[OpenMP] Remove register_requires global constructor
#80460
Merged
Conversation
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
jhuber6
requested review from
carlobertolli,
dhruvachak,
doru1004,
jdoerfert,
JonChesterfield,
jplehr,
jsjodin,
ronlieb,
saiislam and
shiltian
llvmbot
added
clang
|
@llvm/pr-subscribers-mlir-openmp @llvm/pr-subscribers-clang-codegen Author: Joseph Huber (jhuber6) ChangesSummary: This patch changes the approach to no longer use global constructors, This function removes support for the old It is worth noting that the current Patch is 2.09 MiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/80460.diff 173 Files Affected:
diff --git a/clang/lib/CodeGen/CGOpenMPRuntime.cpp b/clang/lib/CodeGen/CGOpenMPRuntime.cpp
index 4855e7410a015..a7b72df6d9f89 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntime.cpp
+++ b/clang/lib/CodeGen/CGOpenMPRuntime.cpp
@@ -10100,44 +10100,6 @@ bool CGOpenMPRuntime::markAsGlobalTarget(GlobalDecl GD) {
return !AlreadyEmittedTargetDecls.insert(D).second;
}
-llvm::Function *CGOpenMPRuntime::emitRequiresDirectiveRegFun() {
- // If we don't have entries or if we are emitting code for the device, we
- // don't need to do anything.
- if (CGM.getLangOpts().OMPTargetTriples.empty() ||
- CGM.getLangOpts().OpenMPSimd || CGM.getLangOpts().OpenMPIsTargetDevice ||
- (OMPBuilder.OffloadInfoManager.empty() &&
- !HasEmittedDeclareTargetRegion && !HasEmittedTargetRegion))
- return nullptr;
-
- // Create and register the function that handles the requires directives.
- ASTContext &C = CGM.getContext();
-
- llvm::Function *RequiresRegFn;
- {
- CodeGenFunction CGF(CGM);
- const auto &FI = CGM.getTypes().arrangeNullaryFunction();
- llvm::FunctionType *FTy = CGM.getTypes().GetFunctionType(FI);
- std::string ReqName = getName({"omp_offloading", "requires_reg"});
- RequiresRegFn = CGM.CreateGlobalInitOrCleanUpFunction(FTy, ReqName, FI);
- CGF.StartFunction(GlobalDecl(), C.VoidTy, RequiresRegFn, FI, {});
- // TODO: check for other requires clauses.
- // The requires directive takes effect only when a target region is
- // present in the compilation unit. Otherwise it is ignored and not
- // passed to the runtime. This avoids the runtime from throwing an error
- // for mismatching requires clauses across compilation units that don't
- // contain at least 1 target region.
- assert((HasEmittedTargetRegion || HasEmittedDeclareTargetRegion ||
- !OMPBuilder.OffloadInfoManager.empty()) &&
- "Target or declare target region expected.");
- CGF.EmitRuntimeCall(OMPBuilder.getOrCreateRuntimeFunction(
- CGM.getModule(), OMPRTL___tgt_register_requires),
- llvm::ConstantInt::get(
- CGM.Int64Ty, OMPBuilder.Config.getRequiresFlags()));
- CGF.FinishFunction();
- }
- return RequiresRegFn;
-}
-
void CGOpenMPRuntime::emitTeamsCall(CodeGenFunction &CGF,
const OMPExecutableDirective &D,
SourceLocation Loc,
diff --git a/clang/lib/CodeGen/CGOpenMPRuntime.h b/clang/lib/CodeGen/CGOpenMPRuntime.h
index b01b39abd1606..c3206427b143e 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntime.h
+++ b/clang/lib/CodeGen/CGOpenMPRuntime.h
@@ -1407,10 +1407,6 @@ class CGOpenMPRuntime {
/// \param GD Global to scan.
virtual bool emitTargetGlobal(GlobalDecl GD);
- /// Creates and returns a registration function for when at least one
- /// requires directives was used in the current module.
- llvm::Function *emitRequiresDirectiveRegFun();
-
/// Creates all the offload entries in the current compilation unit
/// along with the associated metadata.
void createOffloadEntriesAndInfoMetadata();
diff --git a/clang/lib/CodeGen/CodeGenModule.cpp b/clang/lib/CodeGen/CodeGenModule.cpp
index c63e4ecc3dcba..d6d75efbbb2a8 100644
--- a/clang/lib/CodeGen/CodeGenModule.cpp
+++ b/clang/lib/CodeGen/CodeGenModule.cpp
@@ -836,10 +836,6 @@ void CodeGenModule::Release() {
AddGlobalCtor(CudaCtorFunction);
}
if (OpenMPRuntime) {
- if (llvm::Function *OpenMPRequiresDirectiveRegFun =
- OpenMPRuntime->emitRequiresDirectiveRegFun()) {
- AddGlobalCtor(OpenMPRequiresDirectiveRegFun, 0);
- }
OpenMPRuntime->createOffloadEntriesAndInfoMetadata();
OpenMPRuntime->clear();
}
diff --git a/clang/test/OpenMP/bug60602.cpp b/clang/test/OpenMP/bug60602.cpp
index 2fbfdfde07a0c..3ecc70cab778a 100644
--- a/clang/test/OpenMP/bug60602.cpp
+++ b/clang/test/OpenMP/bug60602.cpp
@@ -569,10 +569,3 @@ int kernel_within_loop(int *a, int *b, int N, int num_iters) {
// CHECK: omp.precond.end:
// CHECK-NEXT: ret void
//
-//
-// CHECK-LABEL: define internal void @.omp_offloading.requires_reg
-// CHECK-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK-NEXT: entry:
-// CHECK-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK-NEXT: ret void
-//
diff --git a/clang/test/OpenMP/distribute_codegen.cpp b/clang/test/OpenMP/distribute_codegen.cpp
index e3b43002a0518..31ec6ff911905 100644
--- a/clang/test/OpenMP/distribute_codegen.cpp
+++ b/clang/test/OpenMP/distribute_codegen.cpp
@@ -1037,13 +1037,6 @@ int fint(void) { return ftemplate<int>(); }
// CHECK1-NEXT: ret void
//
//
-// CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK1-SAME: () #[[ATTR3:[0-9]+]] {
-// CHECK1-NEXT: entry:
-// CHECK1-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK1-NEXT: ret void
-//
-//
// CHECK3-LABEL: define {{[^@]+}}@_Z23without_schedule_clausePfS_S_S_
// CHECK3-SAME: (ptr noundef [[A:%.*]], ptr noundef [[B:%.*]], ptr noundef [[C:%.*]], ptr noundef [[D:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK3-NEXT: entry:
@@ -1953,13 +1946,6 @@ int fint(void) { return ftemplate<int>(); }
// CHECK3-NEXT: ret void
//
//
-// CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK3-SAME: () #[[ATTR3:[0-9]+]] {
-// CHECK3-NEXT: entry:
-// CHECK3-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK3-NEXT: ret void
-//
-//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z23without_schedule_clausePfS_S_S__l56
// CHECK17-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], ptr noundef [[A:%.*]], ptr noundef [[B:%.*]], ptr noundef [[C:%.*]], ptr noundef [[D:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK17-NEXT: entry:
diff --git a/clang/test/OpenMP/distribute_firstprivate_codegen.cpp b/clang/test/OpenMP/distribute_firstprivate_codegen.cpp
index 361e26bc2984c..800a002e43968 100644
--- a/clang/test/OpenMP/distribute_firstprivate_codegen.cpp
+++ b/clang/test/OpenMP/distribute_firstprivate_codegen.cpp
@@ -304,13 +304,6 @@ int main() {
// CHECK1-NEXT: ret void
//
//
-// CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK1-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK1-NEXT: entry:
-// CHECK1-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK1-NEXT: ret void
-//
-//
// CHECK3-LABEL: define {{[^@]+}}@main
// CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK3-NEXT: entry:
@@ -476,13 +469,6 @@ int main() {
// CHECK3-NEXT: ret void
//
//
-// CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK3-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK3-NEXT: entry:
-// CHECK3-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK3-NEXT: ret void
-//
-//
// CHECK9-LABEL: define {{[^@]+}}@main
// CHECK9-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK9-NEXT: entry:
@@ -794,7 +780,7 @@ int main() {
//
//
// CHECK9-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
-// CHECK9-SAME: () #[[ATTR5:[0-9]+]] comdat {
+// CHECK9-SAME: () #[[ATTR1]] comdat {
// CHECK9-NEXT: entry:
// CHECK9-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
// CHECK9-NEXT: [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
@@ -1148,13 +1134,6 @@ int main() {
// CHECK9-NEXT: ret void
//
//
-// CHECK9-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK9-SAME: () #[[ATTR6:[0-9]+]] {
-// CHECK9-NEXT: entry:
-// CHECK9-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK9-NEXT: ret void
-//
-//
// CHECK11-LABEL: define {{[^@]+}}@main
// CHECK11-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK11-NEXT: entry:
@@ -1464,7 +1443,7 @@ int main() {
//
//
// CHECK11-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
-// CHECK11-SAME: () #[[ATTR5:[0-9]+]] comdat {
+// CHECK11-SAME: () #[[ATTR1]] comdat {
// CHECK11-NEXT: entry:
// CHECK11-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
// CHECK11-NEXT: [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
@@ -1815,10 +1794,3 @@ int main() {
// CHECK11-NEXT: [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT: ret void
//
-//
-// CHECK11-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK11-SAME: () #[[ATTR6:[0-9]+]] {
-// CHECK11-NEXT: entry:
-// CHECK11-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK11-NEXT: ret void
-//
diff --git a/clang/test/OpenMP/distribute_lastprivate_codegen.cpp b/clang/test/OpenMP/distribute_lastprivate_codegen.cpp
index e005de30e14d1..772372076e947 100644
--- a/clang/test/OpenMP/distribute_lastprivate_codegen.cpp
+++ b/clang/test/OpenMP/distribute_lastprivate_codegen.cpp
@@ -291,13 +291,6 @@ int main() {
// CHECK1-NEXT: ret void
//
//
-// CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK1-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK1-NEXT: entry:
-// CHECK1-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK1-NEXT: ret void
-//
-//
// CHECK3-LABEL: define {{[^@]+}}@main
// CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK3-NEXT: entry:
@@ -460,13 +453,6 @@ int main() {
// CHECK3-NEXT: ret void
//
//
-// CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK3-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK3-NEXT: entry:
-// CHECK3-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK3-NEXT: ret void
-//
-//
// CHECK9-LABEL: define {{[^@]+}}@main
// CHECK9-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK9-NEXT: entry:
@@ -797,7 +783,7 @@ int main() {
//
//
// CHECK9-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
-// CHECK9-SAME: () #[[ATTR5:[0-9]+]] comdat {
+// CHECK9-SAME: () #[[ATTR1]] comdat {
// CHECK9-NEXT: entry:
// CHECK9-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
// CHECK9-NEXT: [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
@@ -1169,13 +1155,6 @@ int main() {
// CHECK9-NEXT: ret void
//
//
-// CHECK9-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK9-SAME: () #[[ATTR6:[0-9]+]] {
-// CHECK9-NEXT: entry:
