[clang] [llvm] [SPIRV] GPU intrinsics (PR #131190)

AlexVlx wrote: > > not everything should end up in Clang, even though it is individually > > "convenient" > > We should at least have some builtins for those SPIR-V intrinsics you listed. > I think right now we only have `reflect`. Probably, I think there's a sense that in general it's desira

[clang] clang: Do not implicitly addrspacecast in EmitAggExprToLValue (PR #129837)

https://github.com/AlexVlx requested changes to this pull request. I'm temporarily blocking this to go through a bit more than OCL - I just want to ensure that this doesn't break some C/C++ corner case. Otherwise LGTM. https://github.com/llvm/llvm-project/pull/129837 ___

[clang] [clang][CodeGen] `sret` args should always point to the `alloca` AS, so use that (PR #114062)

https://github.com/AlexVlx closed https://github.com/llvm/llvm-project/pull/114062 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[clang] [clang][CodeGen] Additional fixes for #114062 (PR #128166)

https://github.com/AlexVlx updated https://github.com/llvm/llvm-project/pull/128166 >From 77db8a15d472e112be93cb7566cf56e26dc80501 Mon Sep 17 00:00:00 2001 From: Alex Voicu Date: Fri, 21 Feb 2025 13:47:42 +0200 Subject: [PATCH 1/2] Handle HLL casts on sret args, remove ill advised cast strip a

[clang] [clang][CodeGen] Additional fixes for #114062 (PR #128166)

https://github.com/AlexVlx updated https://github.com/llvm/llvm-project/pull/128166 >From 77db8a15d472e112be93cb7566cf56e26dc80501 Mon Sep 17 00:00:00 2001 From: Alex Voicu Date: Fri, 21 Feb 2025 13:47:42 +0200 Subject: [PATCH 1/4] Handle HLL casts on sret args, remove ill advised cast strip a

[clang] [clang][CodeGen] Additional fixes for #114062 (PR #128166)

@@ -0,0 +1,38 @@ +// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --include-generated-funcs --version 5 +// RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple amdgcn-amd-amdhsa -fopenmp-targets=amdgcn-amd-amdhsa -emit-llvm %s -fopenmp-is-t

[clang] [clang][CodeGen] Additional fixes for #114062 (PR #128166)

@@ -2352,6 +2353,22 @@ Value *ScalarExprEmitter::VisitCastExpr(CastExpr *CE) { Value *Src = Visit(const_cast(E)); llvm::Type *SrcTy = Src->getType(); llvm::Type *DstTy = ConvertType(DestTy); + +// FIXME: this is a gross but seemingly necessary workaround for an

[clang] [clang][CodeGen] Additional fixes for #114062 (PR #128166)

@@ -2352,6 +2353,22 @@ Value *ScalarExprEmitter::VisitCastExpr(CastExpr *CE) { Value *Src = Visit(const_cast(E)); llvm::Type *SrcTy = Src->getType(); llvm::Type *DstTy = ConvertType(DestTy); + +// FIXME: this is a gross but seemingly necessary workaround for an

[clang] [clang][CodeGen] Additional fixes for #114062 (PR #128166)

@@ -0,0 +1,32 @@ +// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 5 +// RUN: %clang_cc1 -triple amdgcn-amd-amdhsa-gnu -target-cpu gfx900 -emit-llvm -o - %s | FileCheck %s + +struct X { int z[17]; }; + +// CHECK-LABEL: define dso_

[clang] [clang][CodeGen] Additional fixes for #114062 (PR #128166)

https://github.com/AlexVlx edited https://github.com/llvm/llvm-project/pull/128166 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[clang] [clang][CodeGen] Additional fixes for #114062 (PR #128166)

https://github.com/AlexVlx updated https://github.com/llvm/llvm-project/pull/128166 >From 77db8a15d472e112be93cb7566cf56e26dc80501 Mon Sep 17 00:00:00 2001 From: Alex Voicu Date: Fri, 21 Feb 2025 13:47:42 +0200 Subject: [PATCH 1/3] Handle HLL casts on sret args, remove ill advised cast strip a

