https://github.com/ChuanqiXu9 created https://github.com/llvm/llvm-project/pull/92353
Close https://github.com/llvm/llvm-project/issues/91418 Since we load the variable's initializers lazily, it'd be problematic if the initializers dependent on each other. So here we try to load the initializers of static variables to make sure they are passed to code generator by order. If we read any thing interesting, we would consume that before emitting the current declaration. >From 699da64855f147708f153c30177a1d02a4e014f7 Mon Sep 17 00:00:00 2001 From: Chuanqi Xu <yedeng...@linux.alibaba.com> Date: Wed, 15 May 2024 12:37:16 +0800 Subject: [PATCH] [Serialization] Read the initializer for interesting static variables before consuming it Close https://github.com/llvm/llvm-project/issues/91418 Since we load the variable's initializers lazily, it'd be problematic if the initializers dependent on each other. So here we try to load the initializers of static variables to make sure they are passed to code generator by order. If we read any thing interesting, we would consume that before emitting the current declaration. --- clang/lib/Serialization/ASTReaderDecl.cpp | 29 ++- clang/test/Modules/pr91418.cppm | 67 +++++ clang/test/OpenMP/nvptx_lambda_capturing.cpp | 246 +++++++++---------- 3 files changed, 216 insertions(+), 126 deletions(-) create mode 100644 clang/test/Modules/pr91418.cppm diff --git a/clang/lib/Serialization/ASTReaderDecl.cpp b/clang/lib/Serialization/ASTReaderDecl.cpp index 0c647086e304a..a6254b70560c3 100644 --- a/clang/lib/Serialization/ASTReaderDecl.cpp +++ b/clang/lib/Serialization/ASTReaderDecl.cpp @@ -4186,12 +4186,35 @@ void ASTReader::PassInterestingDeclsToConsumer() { GetDecl(ID); EagerlyDeserializedDecls.clear(); - while (!PotentiallyInterestingDecls.empty()) { - Decl *D = PotentiallyInterestingDecls.front(); - PotentiallyInterestingDecls.pop_front(); + auto ConsumingPotentialInterestingDecls = [this]() { + while (!PotentiallyInterestingDecls.empty()) { + Decl *D = PotentiallyInterestingDecls.front(); + PotentiallyInterestingDecls.pop_front(); + if (isConsumerInterestedIn(D)) + PassInterestingDeclToConsumer(D); + } + }; + std::deque<Decl *> MaybeInterestingDecls = + std::move(PotentiallyInterestingDecls); + assert(PotentiallyInterestingDecls.empty()); + while (!MaybeInterestingDecls.empty()) { + Decl *D = MaybeInterestingDecls.front(); + MaybeInterestingDecls.pop_front(); + // Since we load the variable's initializers lazily, it'd be problematic + // if the initializers dependent on each other. So here we try to load the + // initializers of static variables to make sure they are passed to code + // generator by order. If we read anything interesting, we would consume + // that before emitting the current declaration. + if (auto *VD = dyn_cast<VarDecl>(D); + VD && VD->isFileVarDecl() && !VD->isExternallyVisible()) + VD->getInit(); + ConsumingPotentialInterestingDecls(); if (isConsumerInterestedIn(D)) PassInterestingDeclToConsumer(D); } + + // If we add any new potential interesting decl in the last call, consume it. + ConsumingPotentialInterestingDecls(); } void ASTReader::loadDeclUpdateRecords(PendingUpdateRecord &Record) { diff --git a/clang/test/Modules/pr91418.cppm b/clang/test/Modules/pr91418.cppm new file mode 100644 index 0000000000000..33fec992439d6 --- /dev/null +++ b/clang/test/Modules/pr91418.cppm @@ -0,0 +1,67 @@ +// RUN: rm -rf %t +// RUN: mkdir -p %t +// RUN: split-file %s %t +// +// RUN: %clang_cc1 -triple %itanium_abi_triple -std=c++20 -x c++-header %t/foo.h \ +// RUN: -emit-pch -o %t/foo.pch +// RUN: %clang_cc1 -triple %itanium_abi_triple -std=c++20 %t/use.cpp -include-pch \ +// RUN: %t/foo.pch -emit-llvm -o - | FileCheck %t/use.cpp + +//--- foo.h +#ifndef FOO_H +#define FOO_H +typedef float __m128 __attribute__((__vector_size__(16), __aligned__(16))); + +static __inline__ __m128 __attribute__((__always_inline__, __min_vector_width__(128))) +_mm_setr_ps(float __z, float __y, float __x, float __w) +{ + return __extension__ (__m128){ __z, __y, __x, __w }; +} + +typedef __m128 VR; + +inline VR MakeVR( float X, float Y, float Z, float W ) +{ + return _mm_setr_ps( X, Y, Z, W ); +} + +extern "C" float sqrtf(float); + +namespace VectorSinConstantsSSE +{ + float a = (16 * sqrtf(0.225f)); + VR A = MakeVR(a, a, a, a); + static const float b = (16 * sqrtf(0.225f)); + static const VR B = MakeVR(b, b, b, b); +} + +#endif // FOO_H + +//--- use.cpp +#include "foo.h" +float use() { + return VectorSinConstantsSSE::A[0] + VectorSinConstantsSSE::A[1] + + VectorSinConstantsSSE::A[2] + VectorSinConstantsSSE::A[3] + + VectorSinConstantsSSE::B[0] + VectorSinConstantsSSE::B[1] + + VectorSinConstantsSSE::B[2] + VectorSinConstantsSSE::B[3]; +} + +// CHECK: define{{.*}}@__cxx_global_var_init( +// CHECK: store{{.*}}[[a_RESULT:%[a-zA-Z0-9]+]], ptr @_ZN21VectorSinConstantsSSE1aE + +// CHECK: define{{.*}}@__cxx_global_var_init.1( +// CHECK: [[A_CALL:%[a-zA-Z0-9]+]] = call{{.*}}@_Z6MakeVRffff( +// CHECK: store{{.*}}[[A_CALL]], ptr @_ZN21VectorSinConstantsSSE1AE + +// CHECK: define{{.*}}@__cxx_global_var_init.2( +// CHECK: [[B_CALL:%[a-zA-Z0-9]+]] = call{{.*}}@_Z6MakeVRffff( +// CHECK: store{{.*}}[[B_CALL]], ptr @_ZN21VectorSinConstantsSSEL1BE + +// CHECK: define{{.*}}@__cxx_global_var_init.3( +// CHECK: store{{.*}}[[b_RESULT:%[a-zA-Z0-9]+]], ptr @_ZN21VectorSinConstantsSSEL1bE + +// CHECK: @_GLOBAL__sub_I_use.cpp +// CHECK: call{{.*}}@__cxx_global_var_init( +// CHECK: call{{.*}}@__cxx_global_var_init.1( +// CHECK: call{{.*}}@__cxx_global_var_init.3( +// CHECK: call{{.*}}@__cxx_global_var_init.2( diff --git a/clang/test/OpenMP/nvptx_lambda_capturing.cpp b/clang/test/OpenMP/nvptx_lambda_capturing.cpp index 641fbc38dd6bc..efea8d4a05618 100644 --- a/clang/test/OpenMP/nvptx_lambda_capturing.cpp +++ b/clang/test/OpenMP/nvptx_lambda_capturing.cpp @@ -1165,8 +1165,113 @@ int main(int argc, char **argv) { // CHECK2-NEXT: ret void // // +// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN1S3fooEv_l27 +// CHECK3-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], ptr noundef [[THIS:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[L:%.*]]) #[[ATTR0:[0-9]+]] { +// CHECK3-NEXT: entry: +// CHECK3-NEXT: [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8 +// CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca ptr, align 8 +// CHECK3-NEXT: [[L_ADDR:%.*]] = alloca ptr, align 8 +// CHECK3-NEXT: [[TMP:%.*]] = alloca ptr, align 8 +// CHECK3-NEXT: [[L1:%.*]] = alloca [[CLASS_ANON:%.*]], align 8 +// CHECK3-NEXT: [[_TMP2:%.*]] = alloca ptr, align 8 +// CHECK3-NEXT: store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8 +// CHECK3-NEXT: store ptr [[THIS]], ptr [[THIS_ADDR]], align 8 +// CHECK3-NEXT: store ptr [[L]], ptr [[L_ADDR]], align 8 +// CHECK3-NEXT: [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8 +// CHECK3-NEXT: [[TMP1:%.