-// CHECK9-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK9-NEXT: ret void
-//
-//
// CHECK11-LABEL: define {{[^@]+}}@main
// CHECK11-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK11-NEXT: entry:
@@ -1504,7 +1483,7 @@ int main() {
//
//
// CHECK11-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
-// CHECK11-SAME: () #[[ATTR5:[0-9]+]] comdat {
+// CHECK11-SAME: () #[[ATTR1]] comdat {
// CHECK11-NEXT: entry:
// CHECK11-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
// CHECK11-NEXT: [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
@@ -1873,10 +1852,3 @@ int main() {
// CHECK11-NEXT: [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT: ret void
//
-//
-// CHECK11-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK11-SAME: () #[[ATTR6:[0-9]+]] {
-// CHECK11-NEXT: entry:
-// CHECK11-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK11-NEXT: ret void
-//
diff --git a/clang/test/OpenMP/distribute_parallel_for_codegen.cpp b/clang/test/OpenMP/distribute_parallel_for_codegen.cpp
index 7bdc4c5ab21a7..95adefa8020f6 100644
--- a/clang/test/OpenMP/distribute_parallel_for_codegen.cpp
+++ b/clang/test/OpenMP/distribute_parallel_for_codegen.cpp
@@ -2538,13 +2538,6 @@ int main() {
// CHECK1-NEXT: ret void
//
//
-// CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK1-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK1-NEXT: entry:
-// CHECK1-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK1-NEXT: ret void
-//
-//
// CHECK3-LABEL: define {{[^@]+}}@main
// CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK3-NEXT: entry:
@@ -4265,13 +4258,6 @@ int main() {
// CHECK3-NEXT: ret void
//
//
-// CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK3-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK3-NEXT: entry:
-// CHECK3-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK3-NEXT: ret void
-//
-//
// CHECK9-LABEL: define {{[^@]+}}@main
// CHECK9-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK9-NEXT: entry:
@@ -8886,13 +8872,6 @@ int main() {
// CHECK9-NEXT: ret void
//
//
-// CHECK9-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK9-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK9-NEXT: entry:
-// CHECK9-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK9-NEXT: ret void
-//
-//
// CHECK11-LABEL: define {{[^@]+}}@main
// CHECK11-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK11-NEXT: entry:
@@ -13404,10 +13383,3 @@ int main() {
// CHECK11: omp.precond.end:
// CHECK11-NEXT: ret void
//
-//
-// CHECK11-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK11-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK11-NEXT: entry:
-// CHECK11-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK11-NEXT: ret void
-//
diff --git a/clang/test/OpenMP/distribute_parallel_for_firstprivate_codegen.cpp b/clang/test/OpenMP/distribute_parallel_for_firstprivate_codegen.cpp
index 9f900facc6a54..46c115e40e435 100644
--- a/clang/test/OpenMP/distribute_parallel_for_firstprivate_codegen.cpp
+++ b/clang/test/OpenMP/distribute_parallel_for_firstprivate_codegen.cpp
@@ -504,13 +504,6 @@ int main() {
// CHECK1-NEXT: ret void
//
//
-// CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK1-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK1-NEXT: entry:
-// CHECK1-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK1-NEXT: ret void
-//
-//
// CHECK3-LABEL: define {{[^@]+}}@main
// CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK3-NEXT: entry:
@@ -759,13 +752,6 @@ int main() {
// CHECK3-NEXT: ret void
//
//
-// CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK3-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK3-NEXT: entry:
-// CHECK3-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK3-NEXT: ret void
-//
-//
// CHECK8-LABEL: define {{[^@]+}}@main
// CHECK8-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK8-NEXT: entry:
@@ -1207,7 +1193,7 @@ int main() {
//
//
// CHECK8-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
-// CHECK8-SAME: () #[[ATTR5:[0-9]+]] comdat {
+// CHECK8-SAME: () #[[ATTR1]] comdat {
// CHECK8-NEXT: entry:
// CHECK8-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
// CHECK8-NEXT: [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
@@ -1685,13 +1671,6 @@ int main() {
// CHECK8-NEXT: ret void
//
//
-// CHECK8-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK8-SAME: () #[[ATTR6:[0-9]+]] {
-// CHECK8-NEXT: entry:
-// CHECK8-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK8-NEXT: ret void
-//
-//
// CHECK10-LABEL: define {{[^@]+}}@main
// CHECK10-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK10-NEXT: entry:
@@ -2127,7 +2106,7 @@ int main() {
//
//
// CHECK10-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
-// CHECK10-SAME: () #[[ATTR5:[0-9]+]] comdat {
+// CHECK10-SAME: () #[[ATTR1]] comdat {
// CHECK10-NEXT: entry:
// CHECK10-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
// CHECK10-NEXT: [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
@@ -2598,10 +2577,3 @@ int main() {
// CHECK10-NEXT: [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK10-NEXT: ret void
//
-//
-// CHECK10-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK10-SAME: () #[[ATTR6:[0-9]+]] {
-// CHECK10-NEXT: entry:
-// CHECK10-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK10-NEXT: ret void
-//
diff --git a/clang/test/OpenMP/distribute_parallel_for_if_codegen.cpp b/clang/test/OpenMP/distribute_parallel_for_if_codegen.cpp
index 83c9f504ccaca..846e7beb5d92f 100644
--- a/clang/test/OpenMP/distribute_parallel_for_if_codegen.cpp
+++ b/clang/test/OpenMP/distribute_parallel_for_if_codegen.cpp
@@ -1609,10 +1609,3 @@ int main() {
// CHECK1-NEXT: call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP3]])
// CHECK1-NEXT: ret void
//
-//
-// CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK1-SAME: () #[[ATTR5:[0-9]+]] {
-// CHECK1-NEXT: entry:
-// CHECK1-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK1-NEXT: ret void
-//
diff --git a/clang/test/OpenMP/distribute_parallel_for_lastprivate_codegen.cpp b/clang/test/OpenMP/distribute_parallel_for_lastprivate_codegen.cpp
index 8c44a1e71ae79..aa981f606cc87 100644
--- a/clang/test/OpenMP/distribute_parallel_for_lastprivate_codegen.cpp
+++ b/clang/test/OpenMP/distribute_parallel_for_lastprivate_codegen.cpp
@@ -462,13 +462,6 @@ int main() {
// CHECK1-NEXT: ret void
//
//
-// CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK1-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK1-NEXT: entry:
-// CHECK1-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK1-NEXT: ret void
-//
-//
// CHECK3-LABEL: define {{[^@]+}}@main
// CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK3-NEXT: entry:
@@ -734,13 +727,6 @@ int main() {
// CHECK3-NEXT: ret void
//
//
-// CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK3-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK3-NEXT: entry:
-// CHECK3-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK3-NEXT: ret void
-//
-//
// CHECK9-LABEL: define {{[^@]+}}@main
// CHECK9-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK9-NEXT: entry:
@@ -1219,7 +1205,7 @@ int main() {
//
//
// CHECK9-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
-// CHECK9-SAME: () #[[ATTR5:[0-9]+]] comdat {
+// CHECK9-SAME: () #[[ATTR1]] comdat {
// CHECK9-NEXT: entry:
// CHECK9-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
// CHECK9-NEXT: [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
@@ -1733,13 +1719,6 @@ int main() {
// CHECK9-NEXT: ret void
//
//
-// CHECK9-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK9-SAME: () #[[ATTR6:[0-9]+]] {
-// CHECK9-NEXT: entry:
-// CHECK9-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK9-NEXT: ret void
-//
-//
// CHECK11-LABEL: define {{[^@]+}}@main
// CHECK11-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK11-NEXT: entry:
@@ -2212,7 +2191,7 @@ int main() {
//
//
// CHECK11-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
-// CHECK11-SAME: () #[[ATTR5:[0-9]+]] comdat {
+// CHECK11-SAME: () #[[ATTR1]] comdat {
// CHECK11-NEXT: entry:
// CHECK11-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
// CHECK11-NEXT: [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
@@ -2719,10 +2698,3 @@ int main() {
// CHECK11-NEXT: [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT: ret void
//
-//
-// CHECK11-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK11-SAME: () #[[ATTR6:[0-9]+]] {
-// CHECK11-NEXT: entry:
-// CHECK11-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK11-NEXT: ret void
-//
diff --git a/clang/test/OpenMP/distribute_parallel_for_num_threads_codegen.cpp b/clang/test/OpenMP/distribute_parallel_for_num_threads_codegen.cpp
index 9f769ca2886fe..5d9244268d554 100644
--- a/clang/test/OpenMP/distribute_parallel_for_num_threads_codegen.cpp
+++ b/clang/test/OpenMP/distribute_parallel_for_num_threads_codegen.cpp
@@ -112,7 +112,7 @@ int main() {
// CHECK1-NEXT: store i32 0, ptr [[RETVAL]], align 4
// CHECK1-NEXT: call void @_ZN1SC1El(ptr noundef nonnull align 8 dereferenceable(24) [[S]], i64 noundef 0)
// CHECK1-NEXT: [[CALL:%.*]] = invoke noundef signext i8 @_ZN1ScvcEv(ptr noundef nonnull align 8 dereferenceable(24) [[S]])
-// CHECK1-NEXT: to label [[INVOKE_CONT:%.*]] unwind label [[LPAD:%.*]]
+// CHECK1-NEXT: to label [[INVOKE_CONT:%.*]] unwind label [[LPAD:%.*]]
// CHECK1: invoke.cont:
// CHECK1-NEXT: store i8 [[CALL]], ptr [[A]], align 1
// CHECK1-NEXT: [[TMP0:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
@@ -145,16 +145,16 @@ int main() {
// CHECK1-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK1-NEXT: br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK1: omp_offload.failed:
-// CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68() #[[ATTR4:[0-9]+]]
+// CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68() #[[ATTR3:[0-9]+]]
// CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]]
// CHECK1: lpad:
// CHECK1-NEXT: [[TMP15:%.*]] = landingpad { ptr, i32 }
-// CHECK1-NEXT: cleanup
+// CHECK1-NEXT: cleanup
// CHECK1-NEXT: [[TMP16:%.*]] = extractvalue { ptr, i32 } [[TMP15]], 0
// CHECK1-NEXT: store ptr [[TMP16]], ptr [[EXN_SLOT]], align 8
// CHECK1-NEXT: [[TMP17:%.*]] = extractvalue { ptr, i32 } [[TMP15]], 1
// CHECK1-NEXT: store...