[clang] [clang][CodeGen] Additional fixes for #114062 (PR #128166)

https://github.com/AlexVlx created https://github.com/llvm/llvm-project/pull/128166 This addresses two issues introduced by moving indirect args into an explicit AS (please see and

[clang] [clang][HIP] Make some math not not work with AMDGCN SPIR-V (PR #128360)

@@ -14,6 +14,12 @@ #include "hip/hip_version.h" #endif // __has_include("hip/hip_version.h") +#ifdef __SPIRV__ +#define __PRIVATE_AS __attribute__((address_space(0))) AlexVlx wrote: Acknowledged, but the header was already using numbered address spaces (ROCDL

[clang] [spirv][amdgpu] Set atomic size in the clang target info (PR #128569)

https://github.com/AlexVlx approved this pull request. LGTM, thanks. FWIW we've only done HIP bring-up, so there will probably be a few of these lingering. https://github.com/llvm/llvm-project/pull/128569 ___ cfe-commits mailing list cfe-commits@lists

[clang] [clang][CodeGen] Additional fixes for #114062 (PR #128166)

@@ -2352,6 +2353,22 @@ Value *ScalarExprEmitter::VisitCastExpr(CastExpr *CE) { Value *Src = Visit(const_cast(E)); llvm::Type *SrcTy = Src->getType(); llvm::Type *DstTy = ConvertType(DestTy); + +// FIXME: this is a gross but seemingly necessary workaround for an

[clang] [NFC] Fix Sanitizer breakage introduced in #128166 (PR #128990)

https://github.com/AlexVlx closed https://github.com/llvm/llvm-project/pull/128990 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[clang] [clang][CodeGen] Additional fixes for #114062 (PR #128166)

https://github.com/AlexVlx updated https://github.com/llvm/llvm-project/pull/128166 >From 77db8a15d472e112be93cb7566cf56e26dc80501 Mon Sep 17 00:00:00 2001 From: Alex Voicu Date: Fri, 21 Feb 2025 13:47:42 +0200 Subject: [PATCH 1/5] Handle HLL casts on sret args, remove ill advised cast strip a

[clang] [clang][CodeGen] Additional fixes for #114062 (PR #128166)

https://github.com/AlexVlx updated https://github.com/llvm/llvm-project/pull/128166 >From 77db8a15d472e112be93cb7566cf56e26dc80501 Mon Sep 17 00:00:00 2001 From: Alex Voicu Date: Fri, 21 Feb 2025 13:47:42 +0200 Subject: [PATCH 1/5] Handle HLL casts on sret args, remove ill advised cast strip a

[clang] [llvm] [clang][OpenMP][SPIR-V] Fix addrspace of globals and global constants (PR #134399)

@@ -5384,6 +5384,11 @@ LangAS CodeGenModule::GetGlobalVarAddressSpace(const VarDecl *D) { LangAS AS; if (OpenMPRuntime->hasAllocateAttributeForGlobalVar(D, AS)) return AS; +if (LangOpts.OpenMPIsTargetDevice && getTriple().isSPIRV()) AlexVlx w

[clang] [llvm] [clang][OpenMP][SPIR-V] Fix addrspace of globals and global constants (PR #134399)

@@ -5402,6 +5407,10 @@ LangAS CodeGenModule::GetGlobalConstantAddressSpace() const { // UniformConstant storage class is not viable as pointers to it may not be // casted to Generic pointers which are used to model HIP's "flat" pointers. return LangAS::cuda_device

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

@@ -284,6 +284,18 @@ void CodeGenFunction::AddAMDGPUFenceAddressSpaceMMRA(llvm::Instruction *Inst, Inst->setMetadata(LLVMContext::MD_mmra, MMRAMetadata::getMD(Ctx, MMRAs)); } +static Value *GetOrInsertAMDGPUPredicate(CodeGenFunction &CGF, Twine Name) { + auto PTy = Integer