*]] = load ptr, ptr [[L_ADDR]], align 8 +// CHECK3-NEXT: store ptr [[TMP1]], ptr [[TMP]], align 8 +// CHECK3-NEXT: [[TMP2:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN1S3fooEv_l27_kernel_environment, ptr [[DYN_PTR]]) +// CHECK3-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP2]], -1 +// CHECK3-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]] +// CHECK3: user_code.entry: +// CHECK3-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TMP]], align 8 +// CHECK3-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[L1]], ptr align 8 [[TMP3]], i64 8, i1 false) +// CHECK3-NEXT: store ptr [[L1]], ptr [[_TMP2]], align 8 +// CHECK3-NEXT: [[TMP4:%.*]] = load ptr, ptr [[_TMP2]], align 8 +// CHECK3-NEXT: [[TMP5:%.*]] = getelementptr inbounds [[CLASS_ANON]], ptr [[TMP4]], i32 0, i32 0 +// CHECK3-NEXT: store ptr [[TMP0]], ptr [[TMP5]], align 8 +// CHECK3-NEXT: [[TMP6:%.*]] = load ptr, ptr [[_TMP2]], align 8 +// CHECK3-NEXT: [[CALL:%.*]] = call noundef i32 @_ZZN1S3fooEvENKUlvE_clEv(ptr noundef nonnull align 8 dereferenceable(8) [[TMP6]]) #[[ATTR7:[0-9]+]] +// CHECK3-NEXT: call void @__kmpc_target_deinit() +// CHECK3-NEXT: ret void +// CHECK3: worker.exit: +// CHECK3-NEXT: ret void +// +// +// CHECK3-LABEL: define {{[^@]+}}@_ZZN1S3fooEvENKUlvE_clEv +// CHECK3-SAME: (ptr noundef nonnull align 8 dereferenceable(8) [[THIS:%.*]]) #[[ATTR2:[0-9]+]] comdat align 2 { +// CHECK3-NEXT: entry: +// CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca ptr, align 8 +// CHECK3-NEXT: store ptr [[THIS]], ptr [[THIS_ADDR]], align 8 +// CHECK3-NEXT: [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8 +// CHECK3-NEXT: [[TMP0:%.*]] = getelementptr inbounds [[CLASS_ANON:%.*]], ptr [[THIS1]], i32 0, i32 0 +// CHECK3-NEXT: [[TMP1:%.*]] = load ptr, ptr [[TMP0]], align 8 +// CHECK3-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], ptr [[TMP1]], i32 0, i32 0 +// CHECK3-NEXT: [[TMP2:%.*]] = load i32, ptr [[A]], align 4 +// CHECK3-NEXT: ret i32 [[TMP2]] +// +// +// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN1S3fooEv_l29 +// CHECK3-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], ptr noundef [[THIS:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[L:%.*]]) #[[ATTR3:[0-9]+]] { +// CHECK3-NEXT: entry: +// CHECK3-NEXT: [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8 +// CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca ptr, align 8 +// CHECK3-NEXT: [[L_ADDR:%.*]] = alloca ptr, align 8 +// CHECK3-NEXT: [[TMP:%.*]] = alloca ptr, align 8 +// CHECK3-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8 +// CHECK3-NEXT: store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8 +// CHECK3-NEXT: store ptr [[THIS]], ptr [[THIS_ADDR]], align 8 +// CHECK3-NEXT: store ptr [[L]], ptr [[L_ADDR]], align 8 +// CHECK3-NEXT: [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8 +// CHECK3-NEXT: [[TMP1:%.*]] = load ptr, ptr [[L_ADDR]], align 8 +// CHECK3-NEXT: store ptr [[TMP1]], ptr [[TMP]], align 8 +// CHECK3-NEXT: [[TMP2:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN1S3fooEv_l29_kernel_environment, ptr [[DYN_PTR]]) +// CHECK3-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP2]], -1 +// CHECK3-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]] +// CHECK3: user_code.entry: +// CHECK3-NEXT: [[TMP3:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1:[0-9]+]]) +// CHECK3-NEXT: [[TMP4:%.