[truncated]
|
|
@llvm/pr-subscribers-flang-openmp Author: Joseph Huber (jhuber6) ChangesSummary: This patch changes the approach to no longer use global constructors, This function removes support for the old It is worth noting that the current Patch is 2.09 MiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/80460.diff 173 Files Affected:
diff --git a/clang/lib/CodeGen/CGOpenMPRuntime.cpp b/clang/lib/CodeGen/CGOpenMPRuntime.cpp
index 4855e7410a015..a7b72df6d9f89 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntime.cpp
+++ b/clang/lib/CodeGen/CGOpenMPRuntime.cpp
@@ -10100,44 +10100,6 @@ bool CGOpenMPRuntime::markAsGlobalTarget(GlobalDecl GD) {
return !AlreadyEmittedTargetDecls.insert(D).second;
}
-llvm::Function *CGOpenMPRuntime::emitRequiresDirectiveRegFun() {
- // If we don't have entries or if we are emitting code for the device, we
- // don't need to do anything.
- if (CGM.getLangOpts().OMPTargetTriples.empty() ||
- CGM.getLangOpts().OpenMPSimd || CGM.getLangOpts().OpenMPIsTargetDevice ||
- (OMPBuilder.OffloadInfoManager.empty() &&
- !HasEmittedDeclareTargetRegion && !HasEmittedTargetRegion))
- return nullptr;
-
- // Create and register the function that handles the requires directives.
- ASTContext &C = CGM.getContext();
-
- llvm::Function *RequiresRegFn;
- {
- CodeGenFunction CGF(CGM);
- const auto &FI = CGM.getTypes().arrangeNullaryFunction();
- llvm::FunctionType *FTy = CGM.getTypes().GetFunctionType(FI);
- std::string ReqName = getName({"omp_offloading", "requires_reg"});
- RequiresRegFn = CGM.CreateGlobalInitOrCleanUpFunction(FTy, ReqName, FI);
- CGF.StartFunction(GlobalDecl(), C.VoidTy, RequiresRegFn, FI, {});
- // TODO: check for other requires clauses.
- // The requires directive takes effect only when a target region is
- // present in the compilation unit. Otherwise it is ignored and not
- // passed to the runtime. This avoids the runtime from throwing an error
- // for mismatching requires clauses across compilation units that don't
- // contain at least 1 target region.
- assert((HasEmittedTargetRegion || HasEmittedDeclareTargetRegion ||
- !OMPBuilder.OffloadInfoManager.empty()) &&
- "Target or declare target region expected.");
- CGF.EmitRuntimeCall(OMPBuilder.getOrCreateRuntimeFunction(
- CGM.getModule(), OMPRTL___tgt_register_requires),
- llvm::ConstantInt::get(
- CGM.Int64Ty, OMPBuilder.Config.getRequiresFlags()));
- CGF.FinishFunction();
- }
- return RequiresRegFn;
-}
-
void CGOpenMPRuntime::emitTeamsCall(CodeGenFunction &CGF,
const OMPExecutableDirective &D,
SourceLocation Loc,
diff --git a/clang/lib/CodeGen/CGOpenMPRuntime.h b/clang/lib/CodeGen/CGOpenMPRuntime.h
index b01b39abd1606..c3206427b143e 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntime.h
+++ b/clang/lib/CodeGen/CGOpenMPRuntime.h
@@ -1407,10 +1407,6 @@ class CGOpenMPRuntime {
/// \param GD Global to scan.
virtual bool emitTargetGlobal(GlobalDecl GD);
- /// Creates and returns a registration function for when at least one
- /// requires directives was used in the current module.
- llvm::Function *emitRequiresDirectiveRegFun();
-
/// Creates all the offload entries in the current compilation unit
/// along with the associated metadata.
void createOffloadEntriesAndInfoMetadata();
diff --git a/clang/lib/CodeGen/CodeGenModule.cpp b/clang/lib/CodeGen/CodeGenModule.cpp
index c63e4ecc3dcba..d6d75efbbb2a8 100644
--- a/clang/lib/CodeGen/CodeGenModule.cpp
+++ b/clang/lib/CodeGen/CodeGenModule.cpp
@@ -836,10 +836,6 @@ void CodeGenModule::Release() {
AddGlobalCtor(CudaCtorFunction);
}
if (OpenMPRuntime) {
- if (llvm::Function *OpenMPRequiresDirectiveRegFun =
- OpenMPRuntime->emitRequiresDirectiveRegFun()) {
- AddGlobalCtor(OpenMPRequiresDirectiveRegFun, 0);
- }
OpenMPRuntime->createOffloadEntriesAndInfoMetadata();
OpenMPRuntime->clear();
}
diff --git a/clang/test/OpenMP/bug60602.cpp b/clang/test/OpenMP/bug60602.cpp
index 2fbfdfde07a0c..3ecc70cab778a 100644
--- a/clang/test/OpenMP/bug60602.cpp
+++ b/clang/test/OpenMP/bug60602.cpp
@@ -569,10 +569,3 @@ int kernel_within_loop(int *a, int *b, int N, int num_iters) {
// CHECK: omp.precond.end:
// CHECK-NEXT: ret void
//
-//
-// CHECK-LABEL: define internal void @.omp_offloading.requires_reg
-// CHECK-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK-NEXT: entry:
-// CHECK-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK-NEXT: ret void
-//
diff --git a/clang/test/OpenMP/distribute_codegen.cpp b/clang/test/OpenMP/distribute_codegen.cpp
index e3b43002a0518..31ec6ff911905 100644
--- a/clang/test/OpenMP/distribute_codegen.cpp
+++ b/clang/test/OpenMP/distribute_codegen.cpp
@@ -1037,13 +1037,6 @@ int fint(void) { return ftemplate<int>(); }
// CHECK1-NEXT: ret void
//
//
-// CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK1-SAME: () #[[ATTR3:[0-9]+]] {
-// CHECK1-NEXT: entry:
-// CHECK1-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK1-NEXT: ret void
-//
-//
// CHECK3-LABEL: define {{[^@]+}}@_Z23without_schedule_clausePfS_S_S_
// CHECK3-SAME: (ptr noundef [[A:%.*]], ptr noundef [[B:%.*]], ptr noundef [[C:%.*]], ptr noundef [[D:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK3-NEXT: entry:
@@ -1953,13 +1946,6 @@ int fint(void) { return ftemplate<int>(); }
// CHECK3-NEXT: ret void
//
//
-// CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK3-SAME: () #[[ATTR3:[0-9]+]] {
-// CHECK3-NEXT: entry:
-// CHECK3-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK3-NEXT: ret void
-//
-//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z23without_schedule_clausePfS_S_S__l56
// CHECK17-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], ptr noundef [[A:%.*]], ptr noundef [[B:%.*]], ptr noundef [[C:%.*]], ptr noundef [[D:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK17-NEXT: entry:
diff --git a/clang/test/OpenMP/distribute_firstprivate_codegen.cpp b/clang/test/OpenMP/distribute_firstprivate_codegen.cpp
index 361e26bc2984c..800a002e43968 100644
--- a/clang/test/OpenMP/distribute_firstprivate_codegen.cpp
+++ b/clang/test/OpenMP/distribute_firstprivate_codegen.cpp
@@ -304,13 +304,6 @@ int main() {
// CHECK1-NEXT: ret void
//
//
-// CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK1-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK1-NEXT: entry:
-// CHECK1-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK1-NEXT: ret void
-//
-//
// CHECK3-LABEL: define {{[^@]+}}@main
// CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK3-NEXT: entry:
@@ -476,13 +469,6 @@ int main() {
// CHECK3-NEXT: ret void
//
//
-// CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK3-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK3-NEXT: entry:
-// CHECK3-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK3-NEXT: ret void
-//
-//
// CHECK9-LABEL: define {{[^@]+}}@main
// CHECK9-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK9-NEXT: entry:
@@ -794,7 +780,7 @@ int main() {
//
//
// CHECK9-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
-// CHECK9-SAME: () #[[ATTR5:[0-9]+]] comdat {
+// CHECK9-SAME: () #[[ATTR1]] comdat {
// CHECK9-NEXT: entry:
// CHECK9-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
// CHECK9-NEXT: [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
@@ -1148,13 +1134,6 @@ int main() {
// CHECK9-NEXT: ret void
//
//
-// CHECK9-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK9-SAME: () #[[ATTR6:[0-9]+]] {
-// CHECK9-NEXT: entry:
-// CHECK9-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK9-NEXT: ret void
-//
-//
// CHECK11-LABEL: define {{[^@]+}}@main
// CHECK11-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK11-NEXT: entry:
@@ -1464,7 +1443,7 @@ int main() {
//
//
// CHECK11-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
-// CHECK11-SAME: () #[[ATTR5:[0-9]+]] comdat {
+// CHECK11-SAME: () #[[ATTR1]] comdat {
// CHECK11-NEXT: entry:
// CHECK11-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
// CHECK11-NEXT: [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
@@ -1815,10 +1794,3 @@ int main() {
// CHECK11-NEXT: [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT: ret void
//
-//
-// CHECK11-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK11-SAME: () #[[ATTR6:[0-9]+]] {
-// CHECK11-NEXT: entry:
-// CHECK11-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK11-NEXT: ret void
-//
diff --git a/clang/test/OpenMP/distribute_lastprivate_codegen.cpp b/clang/test/OpenMP/distribute_lastprivate_codegen.cpp
index e005de30e14d1..772372076e947 100644
--- a/clang/test/OpenMP/distribute_lastprivate_codegen.cpp
+++ b/clang/test/OpenMP/distribute_lastprivate_codegen.cpp
@@ -291,13 +291,6 @@ int main() {
// CHECK1-NEXT: ret void
//
//
-// CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK1-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK1-NEXT: entry:
-// CHECK1-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK1-NEXT: ret void
-//
-//
// CHECK3-LABEL: define {{[^@]+}}@main
// CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK3-NEXT: entry:
@@ -460,13 +453,6 @@ int main() {
// CHECK3-NEXT: ret void
//
//
-// CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK3-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK3-NEXT: entry:
-// CHECK3-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK3-NEXT: ret void
-//
-//
// CHECK9-LABEL: define {{[^@]+}}@main
// CHECK9-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK9-NEXT: entry:
@@ -797,7 +783,7 @@ int main() {
//
//
// CHECK9-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
-// CHECK9-SAME: () #[[ATTR5:[0-9]+]] comdat {
+// CHECK9-SAME: () #[[ATTR1]] comdat {
// CHECK9-NEXT: entry:
// CHECK9-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
// CHECK9-NEXT: [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
@@ -1169,13 +1155,6 @@ int main() {
// CHECK9-NEXT: ret void
//
//
-// CHECK9-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK9-SAME: () #[[ATTR6:[0-9]+]] {
-// CHECK9-NEXT: entry:
-// CHECK9-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK9-NEXT: ret void
-//
-//
// CHECK11-LABEL: define {{[^@]+}}@main
// CHECK11-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK11-NEXT: entry:
@@ -1504,7 +1483,7 @@ int main() {
//
//
// CHECK11-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
-// CHECK11-SAME: () #[[ATTR5:[0-9]+]] comdat {
+// CHECK11-SAME: () #[[ATTR1]] comdat {
// CHECK11-NEXT: entry:
// CHECK11-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
// CHECK11-NEXT: [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
@@ -1873,10 +1852,3 @@ int main() {
// CHECK11-NEXT: [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT: ret void
//
-//
-// CHECK11-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK11-SAME: () #[[ATTR6:[0-9]+]] {
-// CHECK11-NEXT: entry:
-// CHECK11-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK11-NEXT: ret void
-//
diff --git a/clang/test/OpenMP/distribute_parallel_for_codegen.cpp b/clang/test/OpenMP/distribute_parallel_for_codegen.cpp
index 7bdc4c5ab21a7..95adefa8020f6 100644
--- a/clang/test/OpenMP/distribute_parallel_for_codegen.cpp
+++ b/clang/test/OpenMP/distribute_parallel_for_codegen.cpp
@@ -2538,13 +2538,6 @@ int main() {
// CHECK1-NEXT: ret void
//
//
-// CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK1-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK1-NEXT: entry:
-// CHECK1-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK1-NEXT: ret void
-//
-//
// CHECK3-LABEL: define {{[^@]+}}@main
// CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK3-NEXT: entry:
@@ -4265,13 +4258,6 @@ int main() {
// CHECK3-NEXT: ret void
//
//
-// CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK3-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK3-NEXT: entry:
-// CHECK3-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK3-NEXT: ret void
-//
-//
// CHECK9-LABEL: define {{[^@]+}}@main
// CHECK9-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK9-NEXT: entry:
@@ -8886,13 +8872,6 @@ int main() {
// CHECK9-NEXT: ret void
//
//
-// CHECK9-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK9-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK9-NEXT: entry:
-// CHECK9-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK9-NEXT: ret void
-//
-//
// CHECK11-LABEL: define {{[^@]+}}@main
// CHECK11-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK11-NEXT: entry:
@@ -13404,10 +13383,3 @@ int main() {
// CHECK11: omp.precond.end:
// CHECK11-NEXT: ret void
//
-//
-// CHECK11-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK11-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK11-NEXT: entry:
-// CHECK11-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK11-NEXT: ret void
-//
diff --git a/clang/test/OpenMP/distribute_parallel_for_firstprivate_codegen.cpp b/clang/test/OpenMP/distribute_parallel_for_firstprivate_codegen.cpp
index 9f900facc6a54..46c115e40e435 100644
--- a/clang/test/OpenMP/distribute_parallel_for_firstprivate_codegen.cpp
+++ b/clang/test/OpenMP/distribute_parallel_for_firstprivate_codegen.cpp
@@ -504,13 +504,6 @@ int main() {
// CHECK1-NEXT: ret void
//
//
-// CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK1-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK1-NEXT: entry:
-// CHECK1-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK1-NEXT: ret void
-//
-//
// CHECK3-LABEL: define {{[^@]+}}@main
// CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK3-NEXT: entry:
@@ -759,13 +752,6 @@ int main() {
// CHECK3-NEXT: ret void
//
//
-// CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK3-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK3-NEXT: entry:
-// CHECK3-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK3-NEXT: ret void
-//
-//
// CHECK8-LABEL: define {{[^@]+}}@main
// CHECK8-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK8-NEXT: entry:
@@ -1207,7 +1193,7 @@ int main() {
//
//
// CHECK8-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
-// CHECK8-SAME: () #[[ATTR5:[0-9]+]] comdat {
+// CHECK8-SAME: () #[[ATTR1]] comdat {
// CHECK8-NEXT: entry:
// CHECK8-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
// CHECK8-NEXT: [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
@@ -1685,13 +1671,6 @@ int main() {
// CHECK8-NEXT: ret void
//
//
-// CHECK8-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK8-SAME: () #[[ATTR6:[0-9]+]] {
-// CHECK8-NEXT: entry:
-// CHECK8-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK8-NEXT: ret void
-//
-//
// CHECK10-LABEL: define {{[^@]+}}@main
// CHECK10-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK10-NEXT: entry:
@@ -2127,7 +2106,7 @@ int main() {
//
//
// CHECK10-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
-// CHECK10-SAME: () #[[ATTR5:[0-9]+]] comdat {
+// CHECK10-SAME: () #[[ATTR1]] comdat {
// CHECK10-NEXT: entry:
// CHECK10-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
// CHECK10-NEXT: [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
@@ -2598,10 +2577,3 @@ int main() {
// CHECK10-NEXT: [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK10-NEXT: ret void
//
-//
-// CHECK10-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK10-SAME: () #[[ATTR6:[0-9]+]] {
-// CHECK10-NEXT: entry:
-// CHECK10-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK10-NEXT: ret void
-//
diff --git a/clang/test/OpenMP/distribute_parallel_for_if_codegen.cpp b/clang/test/OpenMP/distribute_parallel_for_if_codegen.cpp
index 83c9f504ccaca..846e7beb5d92f 100644
--- a/clang/test/OpenMP/distribute_parallel_for_if_codegen.cpp
+++ b/clang/test/OpenMP/distribute_parallel_for_if_codegen.cpp
@@ -1609,10 +1609,3 @@ int main() {
// CHECK1-NEXT: call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP3]])
// CHECK1-NEXT: ret void
//
-//
-// CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK1-SAME: () #[[ATTR5:[0-9]+]] {
-// CHECK1-NEXT: entry:
-// CHECK1-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK1-NEXT: ret void
-//
diff --git a/clang/test/OpenMP/distribute_parallel_for_lastprivate_codegen.cpp b/clang/test/OpenMP/distribute_parallel_for_lastprivate_codegen.cpp
index 8c44a1e71ae79..aa981f606cc87 100644
--- a/clang/test/OpenMP/distribute_parallel_for_lastprivate_codegen.cpp
+++ b/clang/test/OpenMP/distribute_parallel_for_lastprivate_codegen.cpp
@@ -462,13 +462,6 @@ int main() {
// CHECK1-NEXT: ret void
//
//
-// CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK1-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK1-NEXT: entry:
-// CHECK1-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK1-NEXT: ret void
-//
-//
// CHECK3-LABEL: define {{[^@]+}}@main
// CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK3-NEXT: entry:
@@ -734,13 +727,6 @@ int main() {
// CHECK3-NEXT: ret void
//
//
-// CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK3-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK3-NEXT: entry:
-// CHECK3-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK3-NEXT: ret void
-//
-//
// CHECK9-LABEL: define {{[^@]+}}@main
// CHECK9-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK9-NEXT: entry:
@@ -1219,7 +1205,7 @@ int main() {
//
//
// CHECK9-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
-// CHECK9-SAME: () #[[ATTR5:[0-9]+]] comdat {
+// CHECK9-SAME: () #[[ATTR1]] comdat {
// CHECK9-NEXT: entry:
// CHECK9-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
// CHECK9-NEXT: [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
@@ -1733,13 +1719,6 @@ int main() {
// CHECK9-NEXT: ret void
//
//
-// CHECK9-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK9-SAME: () #[[ATTR6:[0-9]+]] {
-// CHECK9-NEXT: entry:
-// CHECK9-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK9-NEXT: ret void
-//
-//
// CHECK11-LABEL: define {{[^@]+}}@main
// CHECK11-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK11-NEXT: entry:
@@ -2212,7 +2191,7 @@ int main() {
//
//
// CHECK11-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
-// CHECK11-SAME: () #[[ATTR5:[0-9]+]] comdat {
+// CHECK11-SAME: () #[[ATTR1]] comdat {
// CHECK11-NEXT: entry:
// CHECK11-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
// CHECK11-NEXT: [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
@@ -2719,10 +2698,3 @@ int main() {
// CHECK11-NEXT: [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT: ret void
//
-//
-// CHECK11-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK11-SAME: () #[[ATTR6:[0-9]+]] {
-// CHECK11-NEXT: entry:
-// CHECK11-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK11-NEXT: ret void
-//
diff --git a/clang/test/OpenMP/distribute_parallel_for_num_threads_codegen.cpp b/clang/test/OpenMP/distribute_parallel_for_num_threads_codegen.cpp
index 9f769ca2886fe..5d9244268d554 100644
--- a/clang/test/OpenMP/distribute_parallel_for_num_threads_codegen.cpp
+++ b/clang/test/OpenMP/distribute_parallel_for_num_threads_codegen.cpp
@@ -112,7 +112,7 @@ int main() {
// CHECK1-NEXT: store i32 0, ptr [[RETVAL]], align 4
// CHECK1-NEXT: call void @_ZN1SC1El(ptr noundef nonnull align 8 dereferenceable(24) [[S]], i64 noundef 0)
// CHECK1-NEXT: [[CALL:%.*]] = invoke noundef signext i8 @_ZN1ScvcEv(ptr noundef nonnull align 8 dereferenceable(24) [[S]])
-// CHECK1-NEXT: to label [[INVOKE_CONT:%.*]] unwind label [[LPAD:%.*]]
+// CHECK1-NEXT: to label [[INVOKE_CONT:%.*]] unwind label [[LPAD:%.*]]
// CHECK1: invoke.cont:
// CHECK1-NEXT: store i8 [[CALL]], ptr [[A]], align 1
// CHECK1-NEXT: [[TMP0:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
@@ -145,16 +145,16 @@ int main() {
// CHECK1-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK1-NEXT: br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK1: omp_offload.failed:
-// CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68() #[[ATTR4:[0-9]+]]
+// CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68() #[[ATTR3:[0-9]+]]
// CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]]
// CHECK1: lpad:
// CHECK1-NEXT: [[TMP15:%.*]] = landingpad { ptr, i32 }
-// CHECK1-NEXT: cleanup
+// CHECK1-NEXT: cleanup
// CHECK1-NEXT: [[TMP16:%.*]] = extractvalue { ptr, i32 } [[TMP15]], 0
// CHECK1-NEXT: store ptr [[TMP16]], ptr [[EXN_SLOT]], align 8
// CHECK1-NEXT: [[TMP17:%.*]] = extractvalue { ptr, i32 } [[TMP15]], 1
// CHECK1-NEXT: store...