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

@@ -15576,6 +15609,38 @@ static bool isOverflowingIntegerType(ASTContext &Ctx, QualType T) { return Ctx.getIntWidth(T) >= Ctx.getIntWidth(Ctx.IntTy); } +static Expr *ExpandAMDGPUPredicateBI(ASTContext &Ctx, CallExpr *CE) { + if (!CE->getBuiltinCallee()) +return CXXBool

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

https://github.com/AlexVlx updated https://github.com/llvm/llvm-project/pull/134016 >From 91eeaf02336e539f14dcb0a79ff15dbe8befe6f1 Mon Sep 17 00:00:00 2001 From: Alex Voicu Date: Wed, 2 Apr 2025 02:47:42 +0100 Subject: [PATCH 1/8] Add the functional identity and feature queries. --- clang/doc

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

https://github.com/AlexVlx updated https://github.com/llvm/llvm-project/pull/134016 >From 91eeaf02336e539f14dcb0a79ff15dbe8befe6f1 Mon Sep 17 00:00:00 2001 From: Alex Voicu Date: Wed, 2 Apr 2025 02:47:42 +0100 Subject: [PATCH 1/9] Add the functional identity and feature queries. --- clang/doc

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

https://github.com/AlexVlx updated https://github.com/llvm/llvm-project/pull/134016 >From 91eeaf02336e539f14dcb0a79ff15dbe8befe6f1 Mon Sep 17 00:00:00 2001 From: Alex Voicu Date: Wed, 2 Apr 2025 02:47:42 +0100 Subject: [PATCH 01/10] Add the functional identity and feature queries. --- clang/d

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

AlexVlx wrote: > So in short: what you're trying to prevent is "this was stored in a variable, > then checked later when we are no longer on the device, thus the answer is > different". Not quite, although that is definitely an interesting consideration. What I am trying to address here is t

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

AlexVlx wrote: > > as a must have, it allows featureful AMDGCN flavoured SPIR-V to be > > produced, where target specific capability is guarded and chosen or > > discarded when finalising compilation for a concrete target. > > I understand the reasoning behind providing such mechanisms to guar

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

@@ -0,0 +1,64 @@ +// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --check-globals all --version 5 +// RUN: %clang_cc1 -triple amdgcn-amd-amdhsa -target-cpu gfx900 -emit-llvm %s -o - | FileCheck --check-prefix=AMDGCN-GFX900 %s +// RUN: %cla

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

@@ -585,6 +597,23 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID, llvm::Value *Env = EmitScalarExpr(E->getArg(0)); return Builder.CreateCall(F, {Env}); } + case AMDGPU::BI__builtin_amdgcn_processor_is: { +assert(CGM.getTriple().isSPIRV() &&

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

@@ -4920,6 +4920,116 @@ If no address spaces names are provided, all address spaces are fenced. __builtin_amdgcn_fence(__ATOMIC_SEQ_CST, "workgroup", "local") __builtin_amdgcn_fence(__ATOMIC_SEQ_CST, "workgroup", "local", "global") +__builtin_amdgcn_processor_is and __bui

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

@@ -585,6 +597,23 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID, llvm::Value *Env = EmitScalarExpr(E->getArg(0)); return Builder.CreateCall(F, {Env}); } + case AMDGPU::BI__builtin_amdgcn_processor_is: { +assert(CGM.getTriple().isSPIRV() &&

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

@@ -4920,6 +4920,116 @@ If no address spaces names are provided, all address spaces are fenced. __builtin_amdgcn_fence(__ATOMIC_SEQ_CST, "workgroup", "local") __builtin_amdgcn_fence(__ATOMIC_SEQ_CST, "workgroup", "local", "global") +__builtin_amdgcn_processor_is and __bui

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

@@ -4920,6 +4920,116 @@ If no address spaces names are provided, all address spaces are fenced. __builtin_amdgcn_fence(__ATOMIC_SEQ_CST, "workgroup", "local") __builtin_amdgcn_fence(__ATOMIC_SEQ_CST, "workgroup", "local", "global") +__builtin_amdgcn_processor_is and __bui

