Skip to content

[OpenMP][OMPIRBuilder] Use runtime CC for runtime calls #168608

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
sarnex merged 1 commit into from
Nov 21, 2025
Merged

[OpenMP][OMPIRBuilder] Use runtime CC for runtime calls #168608

sarnex merged 1 commit into from
Nov 21, 2025

Conversation

@sarnex
Copy link
Member

@sarnex sarnex commented Nov 18, 2025 •
edited
Loading

Some targets have a specific calling convention that should be used for generated calls to runtime functions.

Pass that down and use it.

@sarnex sarnex marked this pull request as ready for review November 19, 2025 15:11
@llvmbot llvmbot added clang:codegen IR generation bugs: mangling, exceptions, etc. mlir:llvm mlir flang:openmp clang:openmp OpenMP related changes to Clang labels Nov 19, 2025
@llvmbot
Copy link
Member

llvmbot commented Nov 19, 2025 •
edited
Loading

@llvm/pr-subscribers-clang-codegen

@llvm/pr-subscribers-mlir-llvm

Author: Nick Sarnie (sarnex)

Changes

Some targets have a specific calling convention that should be used for generated calls to runtime functions.

Pass that down and use it.

Patch is 42.09 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/168608.diff

7 Files Affected:

  • (modified) clang/lib/CodeGen/CGOpenMPRuntime.cpp (+1)
  • (modified) clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp (+2)
  • (modified) clang/test/OpenMP/spirv_target_codegen_basic.cpp (+2-1)
  • (modified) llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h (+10)
  • (modified) llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp (+120-109)
  • (modified) mlir/lib/Target/LLVMIR/ModuleTranslation.cpp (+4-1)
  • (added) mlir/test/Target/LLVMIR/omptarget-runtimecc.mlir (+12)
diff --git a/clang/lib/CodeGen/CGOpenMPRuntime.cpp b/clang/lib/CodeGen/CGOpenMPRuntime.cpp
index 1224fa681cdc0..a8255ac74cfcf 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntime.cpp
+++ b/clang/lib/CodeGen/CGOpenMPRuntime.cpp
@@ -1039,6 +1039,7 @@ CGOpenMPRuntime::CGOpenMPRuntime(CodeGenModule &CGM)
       hasRequiresUnifiedSharedMemory(), /*HasRequiresDynamicAllocators*/ false);
   Config.setDefaultTargetAS(
       CGM.getContext().getTargetInfo().getTargetAddressSpace(LangAS::Default));
+  Config.setRuntimeCC(CGM.getRuntimeCC());
 
   OMPBuilder.setConfig(Config);
   OMPBuilder.initialize();
diff --git a/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp b/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp
index fddeba98adccc..2f69a53787f0c 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp
+++ b/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp
@@ -871,6 +871,8 @@ CGOpenMPRuntimeGPU::CGOpenMPRuntimeGPU(CodeGenModule &CGM)
       hasRequiresUnifiedSharedMemory(), /*HasRequiresDynamicAllocators*/ false);
   Config.setDefaultTargetAS(
       CGM.getContext().getTargetInfo().getTargetAddressSpace(LangAS::Default));
+  Config.setRuntimeCC(CGM.getRuntimeCC());
+
   OMPBuilder.setConfig(Config);
 
   if (!CGM.getLangOpts().OpenMPIsTargetDevice)
diff --git a/clang/test/OpenMP/spirv_target_codegen_basic.cpp b/clang/test/OpenMP/spirv_target_codegen_basic.cpp
index 6e029fb93644d..ad0e110c0532a 100644
--- a/clang/test/OpenMP/spirv_target_codegen_basic.cpp
+++ b/clang/test/OpenMP/spirv_target_codegen_basic.cpp
@@ -9,9 +9,10 @@
 // CHECK: @"_gomp_critical_user_$var" = common addrspace(1) global [8 x i32] zeroinitializer, align 8
 
 // CHECK: define weak_odr protected spir_kernel void @__omp_offloading_{{.*}}
-
+// CHECK: call spir_func addrspace(9) i32 @__kmpc_target_init(
 // CHECK: call spir_func addrspace(9) void @__kmpc_critical(ptr addrspace(4) addrspacecast (ptr addrspace(1) @{{.*}} to ptr addrspace(4)), i32 %{{.*}}, ptr addrspace(4) addrspacecast (ptr addrspace(1) @"_gomp_critical_user_$var" to ptr addrspace(4)))
 // CHECK: call spir_func addrspace(9) void @__kmpc_end_critical(ptr addrspace(4) addrspacecast (ptr addrspace(1) @{{.*}} to ptr addrspace(4)), i32 %{{.*}}, ptr addrspace(4) addrspacecast (ptr addrspace(1) @"_gomp_critical_user_$var" to ptr addrspace(4)))
+// CHECK: call spir_func addrspace(9) void @__kmpc_target_deinit(
 
 int main() {
   int ret = 0;
diff --git a/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h b/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h
index 9f77c24d0b27b..f864a895a1259 100644
--- a/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h
+++ b/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h
@@ -17,6 +17,7 @@
 #include "llvm/Frontend/Atomic/Atomic.h"
 #include "llvm/Frontend/OpenMP/OMPConstants.h"
 #include "llvm/Frontend/OpenMP/OMPGridValues.h"
+#include "llvm/IR/CallingConv.h"
 #include "llvm/IR/DebugLoc.h"
 #include "llvm/IR/IRBuilder.h"
 #include "llvm/IR/Module.h"
@@ -138,6 +139,8 @@ class OpenMPIRBuilderConfig {
   // Default address space for the target.
   unsigned DefaultTargetAS = 0;
 
+  CallingConv::ID RuntimeCC = llvm::CallingConv::C;
+
   LLVM_ABI OpenMPIRBuilderConfig();
   LLVM_ABI OpenMPIRBuilderConfig(bool IsTargetDevice, bool IsGPU,
                                  bool OpenMPOffloadMandatory,
@@ -170,6 +173,8 @@ class OpenMPIRBuilderConfig {
 
   unsigned getDefaultTargetAS() const { return DefaultTargetAS; }
 
+  CallingConv::ID getRuntimeCC() const { return RuntimeCC; }
+
   bool hasRequiresFlags() const { return RequiresFlags; }
   LLVM_ABI bool hasRequiresReverseOffload() const;
   LLVM_ABI bool hasRequiresUnifiedAddress() const;
@@ -208,6 +213,7 @@ class OpenMPIRBuilderConfig {
   void setSeparator(StringRef S) { Separator = S; }
   void setGridValue(omp::GV G) { GridValue = G; }
   void setDefaultTargetAS(unsigned AS) { DefaultTargetAS = AS; }
+  void setRuntimeCC(CallingConv::ID CC) { RuntimeCC = CC; }
 
   LLVM_ABI void setHasRequiresReverseOffload(bool Value);
   LLVM_ABI void setHasRequiresUnifiedAddress(bool Value);
@@ -2132,6 +2138,10 @@ class OpenMPIRBuilder {
 
   LLVM_ABI Function *getOrCreateRuntimeFunctionPtr(omp::RuntimeFunction FnID);
 
+  CallInst *createRuntimeFunctionCall(FunctionCallee Callee,
+                                      ArrayRef<Value *> Args,
+                                      StringRef Name = "");
+
   /// Return the (LLVM-IR) string describing the source location \p LocStr.
   LLVM_ABI Constant *getOrCreateSrcLocStr(StringRef LocStr,
                                           uint32_t &SrcLocStrSize);
diff --git a/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp b/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
index ac86fa859967e..5101717526263 100644
--- a/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
+++ b/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
@@ -643,7 +643,7 @@ OpenMPIRBuilder::getOrCreateRuntimeFunction(Module &M, RuntimeFunction FnID) {
     break;
 #include "llvm/Frontend/OpenMP/OMPKinds.def"
     }
-
+    Fn->setCallingConv(Config.getRuntimeCC());
     // Add information if the runtime function takes a callback function
     if (FnID == OMPRTL___kmpc_fork_call || FnID == OMPRTL___kmpc_fork_teams) {
       if (!Fn->hasMetadata(LLVMContext::MD_callback)) {
@@ -682,6 +682,14 @@ Function *OpenMPIRBuilder::getOrCreateRuntimeFunctionPtr(RuntimeFunction FnID) {
   return Fn;
 }
 
+CallInst *OpenMPIRBuilder::createRuntimeFunctionCall(FunctionCallee Callee,
+                                                     ArrayRef<Value *> Args,
+                                                     StringRef Name) {
+  CallInst *Call = Builder.CreateCall(Callee, Args, Name);
+  Call->setCallingConv(Config.getRuntimeCC());
+  return Call;
+}
+
 void OpenMPIRBuilder::initialize() { initializeTypes(M); }
 
 static void raiseUserConstantDataAllocasToEntryBlock(IRBuilderBase &Builder,
@@ -1024,7 +1032,7 @@ Constant *OpenMPIRBuilder::getOrCreateSrcLocStr(const LocationDescription &Loc,
 }
 
 Value *OpenMPIRBuilder::getOrCreateThreadID(Value *Ident) {
-  return Builder.CreateCall(
+  return createRuntimeFunctionCall(
       getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_global_thread_num), Ident,
       "omp_global_thread_num");
 }
@@ -1069,11 +1077,11 @@ OpenMPIRBuilder::createBarrier(const LocationDescription &Loc, Directive Kind,
   bool UseCancelBarrier =
       !ForceSimpleCall && isLastFinalizationInfoCancellable(OMPD_parallel);
 