[truncated]
|
|
@llvm/pr-subscribers-clang Author: Joseph Huber (jhuber6) ChangesSummary: This patch changes the approach to no longer use global constructors, This function removes support for the old It is worth noting that the current Patch is 2.09 MiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/80460.diff 173 Files Affected:
diff --git a/clang/lib/CodeGen/CGOpenMPRuntime.cpp b/clang/lib/CodeGen/CGOpenMPRuntime.cpp
index 4855e7410a015..a7b72df6d9f89 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntime.cpp
+++ b/clang/lib/CodeGen/CGOpenMPRuntime.cpp
@@ -10100,44 +10100,6 @@ bool CGOpenMPRuntime::markAsGlobalTarget(GlobalDecl GD) {
return !AlreadyEmittedTargetDecls.insert(D).second;
}
-llvm::Function *CGOpenMPRuntime::emitRequiresDirectiveRegFun() {
- // If we don't have entries or if we are emitting code for the device, we
- // don't need to do anything.
- if (CGM.getLangOpts().OMPTargetTriples.empty() ||
- CGM.getLangOpts().OpenMPSimd || CGM.getLangOpts().OpenMPIsTargetDevice ||
- (OMPBuilder.OffloadInfoManager.empty() &&
- !HasEmittedDeclareTargetRegion && !HasEmittedTargetRegion))
- return nullptr;
-
- // Create and register the function that handles the requires directives.
- ASTContext &C = CGM.getContext();
-
- llvm::Function *RequiresRegFn;
- {
- CodeGenFunction CGF(CGM);
- const auto &FI = CGM.getTypes().arrangeNullaryFunction();
- llvm::FunctionType *FTy = CGM.getTypes().GetFunctionType(FI);
- std::string ReqName = getName({"omp_offloading", "requires_reg"});
- RequiresRegFn = CGM.CreateGlobalInitOrCleanUpFunction(FTy, ReqName, FI);
- CGF.StartFunction(GlobalDecl(), C.VoidTy, RequiresRegFn, FI, {});
- // TODO: check for other requires clauses.
- // The requires directive takes effect only when a target region is
- // present in the compilation unit. Otherwise it is ignored and not
- // passed to the runtime. This avoids the runtime from throwing an error
- // for mismatching requires clauses across compilation units that don't
- // contain at least 1 target region.
- assert((HasEmittedTargetRegion || HasEmittedDeclareTargetRegion ||
- !OMPBuilder.OffloadInfoManager.empty()) &&
- "Target or declare target region expected.");
- CGF.EmitRuntimeCall(OMPBuilder.getOrCreateRuntimeFunction(
- CGM.getModule(), OMPRTL___tgt_register_requires),
- llvm::ConstantInt::get(
- CGM.Int64Ty, OMPBuilder.Config.getRequiresFlags()));
- CGF.FinishFunction();
- }
- return RequiresRegFn;
-}
-
void CGOpenMPRuntime::emitTeamsCall(CodeGenFunction &CGF,
const OMPExecutableDirective &D,
SourceLocation Loc,
diff --git a/clang/lib/CodeGen/CGOpenMPRuntime.h b/clang/lib/CodeGen/CGOpenMPRuntime.h
index b01b39abd1606..c3206427b143e 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntime.h
+++ b/clang/lib/CodeGen/CGOpenMPRuntime.h
@@ -1407,10 +1407,6 @@ class CGOpenMPRuntime {
/// \param GD Global to scan.
virtual bool emitTargetGlobal(GlobalDecl GD);
- /// Creates and returns a registration function for when at least one
- /// requires directives was used in the current module.
- llvm::Function *emitRequiresDirectiveRegFun();
-
/// Creates all the offload entries in the current compilation unit
/// along with the associated metadata.
void createOffloadEntriesAndInfoMetadata();
diff --git a/clang/lib/CodeGen/CodeGenModule.cpp b/clang/lib/CodeGen/CodeGenModule.cpp
index c63e4ecc3dcba..d6d75efbbb2a8 100644
--- a/clang/lib/CodeGen/CodeGenModule.cpp
+++ b/clang/lib/CodeGen/CodeGenModule.cpp
@@ -836,10 +836,6 @@ void CodeGenModule::Release() {
AddGlobalCtor(CudaCtorFunction);
}
if (OpenMPRuntime) {
- if (llvm::Function *OpenMPRequiresDirectiveRegFun =
- OpenMPRuntime->emitRequiresDirectiveRegFun()) {
- AddGlobalCtor(OpenMPRequiresDirectiveRegFun, 0);
- }
OpenMPRuntime->createOffloadEntriesAndInfoMetadata();
OpenMPRuntime->clear();
}
diff --git a/clang/test/OpenMP/bug60602.cpp b/clang/test/OpenMP/bug60602.cpp
index 2fbfdfde07a0c..3ecc70cab778a 100644
--- a/clang/test/OpenMP/bug60602.cpp
+++ b/clang/test/OpenMP/bug60602.cpp
@@ -569,10 +569,3 @@ int kernel_within_loop(int *a, int *b, int N, int num_iters) {
// CHECK: omp.precond.end:
// CHECK-NEXT: ret void
//
-//
-// CHECK-LABEL: define internal void @.omp_offloading.requires_reg
-// CHECK-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK-NEXT: entry:
-// CHECK-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK-NEXT: ret void
-//
diff --git a/clang/test/OpenMP/distribute_codegen.cpp b/clang/test/OpenMP/distribute_codegen.cpp
index e3b43002a0518..31ec6ff911905 100644
--- a/clang/test/OpenMP/distribute_codegen.cpp
+++ b/clang/test/OpenMP/distribute_codegen.cpp
@@ -1037,13 +1037,6 @@ int fint(void) { return ftemplate<int>(); }
// CHECK1-NEXT: ret void
//
//
-// CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK1-SAME: () #[[ATTR3:[0-9]+]] {
-// CHECK1-NEXT: entry:
-// CHECK1-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK1-NEXT: ret void
-//
-//
// CHECK3-LABEL: define {{[^@]+}}@_Z23without_schedule_clausePfS_S_S_
// CHECK3-SAME: (ptr noundef [[A:%.*]], ptr noundef [[B:%.*]], ptr noundef [[C:%.*]], ptr noundef [[D:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK3-NEXT: entry:
@@ -1953,13 +1946,6 @@ int fint(void) { return ftemplate<int>(); }
// CHECK3-NEXT: ret void
//
//
-// CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK3-SAME: () #[[ATTR3:[0-9]+]] {
-// CHECK3-NEXT: entry:
-// CHECK3-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK3-NEXT: ret void
-//
-//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z23without_schedule_clausePfS_S_S__l56
// CHECK17-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], ptr noundef [[A:%.*]], ptr noundef [[B:%.*]], ptr noundef [[C:%.*]], ptr noundef [[D:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK17-NEXT: entry:
diff --git a/clang/test/OpenMP/distribute_firstprivate_codegen.cpp b/clang/test/OpenMP/distribute_firstprivate_codegen.cpp
index 361e26bc2984c..800a002e43968 100644
--- a/clang/test/OpenMP/distribute_firstprivate_codegen.cpp
+++ b/clang/test/OpenMP/distribute_firstprivate_codegen.cpp
@@ -304,13 +304,6 @@ int main() {
// CHECK1-NEXT: ret void
//
//
-// CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK1-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK1-NEXT: entry:
-// CHECK1-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK1-NEXT: ret void
-//
-//
// CHECK3-LABEL: define {{[^@]+}}@main
// CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK3-NEXT: entry:
@@ -476,13 +469,6 @@ int main() {
// CHECK3-NEXT: ret void
//
//
-// CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK3-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK3-NEXT: entry:
-// CHECK3-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK3-NEXT: ret void
-//
-//
// CHECK9-LABEL: define {{[^@]+}}@main
// CHECK9-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK9-NEXT: entry:
@@ -794,7 +780,7 @@ int main() {
//
//
// CHECK9-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
-// CHECK9-SAME: () #[[ATTR5:[0-9]+]] comdat {
+// CHECK9-SAME: () #[[ATTR1]] comdat {
// CHECK9-NEXT: entry:
// CHECK9-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
// CHECK9-NEXT: [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
@@ -1148,13 +1134,6 @@ int main() {
// CHECK9-NEXT: ret void
//
//
-// CHECK9-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK9-SAME: () #[[ATTR6:[0-9]+]] {
-// CHECK9-NEXT: entry:
-// CHECK9-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK9-NEXT: ret void
-//
-//
// CHECK11-LABEL: define {{[^@]+}}@main
// CHECK11-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK11-NEXT: entry:
@@ -1464,7 +1443,7 @@ int main() {
//
//
// CHECK11-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
-// CHECK11-SAME: () #[[ATTR5:[0-9]+]] comdat {
+// CHECK11-SAME: () #[[ATTR1]] comdat {
// CHECK11-NEXT: entry:
// CHECK11-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
// CHECK11-NEXT: [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
@@ -1815,10 +1794,3 @@ int main() {
// CHECK11-NEXT: [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT: ret void
//
-//
-// CHECK11-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK11-SAME: () #[[ATTR6:[0-9]+]] {
-// CHECK11-NEXT: entry:
-// CHECK11-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK11-NEXT: ret void
-//
diff --git a/clang/test/OpenMP/distribute_lastprivate_codegen.cpp b/clang/test/OpenMP/distribute_lastprivate_codegen.cpp
index e005de30e14d1..772372076e947 100644
--- a/clang/test/OpenMP/distribute_lastprivate_codegen.cpp
+++ b/clang/test/OpenMP/distribute_lastprivate_codegen.cpp
@@ -291,13 +291,6 @@ int main() {
// CHECK1-NEXT: ret void
//
//
-// CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK1-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK1-NEXT: entry:
-// CHECK1-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK1-NEXT: ret void