*]] = load ptr, ptr [[TMP]], align 8 +// CHECK3-NEXT: [[TMP5:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0 +// CHECK3-NEXT: store ptr [[TMP0]], ptr [[TMP5]], align 8 +// CHECK3-NEXT: [[TMP6:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1 +// CHECK3-NEXT: store ptr [[TMP4]], ptr [[TMP6]], align 8 +// CHECK3-NEXT: call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP3]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN1S3fooEv_l29_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 2) +// CHECK3-NEXT: call void @__kmpc_target_deinit() +// CHECK3-NEXT: ret void +// CHECK3: worker.exit: +// CHECK3-NEXT: ret void +// +// +// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN1S3fooEv_l29_omp_outlined +// CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef [[THIS:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[L:%.*]]) #[[ATTR4:[0-9]+]] { +// CHECK3-NEXT: entry: +// CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca ptr, align 8 +// CHECK3-NEXT: [[L_ADDR:%.*]] = alloca ptr, align 8 +// CHECK3-NEXT: [[TMP:%.*]] = alloca ptr, align 8 +// CHECK3-NEXT: [[L1:%.*]] = alloca [[CLASS_ANON:%.*]], align 8 +// CHECK3-NEXT: [[_TMP2:%.*]] = alloca ptr, align 8 +// CHECK3-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK3-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8 +// CHECK3-NEXT: store ptr [[THIS]], ptr [[THIS_ADDR]], align 8 +// CHECK3-NEXT: store ptr [[L]], ptr [[L_ADDR]], align 8 +// CHECK3-NEXT: [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8 +// CHECK3-NEXT: [[TMP1:%.*]] = load ptr, ptr [[L_ADDR]], align 8 +// CHECK3-NEXT: store ptr [[TMP1]], ptr [[TMP]], align 8 +// CHECK3-NEXT: [[TMP2:%.*]] = load ptr, ptr [[TMP]], align 8 +// CHECK3-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[L1]], ptr align 8 [[TMP2]], i64 8, i1 false) +// CHECK3-NEXT: store ptr [[L1]], ptr [[_TMP2]], align 8 +// CHECK3-NEXT: [[TMP3:%.*]] = load ptr, ptr [[_TMP2]], align 8 +// CHECK3-NEXT: [[TMP4:%.*]] = getelementptr inbounds [[CLASS_ANON]], ptr [[TMP3]], i32 0, i32 0 +// CHECK3-NEXT: store ptr [[TMP0]], ptr [[TMP4]], align 8 +// CHECK3-NEXT: [[TMP5:%.*]] = load ptr, ptr [[_TMP2]], align 8 +// CHECK3-NEXT: [[CALL:%.*]] = call noundef i32 @_ZZN1S3fooEvENKUlvE_clEv(ptr noundef nonnull align 8 dereferenceable(8) [[TMP5]]) #[[ATTR7]] +// CHECK3-NEXT: ret void +// +// // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l41 -// CHECK3-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i64 noundef [[ARGC:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[C:%.*]], ptr noundef [[D:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], ptr noundef nonnull align 8 dereferenceable(40) [[L:%.*]]) #[[ATTR0:[0-9]+]] { +// CHECK3-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i64 noundef [[ARGC:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[C:%.*]], ptr noundef [[D:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], ptr noundef nonnull align 8 dereferenceable(40) [[L:%.*]]) #[[ATTR0]] { // CHECK3-NEXT: entry: // CHECK3-NEXT: [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8 // CHECK3-NEXT: [[ARGC_ADDR:%.*]] = alloca i64, align 8 @@ -1178,7 +1283,7 @@ int main(int argc, char **argv) { // CHECK3-NEXT: [[TMP:%.*]] = alloca ptr, align 8 // CHECK3-NEXT: [[_TMP1:%.*]] = alloca ptr, align 8 // CHECK3-NEXT: [[_TMP2:%.*]] = alloca ptr, align 8 -// CHECK3-NEXT: [[L3:%.*]] = alloca [[CLASS_ANON:%.*]], align 8 +// CHECK3-NEXT: [[L3:%.*]] = alloca [[CLASS_ANON_1:%.*]], align 8 // CHECK3-NEXT: [[_TMP4:%.*]] = alloca ptr, align 8 // CHECK3-NEXT: [[B5:%.*]] = alloca i32, align 4 // CHECK3-NEXT: [[_TMP6:%.*]] = alloca ptr, align 8 @@ -1214,20 +1319,20 @@ int main(int argc, char **argv) { // CHECK3-NEXT: store i32 [[TMP9]], ptr [[C7]], align 4 // CHECK3-NEXT: store ptr [[C7]], ptr [[_TMP8]], align 8 // CHECK3-NEXT: [[TMP10:%.