llvmbot wrote:
<!--LLVM PR SUMMARY COMMENT--> @llvm/pr-subscribers-clang-codegen Author: Nathan Gauër (Keenuts) <details> <summary>Changes</summary> Destructor calls were emitted without convergence intrinsics when building for SPIR-V, which means invalid IR since we mixed controlled and non-controlled convergence. --- Full diff: https://github.com/llvm/llvm-project/pull/133469.diff 2 Files Affected: - (modified) clang/lib/CodeGen/CGDeclCXX.cpp (+4-1) - (modified) clang/test/CodeGenHLSL/GlobalDestructors.hlsl (+28-12) ``````````diff diff --git a/clang/lib/CodeGen/CGDeclCXX.cpp b/clang/lib/CodeGen/CGDeclCXX.cpp index 1ad34ae61f96a..b142aaf3fe0ee 100644 --- a/clang/lib/CodeGen/CGDeclCXX.cpp +++ b/clang/lib/CodeGen/CGDeclCXX.cpp @@ -1152,7 +1152,7 @@ void CodeGenFunction::GenerateCXXGlobalCleanUpFunc( llvm::Constant *Arg; std::tie(CalleeTy, Callee, Arg) = DtorsOrStermFinalizers[e - i - 1]; - llvm::CallInst *CI = nullptr; + llvm::CallBase *CI = nullptr; if (Arg == nullptr) { assert( CGM.getCXXABI().useSinitAndSterm() && @@ -1164,6 +1164,9 @@ void CodeGenFunction::GenerateCXXGlobalCleanUpFunc( // Make sure the call and the callee agree on calling convention. if (llvm::Function *F = dyn_cast<llvm::Function>(Callee)) CI->setCallingConv(F->getCallingConv()); + + if (CGM.shouldEmitConvergenceTokens() && CI->isConvergent()) + CI = addConvergenceControlToken(CI); } } diff --git a/clang/test/CodeGenHLSL/GlobalDestructors.hlsl b/clang/test/CodeGenHLSL/GlobalDestructors.hlsl index f98318601134b..9f90971bafd05 100644 --- a/clang/test/CodeGenHLSL/GlobalDestructors.hlsl +++ b/clang/test/CodeGenHLSL/GlobalDestructors.hlsl @@ -1,5 +1,6 @@ -// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.0-compute -emit-llvm -disable-llvm-passes %s -o - | FileCheck %s --check-prefixes=CS,NOINLINE,CHECK -// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library -emit-llvm -disable-llvm-passes %s -o - | FileCheck %s --check-prefixes=LIB,NOINLINE,CHECK +// RUN: %clang_cc1 -triple spirv-unknown-vulkan1.3-compute -emit-llvm -disable-llvm-passes %s -o - | FileCheck %s --check-prefixes=CS,NOINLINE-SPIRV,CHECK +// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.0-compute -emit-llvm -disable-llvm-passes %s -o - | FileCheck %s --check-prefixes=CS,NOINLINE-DXIL,CHECK +// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library -emit-llvm -disable-llvm-passes %s -o - | FileCheck %s --check-prefixes=LIB,NOINLINE-DXIL,CHECK // RUN: %clang_cc1 -triple dxil-pc-shadermodel6.0-compute -emit-llvm -O0 %s -o - | FileCheck %s --check-prefixes=INLINE,CHECK // RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library -emit-llvm -O0 %s -o - | FileCheck %s --check-prefixes=INLINE,CHECK @@ -57,11 +58,19 @@ void main(unsigned GI : SV_GroupIndex) { // CHECK: define void @main() // CHECK-NEXT: entry: // Verify destructor is emitted -// NOINLINE-NEXT: call void @_GLOBAL__sub_I_GlobalDestructors.hlsl() -// NOINLINE-NEXT: %0 = call i32 @llvm.dx.flattened.thread.id.in.group() -// NOINLINE-NEXT: call void @_Z4mainj(i32 %0) -// NOINLINE-NEXT: call void @_GLOBAL__D_a() -// NOINLINE-NEXT: ret void +// NOINLINE-DXIL-NEXT: call void @_GLOBAL__sub_I_GlobalDestructors.hlsl() +// NOINLINE-DXIL-NEXT: %0 = call i32 @llvm.dx.flattened.thread.id.in.group() +// NOINLINE-DXIL-NEXT: call void @_Z4mainj(i32 %0) +// NOINLINE-DXIL-NEXT: call void @_GLOBAL__D_a() +// NOINLINE-DXIL-NEXT: ret void + +// NOINLINE-SPIRV-NEXT: %0 = call token @llvm.experimental.convergence.entry() +// NOINLINE-SPIRV-NEXT: call spir_func void @_GLOBAL__sub_I_GlobalDestructors.hlsl() [ "convergencectrl"(token %0) ] +// NOINLINE-SPIRV-NEXT: %1 = call i32 @llvm.spv.flattened.thread.id.in.group() +// NOINLINE-SPIRV-NEXT: call spir_func void @_Z4mainj(i32 %1) [ "convergencectrl"(token %0) ] +// NOINLINE-SPIRV-NEXT: call spir_func void @_GLOBAL__D_a() [ "convergencectrl"(token %0) ] +// NOINLINE-SPIRV-NEXT: ret void + // Verify inlining leaves only calls to "llvm." intrinsics // INLINE-NOT: call {{[^@]*}} @{{[^l][^l][^v][^m][^\.]}} // INLINE: ret void @@ -69,10 +78,17 @@ void main(unsigned GI : SV_GroupIndex) { // This is really just a sanity check I needed for myself to verify that // function scope static variables also get destroyed properly. -// NOINLINE: define internal void @_GLOBAL__D_a() [[IntAttr:\#[0-9]+]] -// NOINLINE-NEXT: entry: -// NOINLINE-NEXT: call void @_ZN4TailD1Ev(ptr @_ZZ3WagvE1T) -// NOINLINE-NEXT: call void @_ZN6PupperD1Ev(ptr @GlobalPup) -// NOINLINE-NEXT: ret void +// NOINLINE-DXIL: define internal void @_GLOBAL__D_a() [[IntAttr:\#[0-9]+]] +// NOINLINE-DXIL-NEXT: entry: +// NOINLINE-DXIL-NEXT: call void @_ZN4TailD1Ev(ptr @_ZZ3WagvE1T) +// NOINLINE-DXIL-NEXT: call void @_ZN6PupperD1Ev(ptr @GlobalPup) +// NOINLINE-DXIL-NEXT: ret void + +// NOINLINE-SPIRV: define internal spir_func void @_GLOBAL__D_a() [[IntAttr:\#[0-9]+]] +// NOINLINE-SPIRV-NEXT: entry: +// NOINLINE-SPIRV-NEXT: %0 = call token @llvm.experimental.convergence.entry() +// NOINLINE-SPIRV-NEXT: call spir_func void @_ZN4TailD1Ev(ptr @_ZZ3WagvE1T) [ "convergencectrl"(token %0) ] +// NOINLINE-SPIRV-NEXT: call spir_func void @_ZN6PupperD1Ev(ptr @GlobalPup) [ "convergencectrl"(token %0) ] +// NOINLINE-SPIRV-NEXT: ret void // NOINLINE: attributes [[IntAttr]] = {{.*}} alwaysinline `````````` </details> https://github.com/llvm/llvm-project/pull/133469 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