-//
-//
// CHECK3-LABEL: define {{[^@]+}}@main
// CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK3-NEXT: entry:
@@ -460,13 +453,6 @@ int main() {
// CHECK3-NEXT: ret void
//
//
-// CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK3-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK3-NEXT: entry:
-// CHECK3-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK3-NEXT: ret void
-//
-//
// CHECK9-LABEL: define {{[^@]+}}@main
// CHECK9-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK9-NEXT: entry:
@@ -797,7 +783,7 @@ int main() {
//
//
// CHECK9-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
-// CHECK9-SAME: () #[[ATTR5:[0-9]+]] comdat {
+// CHECK9-SAME: () #[[ATTR1]] comdat {
// CHECK9-NEXT: entry:
// CHECK9-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
// CHECK9-NEXT: [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
@@ -1169,13 +1155,6 @@ int main() {
// CHECK9-NEXT: ret void
//
//
-// CHECK9-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK9-SAME: () #[[ATTR6:[0-9]+]] {
-// CHECK9-NEXT: entry:
-// CHECK9-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK9-NEXT: ret void
-//
-//
// CHECK11-LABEL: define {{[^@]+}}@main
// CHECK11-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK11-NEXT: entry:
@@ -1504,7 +1483,7 @@ int main() {
//
//
// CHECK11-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
-// CHECK11-SAME: () #[[ATTR5:[0-9]+]] comdat {
+// CHECK11-SAME: () #[[ATTR1]] comdat {
// CHECK11-NEXT: entry:
// CHECK11-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
// CHECK11-NEXT: [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
@@ -1873,10 +1852,3 @@ int main() {
// CHECK11-NEXT: [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT: ret void
//
-//
-// CHECK11-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK11-SAME: () #[[ATTR6:[0-9]+]] {
-// CHECK11-NEXT: entry:
-// CHECK11-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK11-NEXT: ret void
-//
diff --git a/clang/test/OpenMP/distribute_parallel_for_codegen.cpp b/clang/test/OpenMP/distribute_parallel_for_codegen.cpp
index 7bdc4c5ab21a7..95adefa8020f6 100644
--- a/clang/test/OpenMP/distribute_parallel_for_codegen.cpp
+++ b/clang/test/OpenMP/distribute_parallel_for_codegen.cpp
@@ -2538,13 +2538,6 @@ int main() {
// CHECK1-NEXT: ret void
//
//
-// CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK1-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK1-NEXT: entry:
-// CHECK1-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK1-NEXT: ret void
-//
-//
// CHECK3-LABEL: define {{[^@]+}}@main
// CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK3-NEXT: entry:
@@ -4265,13 +4258,6 @@ int main() {
// CHECK3-NEXT: ret void
//
//
-// CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK3-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK3-NEXT: entry:
-// CHECK3-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK3-NEXT: ret void
-//
-//
// CHECK9-LABEL: define {{[^@]+}}@main
// CHECK9-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK9-NEXT: entry:
@@ -8886,13 +8872,6 @@ int main() {
// CHECK9-NEXT: ret void
//
//
-// CHECK9-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK9-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK9-NEXT: entry:
-// CHECK9-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK9-NEXT: ret void
-//
-//
// CHECK11-LABEL: define {{[^@]+}}@main
// CHECK11-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK11-NEXT: entry:
@@ -13404,10 +13383,3 @@ int main() {
// CHECK11: omp.precond.end:
// CHECK11-NEXT: ret void
//
-//
-// CHECK11-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK11-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK11-NEXT: entry:
-// CHECK11-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK11-NEXT: ret void
-//
diff --git a/clang/test/OpenMP/distribute_parallel_for_firstprivate_codegen.cpp b/clang/test/OpenMP/distribute_parallel_for_firstprivate_codegen.cpp
index 9f900facc6a54..46c115e40e435 100644
--- a/clang/test/OpenMP/distribute_parallel_for_firstprivate_codegen.cpp
+++ b/clang/test/OpenMP/distribute_parallel_for_firstprivate_codegen.cpp
@@ -504,13 +504,6 @@ int main() {
// CHECK1-NEXT: ret void
//
//
-// CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK1-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK1-NEXT: entry:
-// CHECK1-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK1-NEXT: ret void
-//
-//
// CHECK3-LABEL: define {{[^@]+}}@main
// CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK3-NEXT: entry:
@@ -759,13 +752,6 @@ int main() {
// CHECK3-NEXT: ret void
//
//
-// CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK3-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK3-NEXT: entry:
-// CHECK3-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK3-NEXT: ret void
-//
-//
// CHECK8-LABEL: define {{[^@]+}}@main
// CHECK8-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK8-NEXT: entry:
@@ -1207,7 +1193,7 @@ int main() {
//
//
// CHECK8-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
-// CHECK8-SAME: () #[[ATTR5:[0-9]+]] comdat {
+// CHECK8-SAME: () #[[ATTR1]] comdat {
// CHECK8-NEXT: entry:
// CHECK8-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
// CHECK8-NEXT: [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
@@ -1685,13 +1671,6 @@ int main() {
// CHECK8-NEXT: ret void
//
//
-// CHECK8-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK8-SAME: () #[[ATTR6:[0-9]+]] {
-// CHECK8-NEXT: entry:
-// CHECK8-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK8-NEXT: ret void
-//
-//
// CHECK10-LABEL: define {{[^@]+}}@main
// CHECK10-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK10-NEXT: entry:
@@ -2127,7 +2106,7 @@ int main() {
//
//
// CHECK10-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
-// CHECK10-SAME: () #[[ATTR5:[0-9]+]] comdat {
+// CHECK10-SAME: () #[[ATTR1]] comdat {
// CHECK10-NEXT: entry:
// CHECK10-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
// CHECK10-NEXT: [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
@@ -2598,10 +2577,3 @@ int main() {
// CHECK10-NEXT: [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK10-NEXT: ret void
//
-//
-// CHECK10-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK10-SAME: () #[[ATTR6:[0-9]+]] {
-// CHECK10-NEXT: entry:
-// CHECK10-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK10-NEXT: ret void
-//
diff --git a/clang/test/OpenMP/distribute_parallel_for_if_codegen.cpp b/clang/test/OpenMP/distribute_parallel_for_if_codegen.cpp
index 83c9f504ccaca..846e7beb5d92f 100644
--- a/clang/test/OpenMP/distribute_parallel_for_if_codegen.cpp
+++ b/clang/test/OpenMP/distribute_parallel_for_if_codegen.cpp
@@ -1609,10 +1609,3 @@ int main() {
// CHECK1-NEXT: call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP3]])
// CHECK1-NEXT: ret void
//
-//
-// CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK1-SAME: () #[[ATTR5:[0-9]+]] {
-// CHECK1-NEXT: entry:
-// CHECK1-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK1-NEXT: ret void
-//
diff --git a/clang/test/OpenMP/distribute_parallel_for_lastprivate_codegen.cpp b/clang/test/OpenMP/distribute_parallel_for_lastprivate_codegen.cpp
index 8c44a1e71ae79..aa981f606cc87 100644
--- a/clang/test/OpenMP/distribute_parallel_for_lastprivate_codegen.cpp
+++ b/clang/test/OpenMP/distribute_parallel_for_lastprivate_codegen.cpp
@@ -462,13 +462,6 @@ int main() {
// CHECK1-NEXT: ret void
//
//
-// CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK1-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK1-NEXT: entry:
-// CHECK1-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK1-NEXT: ret void
-//
-//
// CHECK3-LABEL: define {{[^@]+}}@main
// CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK3-NEXT: entry:
@@ -734,13 +727,6 @@ int main() {
// CHECK3-NEXT: ret void
//
//
-// CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK3-SAME: () #[[ATTR4:[0-9]+]] {
-// CHECK3-NEXT: entry:
-// CHECK3-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK3-NEXT: ret void
-//
-//
// CHECK9-LABEL: define {{[^@]+}}@main
// CHECK9-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK9-NEXT: entry:
@@ -1219,7 +1205,7 @@ int main() {
//
//
// CHECK9-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
-// CHECK9-SAME: () #[[ATTR5:[0-9]+]] comdat {
+// CHECK9-SAME: () #[[ATTR1]] comdat {
// CHECK9-NEXT: entry:
// CHECK9-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
// CHECK9-NEXT: [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
@@ -1733,13 +1719,6 @@ int main() {
// CHECK9-NEXT: ret void
//
//
-// CHECK9-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK9-SAME: () #[[ATTR6:[0-9]+]] {
-// CHECK9-NEXT: entry:
-// CHECK9-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK9-NEXT: ret void
-//
-//