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

https://github.com/AlexVlx updated https://github.com/llvm/llvm-project/pull/134016 >From 91eeaf02336e539f14dcb0a79ff15dbe8befe6f1 Mon Sep 17 00:00:00 2001 From: Alex Voicu Date: Wed, 2 Apr 2025 02:47:42 +0100 Subject: [PATCH 01/11] Add the functional identity and feature queries. --- clang/d

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

@@ -4920,6 +4920,116 @@ If no address spaces names are provided, all address spaces are fenced. __builtin_amdgcn_fence(__ATOMIC_SEQ_CST, "workgroup", "local") __builtin_amdgcn_fence(__ATOMIC_SEQ_CST, "workgroup", "local", "global") +__builtin_amdgcn_processor_is and __bui

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

https://github.com/AlexVlx updated https://github.com/llvm/llvm-project/pull/134016 >From 91eeaf02336e539f14dcb0a79ff15dbe8befe6f1 Mon Sep 17 00:00:00 2001 From: Alex Voicu Date: Wed, 2 Apr 2025 02:47:42 +0100 Subject: [PATCH 1/7] Add the functional identity and feature queries. --- clang/doc

[clang] [llvm] [HIP][HIPSTDPAR] Re-work allocation interposition for `hipstdpar` (PR #138790)

https://github.com/AlexVlx closed https://github.com/llvm/llvm-project/pull/138790 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

https://github.com/AlexVlx updated https://github.com/llvm/llvm-project/pull/134016 Rate limit · GitHub body { background-color: #f6f8fa; color: #24292e; font-family: -apple-system,BlinkMacSystemFont,Segoe UI,Helvetica,Arial,sans-s

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

https://github.com/AlexVlx commented: > A side question, is it legal to use the builtin in unstructured control flow, > like here: https://godbolt.org/z/no7Kzv19r ? Note, if the answer is "no", > then enforcing the builtin to initialize something would (probably) > automatically prevent this c

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

https://github.com/AlexVlx edited https://github.com/llvm/llvm-project/pull/134016 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

https://github.com/AlexVlx updated https://github.com/llvm/llvm-project/pull/134016 Rate limit · GitHub body { background-color: #f6f8fa; color: #24292e; font-family: -apple-system,BlinkMacSystemFont,Segoe UI,Helvetica,Arial,sans-s

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

AlexVlx wrote: @erichkeane @AaronBallman if, when you have time, you could please indicate if the new direction is at least generally aligned with what you had in mind, it'd be appreciated! https://github.com/llvm/llvm-project/pull/134016 ___ cfe-com

[clang] [HIP] change default offload archs (PR #139281)

AlexVlx wrote: > The main obstacle of letting clang emit error when `--offload-arch` is not > specified is HIP apps using hipcc as CMAKE_CXX_COMPILER. hipcc adds -xhip by > default for .cpp programs. This is a known and long existing issue. > > Another option is to have multiple `--offload-arc

[clang] [llvm] [HIP][HIPSTDPAR] Re-work allocation interposition for `hipstdpar` (PR #138790)

https://github.com/AlexVlx created https://github.com/llvm/llvm-project/pull/138790 The allocation interposition mode had a number of issues, which are primarily addressed in the library component via . However, it is necessary to interpose some add

[clang] [llvm] [HIP][HIPSTDPAR] Re-work allocation interposition for `hipstdpar` (PR #138790)

https://github.com/AlexVlx updated https://github.com/llvm/llvm-project/pull/138790 >From 865ff3dff1833607f0d546ab0ebd95b98a8ed71b Mon Sep 17 00:00:00 2001 From: Alex Voicu Date: Wed, 7 May 2025 01:25:17 +0100 Subject: [PATCH 1/2] Re-work allocation interposition for `hipstdpar`. --- clang/do

[clang] [HIP] change default offload archs (PR #139281)

AlexVlx wrote: > > > The main obstacle of letting clang emit error when `--offload-arch` is > > > not specified is HIP apps using hipcc as CMAKE_CXX_COMPILER. hipcc adds > > > -xhip by default for .cpp programs. This is a known and long existing > > > issue. > > > Another option is to have mul

[clang] [HIP] change default offload archs (PR #139281)