*]] = load ptr, ptr [[_TMP4]], align 8 -// CHECK3-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[CLASS_ANON]], ptr [[TMP10]], i32 0, i32 0 +// CHECK3-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[CLASS_ANON_1]], ptr [[TMP10]], i32 0, i32 0 // CHECK3-NEXT: store ptr [[ARGC_ADDR]], ptr [[TMP11]], align 8 -// CHECK3-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[CLASS_ANON]], ptr [[TMP10]], i32 0, i32 1 +// CHECK3-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[CLASS_ANON_1]], ptr [[TMP10]], i32 0, i32 1 // CHECK3-NEXT: [[TMP13:%.*]] = load ptr, ptr [[_TMP6]], align 8 // CHECK3-NEXT: store ptr [[TMP13]], ptr [[TMP12]], align 8 -// CHECK3-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[CLASS_ANON]], ptr [[TMP10]], i32 0, i32 2 +// CHECK3-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[CLASS_ANON_1]], ptr [[TMP10]], i32 0, i32 2 // CHECK3-NEXT: [[TMP15:%.*]] = load ptr, ptr [[_TMP8]], align 8 // CHECK3-NEXT: store ptr [[TMP15]], ptr [[TMP14]], align 8 -// CHECK3-NEXT: [[TMP16:%.*]] = getelementptr inbounds [[CLASS_ANON]], ptr [[TMP10]], i32 0, i32 3 +// CHECK3-NEXT: [[TMP16:%.*]] = getelementptr inbounds [[CLASS_ANON_1]], ptr [[TMP10]], i32 0, i32 3 // CHECK3-NEXT: store ptr [[D_ADDR]], ptr [[TMP16]], align 8 -// CHECK3-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[CLASS_ANON]], ptr [[TMP10]], i32 0, i32 4 +// CHECK3-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[CLASS_ANON_1]], ptr [[TMP10]], i32 0, i32 4 // CHECK3-NEXT: store ptr [[TMP2]], ptr [[TMP17]], align 8 // CHECK3-NEXT: [[TMP18:%.*]] = load ptr, ptr [[_TMP4]], align 8 -// CHECK3-NEXT: [[CALL:%.*]] = call noundef i64 @"_ZZ4mainENK3$_0clEv"(ptr noundef nonnull align 8 dereferenceable(40) [[TMP18]]) #[[ATTR7:[0-9]+]] +// CHECK3-NEXT: [[CALL:%.*]] = call noundef i64 @"_ZZ4mainENK3$_0clEv"(ptr noundef nonnull align 8 dereferenceable(40) [[TMP18]]) #[[ATTR7]] // CHECK3-NEXT: call void @__kmpc_target_deinit() // CHECK3-NEXT: ret void // CHECK3: worker.exit: @@ -1235,7 +1340,7 @@ int main(int argc, char **argv) { // // // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l43 -// CHECK3-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[ARGC:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[C:%.*]], ptr noundef [[D:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], ptr noundef nonnull align 8 dereferenceable(40) [[L:%.*]]) #[[ATTR3:[0-9]+]] { +// CHECK3-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[ARGC:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[C:%.*]], ptr noundef [[D:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], ptr noundef nonnull align 8 dereferenceable(40) [[L:%.*]]) #[[ATTR3]] { // CHECK3-NEXT: entry: // CHECK3-NEXT: [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8 // CHECK3-NEXT: [[ARGC_ADDR:%.*]] = alloca ptr, align 8 @@ -1267,7 +1372,7 @@ int main(int argc, char **argv) { // CHECK3-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP5]], -1 // CHECK3-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]] // CHECK3: user_code.entry: -// CHECK3-NEXT: [[TMP6:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1:[0-9]+]]) +// CHECK3-NEXT: [[TMP6:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) // CHECK3-NEXT: [[TMP7:%.*]] = load ptr, ptr [[TMP]], align 8 // CHECK3-NEXT: [[TMP8:%.*]] = load ptr, ptr [[_TMP1]], align 8 // CHECK3-NEXT: [[TMP9:%.*]] = load ptr, ptr [[D_ADDR]], align 8 @@ -1292,7 +1397,7 @@ int main(int argc, char **argv) { // // // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l43_omp_outlined -// CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[ARGC:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[C:%.