| bf16x8 pfmax16b(bf16x8 a, bf16x8 b) { | ||
| return __builtin_elementwise_maximumnum(a, b); | ||
| } | ||
| // CHECK-LABEL: define dso_local noundef <4 x float> @_Z7pfmax32Dv4_fS_( | ||
| // CHECK-SAME: <4 x float> noundef [[A:%.*]], <4 x float> noundef [[B:%.*]]) #[[ATTR0]] { | ||
| // CHECK-NEXT: [[ENTRY:.*:]] | ||
| // CHECK-NEXT: [[A_ADDR:%.*]] = alloca <4 x float>, align 16 | ||
| // CHECK-NEXT: [[B_ADDR:%.*]] = alloca <4 x float>, align 16 | ||
| // CHECK-NEXT: store <4 x float> [[A]], ptr [[A_ADDR]], align 16, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: store <4 x float> [[B]], ptr [[B_ADDR]], align 16, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: [[TMP0:%.*]] = load <4 x float>, ptr [[A_ADDR]], align 16, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: [[TMP1:%.*]] = load <4 x float>, ptr [[B_ADDR]], align 16, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: [[ELT_MAXIMUMNUM:%.*]] = call <4 x float> @llvm.maximumnum.v4f32(<4 x float> [[TMP0]], <4 x float> [[TMP1]]) | ||
| // CHECK-NEXT: ret <4 x float> [[ELT_MAXIMUMNUM]] | ||
| // | ||
| float4 pfmax32(float4 a, float4 b) { | ||
| return __builtin_elementwise_maximumnum(a, b); | ||
| } | ||
| // CHECK-LABEL: define dso_local noundef <2 x double> @_Z7pfmax64Dv2_dS_( | ||
| // CHECK-SAME: <2 x double> noundef [[A:%.*]], <2 x double> noundef [[B:%.*]]) #[[ATTR0]] { | ||
| // CHECK-NEXT: [[ENTRY:.*:]] | ||
| // CHECK-NEXT: [[A_ADDR:%.*]] = alloca <2 x double>, align 16 | ||
| // CHECK-NEXT: [[B_ADDR:%.*]] = alloca <2 x double>, align 16 | ||
| // CHECK-NEXT: store <2 x double> [[A]], ptr [[A_ADDR]], align 16, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: store <2 x double> [[B]], ptr [[B_ADDR]], align 16, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: [[TMP0:%.*]] = load <2 x double>, ptr [[A_ADDR]], align 16, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: [[TMP1:%.*]] = load <2 x double>, ptr [[B_ADDR]], align 16, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: [[ELT_MAXIMUMNUM:%.*]] = call <2 x double> @llvm.maximumnum.v2f64(<2 x double> [[TMP0]], <2 x double> [[TMP1]]) | ||
| // CHECK-NEXT: ret <2 x double> [[ELT_MAXIMUMNUM]] | ||
| // | ||
| double2 pfmax64(double2 a, double2 b) { | ||
| return __builtin_elementwise_maximumnum(a, b); | ||
| } | ||
|
|
||
| // CHECK-LABEL: define dso_local noundef <2 x x86_fp80> @_Z7pfmax80Dv2_eS_( | ||
| // CHECK-SAME: ptr noundef byval(<2 x x86_fp80>) align 32 [[TMP0:%.*]], ptr noundef byval(<2 x x86_fp80>) align 32 [[TMP1:%.*]]) #[[ATTR2]] { | ||
| // CHECK-NEXT: [[ENTRY:.*:]] | ||
| // CHECK-NEXT: [[A_ADDR:%.*]] = alloca <2 x x86_fp80>, align 32 | ||
| // CHECK-NEXT: [[B_ADDR:%.*]] = alloca <2 x x86_fp80>, align 32 | ||
| // CHECK-NEXT: [[A:%.*]] = load <2 x x86_fp80>, ptr [[TMP0]], align 32, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: [[B:%.*]] = load <2 x x86_fp80>, ptr [[TMP1]], align 32, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: store <2 x x86_fp80> [[A]], ptr [[A_ADDR]], align 32, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: store <2 x x86_fp80> [[B]], ptr [[B_ADDR]], align 32, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: [[TMP2:%.*]] = load <2 x x86_fp80>, ptr [[A_ADDR]], align 32, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: [[TMP3:%.*]] = load <2 x x86_fp80>, ptr [[B_ADDR]], align 32, !tbaa [[TBAA2]] | ||
| // CHECK-NEXT: [[ELT_MINIMUMNUM:%.*]] = call <2 x x86_fp80> @llvm.minimumnum.v2f80(<2 x x86_fp80> [[TMP2]], <2 x x86_fp80> [[TMP3]]) | ||
| // CHECK-NEXT: ret <2 x x86_fp80> [[ELT_MINIMUMNUM]] | ||
| // | ||
| ldouble2 pfmax80(ldouble2 a, ldouble2 b) { | ||
| return __builtin_elementwise_minimumnum(a, b); | ||
| } | ||
|
|
||
| //. | ||
| // CHECK: [[TBAA2]] = !{[[META3:![0-9]+]], [[META3]], i64 0} | ||
| // CHECK: [[META3]] = !{!"omnipotent char", [[META4:![0-9]+]], i64 0} | ||
| // CHECK: [[META4]] = !{!"Simple C++ TBAA"} | ||
| //. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I'm confused. :-)
We have
__builtin_elementwise_min,__builtin_elementwise_minnum,__builtin_elementwise_minimum, and now we're adding__builtin_elementwise_minimumnum?I think the docs need to be expanded a bit to help understand what the difference is between the four choices. Same for maximum.
Why do we need four different builtins to select the lesser of two values?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
ugh, I see now, these are the elementwise variants of the existing C math library functions. Carry on. :-D
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
yes, the difference is explained in the link http://llvm.org/docs/LangRef.html#llvm-min-intrinsics-comparation