AlexVlx wrote: I think that in general we also need to decide on what happens when you pick an amdgcn— triple. IMHO for that case we should probably error out if no mcpu is provided, since there’s no reasonable default, except for “all”, but that would be incredibly disruptive. https://github

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

AlexVlx wrote: Gentle ping. https://github.com/llvm/llvm-project/pull/134016 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

@@ -13338,4 +13338,23 @@ def err_acc_device_type_multiple_archs // AMDGCN builtins diagnostics def err_amdgcn_load_lds_size_invalid_value : Error<"invalid size value">; def note_amdgcn_load_lds_size_valid_value : Note<"size must be %select{1, 2, or 4|1, 2, 4, 12 or 16}0">; +de

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

@@ -29,6 +29,8 @@ MODULE_PASS("amdgpu-printf-runtime-binding", AMDGPUPrintfRuntimeBindingPass()) MODULE_PASS("amdgpu-remove-incompatible-functions", AMDGPURemoveIncompatibleFunctionsPass(*this)) MODULE_PASS("amdgpu-sw-lower-lds", AMDGPUSwLowerLDSPass(*this)) MODULE_PASS("amdg

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

@@ -4966,6 +4966,89 @@ If no address spaces names are provided, all address spaces are fenced. __builtin_amdgcn_fence(__ATOMIC_SEQ_CST, "workgroup", "local") __builtin_amdgcn_fence(__ATOMIC_SEQ_CST, "workgroup", "local", "global") +__builtin_amdgcn_processor_is and __buil

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

https://github.com/AlexVlx commented: > First, please take a look at the LLVM coding standard re the use of 'auto'. > > Second: The use of a special type for these builtins is a little novel > (though I see the predicate type already exists?), but I guess I'm ok with > it. I have some concern

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

https://github.com/AlexVlx edited https://github.com/llvm/llvm-project/pull/134016 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

@@ -9102,6 +9102,15 @@ bool InitializationSequence::Diagnose(Sema &S, case FK_ConversionFailed: { QualType FromType = OnlyArg->getType(); +// __amdgpu_feature_predicate_t can be explicitly cast to the logical op +// type, although this is almost always an error a

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

@@ -9102,6 +9102,15 @@ bool InitializationSequence::Diagnose(Sema &S, case FK_ConversionFailed: { QualType FromType = OnlyArg->getType(); +// __amdgpu_feature_predicate_t can be explicitly cast to the logical op +// type, although this is almost always an error a

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

@@ -9102,6 +9102,15 @@ bool InitializationSequence::Diagnose(Sema &S, case FK_ConversionFailed: { QualType FromType = OnlyArg->getType(); +// __amdgpu_feature_predicate_t can be explicitly cast to the logical op +// type, although this is almost always an error a

[clang] [HIP] Emit the CUID value in the module with the new driver (PR #144570)

https://github.com/AlexVlx approved this pull request. LGTM, thanks. https://github.com/llvm/llvm-project/pull/144570 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

@@ -0,0 +1,157 @@ +//===- AMDGPUExpandFeaturePredicates.cpp - Feature Predicate Expander Pass ===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Ap

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

@@ -0,0 +1,157 @@ +//===- AMDGPUExpandFeaturePredicates.cpp - Feature Predicate Expander Pass ===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Ap

[clang] [Clang][SPIRV][AMDGPU] Override `supportsLibCall` for AMDGCNSPIRV (PR #143814)

https://github.com/AlexVlx edited https://github.com/llvm/llvm-project/pull/143814 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

@@ -0,0 +1,157 @@ +//===- AMDGPUExpandFeaturePredicates.cpp - Feature Predicate Expander Pass ===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Ap

[clang] [Clang][SPIRV][AMDGPU] Override `supportsLibCall` for AMDGCNSPIRV (PR #143814)

https://github.com/AlexVlx closed https://github.com/llvm/llvm-project/pull/143814 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[clang] [Clang][SPIRV][AMDGPU] Override `supportsLibCall` for AMDGCNSPIRV (PR #143814)

https://github.com/AlexVlx commented: > Well, I'm not sure if my review is gating here, if "AMDGPU does not support > any libcalls at the moment" - it is what it is, no objections from my side. Since I am modifying the shared CodeGenInfo bit, it made sense to ask:) Also, I wasn't entirely cer