*]], ptr noundef [[D:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], ptr noundef nonnull align 8 dereferenceable(40) [[L:%.*]]) #[[ATTR4:[0-9]+]] { +// CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[ARGC:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[C:%.*]], ptr noundef [[D:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], ptr noundef nonnull align 8 dereferenceable(40) [[L:%.*]]) #[[ATTR4]] { // CHECK3-NEXT: entry: // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8 @@ -1305,7 +1410,7 @@ int main(int argc, char **argv) { // CHECK3-NEXT: [[TMP:%.*]] = alloca ptr, align 8 // CHECK3-NEXT: [[_TMP1:%.*]] = alloca ptr, align 8 // CHECK3-NEXT: [[_TMP2:%.*]] = alloca ptr, align 8 -// CHECK3-NEXT: [[L3:%.*]] = alloca [[CLASS_ANON:%.*]], align 8 +// CHECK3-NEXT: [[L3:%.*]] = alloca [[CLASS_ANON_1:%.*]], align 8 // CHECK3-NEXT: [[_TMP4:%.*]] = alloca ptr, align 8 // CHECK3-NEXT: [[ARGC5:%.*]] = alloca i32, align 4 // CHECK3-NEXT: [[B6:%.*]] = alloca i32, align 4 @@ -1345,128 +1450,23 @@ int main(int argc, char **argv) { // CHECK3-NEXT: [[TMP11:%.*]] = load i32, ptr [[TMP3]], align 4 // CHECK3-NEXT: store i32 [[TMP11]], ptr [[A10]], align 4 // CHECK3-NEXT: [[TMP12:%.*]] = load ptr, ptr [[_TMP4]], align 8 -// CHECK3-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[CLASS_ANON]], ptr [[TMP12]], i32 0, i32 0 +// CHECK3-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[CLASS_ANON_1]], ptr [[TMP12]], i32 0, i32 0 // CHECK3-NEXT: store ptr [[ARGC5]], ptr [[TMP13]], align 8 -// CHECK3-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[CLASS_ANON]], ptr [[TMP12]], i32 0, i32 1 +// CHECK3-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[CLASS_ANON_1]], ptr [[TMP12]], i32 0, i32 1 // CHECK3-NEXT: [[TMP15:%.*]] = load ptr, ptr [[_TMP7]], align 8 // CHECK3-NEXT: store ptr [[TMP15]], ptr [[TMP14]], align 8 -// CHECK3-NEXT: [[TMP16:%.*]] = getelementptr inbounds [[CLASS_ANON]], ptr [[TMP12]], i32 0, i32 2 +// CHECK3-NEXT: [[TMP16:%.*]] = getelementptr inbounds [[CLASS_ANON_1]], ptr [[TMP12]], i32 0, i32 2 // CHECK3-NEXT: [[TMP17:%.*]] = load ptr, ptr [[_TMP9]], align 8 // CHECK3-NEXT: store ptr [[TMP17]], ptr [[TMP16]], align 8 -// CHECK3-NEXT: [[TMP18:%.*]] = getelementptr inbounds [[CLASS_ANON]], ptr [[TMP12]], i32 0, i32 3 +// CHECK3-NEXT: [[TMP18:%.*]] = getelementptr inbounds [[CLASS_ANON_1]], ptr [[TMP12]], i32 0, i32 3 // CHECK3-NEXT: store ptr [[D_ADDR]], ptr [[TMP18]], align 8 -// CHECK3-NEXT: [[TMP19:%.*]] = getelementptr inbounds [[CLASS_ANON]], ptr [[TMP12]], i32 0, i32 4 +// CHECK3-NEXT: [[TMP19:%.*]] = getelementptr inbounds [[CLASS_ANON_1]], ptr [[TMP12]], i32 0, i32 4 // CHECK3-NEXT: store ptr [[A10]], ptr [[TMP19]], align 8 // CHECK3-NEXT: [[TMP20:%.*]] = load ptr, ptr [[_TMP4]], align 8 // CHECK3-NEXT: [[CALL:%.*]] = call noundef i64 @"_ZZ4mainENK3$_0clEv"(ptr noundef nonnull align 8 dereferenceable(40) [[TMP20]]) #[[ATTR7]] // CHECK3-NEXT: ret void // // -// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN1S3fooEv_l27 -// CHECK3-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], ptr noundef [[THIS:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[L:%.*]]) #[[ATTR0]] { -// CHECK3-NEXT: entry: -// CHECK3-NEXT: [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8 -// CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca ptr, align 8 -// CHECK3-NEXT: [[L_ADDR:%.