// CHECK11-LABEL: define {{[^@]+}}@main
// CHECK11-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK11-NEXT: entry:
@@ -2212,7 +2191,7 @@ int main() {
//
//
// CHECK11-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
-// CHECK11-SAME: () #[[ATTR5:[0-9]+]] comdat {
+// CHECK11-SAME: () #[[ATTR1]] comdat {
// CHECK11-NEXT: entry:
// CHECK11-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
// CHECK11-NEXT: [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
@@ -2719,10 +2698,3 @@ int main() {
// CHECK11-NEXT: [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT: ret void
//
-//
-// CHECK11-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
-// CHECK11-SAME: () #[[ATTR6:[0-9]+]] {
-// CHECK11-NEXT: entry:
-// CHECK11-NEXT: call void @__tgt_register_requires(i64 1)
-// CHECK11-NEXT: ret void
-//
diff --git a/clang/test/OpenMP/distribute_parallel_for_num_threads_codegen.cpp b/clang/test/OpenMP/distribute_parallel_for_num_threads_codegen.cpp
index 9f769ca2886fe..5d9244268d554 100644
--- a/clang/test/OpenMP/distribute_parallel_for_num_threads_codegen.cpp
+++ b/clang/test/OpenMP/distribute_parallel_for_num_threads_codegen.cpp
@@ -112,7 +112,7 @@ int main() {
// CHECK1-NEXT: store i32 0, ptr [[RETVAL]], align 4
// CHECK1-NEXT: call void @_ZN1SC1El(ptr noundef nonnull align 8 dereferenceable(24) [[S]], i64 noundef 0)
// CHECK1-NEXT: [[CALL:%.*]] = invoke noundef signext i8 @_ZN1ScvcEv(ptr noundef nonnull align 8 dereferenceable(24) [[S]])
-// CHECK1-NEXT: to label [[INVOKE_CONT:%.*]] unwind label [[LPAD:%.*]]
+// CHECK1-NEXT: to label [[INVOKE_CONT:%.*]] unwind label [[LPAD:%.*]]
// CHECK1: invoke.cont:
// CHECK1-NEXT: store i8 [[CALL]], ptr [[A]], align 1
// CHECK1-NEXT: [[TMP0:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
@@ -145,16 +145,16 @@ int main() {
// CHECK1-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK1-NEXT: br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK1: omp_offload.failed:
-// CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68() #[[ATTR4:[0-9]+]]
+// CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68() #[[ATTR3:[0-9]+]]
// CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]]
// CHECK1: lpad:
// CHECK1-NEXT: [[TMP15:%.*]] = landingpad { ptr, i32 }
-// CHECK1-NEXT: cleanup
+// CHECK1-NEXT: cleanup
// CHECK1-NEXT: [[TMP16:%.*]] = extractvalue { ptr, i32 } [[TMP15]], 0
// CHECK1-NEXT: store ptr [[TMP16]], ptr [[EXN_SLOT]], align 8
// CHECK1-NEXT: [[TMP17:%.*]] = extractvalue { ptr, i32 } [[TMP15]], 1
// CHECK1-NEXT: store...
[truncated]
|
|
✅ With the latest revision this PR passed the C/C++ code formatter. |
jhuber6
force-pushed
the
RegisterRequires
branch
from
6b102a9 to
ae8a117
Compare
doru1004
reviewed
Feb 2, 2024
| @@ -199,7 +199,7 @@ static int initLibrary(DeviceTy &Device) { | |||
| Entry.size) != OFFLOAD_SUCCESS) | |||
| REPORT("Failed to write symbol for USM %s\n", Entry.name); | |||
| } | |||
| } else { | |||
| } else if (Entry.addr) { | |||
There was a problem hiding this comment.
So now we don't have a "default" else branch here. Is it not needed?
There was a problem hiding this comment.
Also, could you explain in a comment why Entry.addr is used here?
There was a problem hiding this comment.
This is just hacking around the offloading entires. I wanted to use the same struct and handling for convenience, but right now there's no basic flag for what the kind of the thing is. Previously we just had functions and globals, if the size is zero it's a global. I added a third thing which can also be checked if the address is nullptr currently. I want to overhaul this in the future but I'm hoping to get this stuff working first.
There was a problem hiding this comment.
If the address is not-null, then it's a kernel waiting to be registered. It's a little hacky but in the current construction a kernel will always have a stub address pass in by the compiler. So one of these "requires" entries will have the addr be nullptr and the size be zero.
There was a problem hiding this comment.
Should we check for size > 0 too?
There was a problem hiding this comment.
Or is that too restrictive
There was a problem hiding this comment.
Isn't that already done in the if above?
There was a problem hiding this comment.
So it will only enter this branch if size is 0 and then if the address is not nullptr.
There was a problem hiding this comment.
Yes, I'll make the offloading entries less dumb in the future. I don't like how for working w/ CUDA and stuff we need cuda_offloading_entries and whatnot as well. I've been planning on merging them and simply having a field for the language kind and stuff. Been planning on doing that for awhile.
jhuber6
added a commit
to jhuber6/llvm-project
that referenced
this pull request
Feb 2, 2024
Summary: Currently we rely on global constructors to initialize and shut down the OpenMP runtime library and plugin manager. This causes some issues because we do not have a defined lifetime that we can rely on to release and allocate resources. This patch instead adds some simple reference counted initialization and deinitialization function. A future patch will use the `deinit` interface to more intelligently handle plugin deinitilization. Right now we do nothing and rely on `atexit` inside of the plugins to tear them down. This isn't great because it limits our ability to control these things. Note that I made the `__tgt_register_lib` functions do the initialization instead of adding calls to the new runtime functions in the linker wrapper. The reason for this is because in the past it's been easier to not introduce a new function call, since sometimes the user's compiler will link against an older `libomptarget`. Maybe if we change the name with offloading in the future we can simplify this. Depends on llvm#80460
skatrak
reviewed
Feb 5, 2024
| @@ -6872,35 +6883,6 @@ void OpenMPIRBuilder::loadOffloadInfoMetadata(StringRef HostFilePath) { | |||
| loadOffloadInfoMetadata(*M.get()); | |||
| } | |||
|
|
|||
| Function *OpenMPIRBuilder::createRegisterRequires(StringRef Name) { | |||
There was a problem hiding this comment.
Removing this function breaks the compilation of the OpenMP MLIR dialect, as it's used there (mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp: convertRequiresAttr()).
My understanding is that creating this function would no longer be necessary, so the solution to that should be to remove convertRequiresAttr() and replace the call to it in OpenMPDialectLLVMIRTranslationInterface::amendOperation() to return success() in the same file. Flang should already pick up your other changes, so REQUIRES information should still work in Fortran after this.
There was a problem hiding this comment.
Thanks for the heads up. Do you know if there will be any other changes required to make the requires information use the "new" format?
There was a problem hiding this comment.
Flang already picks up your changes to createOffloadEntriesAndInfoMetadata() and it appears to be putting the corresponding flags into a structure, though I'm not sure it's the right one.
I ran flang-new -fc1 -emit-llvm -fopenmp flang/test/Lower/OpenMP/requires.f90 -o - for a simple test with a single target region and I'm seeing two offloading entries being produced, which I'm not sure whether it's expected behavior. It looks like only one of them has the requires flags and the other one is the one that's properly linked with the kernel.
@.omp_offloading.entry_name = internal unnamed_addr constant [43 x i8] c"__omp_offloading_10307_d2a215c__QQmain_l12\00"
@.omp_offloading.entry.__omp_offloading_10307_d2a215c__QQmain_l12 = weak constant %struct.__tgt_offload_entry { ptr @.__omp_offloading_10307_d2a215c__QQmain_l12.region_id, ptr @.omp_offloading.entry_name, i64 0, i32 0, i32 0 }, section "omp_offloading_entries", align 1
@.omp_offloading.entry_name.1 = internal unnamed_addr constant [1 x i8] zeroinitializer
@.omp_offloading.entry. = weak constant %struct.__tgt_offload_entry { ptr null, ptr @.omp_offloading.entry_name.1, i64 0, i32 22, i32 10 }, section "omp_offloading_entries", align 1
...
%14 = call i32 @__tgt_target_kernel(ptr @1, i64 -1, i32 -1, i32 0, ptr @.__omp_offloading_10307_d2a215c__QQmain_l12.region_id, ptr %kernel_args)
There was a problem hiding this comment.
That looks a little weird, the i32 value is 22 which I don't know what that corresponds to with the flags. The value is then 10 which would be two requires flags merged together maybe? 22 is 16, 4, and 2, so I don't know why all those flags would be set at the same time.
There was a problem hiding this comment.
The test itself has "requires unified_shared_memory reverse_offload" (it originally just checks that the flags are passed to MLIR), so that's where the "10" is coming from. The "22" I don't know what it's supposed represent.