[clang] [Clang][SPIRV][AMDGPU] Override `supportsLibCall` for AMDGCNSPIRV (PR #143814)

https://github.com/AlexVlx updated https://github.com/llvm/llvm-project/pull/143814 >From a976da9d6033a252fab75de216fe6580d6e3e6ff Mon Sep 17 00:00:00 2001 From: Alex Voicu Date: Thu, 12 Jun 2025 00:47:43 +0100 Subject: [PATCH 1/2] Override `SPIRVTargetCodeGenInfo::supportsLibCall()` in order

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

AlexVlx wrote: Gentle ping. https://github.com/llvm/llvm-project/pull/134016 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

@@ -9102,6 +9102,15 @@ bool InitializationSequence::Diagnose(Sema &S, case FK_ConversionFailed: { QualType FromType = OnlyArg->getType(); +// __amdgpu_feature_predicate_t can be explicitly cast to the logical op +// type, although this is almost always an error a

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

@@ -9102,6 +9102,15 @@ bool InitializationSequence::Diagnose(Sema &S, case FK_ConversionFailed: { QualType FromType = OnlyArg->getType(); +// __amdgpu_feature_predicate_t can be explicitly cast to the logical op +// type, although this is almost always an error a

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

@@ -0,0 +1,157 @@ +//===- AMDGPUExpandFeaturePredicates.cpp - Feature Predicate Expander Pass ===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Ap

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

@@ -0,0 +1,157 @@ +//===- AMDGPUExpandFeaturePredicates.cpp - Feature Predicate Expander Pass ===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Ap

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

AlexVlx wrote: > I'm generally very unhappy about any kind of functionality that can cause > compilation failures either because the optimizer did not optimize enough > (including at O0) or because it optimized too much (producing code patterns > that are no longer recognized as trivially dead

[clang] [Clang][SPIRV][AMDGPU] Override `supportsLibCall` for AMDGCNSPIRV (PR #143814)

https://github.com/AlexVlx commented: #143431 is related to this, but it proposes a more pervasive change, which would require further discussion and clarification around what the SPIR-V BE wants to do / support in term of libcalls. https://github.com/llvm/llvm-project/pull/143814

[clang] [Clang][SPIRV][AMDGPU] Override `supportsLibCall` for AMDGCNSPIRV (PR #143814)

https://github.com/AlexVlx edited https://github.com/llvm/llvm-project/pull/143814 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[clang] [Clang][SPIRV][AMDGPU] Override `supportsLibCall` for AMDGCNSPIRV (PR #143814)

https://github.com/AlexVlx created https://github.com/llvm/llvm-project/pull/143814 The `supportsLibCall` predicate is used to select whether some math builtins get expanded in the FE or they get lowered into libcalls. The default implementation unconditionally returns true, which is problemat

[clang] [Clang][SPIRV][AMDGPU] Override `supportsLibCall` for AMDGCNSPIRV (PR #143814)

https://github.com/AlexVlx updated https://github.com/llvm/llvm-project/pull/143814 >From a976da9d6033a252fab75de216fe6580d6e3e6ff Mon Sep 17 00:00:00 2001 From: Alex Voicu Date: Thu, 12 Jun 2025 00:47:43 +0100 Subject: [PATCH 1/2] Override `SPIRVTargetCodeGenInfo::supportsLibCall()` in order

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

@@ -0,0 +1,21 @@ +// REQUIRES: amdgpu-registered-target +// REQUIRES: spirv-registered-target +// RUN: %clang_cc1 -fsyntax-only -verify -triple amdgcn -Wno-unused-value %s +// RUN: %clang_cc1 -fsyntax-only -verify -triple spirv64-amd-amdhsa -Wno-unused-value %s +// RUN: %clang_cc

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

AlexVlx wrote: Gentle ping. https://github.com/llvm/llvm-project/pull/134016 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