*]] = alloca ptr, align 8 -// CHECK3-NEXT: [[TMP:%.*]] = alloca ptr, align 8 -// CHECK3-NEXT: [[L1:%.*]] = alloca [[CLASS_ANON_1:%.*]], align 8 -// CHECK3-NEXT: [[_TMP2:%.*]] = alloca ptr, align 8 -// CHECK3-NEXT: store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8 -// CHECK3-NEXT: store ptr [[THIS]], ptr [[THIS_ADDR]], align 8 -// CHECK3-NEXT: store ptr [[L]], ptr [[L_ADDR]], align 8 -// CHECK3-NEXT: [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8 -// CHECK3-NEXT: [[TMP1:%.*]] = load ptr, ptr [[L_ADDR]], align 8 -// CHECK3-NEXT: store ptr [[TMP1]], ptr [[TMP]], align 8 -// CHECK3-NEXT: [[TMP2:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN1S3fooEv_l27_kernel_environment, ptr [[DYN_PTR]]) -// CHECK3-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP2]], -1 -// CHECK3-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]] -// CHECK3: user_code.entry: -// CHECK3-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TMP]], align 8 -// CHECK3-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[L1]], ptr align 8 [[TMP3]], i64 8, i1 false) -// CHECK3-NEXT: store ptr [[L1]], ptr [[_TMP2]], align 8 -// CHECK3-NEXT: [[TMP4:%.*]] = load ptr, ptr [[_TMP2]], align 8 -// CHECK3-NEXT: [[TMP5:%.*]] = getelementptr inbounds [[CLASS_ANON_1]], ptr [[TMP4]], i32 0, i32 0 -// CHECK3-NEXT: store ptr [[TMP0]], ptr [[TMP5]], align 8 -// CHECK3-NEXT: [[TMP6:%.*]] = load ptr, ptr [[_TMP2]], align 8 -// CHECK3-NEXT: [[CALL:%.*]] = call noundef i32 @_ZZN1S3fooEvENKUlvE_clEv(ptr noundef nonnull align 8 dereferenceable(8) [[TMP6]]) #[[ATTR7]] -// CHECK3-NEXT: call void @__kmpc_target_deinit() -// CHECK3-NEXT: ret void -// CHECK3: worker.exit: -// CHECK3-NEXT: ret void -// -// -// CHECK3-LABEL: define {{[^@]+}}@_ZZN1S3fooEvENKUlvE_clEv -// CHECK3-SAME: (ptr noundef nonnull align 8 dereferenceable(8) [[THIS:%.*]]) #[[ATTR2:[0-9]+]] comdat align 2 { -// CHECK3-NEXT: entry: -// CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca ptr, align 8 -// CHECK3-NEXT: store ptr [[THIS]], ptr [[THIS_ADDR]], align 8 -// CHECK3-NEXT: [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8 -// CHECK3-NEXT: [[TMP0:%.*]] = getelementptr inbounds [[CLASS_ANON_1:%.*]], ptr [[THIS1]], i32 0, i32 0 -// CHECK3-NEXT: [[TMP1:%.*]] = load ptr, ptr [[TMP0]], align 8 -// CHECK3-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], ptr [[TMP1]], i32 0, i32 0 -// CHECK3-NEXT: [[TMP2:%.*]] = load i32, ptr [[A]], align 4 -// CHECK3-NEXT: ret i32 [[TMP2]] -// -// -// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN1S3fooEv_l29 -// CHECK3-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], ptr noundef [[THIS:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[L:%.*]]) #[[ATTR3]] { -// CHECK3-NEXT: entry: -// CHECK3-NEXT: [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8 -// CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca ptr, align 8 -// CHECK3-NEXT: [[L_ADDR:%.*]] = alloca ptr, align 8 -// CHECK3-NEXT: [[TMP:%.*]] = alloca ptr, align 8 -// CHECK3-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8 -// CHECK3-NEXT: store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8 -// CHECK3-NEXT: store ptr [[THIS]], ptr [[THIS_ADDR]], align 8 -// CHECK3-NEXT: store ptr [[L]], ptr [[L_ADDR]], align 8 -// CHECK3-NEXT: [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8 -// CHECK3-NEXT: [[TMP1:%.