-  Value *Result =
-      Builder.CreateCall(getOrCreateRuntimeFunctionPtr(
-                             UseCancelBarrier ? OMPRTL___kmpc_cancel_barrier
-                                              : OMPRTL___kmpc_barrier),
-                         Args);
+  Value *Result = createRuntimeFunctionCall(
+      getOrCreateRuntimeFunctionPtr(UseCancelBarrier
+                                        ? OMPRTL___kmpc_cancel_barrier
+                                        : OMPRTL___kmpc_barrier),
+      Args);
 
   if (UseCancelBarrier && CheckCancelFlag)
     if (Error Err = emitCancelationCheckImpl(Result, OMPD_parallel))
@@ -1112,7 +1120,7 @@ OpenMPIRBuilder::createCancel(const LocationDescription &Loc,
   Constant *SrcLocStr = getOrCreateSrcLocStr(Loc, SrcLocStrSize);
   Value *Ident = getOrCreateIdent(SrcLocStr, SrcLocStrSize);
   Value *Args[] = {Ident, getOrCreateThreadID(Ident), CancelKind};
-  Value *Result = Builder.CreateCall(
+  Value *Result = createRuntimeFunctionCall(
       getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_cancel), Args);
   auto ExitCB = [this, CanceledDirective, Loc](InsertPointTy IP) -> Error {
     if (CanceledDirective == OMPD_parallel) {
@@ -1163,7 +1171,7 @@ OpenMPIRBuilder::createCancellationPoint(const LocationDescription &Loc,
   Constant *SrcLocStr = getOrCreateSrcLocStr(Loc, SrcLocStrSize);
   Value *Ident = getOrCreateIdent(SrcLocStr, SrcLocStrSize);
   Value *Args[] = {Ident, getOrCreateThreadID(Ident), CancelKind};
-  Value *Result = Builder.CreateCall(
+  Value *Result = createRuntimeFunctionCall(
       getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_cancellationpoint), Args);
   auto ExitCB = [this, CanceledDirective, Loc](InsertPointTy IP) -> Error {
     if (CanceledDirective == OMPD_parallel) {
@@ -1212,7 +1220,7 @@ OpenMPIRBuilder::InsertPointTy OpenMPIRBuilder::emitTargetKernel(
   SmallVector<Value *> OffloadingArgs{Ident,      DeviceID, NumTeams,
                                       NumThreads, HostPtr,  KernelArgsPtr};
 
-  Return = Builder.CreateCall(
+  Return = createRuntimeFunctionCall(
       getOrCreateRuntimeFunction(M, OMPRTL___tgt_target_kernel),
       OffloadingArgs);
 
@@ -1392,7 +1400,7 @@ static void targetParallelCallback(
   FunctionCallee RTLFn =
       OMPIRBuilder->getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_parallel_51);
 
-  Builder.CreateCall(RTLFn, Parallel51CallArgs);
+  OMPIRBuilder->createRuntimeFunctionCall(RTLFn, Parallel51CallArgs);
 
   LLVM_DEBUG(dbgs() << "With kmpc_parallel_51 placed: "
                     << *Builder.GetInsertBlock()->getParent() << "\n");
@@ -1477,7 +1485,7 @@ hostParallelCallback(OpenMPIRBuilder *OMPIRBuilder, Function &OutlinedFn,
     RealArgs.push_back(NullPtrValue);
   }
 
-  Builder.CreateCall(RTLFn, RealArgs);
+  OMPIRBuilder->createRuntimeFunctionCall(RTLFn, RealArgs);
 
   LLVM_DEBUG(dbgs() << "With fork_call placed: "
                     << *Builder.GetInsertBlock()->getParent() << "\n");
@@ -1523,7 +1531,7 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createParallel(
     Value *Args[] = {
         Ident, ThreadID,
         Builder.CreateIntCast(NumThreads, Int32, /*isSigned*/ false)};
-    Builder.CreateCall(
+    createRuntimeFunctionCall(
         getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_push_num_threads), Args);
   }
 
@@ -1532,7 +1540,7 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createParallel(
     Value *Args[] = {
         Ident, ThreadID,
         ConstantInt::get(Int32, unsigned(ProcBind), /*isSigned=*/true)};
-    Builder.CreateCall(
+    createRuntimeFunctionCall(
         getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_push_proc_bind), Args);
   }
 
@@ -1824,7 +1832,8 @@ void OpenMPIRBuilder::emitFlush(const LocationDescription &Loc) {
   Constant *SrcLocStr = getOrCreateSrcLocStr(Loc, SrcLocStrSize);
   Value *Args[] = {getOrCreateIdent(SrcLocStr, SrcLocStrSize)};
 
-  Builder.CreateCall(getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_flush), Args);
+  createRuntimeFunctionCall(getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_flush),
+                            Args);
 }
 
 void OpenMPIRBuilder::createFlush(const LocationDescription &Loc) {
@@ -1842,8 +1851,8 @@ void OpenMPIRBuilder::emitTaskwaitImpl(const LocationDescription &Loc) {
   Value *Args[] = {Ident, getOrCreateThreadID(Ident)};
 
   // Ignore return result until untied tasks are supported.
-  Builder.CreateCall(getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_omp_taskwait),
-                     Args);
+  createRuntimeFunctionCall(
+      getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_omp_taskwait), Args);
 }
 
 void OpenMPIRBuilder::createTaskwait(const LocationDescription &Loc) {
@@ -1860,8 +1869,8 @@ void OpenMPIRBuilder::emitTaskyieldImpl(const LocationDescription &Loc) {
   Constant *I32Null = ConstantInt::getNullValue(Int32);
   Value *Args[] = {Ident, getOrCreateThreadID(Ident), I32Null};
 
-  Builder.CreateCall(getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_omp_taskyield),
-                     Args);
+  createRuntimeFunctionCall(
+      getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_omp_taskyield), Args);
 }
 
 void OpenMPIRBuilder::createTaskyield(const LocationDescription &Loc) {
@@ -2059,7 +2068,7 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createTask(
     // Emit the @__kmpc_omp_task_alloc runtime call
     // The runtime call returns a pointer to an area where the task captured
     // variables must be copied before the task is run (TaskData)
-    CallInst *TaskData = Builder.CreateCall(
+    CallInst *TaskData = createRuntimeFunctionCall(
         TaskAllocFn, {/*loc_ref=*/Ident, /*gtid=*/ThreadID, /*flags=*/Flags,
                       /*sizeof_task=*/TaskSize, /*sizeof_shared=*/SharedsSize,
                       /*task_func=*/&OutlinedFn});
@@ -2071,7 +2080,7 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createTask(
       Function *TaskDetachFn = getOrCreateRuntimeFunctionPtr(
           OMPRTL___kmpc_task_allow_completion_event);
       llvm::Value *EventVal =
-          Builder.CreateCall(TaskDetachFn, {Ident, ThreadID, TaskData});
+          createRuntimeFunctionCall(TaskDetachFn, {Ident, ThreadID, TaskData});
       llvm::Value *EventHandleAddr =
           Builder.CreatePointerBitCastOrAddrSpaceCast(EventHandle,
                                                       Builder.getPtrTy(0));
@@ -2148,7 +2157,7 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createTask(
       if (Dependencies.size()) {
         Function *TaskWaitFn =
             getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_omp_wait_deps);
-        Builder.CreateCall(
+        createRuntimeFunctionCall(
             TaskWaitFn,
             {Ident, ThreadID, Builder.getInt32(Dependencies.size()), DepArray,
              ConstantInt::get(Builder.getInt32Ty(), 0),
@@ -2158,21 +2167,21 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createTask(
           getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_omp_task_begin_if0);
       Function *TaskCompleteFn =
           getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_omp_task_complete_if0);
-      Builder.CreateCall(TaskBeginFn, {Ident, ThreadID, TaskData});
+      createRuntimeFunctionCall(TaskBeginFn, {Ident, ThreadID, TaskData});
       CallInst *CI = nullptr;
       if (HasShareds)
-        CI = Builder.CreateCall(&OutlinedFn, {ThreadID, TaskData});
+        CI = createRuntimeFunctionCall(&OutlinedFn, {ThreadID, TaskData});
       else
-        CI = Builder.CreateCall(&OutlinedFn, {ThreadID});
+        CI = createRuntimeFunctionCall(&OutlinedFn, {ThreadID});
       CI->setDebugLoc(StaleCI->getDebugLoc());
-      Builder.CreateCall(TaskCompleteFn, {Ident, ThreadID, TaskData});
+      createRuntimeFunctionCall(TaskCompleteFn, {Ident, ThreadID, TaskData});
       Builder.SetInsertPoint(ThenTI);
     }
 
     if (Dependencies.size()) {
       Function *TaskFn =
           getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_omp_task_with_deps);
-      Builder.CreateCall(
+      createRuntimeFunctionCall(
           TaskFn,
           {Ident, ThreadID, TaskData, Builder.getInt32(Dependencies.size()),
            DepArray, ConstantInt::get(Builder.getInt32Ty(), 0),
@@ -2181,7 +2190,7 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createTask(
     } else {
       // Emit the @__kmpc_omp_task runtime call to spawn the task
       Function *TaskFn = getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_omp_task);
-      Builder.CreateCall(TaskFn, {Ident, ThreadID, TaskData});
+      createRuntimeFunctionCall(TaskFn, {Ident, ThreadID, TaskData});
     }
 
     StaleCI->eraseFromParent();
@@ -2218,7 +2227,7 @@ OpenMPIRBuilder::createTaskgroup(const LocationDescription &Loc,
   // Emit the @__kmpc_taskgroup runtime call to start the taskgroup
   Function *TaskgroupFn =
       getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_taskgroup);
-  Builder.CreateCall(TaskgroupFn, {Ident, ThreadID});
+  createRuntimeFunctionCall(TaskgroupFn, {Ident, ThreadID});
 