There was a problem hiding this comment.
I encoded the fact that this is a "requires" entry into the global with 16. Realistically it shouldn't make a difference in the runtime since as long as 16 is set and the address is null it will get counted as a requires entry. It's just a little weird that it wouldn't just be 16. I'll double check. Changing these entries is also on the list.
There was a problem hiding this comment.
It was a very obvious problem. I mixed up hexadecimal and did 0x16 when I meant either 0x10 or 16. Fixed now.
There was a problem hiding this comment.
Oh, then the "16" bit just happened to be part of the number that was being set. Glad to hear this wasn't some unexpected interaction specific to Flang.
jhuber6
force-pushed
the
RegisterRequires
branch
from
ae8a117 to
9535513
Compare
jhuber6
force-pushed
the
RegisterRequires
branch
from
9535513 to
28e7af9
Compare
jhuber6
added a commit
to jhuber6/llvm-project
that referenced
this pull request
Feb 6, 2024
Summary: Currently we rely on global constructors to initialize and shut down the OpenMP runtime library and plugin manager. This causes some issues because we do not have a defined lifetime that we can rely on to release and allocate resources. This patch instead adds some simple reference counted initialization and deinitialization function. A future patch will use the `deinit` interface to more intelligently handle plugin deinitilization. Right now we do nothing and rely on `atexit` inside of the plugins to tear them down. This isn't great because it limits our ability to control these things. Note that I made the `__tgt_register_lib` functions do the initialization instead of adding calls to the new runtime functions in the linker wrapper. The reason for this is because in the past it's been easier to not introduce a new function call, since sometimes the user's compiler will link against an older `libomptarget`. Maybe if we change the name with offloading in the future we can simplify this. Depends on llvm#80460
|
ping |
JonChesterfield
approved these changes
Feb 21, 2024
jhuber6
force-pushed
the
RegisterRequires
branch
3 times, most recently
from
2fa7b81 to
1e5d2dc
Compare
Summary: Currently, OpenMP handles the `omp requires` clause by emitting a global constructor into the runtime for every translation unit that requires it. However, this is not a great solution because it prevents us from having a defined order in which the runtime is accessed and used. This patch changes the approach to no longer use global constructors, but to instead group the flag with the other offloading entires that we already handle. This has the effect of still registering each flag per requires TU, but now we have a single constructor that handles everything. This function removes support for the old `__tgt_register_requires` and replaces it with a warning message. We just had a recent release, and the OpenMP policy for the past four releases since we switched to LLVM is that we do not provide strict backwards compatibility between major LLVM releases now that the library is versioned. This means that a user will need to recompile if they have an old binary that relied on `register_requires` having the old behavior. It is important that we actively deprecate this, as otherwise it would not solve the problem of having no defined init and shutdown order for `libomptarget`. The problem of `libomptarget` not having a define init and shutdown order cascades into a lot of other issues so I have a strong incentive to be rid of it. It is worth noting that the current `__tgt_offload_entry` only has space for a 32-bit integer here. I am planning to overhaul these at some point as well.
jhuber6
force-pushed
the
RegisterRequires
branch
from
1e5d2dc to
76039ed
Compare
jhuber6
added a commit
to jhuber6/llvm-project
that referenced
this pull request
Feb 22, 2024
Summary: Currently we rely on global constructors to initialize and shut down the OpenMP runtime library and plugin manager. This causes some issues because we do not have a defined lifetime that we can rely on to release and allocate resources. This patch instead adds some simple reference counted initialization and deinitialization function. A future patch will use the `deinit` interface to more intelligently handle plugin deinitilization. Right now we do nothing and rely on `atexit` inside of the plugins to tear them down. This isn't great because it limits our ability to control these things. Note that I made the `__tgt_register_lib` functions do the initialization instead of adding calls to the new runtime functions in the linker wrapper. The reason for this is because in the past it's been easier to not introduce a new function call, since sometimes the user's compiler will link against an older `libomptarget`. Maybe if we change the name with offloading in the future we can simplify this. Depends on llvm#80460
jhuber6
added a commit
to jhuber6/llvm-project
that referenced
this pull request
Feb 22, 2024
Summary: Currently we rely on global constructors to initialize and shut down the OpenMP runtime library and plugin manager. This causes some issues because we do not have a defined lifetime that we can rely on to release and allocate resources. This patch instead adds some simple reference counted initialization and deinitialization function. A future patch will use the `deinit` interface to more intelligently handle plugin deinitilization. Right now we do nothing and rely on `atexit` inside of the plugins to tear them down. This isn't great because it limits our ability to control these things. Note that I made the `__tgt_register_lib` functions do the initialization instead of adding calls to the new runtime functions in the linker wrapper. The reason for this is because in the past it's been easier to not introduce a new function call, since sometimes the user's compiler will link against an older `libomptarget`. Maybe if we change the name with offloading in the future we can simplify this. Depends on llvm#80460 Fix decrement
jhuber6
added a commit
to jhuber6/llvm-project
that referenced
this pull request
Feb 22, 2024
Summary: Currently we rely on global constructors to initialize and shut down the OpenMP runtime library and plugin manager. This causes some issues because we do not have a defined lifetime that we can rely on to release and allocate resources. This patch instead adds some simple reference counted initialization and deinitialization function. A future patch will use the `deinit` interface to more intelligently handle plugin deinitilization. Right now we do nothing and rely on `atexit` inside of the plugins to tear them down. This isn't great because it limits our ability to control these things. Note that I made the `__tgt_register_lib` functions do the initialization instead of adding calls to the new runtime functions in the linker wrapper. The reason for this is because in the past it's been easier to not introduce a new function call, since sometimes the user's compiler will link against an older `libomptarget`. Maybe if we change the name with offloading in the future we can simplify this. Depends on llvm#80460 Fix decrement
jhuber6
added a commit
that referenced
this pull request
Feb 22, 2024
Summary: Currently we rely on global constructors to initialize and shut down the OpenMP runtime library and plugin manager. This causes some issues because we do not have a defined lifetime that we can rely on to release and allocate resources. This patch instead adds some simple reference counted initialization and deinitialization function. A future patch will use the `deinit` interface to more intelligently handle plugin deinitilization. Right now we do nothing and rely on `atexit` inside of the plugins to tear them down. This isn't great because it limits our ability to control these things. Note that I made the `__tgt_register_lib` functions do the initialization instead of adding calls to the new runtime functions in the linker wrapper. The reason for this is because in the past it's been easier to not introduce a new function call, since sometimes the user's compiler will link against an older `libomptarget`. Maybe if we change the name with offloading in the future we can simplify this. Depends on #80460
searlmc1
pushed a commit
to ROCm/llvm-project
that referenced
this pull request
Apr 26, 2024
Summary: Currently, OpenMP handles the `omp requires` clause by emitting a global constructor into the runtime for every translation unit that requires it. However, this is not a great solution because it prevents us from having a defined order in which the runtime is accessed and used. This patch changes the approach to no longer use global constructors, but to instead group the flag with the other offloading entires that we already handle. This has the effect of still registering each flag per requires TU, but now we have a single constructor that handles everything. This function removes support for the old `__tgt_register_requires` and replaces it with a warning message. We just had a recent release, and the OpenMP policy for the past four releases since we switched to LLVM is that we do not provide strict backwards compatibility between major LLVM releases now that the library is versioned. This means that a user will need to recompile if they have an old binary that relied on `register_requires` having the old behavior. It is important that we actively deprecate this, as otherwise it would not solve the problem of having no defined init and shutdown order for `libomptarget`. The problem of `libomptarget` not having a define init and shutdown order cascades into a lot of other issues so I have a strong incentive to be rid of it. It is worth noting that the current `__tgt_offload_entry` only has space for a 32-bit integer here. I am planning to overhaul these at some point as well. Change-Id: If8eb1fbb0eed26ab65fbea9d490196f35d05f1d8
searlmc1
pushed a commit
to ROCm/llvm-project
that referenced
this pull request
Apr 26, 2024
) Summary: Currently we rely on global constructors to initialize and shut down the OpenMP runtime library and plugin manager. This causes some issues because we do not have a defined lifetime that we can rely on to release and allocate resources. This patch instead adds some simple reference counted initialization and deinitialization function. A future patch will use the `deinit` interface to more intelligently handle plugin deinitilization. Right now we do nothing and rely on `atexit` inside of the plugins to tear them down. This isn't great because it limits our ability to control these things. Note that I made the `__tgt_register_lib` functions do the initialization instead of adding calls to the new runtime functions in the linker wrapper. The reason for this is because in the past it's been easier to not introduce a new function call, since sometimes the user's compiler will link against an older `libomptarget`. Maybe if we change the name with offloading in the future we can simplify this. Depends on llvm#80460 Change-Id: I70815457fab9b5d68db8e48b3b5e1c75951c05f5
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Summary:
Currently, OpenMP handles the
omp requiresclause by emitting a globalconstructor into the runtime for every translation unit that requires
it. However, this is not a great solution because it prevents us from
having a defined order in which the runtime is accessed and used.
This patch changes the approach to no longer use global constructors,
but to instead group the flag with the other offloading entires that we
already handle. This has the effect of still registering each flag per
requires TU, but now we have a single constructor that handles
everything.
This function removes support for the old
__tgt_register_requiresandreplaces it with a warning message. We just had a recent release, and
the OpenMP policy for the past four releases since we switched to LLVM
is that we do not provide strict backwards compatibility between major
LLVM releases now that the library is versioned. This means that a user
will need to recompile if they have an old binary that relied on
register_requireshaving the old behavior. It is important that weactively deprecate this, as otherwise it would not solve the problem of
having no defined init and shutdown order for
libomptarget. Theproblem of
libomptargetnot having a define init and shutdown ordercascades into a lot of other issues so I have a strong incentive to be
rid of it.
It is worth noting that the current
__tgt_offload_entryonly has spacefor a 32-bit integer here. I am planning to overhaul these at some point
as well.