AlexVlx wrote: > 1. Have what is proposed here as an initial step, with the addition that we > issue warnings on unguarded uses of builtins / ASM (similar to what > `__builtin_available` / `@available` do), and we clean-up non-extern > functions that become unreachable as a consequence of pred

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

AlexVlx wrote: > So your users today are building for generic AMDGPU but using builtins that > are only available on a specific processor release? Presumably that code is > protected _somehow_ and their programs are not simply crashing at runtime. Is > that something you'd be able to leverage

[clang] [llvm] [HIPSTDPAR] Add support for globals (PR #146813)

https://github.com/AlexVlx updated https://github.com/llvm/llvm-project/pull/146813 >From d98e3785a144ada9881cdbe24c86f273850eca20 Mon Sep 17 00:00:00 2001 From: Alex Voicu Date: Thu, 3 Jul 2025 02:02:04 +0100 Subject: [PATCH 1/6] Add support for true globals. --- llvm/lib/Transforms/HipStdPa

[clang] [llvm] [HIPSTDPAR] Add support for globals (PR #146813)

@@ -114,24 +115,221 @@ static inline void clearModule(Module &M) { // TODO: simplify. eraseFromModule(*M.ifuncs().begin()); } +static inline SmallVector> +collectIndirectableUses(GlobalVariable *G) { + // We are interested only in use chains that end in an Instruction. +

[clang] [llvm] [HIPSTDPAR] Add support for globals (PR #146813)

https://github.com/AlexVlx updated https://github.com/llvm/llvm-project/pull/146813 >From d98e3785a144ada9881cdbe24c86f273850eca20 Mon Sep 17 00:00:00 2001 From: Alex Voicu Date: Thu, 3 Jul 2025 02:02:04 +0100 Subject: [PATCH 1/5] Add support for true globals. --- llvm/lib/Transforms/HipStdPa

[clang] [llvm] [HIPSTDPAR] Add support for globals (PR #146813)

https://github.com/AlexVlx updated https://github.com/llvm/llvm-project/pull/146813 >From d98e3785a144ada9881cdbe24c86f273850eca20 Mon Sep 17 00:00:00 2001 From: Alex Voicu Date: Thu, 3 Jul 2025 02:02:04 +0100 Subject: [PATCH 1/4] Add support for true globals. --- llvm/lib/Transforms/HipStdPa

[clang] [llvm] [HIPSTDPAR] Add support for globals (PR #146813)

@@ -114,24 +115,221 @@ static inline void clearModule(Module &M) { // TODO: simplify. eraseFromModule(*M.ifuncs().begin()); } +static inline SmallVector> +collectIndirectableUses(GlobalVariable *G) { + // We are interested only in use chains that end in an Instruction. +

[clang] [llvm] [HIPSTDPAR] Add support for globals (PR #146813)

@@ -114,24 +115,221 @@ static inline void clearModule(Module &M) { // TODO: simplify. eraseFromModule(*M.ifuncs().begin()); } +static inline SmallVector> +collectIndirectableUses(GlobalVariable *G) { + // We are interested only in use chains that end in an Instruction. +

[clang] [llvm] [HIPSTDPAR] Add support for globals (PR #146813)

@@ -114,24 +115,221 @@ static inline void clearModule(Module &M) { // TODO: simplify. eraseFromModule(*M.ifuncs().begin()); } +static inline SmallVector> +collectIndirectableUses(GlobalVariable *G) { + // We are interested only in use chains that end in an Instruction. +

[clang] [llvm] [HIPSTDPAR] Add support for globals (PR #146813)

@@ -114,24 +115,221 @@ static inline void clearModule(Module &M) { // TODO: simplify. eraseFromModule(*M.ifuncs().begin()); } +static inline SmallVector> +collectIndirectableUses(GlobalVariable *G) { + // We are interested only in use chains that end in an Instruction. +

[clang] [llvm] [HIPSTDPAR] Add support for globals (PR #146813)

AlexVlx wrote: This is the run time part needed to make things work: https://github.com/llvm/llvm-project/pull/146813 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-