   BasicBlock *TaskgroupExitBB = splitBB(Builder, true, "taskgroup.exit");
   if (Error Err = BodyGenCB(AllocaIP, Builder.saveIP()))
@@ -2228,7 +2237,7 @@ OpenMPIRBuilder::createTaskgroup(const LocationDescription &Loc,
   // Emit the @__kmpc_end_taskgroup runtime call to end the taskgroup
   Function *EndTaskgroupFn =
       getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_end_taskgroup);
-  Builder.CreateCall(EndTaskgroupFn, {Ident, ThreadID});
+  createRuntimeFunctionCall(EndTaskgroupFn, {Ident, ThreadID});
 
   return Builder.saveIP();
 }
@@ -2384,14 +2393,14 @@ static OpenMPIRBuilder::InsertPointTy getInsertPointAfterInstr(Instruction *I) {
 }
 
 Value *OpenMPIRBuilder::getGPUThreadID() {
-  return Builder.CreateCall(
+  return createRuntimeFunctionCall(
       getOrCreateRuntimeFunction(M,
                                  OMPRTL___kmpc_get_hardware_thread_id_in_block),
       {});
 }
 
 Value *OpenMPIRBuilder::getGPUWarpSize() {
-  return Builder.CreateCall(
+  return createRuntimeFunctionCall(
       getOrCreateRuntimeFunction(M, OMPRTL___kmpc_get_warp_size), {});
 }
 
@@ -2450,7 +2459,7 @@ Value *OpenMPIRBuilder::createRuntimeShuffleFunction(InsertPointTy AllocaIP,
   Value *WarpSizeCast =
       Builder.CreateIntCast(WarpSize, Builder.getInt16Ty(), /*isSigned=*/true);
   Value *ShuffleCall =
-      Builder.CreateCall(ShuffleFunc, {ElemCast, Offset, WarpSizeCast});
+      createRuntimeFunctionCall(ShuffleFunc, {ElemCast, Offset, WarpSizeCast});
   return castValueToType(AllocaIP, ShuffleCall, CastTy);
 }
 
@@ -3016,7 +3025,7 @@ Function *OpenMPIRBuilder::emitShuffleAndReduceFunction(
       ReduceList, Builder.getPtrTy());
   Value *RemoteReduceListPtr = Builder.CreatePointerBitCastOrAddrSpaceCast(
       RemoteListAddrCast, Builder.getPtrTy());
-  Builder.CreateCall(ReduceFn, {LocalReduceListPtr, RemoteReduceListPtr})
+  createRuntimeFunctionCall(ReduceFn, {LocalReduceListPtr, RemoteReduceListPtr})
       ->addFnAttr(Attribute::NoUnwind);
   Builder.CreateBr(MergeBB);
 
@@ -3237,7 +3246,7 @@ Function *OpenMPIRBuilder::emitListToGlobalReduceFunction(
   // Call reduce_function(GlobalReduceList, ReduceList)
   Value *ReduceList =
       Builder.CreateLoad(Builder.getPtrTy(), ReduceListArgAddrCast);
-  Builder.CreateCall(ReduceFn, {LocalReduceListAddrCast, ReduceList})
+  createRuntimeFunctionCall(ReduceFn, {LocalReduceListAddrCast, ReduceList})
       ->addFnAttr(Attribute::NoUnwind);
   Builder.CreateRetVoid();
   Builder.restoreIP(OldIP);
@@ -3428,7 +3437,7 @@ Function *OpenMPIRBuilder::emitGlobalToListReduceFunction(
   // Call reduce_function(ReduceList, GlobalReduceList)
   Value *ReduceList =
       Builder.CreateLoad(Builder.getPtrTy(), ReduceListArgAddrCast);
-  Builder.CreateCall(ReduceFn, {ReduceList, ReductionList})
+  createRuntimeFunctionCall(ReduceFn, {ReduceList, ReductionList})
       ->addFnAttr(Attribute::NoUnwind);
   Builder.CreateRetVoid();
   Builder.restoreIP(OldIP);
@@ -3682,7 +3691,7 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createReductionsGPU(
                      WcFuncCast};
     Function *Pv2Ptr = getOrCreateRuntimeFunctionPtr(
         RuntimeFunction::OMPRTL___kmpc_nvptx_parallel_reduce_nowait_v2);
-    Res = Builder.CreateCall(Pv2Ptr, Args);
+    Res = createRuntimeFunctionCall(Pv2Ptr, Args);
   } else {
     CodeGenIP = Builder.saveIP();
     StructType *ReductionsBufferTy = StructType::create(
@@ -3699,7 +3708,7 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createReductionsGPU(
         ReductionInfos, ReductionFunc, ReductionsBufferTy, FuncAttrs);
     Builder.restoreIP(CodeGenIP);
 
-    Value *KernelTeamsReductionPtr = Builder.CreateCall(
+    Value *KernelTeamsReductionPtr = createRuntimeFunctionCall(
         RedFixedBuferFn, {}, "_openmp_teams_reductions_buffer_$_$ptr");
 
     Value *Args3[] = {SrcLocInfo,
@@ -3716,7 +3725,7 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createReductionsGPU(
 
     Function *TeamsReduceFn = getOrCreateRuntimeFunctionPtr(
         RuntimeFunction::OMPRTL___kmpc_nvptx_teams_reduce_nowait_v2);
-    Res = Builder.CreateCall(TeamsReduceFn, Args3);
+    Res = createRuntimeFunctionCall(TeamsReduceFn, Args3);
   }
 
   // 5. Build if (res == 1)
@@ -3921,10 +3930,10 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createReductions(
       IsNoWait ? RuntimeFunction::OMPRTL___kmpc_reduce_nowait
                : RuntimeFunction::OMPRTL___kmpc_reduce);
   CallInst *ReduceCall =
-      Builder.CreateCall(ReduceFunc,
-                         {Ident, ThreadId, NumVariables, RedArraySize, RedArray,
-                          ReductionFunc, Lock},
-                         "reduce");
+      createRuntimeFunctionCall(ReduceFunc,
+                                {Ident, ThreadId, NumVariables, RedArraySize,
+                                 RedArray, ReductionFunc, Lock},
+                                "reduce");
 
   // Create final reduction entry blocks for the atomic and non-atomic case.
   // Emit IR that dispatches control flow to one of the blocks based on the
@@ -3971,7 +3980,7 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenM...
[truncated]

@llvmbot
Copy link
Member

llvmbot commented Nov 19, 2025

@llvm/pr-subscribers-mlir

Author: Nick Sarnie (sarnex)

Changes

Some targets have a specific calling convention that should be used for generated calls to runtime functions.

Pass that down and use it.

Patch is 42.09 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/168608.diff

7 Files Affected:

  • (modified) clang/lib/CodeGen/CGOpenMPRuntime.cpp (+1)
  • (modified) clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp (+2)
  • (modified) clang/test/OpenMP/spirv_target_codegen_basic.cpp (+2-1)
  • (modified) llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h (+10)
  • (modified) llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp (+120-109)
  • (modified) mlir/lib/Target/LLVMIR/ModuleTranslation.cpp (+4-1)
  • (added) mlir/test/Target/LLVMIR/omptarget-runtimecc.mlir (+12)
diff --git a/clang/lib/CodeGen/CGOpenMPRuntime.cpp b/clang/lib/CodeGen/CGOpenMPRuntime.cpp
index 1224fa681cdc0..a8255ac74cfcf 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntime.cpp
+++ b/clang/lib/CodeGen/CGOpenMPRuntime.cpp
@@ -1039,6 +1039,7 @@ CGOpenMPRuntime::CGOpenMPRuntime(CodeGenModule &CGM)
       hasRequiresUnifiedSharedMemory(), /*HasRequiresDynamicAllocators*/ false);
   Config.setDefaultTargetAS(
       CGM.getContext().getTargetInfo().getTargetAddressSpace(LangAS::Default));
+  Config.setRuntimeCC(CGM.getRuntimeCC());
 
   OMPBuilder.setConfig(Config);
   OMPBuilder.initialize();
diff --git a/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp b/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp
index fddeba98adccc..2f69a53787f0c 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp
+++ b/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp
@@ -871,6 +871,8 @@ CGOpenMPRuntimeGPU::CGOpenMPRuntimeGPU(CodeGenModule &CGM)
       hasRequiresUnifiedSharedMemory(), /*HasRequiresDynamicAllocators*/ false);
   Config.setDefaultTargetAS(
       CGM.getContext().getTargetInfo().getTargetAddressSpace(LangAS::Default));
+  Config.setRuntimeCC(CGM.getRuntimeCC());
+
   OMPBuilder.setConfig(Config);
 
   if (!CGM.getLangOpts().OpenMPIsTargetDevice)
diff --git a/clang/test/OpenMP/spirv_target_codegen_basic.cpp b/clang/test/OpenMP/spirv_target_codegen_basic.cpp
index 6e029fb93644d..ad0e110c0532a 100644
--- a/clang/test/OpenMP/spirv_target_codegen_basic.cpp
+++ b/clang/test/OpenMP/spirv_target_codegen_basic.cpp
@@ -9,9 +9,10 @@
 // CHECK: @"_gomp_critical_user_$var" = common addrspace(1) global [8 x i32] zeroinitializer, align 8
 