*]] = load ptr, ptr [[L_ADDR]], align 8 -// CHECK3-NEXT: store ptr [[TMP1]], ptr [[TMP]], align 8 -// CHECK3-NEXT: [[TMP2:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN1S3fooEv_l29_kernel_environment, ptr [[DYN_PTR]]) -// CHECK3-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP2]], -1 -// CHECK3-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]] -// CHECK3: user_code.entry: -// CHECK3-NEXT: [[TMP3:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) -// CHECK3-NEXT: [[TMP4:%.*]] = load ptr, ptr [[TMP]], align 8 -// CHECK3-NEXT: [[TMP5:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0 -// CHECK3-NEXT: store ptr [[TMP0]], ptr [[TMP5]], align 8 -// CHECK3-NEXT: [[TMP6:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1 -// CHECK3-NEXT: store ptr [[TMP4]], ptr [[TMP6]], align 8 -// CHECK3-NEXT: call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP3]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN1S3fooEv_l29_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 2) -// CHECK3-NEXT: call void @__kmpc_target_deinit() -// CHECK3-NEXT: ret void -// CHECK3: worker.exit: -// CHECK3-NEXT: ret void -// -// -// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN1S3fooEv_l29_omp_outlined -// CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef [[THIS:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[L:%.*]]) #[[ATTR4]] { -// CHECK3-NEXT: entry: -// CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8 -// CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8 -// CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca ptr, align 8 -// CHECK3-NEXT: [[L_ADDR:%.*]] = alloca ptr, align 8 -// CHECK3-NEXT: [[TMP:%.*]] = alloca ptr, align 8 -// CHECK3-NEXT: [[L1:%.*]] = alloca [[CLASS_ANON_1:%.*]], align 8 -// CHECK3-NEXT: [[_TMP2:%.*]] = alloca ptr, align 8 -// CHECK3-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8 -// CHECK3-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8 -// CHECK3-NEXT: store ptr [[THIS]], ptr [[THIS_ADDR]], align 8 -// CHECK3-NEXT: store ptr [[L]], ptr [[L_ADDR]], align 8 -// CHECK3-NEXT: [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8 -// CHECK3-NEXT: [[TMP1:%.*]] = load ptr, ptr [[L_ADDR]], align 8 -// CHECK3-NEXT: store ptr [[TMP1]], ptr [[TMP]], align 8 -// CHECK3-NEXT: [[TMP2:%.*]] = load ptr, ptr [[TMP]], align 8 -// CHECK3-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[L1]], ptr align 8 [[TMP2]], i64 8, i1 false) -// CHECK3-NEXT: store ptr [[L1]], ptr [[_TMP2]], align 8 -// CHECK3-NEXT: [[TMP3:%.*]] = load ptr, ptr [[_TMP2]], align 8 -// CHECK3-NEXT: [[TMP4:%.*]] = getelementptr inbounds [[CLASS_ANON_1]], ptr [[TMP3]], i32 0, i32 0 -// CHECK3-NEXT: store ptr [[TMP0]], ptr [[TMP4]], align 8 -// CHECK3-NEXT: [[TMP5:%.*]] = load ptr, ptr [[_TMP2]], align 8 -// CHECK3-NEXT: [[CALL:%.*]] = call noundef i32 @_ZZN1S3fooEvENKUlvE_clEv(ptr noundef nonnull align 8 dereferenceable(8) [[TMP5]]) #[[ATTR7]] -// CHECK3-NEXT: ret void -// -// // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooIZN1S3fooEvEUlvE_EiRKT__l18 // CHECK3-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[T:%.*]]) #[[ATTR3]] { // CHECK3-NEXT: entry: @@ -1500,7 +1500,7 @@ int main(int argc, char **argv) { // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8 // CHECK3-NEXT: [[T_ADDR:%.*]] = alloca ptr, align 8 // CHECK3-NEXT: [[TMP:%.*]] = alloca ptr, align 8 -// CHECK3-NEXT: [[T1:%.*]] = alloca [[CLASS_ANON_1:%.*]], align 8 +// CHECK3-NEXT: [[T1:%.*]] = alloca [[CLASS_ANON:%.*]], align 8 // CHECK3-NEXT: [[_TMP2:%.*]] = alloca ptr, align 8 // CHECK3-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8 // CHECK3-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