[clang] [llvm] [HIPSTDPAR] Add support for globals (PR #146813)

https://github.com/AlexVlx updated https://github.com/llvm/llvm-project/pull/146813 >From d98e3785a144ada9881cdbe24c86f273850eca20 Mon Sep 17 00:00:00 2001 From: Alex Voicu Date: Thu, 3 Jul 2025 02:02:04 +0100 Subject: [PATCH 1/6] Add support for true globals. --- llvm/lib/Transforms/HipStdPa

[clang] [llvm] [HIPSTDPAR] Add support for globals (PR #146813)

https://github.com/AlexVlx updated https://github.com/llvm/llvm-project/pull/146813 >From d98e3785a144ada9881cdbe24c86f273850eca20 Mon Sep 17 00:00:00 2001 From: Alex Voicu Date: Thu, 3 Jul 2025 02:02:04 +0100 Subject: [PATCH 1/4] Add support for true globals. --- llvm/lib/Transforms/HipStdPa

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

AlexVlx wrote: > @efriedma-quic was kind enough to have a call where we discussed this a bit > more. I'll update tomorrow with a potential way forward, for the group's > consideration. Following up, here's a possible approach to making progress, broken down in phases, (@efriedma-quic can corr

[clang] [llvm] [HIPSTDPAR] Add support for globals (PR #146813)

https://github.com/AlexVlx created https://github.com/llvm/llvm-project/pull/146813 This (mostly) removes one of the largest remaining limitations of `hipstdpar` based algorithm acceleration, by adding support for global variable usage in offloaded algorithms. It is mean to compose with a run

[clang] [llvm] [HIPSTDPAR] Add support for globals (PR #146813)

https://github.com/AlexVlx updated https://github.com/llvm/llvm-project/pull/146813 >From d98e3785a144ada9881cdbe24c86f273850eca20 Mon Sep 17 00:00:00 2001 From: Alex Voicu Date: Thu, 3 Jul 2025 02:02:04 +0100 Subject: [PATCH 1/3] Add support for true globals. --- llvm/lib/Transforms/HipStdPa

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

AlexVlx wrote: > I think Eli is suggesting something like the rule for > [@available](https://clang.llvm.org/docs/LanguageExtensions.html#objective-c-available): > > * If a builtin is unconditionally available, you can always use it without > any diagnostics. > * If a builtin is only available

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

AlexVlx wrote: > > Definitely, more than happy to have a 1-on-1 (2-on-1 even, since I think > > @AaronBallman also suggested something along these lines as well :) ). > > Please email me with some times that will work for you. > > > We've just made the call to foo() illegal on anything that is

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

AlexVlx wrote: > I mean, I'm not particularly attached to the syntax of the "if". I guess we > could designate `if (__builtin_amdgcn_processor_is("gfx900")) {}` as a > "processor-feature-if". The point is that we need to know at the AST level > which processor features are available for each s

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

AlexVlx wrote: > On a different point: I don't think this builtin is actually semantically > different from `__builtin_cpu_is`. As long as we're not treating it as > `constexpr`, the fact that it's lowered by the compiler and doesn't need a > runtime check is just a happy property of GPU targe

[clang] [llvm] [SPIRV] Add more id and range builtIns (PR #143909)

https://github.com/AlexVlx approved this pull request. LGTM, thanks! https://github.com/llvm/llvm-project/pull/143909 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

AlexVlx wrote: > We briefly discussed this in the clang area team meeting, and we weren't > really happy with the design as-is. The underlying idea behind the feature > makes sense, but semantics of the actual builtin is ugly: there's a loose > connection between the condition check, and the r

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

AlexVlx wrote: > An intrinsic seems like the right IR model for CPU recognition, even for > targets that don't specifically need to late-resolve it. That should be much > easier for passes to optimize based on CPU settings than directly emitting > the compiler-rt reference in the frontend. I k

[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

AlexVlx wrote: > Right, I don't see any semantic reason why `__builtin_cpu_is` or > `__builtin_cpu_supports` shouldn't be resolved statically if we have that > information on hand. `-mcpu` / `-march` are not usually sufficient for > folding `__builtin_cpu_is`, since those attributes just speci

<    2   3   4   5   6   7   8   >