 // CHECK: define weak_odr protected spir_kernel void @__omp_offloading_{{.*}}
-
+// CHECK: call spir_func addrspace(9) i32 @__kmpc_target_init(
 // CHECK: call spir_func addrspace(9) void @__kmpc_critical(ptr addrspace(4) addrspacecast (ptr addrspace(1) @{{.*}} to ptr addrspace(4)), i32 %{{.*}}, ptr addrspace(4) addrspacecast (ptr addrspace(1) @"_gomp_critical_user_$var" to ptr addrspace(4)))
 // CHECK: call spir_func addrspace(9) void @__kmpc_end_critical(ptr addrspace(4) addrspacecast (ptr addrspace(1) @{{.*}} to ptr addrspace(4)), i32 %{{.*}}, ptr addrspace(4) addrspacecast (ptr addrspace(1) @"_gomp_critical_user_$var" to ptr addrspace(4)))
+// CHECK: call spir_func addrspace(9) void @__kmpc_target_deinit(
 
 int main() {
   int ret = 0;
diff --git a/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h b/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h
index 9f77c24d0b27b..f864a895a1259 100644
--- a/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h
+++ b/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h
@@ -17,6 +17,7 @@
 #include "llvm/Frontend/Atomic/Atomic.h"
 #include "llvm/Frontend/OpenMP/OMPConstants.h"
 #include "llvm/Frontend/OpenMP/OMPGridValues.h"
+#include "llvm/IR/CallingConv.h"
 #include "llvm/IR/DebugLoc.h"
 #include "llvm/IR/IRBuilder.h"
 #include "llvm/IR/Module.h"
@@ -138,6 +139,8 @@ class OpenMPIRBuilderConfig {
   // Default address space for the target.
   unsigned DefaultTargetAS = 0;
 
+  CallingConv::ID RuntimeCC = llvm::CallingConv::C;
+
   LLVM_ABI OpenMPIRBuilderConfig();
   LLVM_ABI OpenMPIRBuilderConfig(bool IsTargetDevice, bool IsGPU,
                                  bool OpenMPOffloadMandatory,
@@ -170,6 +173,8 @@ class OpenMPIRBuilderConfig {
 
   unsigned getDefaultTargetAS() const { return DefaultTargetAS; }
 
+  CallingConv::ID getRuntimeCC() const { return RuntimeCC; }
+
   bool hasRequiresFlags() const { return RequiresFlags; }
   LLVM_ABI bool hasRequiresReverseOffload() const;
   LLVM_ABI bool hasRequiresUnifiedAddress() const;
@@ -208,6 +213,7 @@ class OpenMPIRBuilderConfig {
   void setSeparator(StringRef S) { Separator = S; }
   void setGridValue(omp::GV G) { GridValue = G; }
   void setDefaultTargetAS(unsigned AS) { DefaultTargetAS = AS; }
+  void setRuntimeCC(CallingConv::ID CC) { RuntimeCC = CC; }
 
   LLVM_ABI void setHasRequiresReverseOffload(bool Value);
   LLVM_ABI void setHasRequiresUnifiedAddress(bool Value);
@@ -2132,6 +2138,10 @@ class OpenMPIRBuilder {
 
   LLVM_ABI Function *getOrCreateRuntimeFunctionPtr(omp::RuntimeFunction FnID);
 
+  CallInst *createRuntimeFunctionCall(FunctionCallee Callee,
+                                      ArrayRef<Value *> Args,
+                                      StringRef Name = "");
+
   /// Return the (LLVM-IR) string describing the source location \p LocStr.
   LLVM_ABI Constant *getOrCreateSrcLocStr(StringRef LocStr,
                                           uint32_t &SrcLocStrSize);
diff --git a/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp b/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
index ac86fa859967e..5101717526263 100644
--- a/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
+++ b/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
@@ -643,7 +643,7 @@ OpenMPIRBuilder::getOrCreateRuntimeFunction(Module &M, RuntimeFunction FnID) {
     break;
 #include "llvm/Frontend/OpenMP/OMPKinds.def"
     }
-
+    Fn->setCallingConv(Config.getRuntimeCC());
     // Add information if the runtime function takes a callback function
     if (FnID == OMPRTL___kmpc_fork_call || FnID == OMPRTL___kmpc_fork_teams) {
       if (!Fn->hasMetadata(LLVMContext::MD_callback)) {
@@ -682,6 +682,14 @@ Function *OpenMPIRBuilder::getOrCreateRuntimeFunctionPtr(RuntimeFunction FnID) {
   return Fn;
 }
 
+CallInst *OpenMPIRBuilder::createRuntimeFunctionCall(FunctionCallee Callee,
+                                                     ArrayRef<Value *> Args,
+                                                     StringRef Name) {
+  CallInst *Call = Builder.CreateCall(Callee, Args, Name);
+  Call->setCallingConv(Config.getRuntimeCC());
+  return Call;
+}
+
 void OpenMPIRBuilder::initialize() { initializeTypes(M); }
 
 static void raiseUserConstantDataAllocasToEntryBlock(IRBuilderBase &Builder,
@@ -1024,7 +1032,7 @@ Constant *OpenMPIRBuilder::getOrCreateSrcLocStr(const LocationDescription &Loc,
 }
 
 Value *OpenMPIRBuilder::getOrCreateThreadID(Value *Ident) {
-  return Builder.CreateCall(
+  return createRuntimeFunctionCall(
       getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_global_thread_num), Ident,
       "omp_global_thread_num");
 }
@@ -1069,11 +1077,11 @@ OpenMPIRBuilder::createBarrier(const LocationDescription &Loc, Directive Kind,
   bool UseCancelBarrier =
       !ForceSimpleCall && isLastFinalizationInfoCancellable(OMPD_parallel);
 
-  Value *Result =
-      Builder.CreateCall(getOrCreateRuntimeFunctionPtr(
-                             UseCancelBarrier ? OMPRTL___kmpc_cancel_barrier
-                                              : OMPRTL___kmpc_barrier),
-                         Args);
+  Value *Result = createRuntimeFunctionCall(
+      getOrCreateRuntimeFunctionPtr(UseCancelBarrier
+                                        ? OMPRTL___kmpc_cancel_barrier
+                                        : OMPRTL___kmpc_barrier),
+      Args);
 
   if (UseCancelBarrier && CheckCancelFlag)
     if (Error Err = emitCancelationCheckImpl(Result, OMPD_parallel))
@@ -1112,7 +1120,7 @@ OpenMPIRBuilder::createCancel(const LocationDescription &Loc,
   Constant *SrcLocStr = getOrCreateSrcLocStr(Loc, SrcLocStrSize);
   Value *Ident = getOrCreateIdent(SrcLocStr, SrcLocStrSize);
   Value *Args[] = {Ident, getOrCreateThreadID(Ident), CancelKind};
-  Value *Result = Builder.CreateCall(
+  Value *Result = createRuntimeFunctionCall(
       getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_cancel), Args);
   auto ExitCB = [this, CanceledDirective, Loc](InsertPointTy IP) -> Error {
     if (CanceledDirective == OMPD_parallel) {
@@ -1163,7 +1171,7 @@ OpenMPIRBuilder::createCancellationPoint(const LocationDescription &Loc,
   Constant *SrcLocStr = getOrCreateSrcLocStr(Loc, SrcLocStrSize);
   Value *Ident = getOrCreateIdent(SrcLocStr, SrcLocStrSize);
   Value *Args[] = {Ident, getOrCreateThreadID(Ident), CancelKind};
-  Value *Result = Builder.CreateCall(
+  Value *Result = createRuntimeFunctionCall(
       getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_cancellationpoint), Args);
   auto ExitCB = [this, CanceledDirective, Loc](InsertPointTy IP) -> Error {
     if (CanceledDirective == OMPD_parallel) {
@@ -1212,7 +1220,7 @@ OpenMPIRBuilder::InsertPointTy OpenMPIRBuilder::emitTargetKernel(
   SmallVector<Value *> OffloadingArgs{Ident,      DeviceID, NumTeams,
                                       NumThreads, HostPtr,  KernelArgsPtr};
 
-  Return = Builder.CreateCall(
+  Return = createRuntimeFunctionCall(
       getOrCreateRuntimeFunction(M, OMPRTL___tgt_target_kernel),
       OffloadingArgs);
 
@@ -1392,7 +1400,7 @@ static void targetParallelCallback(
   FunctionCallee RTLFn =
       OMPIRBuilder->getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_parallel_51);
 
-  Builder.CreateCall(RTLFn, Parallel51CallArgs);
+  OMPIRBuilder->createRuntimeFunctionCall(RTLFn, Parallel51CallArgs);
 
   LLVM_DEBUG(dbgs() << "With kmpc_parallel_51 placed: "
                     << *Builder.GetInsertBlock()->getParent() << "\n");
@@ -1477,7 +1485,7 @@ hostParallelCallback(OpenMPIRBuilder *OMPIRBuilder, Function &OutlinedFn,
     RealArgs.push_back(NullPtrValue);
   }
 
-  Builder.CreateCall(RTLFn, RealArgs);
+  OMPIRBuilder->createRuntimeFunctionCall(RTLFn, RealArgs);
 
   LLVM_DEBUG(dbgs() << "With fork_call placed: "
                     << *Builder.GetInsertBlock()->getParent() << "\n");
@@ -1523,7 +1531,7 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createParallel(
     Value *Args[] = {
         Ident, ThreadID,
         Builder.CreateIntCast(NumThreads, Int32, /*isSigned*/ false)};
-    Builder.CreateCall(
+    createRuntimeFunctionCall(
         getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_push_num_threads), Args);
   }
 
@@ -1532,7 +1540,7 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createParallel(
     Value *Args[] = {
         Ident, ThreadID,
         ConstantInt::get(Int32, unsigned(ProcBind), /*isSigned=*/true)};
-    Builder.CreateCall(
+    createRuntimeFunctionCall(
         getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_push_proc_bind), Args);
   }
 
@@ -1824,7 +1832,8 @@ void OpenMPIRBuilder::emitFlush(const LocationDescription &Loc) {
   Constant *SrcLocStr = getOrCreateSrcLocStr(Loc, SrcLocStrSize);
   Value *Args[] = {getOrCreateIdent(SrcLocStr, SrcLocStrSize)};
 
-  Builder.CreateCall(getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_flush), Args);
+  createRuntimeFunctionCall(getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_flush),
+                            Args);
 }
 
 void OpenMPIRBuilder::createFlush(const LocationDescription &Loc) {
@@ -1842,8 +1851,8 @@ void OpenMPIRBuilder::emitTaskwaitImpl(const LocationDescription &Loc) {
   Value *Args[] = {Ident, getOrCreateThreadID(Ident)};
 
   // Ignore return result until untied tasks are supported.
-  Builder.CreateCall(getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_omp_taskwait),
-                     Args);
+  createRuntimeFunctionCall(
+      getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_omp_taskwait), Args);
 }
 
 void OpenMPIRBuilder::createTaskwait(const LocationDescription &Loc) {
@@ -1860,8 +1869,8 @@ void OpenMPIRBuilder::emitTaskyieldImpl(const LocationDescription &Loc) {
   Constant *I32Null = ConstantInt::getNullValue(Int32);
   Value *Args[] = {Ident, getOrCreateThreadID(Ident), I32Null};
 
-  Builder.CreateCall(getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_omp_taskyield),
-                     Args);
+  createRuntimeFunctionCall(
+      getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_omp_taskyield), Args);
 }
 
 void OpenMPIRBuilder::createTaskyield(const LocationDescription &Loc) {
@@ -2059,7 +2068,7 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createTask(
     // Emit the @__kmpc_omp_task_alloc runtime call
     // The runtime call returns a pointer to an area where the task captured
     // variables must be copied before the task is run (TaskData)
-    CallInst *TaskData = Builder.CreateCall(
+    CallInst *TaskData = createRuntimeFunctionCall(
         TaskAllocFn, {/*loc_ref=*/Ident, /*gtid=*/ThreadID, /*flags=*/Flags,
                       /*sizeof_task=*/TaskSize, /*sizeof_shared=*/SharedsSize,
                       /*task_func=*/&OutlinedFn});
@@ -2071,7 +2080,7 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createTask(
       Function *TaskDetachFn = getOrCreateRuntimeFunctionPtr(
           OMPRTL___kmpc_task_allow_completion_event);
       llvm::Value *EventVal =
-          Builder.CreateCall(TaskDetachFn, {Ident, ThreadID, TaskData});
+          createRuntimeFunctionCall(TaskDetachFn, {Ident, ThreadID, TaskData});
       llvm::Value *EventHandleAddr =
           Builder.CreatePointerBitCastOrAddrSpaceCast(EventHandle,
                                                       Builder.getPtrTy(0));
@@ -2148,7 +2157,7 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createTask(
       if (Dependencies.size()) {
         Function *TaskWaitFn =
             getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_omp_wait_deps);
-        Builder.CreateCall(
+        createRuntimeFunctionCall(
             TaskWaitFn,
             {Ident, ThreadID, Builder.getInt32(Dependencies.size()), DepArray,
              ConstantInt::get(Builder.getInt32Ty(), 0),
@@ -2158,21 +2167,21 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createTask(
           getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_omp_task_begin_if0);
       Function *TaskCompleteFn =
           getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_omp_task_complete_if0);
-      Builder.CreateCall(TaskBeginFn, {Ident, ThreadID, TaskData});
+      createRuntimeFunctionCall(TaskBeginFn, {Ident, ThreadID, TaskData});
       CallInst *CI = nullptr;
       if (HasShareds)
-        CI = Builder.CreateCall(&OutlinedFn, {ThreadID, TaskData});
+        CI = createRuntimeFunctionCall(&OutlinedFn, {ThreadID, TaskData});
       else
-        CI = Builder.CreateCall(&OutlinedFn, {ThreadID});
+        CI = createRuntimeFunctionCall(&OutlinedFn, {ThreadID});
       CI->setDebugLoc(StaleCI->getDebugLoc());
-      Builder.CreateCall(TaskCompleteFn, {Ident, ThreadID, TaskData});
+      createRuntimeFunctionCall(TaskCompleteFn, {Ident, ThreadID, TaskData});
       Builder.SetInsertPoint(ThenTI);
     }
 
     if (Dependencies.size()) {
       Function *TaskFn =
           getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_omp_task_with_deps);
-      Builder.CreateCall(
+      createRuntimeFunctionCall(
           TaskFn,
           {Ident, ThreadID, TaskData, Builder.getInt32(Dependencies.size()),
            DepArray, ConstantInt::get(Builder.getInt32Ty(), 0),
@@ -2181,7 +2190,7 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createTask(
     } else {
       // Emit the @__kmpc_omp_task runtime call to spawn the task
       Function *TaskFn = getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_omp_task);
-      Builder.CreateCall(TaskFn, {Ident, ThreadID, TaskData});
+      createRuntimeFunctionCall(TaskFn, {Ident, ThreadID, TaskData});
     }
 
     StaleCI->eraseFromParent();
@@ -2218,7 +2227,7 @@ OpenMPIRBuilder::createTaskgroup(const LocationDescription &Loc,
   // Emit the @__kmpc_taskgroup runtime call to start the taskgroup
   Function *TaskgroupFn =
       getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_taskgroup);
-  Builder.CreateCall(TaskgroupFn, {Ident, ThreadID});
+  createRuntimeFunctionCall(TaskgroupFn, {Ident, ThreadID});
 
   BasicBlock *TaskgroupExitBB = splitBB(Builder, true, "taskgroup.exit");
   if (Error Err = BodyGenCB(AllocaIP, Builder.saveIP()))
@@ -2228,7 +2237,7 @@ OpenMPIRBuilder::createTaskgroup(const LocationDescription &Loc,
   // Emit the @__kmpc_end_taskgroup runtime call to end the taskgroup
   Function *EndTaskgroupFn =
       getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_end_taskgroup);
-  Builder.CreateCall(EndTaskgroupFn, {Ident, ThreadID});
+  createRuntimeFunctionCall(EndTaskgroupFn, {Ident, ThreadID});
 
   return Builder.saveIP();
 }
@@ -2384,14 +2393,14 @@ static OpenMPIRBuilder::InsertPointTy getInsertPointAfterInstr(Instruction *I) {
 }
 
 Value *OpenMPIRBuilder::getGPUThreadID() {
-  return Builder.CreateCall(
+  return createRuntimeFunctionCall(
       getOrCreateRuntimeFunction(M,
                                  OMPRTL___kmpc_get_hardware_thread_id_in_block),
       {});
 }
 
 Value *OpenMPIRBuilder::getGPUWarpSize() {
-  return Builder.CreateCall(
+  return createRuntimeFunctionCall(
       getOrCreateRuntimeFunction(M, OMPRTL___kmpc_get_warp_size), {});
 }
 
@@ -2450,7 +2459,7 @@ Value *OpenMPIRBuilder::createRuntimeShuffleFunction(InsertPointTy AllocaIP,
   Value *WarpSizeCast =
       Builder.CreateIntCast(WarpSize, Builder.getInt16Ty(), /*isSigned=*/true);
   Value *ShuffleCall =
-      Builder.CreateCall(ShuffleFunc, {ElemCast, Offset, WarpSizeCast});
+      createRuntimeFunctionCall(ShuffleFunc, {ElemCast, Offset, WarpSizeCast});
   return castValueToType(AllocaIP, ShuffleCall, CastTy);
 }
 
@@ -3016,7 +3025,7 @@ Function *OpenMPIRBuilder::emitShuffleAndReduceFunction(
       ReduceList, Builder.getPtrTy());
   Value *RemoteReduceListPtr = Builder.CreatePointerBitCastOrAddrSpaceCast(
       RemoteListAddrCast, Builder.getPtrTy());
-  Builder.CreateCall(ReduceFn, {LocalReduceListPtr, RemoteReduceListPtr})
+  createRuntimeFunctionCall(ReduceFn, {LocalReduceListPtr, RemoteReduceListPtr})
       ->addFnAttr(Attribute::NoUnwind);
   Builder.CreateBr(MergeBB);
 
@@ -3237,7 +3246,7 @@ Function *OpenMPIRBuilder::emitListToGlobalReduceFunction(
   // Call reduce_function(GlobalReduceList, ReduceList)
   Value *ReduceList =
       Builder.CreateLoad(Builder.getPtrTy(), ReduceListArgAddrCast);
-  Builder.CreateCall(ReduceFn, {LocalReduceListAddrCast, ReduceList})
+  createRuntimeFunctionCall(ReduceFn, {LocalReduceListAddrCast, ReduceList})
       ->addFnAttr(Attribute::NoUnwind);
   Builder.CreateRetVoid();
   Builder.restoreIP(OldIP);
@@ -3428,7 +3437,7 @@ Function *OpenMPIRBuilder::emitGlobalToListReduceFunction(
   // Call reduce_function(ReduceList, GlobalReduceList)
   Value *ReduceList =
       Builder.CreateLoad(Builder.getPtrTy(), ReduceListArgAddrCast);
-  Builder.CreateCall(ReduceFn, {ReduceList, ReductionList})
+  createRuntimeFunctionCall(ReduceFn, {ReduceList, ReductionList})
       ->addFnAttr(Attribute::NoUnwind);
   Builder.CreateRetVoid();
   Builder.restoreIP(OldIP);
@@ -3682,7 +3691,7 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createReductionsGPU(
                      WcFuncCast};
     Function *Pv2Ptr = getOrCreateRuntimeFunctionPtr(
         RuntimeFunction::OMPRTL___kmpc_nvptx_parallel_reduce_nowait_v2);
-    Res = Builder.CreateCall(Pv2Ptr, Args);
+    Res = createRuntimeFunctionCall(Pv2Ptr, Args);
   } else {
     CodeGenIP = Builder.saveIP();
     StructType *ReductionsBufferTy = StructType::create(
@@ -3699,7 +3708,7 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createReductionsGPU(
         ReductionInfos, ReductionFunc, ReductionsBufferTy, FuncAttrs);
     Builder.restoreIP(CodeGenIP);
 
-    Value *KernelTeamsReductionPtr = Builder.CreateCall(
+    Value *KernelTeamsReductionPtr = createRuntimeFunctionCall(
         RedFixedBuferFn, {}, "_openmp_teams_reductions_buffer_$_$ptr");
 
     Value *Args3[] = {SrcLocInfo,
@@ -3716,7 +3725,7 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createReductionsGPU(
 
     Function *TeamsReduceFn = getOrCreateRuntimeFunctionPtr(
         RuntimeFunction::OMPRTL___kmpc_nvptx_teams_reduce_nowait_v2);
-    Res = Builder.CreateCall(TeamsReduceFn, Args3);
+    Res = createRuntimeFunctionCall(TeamsReduceFn, Args3);
   }
 
   // 5. Build if (res == 1)
@@ -3921,10 +3930,10 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createReductions(
       IsNoWait ? RuntimeFunction::OMPRTL___kmpc_reduce_nowait
                : RuntimeFunction::OMPRTL___kmpc_reduce);
   CallInst *ReduceCall =
-      Builder.CreateCall(ReduceFunc,
-                         {Ident, ThreadId, NumVariables, RedArraySize, RedArray,
-                          ReductionFunc, Lock},
-                         "reduce");
+      createRuntimeFunctionCall(ReduceFunc,
+                                {Ident, ThreadId, NumVariables, RedArraySize,
+                                 RedArray, ReductionFunc, Lock},
+                                "reduce");
 
   // Create final reduction entry blocks for the atomic and non-atomic case.
   // Emit IR that dispatches control flow to one of the blocks based on the
@@ -3971,7 +3980,7 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenM...
[truncated]

@llvmbot
Copy link
Member

llvmbot commented Nov 19, 2025

@llvm/pr-subscribers-flang-openmp

Author: Nick Sarnie (sarnex)

Changes

Some targets have a specific calling convention that should be used for generated calls to runtime functions.

Pass that down and use it.

Patch is 42.09 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/168608.diff

7 Files Affected:

  • (modified) clang/lib/CodeGen/CGOpenMPRuntime.cpp (+1)
  • (modified) clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp (+2)
  • (modified) clang/test/OpenMP/spirv_target_codegen_basic.cpp (+2-1)
  • (modified) llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h (+10)
  • (modified) llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp (+120-109)
  • (modified) mlir/lib/Target/LLVMIR/ModuleTranslation.cpp (+4-1)
  • (added) mlir/test/Target/LLVMIR/omptarget-runtimecc.mlir (+12)
diff --git a/clang/lib/CodeGen/CGOpenMPRuntime.cpp b/clang/lib/CodeGen/CGOpenMPRuntime.cpp
index 1224fa681cdc0..a8255ac74cfcf 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntime.cpp
+++ b/clang/lib/CodeGen/CGOpenMPRuntime.cpp
@@ -1039,6 +1039,7 @@ CGOpenMPRuntime::CGOpenMPRuntime(CodeGenModule &CGM)
       hasRequiresUnifiedSharedMemory(), /*HasRequiresDynamicAllocators*/ false);
   Config.setDefaultTargetAS(
       CGM.getContext().getTargetInfo().getTargetAddressSpace(LangAS::Default));
+  Config.setRuntimeCC(CGM.getRuntimeCC());
 
   OMPBuilder.setConfig(Config);
   OMPBuilder.initialize();
diff --git a/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp b/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp
index fddeba98adccc..2f69a53787f0c 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp
+++ b/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp
@@ -871,6 +871,8 @@ CGOpenMPRuntimeGPU::CGOpenMPRuntimeGPU(CodeGenModule &CGM)
       hasRequiresUnifiedSharedMemory(), /*HasRequiresDynamicAllocators*/ false);
   Config.setDefaultTargetAS(
       CGM.getContext().getTargetInfo().getTargetAddressSpace(LangAS::Default));
+  Config.setRuntimeCC(CGM.getRuntimeCC());
+
   OMPBuilder.setConfig(Config);
 
   if (!CGM.getLangOpts().OpenMPIsTargetDevice)
diff --git a/clang/test/OpenMP/spirv_target_codegen_basic.cpp b/clang/test/OpenMP/spirv_target_codegen_basic.cpp
index 6e029fb93644d..ad0e110c0532a 100644
--- a/clang/test/OpenMP/spirv_target_codegen_basic.cpp
+++ b/clang/test/OpenMP/spirv_target_codegen_basic.cpp
@@ -9,9 +9,10 @@
 // CHECK: @"_gomp_critical_user_$var" = common addrspace(1) global [8 x i32] zeroinitializer, align 8
 
 // CHECK: define weak_odr protected spir_kernel void @__omp_offloading_{{.*}}
-
+// CHECK: call spir_func addrspace(9) i32 @__kmpc_target_init(
 // CHECK: call spir_func addrspace(9) void @__kmpc_critical(ptr addrspace(4) addrspacecast (ptr addrspace(1) @{{.*}} to ptr addrspace(4)), i32 %{{.*}}, ptr addrspace(4) addrspacecast (ptr addrspace(1) @"_gomp_critical_user_$var" to ptr addrspace(4)))
 // CHECK: call spir_func addrspace(9) void @__kmpc_end_critical(ptr addrspace(4) addrspacecast (ptr addrspace(1) @{{.*}} to ptr addrspace(4)), i32 %{{.*}}, ptr addrspace(4) addrspacecast (ptr addrspace(1) @"_gomp_critical_user_$var" to ptr addrspace(4)))
+// CHECK: call spir_func addrspace(9) void @__kmpc_target_deinit(
 
 int main() {
   int ret = 0;
diff --git a/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h b/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h
index 9f77c24d0b27b..f864a895a1259 100644
--- a/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h
+++ b/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h
@@ -17,6 +17,7 @@
 #include "llvm/Frontend/Atomic/Atomic.h"
 #include "llvm/Frontend/OpenMP/OMPConstants.h"
 #include "llvm/Frontend/OpenMP/OMPGridValues.h"
+#include "llvm/IR/CallingConv.h"
 #include "llvm/IR/DebugLoc.h"
 #include "llvm/IR/IRBuilder.h"
 #include "llvm/IR/Module.h"
@@ -138,6 +139,8 @@ class OpenMPIRBuilderConfig {
   // Default address space for the target.
   unsigned DefaultTargetAS = 0;
 
+  CallingConv::ID RuntimeCC = llvm::CallingConv::C;
+
   LLVM_ABI OpenMPIRBuilderConfig();
   LLVM_ABI OpenMPIRBuilderConfig(bool IsTargetDevice, bool IsGPU,
                                  bool OpenMPOffloadMandatory,
@@ -170,6 +173,8 @@ class OpenMPIRBuilderConfig {
 
   unsigned getDefaultTargetAS() const { return DefaultTargetAS; }
 
+  CallingConv::ID getRuntimeCC() const { return RuntimeCC; }
+
   bool hasRequiresFlags() const { return RequiresFlags; }
   LLVM_ABI bool hasRequiresReverseOffload() const;
   LLVM_ABI bool hasRequiresUnifiedAddress() const;
@@ -208,6 +213,7 @@ class OpenMPIRBuilderConfig {
   void setSeparator(StringRef S) { Separator = S; }
   void setGridValue(omp::GV G) { GridValue = G; }
   void setDefaultTargetAS(unsigned AS) { DefaultTargetAS = AS; }
+  void setRuntimeCC(CallingConv::ID CC) { RuntimeCC = CC; }
 
   LLVM_ABI void setHasRequiresReverseOffload(bool Value);
   LLVM_ABI void setHasRequiresUnifiedAddress(bool Value);
@@ -2132,6 +2138,10 @@ class OpenMPIRBuilder {
 
   LLVM_ABI Function *getOrCreateRuntimeFunctionPtr(omp::RuntimeFunction FnID);
 
+  CallInst *createRuntimeFunctionCall(FunctionCallee Callee,
+                                      ArrayRef<Value *> Args,
+                                      StringRef Name = "");
+
   /// Return the (LLVM-IR) string describing the source location \p LocStr.
   LLVM_ABI Constant *getOrCreateSrcLocStr(StringRef LocStr,
                                           uint32_t &SrcLocStrSize);
diff --git a/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp b/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
index ac86fa859967e..5101717526263 100644
--- a/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
+++ b/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
@@ -643,7 +643,7 @@ OpenMPIRBuilder::getOrCreateRuntimeFunction(Module &M, RuntimeFunction FnID) {
     break;
 #include "llvm/Frontend/OpenMP/OMPKinds.def"
     }
-
+    Fn->setCallingConv(Config.getRuntimeCC());
     // Add information if the runtime function takes a callback function
     if (FnID == OMPRTL___kmpc_fork_call || FnID == OMPRTL___kmpc_fork_teams) {
       if (!Fn->hasMetadata(LLVMContext::MD_callback)) {
@@ -682,6 +682,14 @@ Function *OpenMPIRBuilder::getOrCreateRuntimeFunctionPtr(RuntimeFunction FnID) {
   return Fn;
 }
 
+CallInst *OpenMPIRBuilder::createRuntimeFunctionCall(FunctionCallee Callee,
+                                                     ArrayRef<Value *> Args,
+                                                     StringRef Name) {
+  CallInst *Call = Builder.CreateCall(Callee, Args, Name);
+  Call->setCallingConv(Config.getRuntimeCC());
+  return Call;
+}
+
 void OpenMPIRBuilder::initialize() { initializeTypes(M); }
 
 static void raiseUserConstantDataAllocasToEntryBlock(IRBuilderBase &Builder,
@@ -1024,7 +1032,7 @@ Constant *OpenMPIRBuilder::getOrCreateSrcLocStr(const LocationDescription &Loc,
 }
 
 Value *OpenMPIRBuilder::getOrCreateThreadID(Value *Ident) {
-  return Builder.CreateCall(
+  return createRuntimeFunctionCall(
       getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_global_thread_num), Ident,
       "omp_global_thread_num");
 }
@@ -1069,11 +1077,11 @@ OpenMPIRBuilder::createBarrier(const LocationDescription &Loc, Directive Kind,
   bool UseCancelBarrier =
       !ForceSimpleCall && isLastFinalizationInfoCancellable(OMPD_parallel);
 
-  Value *Result =
-      Builder.CreateCall(getOrCreateRuntimeFunctionPtr(
-                             UseCancelBarrier ? OMPRTL___kmpc_cancel_barrier
-                                              : OMPRTL___kmpc_barrier),
-                         Args);
+  Value *Result = createRuntimeFunctionCall(
+      getOrCreateRuntimeFunctionPtr(UseCancelBarrier
+                                        ? OMPRTL___kmpc_cancel_barrier
+                                        : OMPRTL___kmpc_barrier),
+      Args);
 
   if (UseCancelBarrier && CheckCancelFlag)
     if (Error Err = emitCancelationCheckImpl(Result, OMPD_parallel))
@@ -1112,7 +1120,7 @@ OpenMPIRBuilder::createCancel(const LocationDescription &Loc,
   Constant *SrcLocStr = getOrCreateSrcLocStr(Loc, SrcLocStrSize);
   Value *Ident = getOrCreateIdent(SrcLocStr, SrcLocStrSize);
   Value *Args[] = {Ident, getOrCreateThreadID(Ident), CancelKind};
-  Value *Result = Builder.CreateCall(
+  Value *Result = createRuntimeFunctionCall(
       getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_cancel), Args);
   auto ExitCB = [this, CanceledDirective, Loc](InsertPointTy IP) -> Error {
     if (CanceledDirective == OMPD_parallel) {
@@ -1163,7 +1171,7 @@ OpenMPIRBuilder::createCancellationPoint(const LocationDescription &Loc,
   Constant *SrcLocStr = getOrCreateSrcLocStr(Loc, SrcLocStrSize);
   Value *Ident = getOrCreateIdent(SrcLocStr, SrcLocStrSize);
   Value *Args[] = {Ident, getOrCreateThreadID(Ident), CancelKind};
-  Value *Result = Builder.CreateCall(
+  Value *Result = createRuntimeFunctionCall(
       getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_cancellationpoint), Args);
   auto ExitCB = [this, CanceledDirective, Loc](InsertPointTy IP) -> Error {
     if (CanceledDirective == OMPD_parallel) {
@@ -1212,7 +1220,7 @@ OpenMPIRBuilder::InsertPointTy OpenMPIRBuilder::emitTargetKernel(
   SmallVector<Value *> OffloadingArgs{Ident,      DeviceID, NumTeams,
                                       NumThreads, HostPtr,  KernelArgsPtr};
 
-  Return = Builder.CreateCall(
+  Return = createRuntimeFunctionCall(
       getOrCreateRuntimeFunction(M, OMPRTL___tgt_target_kernel),
       OffloadingArgs);
 
@@ -1392,7 +1400,7 @@ static void targetParallelCallback(
   FunctionCallee RTLFn =
       OMPIRBuilder->getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_parallel_51);
 
-  Builder.CreateCall(RTLFn, Parallel51CallArgs);
+  OMPIRBuilder->createRuntimeFunctionCall(RTLFn, Parallel51CallArgs);
 
   LLVM_DEBUG(dbgs() << "With kmpc_parallel_51 placed: "
                     << *Builder.GetInsertBlock()->getParent() << "\n");
@@ -1477,7 +1485,7 @@ hostParallelCallback(OpenMPIRBuilder *OMPIRBuilder, Function &OutlinedFn,
     RealArgs.push_back(NullPtrValue);
   }
 
-  Builder.CreateCall(RTLFn, RealArgs);
+  OMPIRBuilder->createRuntimeFunctionCall(RTLFn, RealArgs);
 
   LLVM_DEBUG(dbgs() << "With fork_call placed: "
                     << *Builder.GetInsertBlock()->getParent() << "\n");
@@ -1523,7 +1531,7 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createParallel(
     Value *Args[] = {
         Ident, ThreadID,
         Builder.CreateIntCast(NumThreads, Int32, /*isSigned*/ false)};
-    Builder.CreateCall(
+    createRuntimeFunctionCall(
         getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_push_num_threads), Args);
   }
 
@@ -1532,7 +1540,7 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createParallel(
     Value *Args[] = {
         Ident, ThreadID,
         ConstantInt::get(Int32, unsigned(ProcBind), /*isSigned=*/true)};
-    Builder.CreateCall(
+    createRuntimeFunctionCall(
         getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_push_proc_bind), Args);
   }
 
@@ -1824,7 +1832,8 @@ void OpenMPIRBuilder::emitFlush(const LocationDescription &Loc) {
   Constant *SrcLocStr = getOrCreateSrcLocStr(Loc, SrcLocStrSize);
   Value *Args[] = {getOrCreateIdent(SrcLocStr, SrcLocStrSize)};
 
-  Builder.CreateCall(getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_flush), Args);
+  createRuntimeFunctionCall(getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_flush),
+                            Args);
 }
 
 void OpenMPIRBuilder::createFlush(const LocationDescription &Loc) {
@@ -1842,8 +1851,8 @@ void OpenMPIRBuilder::emitTaskwaitImpl(const LocationDescription &Loc) {
   Value *Args[] = {Ident, getOrCreateThreadID(Ident)};
 
   // Ignore return result until untied tasks are supported.
-  Builder.CreateCall(getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_omp_taskwait),
-                     Args);
+  createRuntimeFunctionCall(
+      getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_omp_taskwait), Args);
 }
 
 void OpenMPIRBuilder::createTaskwait(const LocationDescription &Loc) {
@@ -1860,8 +1869,8 @@ void OpenMPIRBuilder::emitTaskyieldImpl(const LocationDescription &Loc) {
   Constant *I32Null = ConstantInt::getNullValue(Int32);
   Value *Args[] = {Ident, getOrCreateThreadID(Ident), I32Null};
 
-  Builder.CreateCall(getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_omp_taskyield),
-                     Args);
+  createRuntimeFunctionCall(
+      getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_omp_taskyield), Args);
 }
 
 void OpenMPIRBuilder::createTaskyield(const LocationDescription &Loc) {
@@ -2059,7 +2068,7 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createTask(
     // Emit the @__kmpc_omp_task_alloc runtime call
     // The runtime call returns a pointer to an area where the task captured
     // variables must be copied before the task is run (TaskData)
-    CallInst *TaskData = Builder.CreateCall(
+    CallInst *TaskData = createRuntimeFunctionCall(
         TaskAllocFn, {/*loc_ref=*/Ident, /*gtid=*/ThreadID, /*flags=*/Flags,
                       /*sizeof_task=*/TaskSize, /*sizeof_shared=*/SharedsSize,
                       /*task_func=*/&OutlinedFn});
@@ -2071,7 +2080,7 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createTask(
       Function *TaskDetachFn = getOrCreateRuntimeFunctionPtr(
           OMPRTL___kmpc_task_allow_completion_event);
       llvm::Value *EventVal =
-          Builder.CreateCall(TaskDetachFn, {Ident, ThreadID, TaskData});
+          createRuntimeFunctionCall(TaskDetachFn, {Ident, ThreadID, TaskData});
       llvm::Value *EventHandleAddr =
           Builder.CreatePointerBitCastOrAddrSpaceCast(EventHandle,
                                                       Builder.getPtrTy(0));
@@ -2148,7 +2157,7 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createTask(
       if (Dependencies.size()) {
         Function *TaskWaitFn =
             getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_omp_wait_deps);
-        Builder.CreateCall(
+        createRuntimeFunctionCall(
             TaskWaitFn,
             {Ident, ThreadID, Builder.getInt32(Dependencies.size()), DepArray,
              ConstantInt::get(Builder.getInt32Ty(), 0),
@@ -2158,21 +2167,21 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createTask(
           getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_omp_task_begin_if0);
       Function *TaskCompleteFn =
           getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_omp_task_complete_if0);
-      Builder.CreateCall(TaskBeginFn, {Ident, ThreadID, TaskData});
+      createRuntimeFunctionCall(TaskBeginFn, {Ident, ThreadID, TaskData});
       CallInst *CI = nullptr;
       if (HasShareds)
-        CI = Builder.CreateCall(&OutlinedFn, {ThreadID, TaskData});
+        CI = createRuntimeFunctionCall(&OutlinedFn, {ThreadID, TaskData});
       else
-        CI = Builder.CreateCall(&OutlinedFn, {ThreadID});
+        CI = createRuntimeFunctionCall(&OutlinedFn, {ThreadID});
       CI->setDebugLoc(StaleCI->getDebugLoc());
-      Builder.CreateCall(TaskCompleteFn, {Ident, ThreadID, TaskData});
+      createRuntimeFunctionCall(TaskCompleteFn, {Ident, ThreadID, TaskData});
       Builder.SetInsertPoint(ThenTI);
     }
 
     if (Dependencies.size()) {
       Function *TaskFn =
           getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_omp_task_with_deps);
-      Builder.CreateCall(
+      createRuntimeFunctionCall(
           TaskFn,
           {Ident, ThreadID, TaskData, Builder.getInt32(Dependencies.size()),
            DepArray, ConstantInt::get(Builder.getInt32Ty(), 0),
@@ -2181,7 +2190,7 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createTask(
     } else {
       // Emit the @__kmpc_omp_task runtime call to spawn the task
       Function *TaskFn = getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_omp_task);
-      Builder.CreateCall(TaskFn, {Ident, ThreadID, TaskData});
+      createRuntimeFunctionCall(TaskFn, {Ident, ThreadID, TaskData});
     }
 
     StaleCI->eraseFromParent();
@@ -2218,7 +2227,7 @@ OpenMPIRBuilder::createTaskgroup(const LocationDescription &Loc,
   // Emit the @__kmpc_taskgroup runtime call to start the taskgroup
   Function *TaskgroupFn =
       getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_taskgroup);
-  Builder.CreateCall(TaskgroupFn, {Ident, ThreadID});
+  createRuntimeFunctionCall(TaskgroupFn, {Ident, ThreadID});
 
   BasicBlock *TaskgroupExitBB = splitBB(Builder, true, "taskgroup.exit");
   if (Error Err = BodyGenCB(AllocaIP, Builder.saveIP()))
@@ -2228,7 +2237,7 @@ OpenMPIRBuilder::createTaskgroup(const LocationDescription &Loc,
   // Emit the @__kmpc_end_taskgroup runtime call to end the taskgroup
   Function *EndTaskgroupFn =
       getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_end_taskgroup);
-  Builder.CreateCall(EndTaskgroupFn, {Ident, ThreadID});
+  createRuntimeFunctionCall(EndTaskgroupFn, {Ident, ThreadID});
 
   return Builder.saveIP();
 }
@@ -2384,14 +2393,14 @@ static OpenMPIRBuilder::InsertPointTy getInsertPointAfterInstr(Instruction *I) {
 }
 
 Value *OpenMPIRBuilder::getGPUThreadID() {
-  return Builder.CreateCall(
+  return createRuntimeFunctionCall(
       getOrCreateRuntimeFunction(M,
                                  OMPRTL___kmpc_get_hardware_thread_id_in_block),
       {});
 }
 
 Value *OpenMPIRBuilder::getGPUWarpSize() {
-  return Builder.CreateCall(
+  return createRuntimeFunctionCall(
       getOrCreateRuntimeFunction(M, OMPRTL___kmpc_get_warp_size), {});
 }
 
@@ -2450,7 +2459,7 @@ Value *OpenMPIRBuilder::createRuntimeShuffleFunction(InsertPointTy AllocaIP,
   Value *WarpSizeCast =
       Builder.CreateIntCast(WarpSize, Builder.getInt16Ty(), /*isSigned=*/true);
   Value *ShuffleCall =
-      Builder.CreateCall(ShuffleFunc, {ElemCast, Offset, WarpSizeCast});
+      createRuntimeFunctionCall(ShuffleFunc, {ElemCast, Offset, WarpSizeCast});
   return castValueToType(AllocaIP, ShuffleCall, CastTy);
 }
 
@@ -3016,7 +3025,7 @@ Function *OpenMPIRBuilder::emitShuffleAndReduceFunction(
       ReduceList, Builder.getPtrTy());
   Value *RemoteReduceListPtr = Builder.CreatePointerBitCastOrAddrSpaceCast(
       RemoteListAddrCast, Builder.getPtrTy());
-  Builder.CreateCall(ReduceFn, {LocalReduceListPtr, RemoteReduceListPtr})
+  createRuntimeFunctionCall(ReduceFn, {LocalReduceListPtr, RemoteReduceListPtr})
       ->addFnAttr(Attribute::NoUnwind);
   Builder.CreateBr(MergeBB);
 
@@ -3237,7 +3246,7 @@ Function *OpenMPIRBuilder::emitListToGlobalReduceFunction(
   // Call reduce_function(GlobalReduceList, ReduceList)
   Value *ReduceList =
       Builder.CreateLoad(Builder.getPtrTy(), ReduceListArgAddrCast);
-  Builder.CreateCall(ReduceFn, {LocalReduceListAddrCast, ReduceList})
+  createRuntimeFunctionCall(ReduceFn, {LocalReduceListAddrCast, ReduceList})
       ->addFnAttr(Attribute::NoUnwind);
   Builder.CreateRetVoid();
   Builder.restoreIP(OldIP);
@@ -3428,7 +3437,7 @@ Function *OpenMPIRBuilder::emitGlobalToListReduceFunction(
   // Call reduce_function(ReduceList, GlobalReduceList)
   Value *ReduceList =
       Builder.CreateLoad(Builder.getPtrTy(), ReduceListArgAddrCast);
-  Builder.CreateCall(ReduceFn, {ReduceList, ReductionList})
+  createRuntimeFunctionCall(ReduceFn, {ReduceList, ReductionList})
       ->addFnAttr(Attribute::NoUnwind);
   Builder.CreateRetVoid();
   Builder.restoreIP(OldIP);
@@ -3682,7 +3691,7 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createReductionsGPU(
                      WcFuncCast};
     Function *Pv2Ptr = getOrCreateRuntimeFunctionPtr(
         RuntimeFunction::OMPRTL___kmpc_nvptx_parallel_reduce_nowait_v2);
-    Res = Builder.CreateCall(Pv2Ptr, Args);
+    Res = createRuntimeFunctionCall(Pv2Ptr, Args);
   } else {
     CodeGenIP = Builder.saveIP();
     StructType *ReductionsBufferTy = StructType::create(
@@ -3699,7 +3708,7 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createReductionsGPU(
         ReductionInfos, ReductionFunc, ReductionsBufferTy, FuncAttrs);
     Builder.restoreIP(CodeGenIP);
 
-    Value *KernelTeamsReductionPtr = Builder.CreateCall(
+    Value *KernelTeamsReductionPtr = createRuntimeFunctionCall(
         RedFixedBuferFn, {}, "_openmp_teams_reductions_buffer_$_$ptr");
 
     Value *Args3[] = {SrcLocInfo,
@@ -3716,7 +3725,7 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createReductionsGPU(
 
     Function *TeamsReduceFn = getOrCreateRuntimeFunctionPtr(
         RuntimeFunction::OMPRTL___kmpc_nvptx_teams_reduce_nowait_v2);
-    Res = Builder.CreateCall(TeamsReduceFn, Args3);
+    Res = createRuntimeFunctionCall(TeamsReduceFn, Args3);
   }
 
   // 5. Build if (res == 1)
@@ -3921,10 +3930,10 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createReductions(
       IsNoWait ? RuntimeFunction::OMPRTL___kmpc_reduce_nowait
                : RuntimeFunction::OMPRTL___kmpc_reduce);
   CallInst *ReduceCall =
-      Builder.CreateCall(ReduceFunc,
-                         {Ident, ThreadId, NumVariables, RedArraySize, RedArray,
-                          ReductionFunc, Lock},
-                         "reduce");
+      createRuntimeFunctionCall(ReduceFunc,
+                                {Ident, ThreadId, NumVariables, RedArraySize,
+                                 RedArray, ReductionFunc, Lock},
+                                "reduce");
 
   // Create final reduction entry blocks for the atomic and non-atomic case.
   // Emit IR that dispatches control flow to one of the blocks based on the
@@ -3971,7 +3980,7 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenM...
[truncated]

@sarnex sarnex requested review from jhuber6 and tblah November 19, 2025 15:12
jhuber6
jhuber6 approved these changes Nov 19, 2025
View reviewed changes
Copy link
Contributor

@jhuber6 jhuber6 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Seems reasonable to me, maybe some of the Flang people have opinions

@sarnex sarnex merged commit 4538818 into llvm:main Nov 21, 2025
17 checks passed
aadeshps-mcw pushed a commit to aadeshps-mcw/llvm-project that referenced this pull request Nov 26, 2025
@sarnex @aadeshps-mcw
[OpenMP][OMPIRBuilder] Use runtime CC for runtime calls (llvm#168608)
ea2d28d 
Some targets have a specific calling convention that should be used for
generated calls to runtime functions.

Pass that down and use it.

Signed-off-by: Nick Sarnie <[email protected]>
Priyanshu3820 pushed a commit to Priyanshu3820/llvm-project that referenced this pull request Nov 26, 2025
@sarnex @Priyanshu3820
[OpenMP][OMPIRBuilder] Use runtime CC for runtime calls (llvm#168608)
323089b 
Some targets have a specific calling convention that should be used for
generated calls to runtime functions.

Pass that down and use it.

Signed-off-by: Nick Sarnie <[email protected]>
augusto2112 pushed a commit to augusto2112/llvm-project that referenced this pull request Dec 3, 2025
@sarnex @augusto2112
[OpenMP][OMPIRBuilder] Use runtime CC for runtime calls (llvm#168608)
e429d7a 
Some targets have a specific calling convention that should be used for
generated calls to runtime functions.

Pass that down and use it.

Signed-off-by: Nick Sarnie <[email protected]>
@Saieiei Saieiei mentioned this pull request Dec 5, 2025
Merged
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@jhuber6 jhuber6 jhuber6 approved these changes

@tblah tblah Awaiting requested review from tblah

Assignees

No one assigned

Labels

clang:codegen IR generation bugs: mangling, exceptions, etc. clang:openmp OpenMP related changes to Clang flang:openmp mlir:llvm mlir

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

3 participants

@sarnex @llvmbot @jhuber6