Differential D113623 Diff 393274 clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp

Changeset View

Standalone View

clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp

Show First 20 Lines • Show All 1,215 Lines • ▼ Show 20 Lines	if (getExecutionMode() == CGOpenMPRuntimeGPU::EM_SPMD)
return;		return;

CGOpenMPRuntime::emitProcBindClause(CGF, ProcBind, Loc);		CGOpenMPRuntime::emitProcBindClause(CGF, ProcBind, Loc);
}		}

void CGOpenMPRuntimeGPU::emitNumThreadsClause(CodeGenFunction &CGF,		void CGOpenMPRuntimeGPU::emitNumThreadsClause(CodeGenFunction &CGF,
llvm::Value *NumThreads,		llvm::Value *NumThreads,
SourceLocation Loc) {		SourceLocation Loc) {
// Do nothing in case of SPMD mode and L0 parallel.		// Nothing to do.
if (getExecutionMode() == CGOpenMPRuntimeGPU::EM_SPMD)
return;

CGOpenMPRuntime::emitNumThreadsClause(CGF, NumThreads, Loc);
}		}

void CGOpenMPRuntimeGPU::emitNumTeamsClause(CodeGenFunction &CGF,		void CGOpenMPRuntimeGPU::emitNumTeamsClause(CodeGenFunction &CGF,
const Expr *NumTeams,		const Expr *NumTeams,
const Expr *ThreadLimit,		const Expr *ThreadLimit,
SourceLocation Loc) {}		SourceLocation Loc) {}

llvm::Function *CGOpenMPRuntimeGPU::emitParallelOutlinedFunction(		llvm::Function *CGOpenMPRuntimeGPU::emitParallelOutlinedFunction(
▲ Show 20 Lines • Show All 268 Lines • ▼ Show 20 Lines	void CGOpenMPRuntimeGPU::emitTeamsCall(CodeGenFunction &CGF,
OutlinedFnArgs.append(CapturedVars.begin(), CapturedVars.end());		OutlinedFnArgs.append(CapturedVars.begin(), CapturedVars.end());
emitOutlinedFunctionCall(CGF, Loc, OutlinedFn, OutlinedFnArgs);		emitOutlinedFunctionCall(CGF, Loc, OutlinedFn, OutlinedFnArgs);
}		}

void CGOpenMPRuntimeGPU::emitParallelCall(CodeGenFunction &CGF,		void CGOpenMPRuntimeGPU::emitParallelCall(CodeGenFunction &CGF,
SourceLocation Loc,		SourceLocation Loc,
llvm::Function *OutlinedFn,		llvm::Function *OutlinedFn,
ArrayRef<llvm::Value *> CapturedVars,		ArrayRef<llvm::Value *> CapturedVars,
const Expr *IfCond) {		const Expr *IfCond,
		llvm::Value *NumThreads) {
if (!CGF.HaveInsertPoint())		if (!CGF.HaveInsertPoint())
return;		return;

auto &&ParallelGen = [this, Loc, OutlinedFn, CapturedVars,		auto &&ParallelGen = [this, Loc, OutlinedFn, CapturedVars, IfCond,
IfCond](CodeGenFunction &CGF, PrePostActionTy &Action) {		NumThreads](CodeGenFunction &CGF,
		PrePostActionTy &Action) {
CGBuilderTy &Bld = CGF.Builder;		CGBuilderTy &Bld = CGF.Builder;
		llvm::Value *NumThreadsVal = NumThreads;
llvm::Function *WFn = WrapperFunctionsMap[OutlinedFn];		llvm::Function *WFn = WrapperFunctionsMap[OutlinedFn];
llvm::Value *ID = llvm::ConstantPointerNull::get(CGM.Int8PtrTy);		llvm::Value *ID = llvm::ConstantPointerNull::get(CGM.Int8PtrTy);
if (WFn)		if (WFn)
ID = Bld.CreateBitOrPointerCast(WFn, CGM.Int8PtrTy);		ID = Bld.CreateBitOrPointerCast(WFn, CGM.Int8PtrTy);
llvm::Value *FnPtr = Bld.CreateBitOrPointerCast(OutlinedFn, CGM.Int8PtrTy);		llvm::Value *FnPtr = Bld.CreateBitOrPointerCast(OutlinedFn, CGM.Int8PtrTy);

// Create a private scope that will globalize the arguments		// Create a private scope that will globalize the arguments
// passed from the outside of the target region.		// passed from the outside of the target region.
Show All 23 Lines	auto &&ParallelGen = [this, Loc, OutlinedFn, CapturedVars, IfCond,

llvm::Value *IfCondVal = nullptr;		llvm::Value *IfCondVal = nullptr;
if (IfCond)		if (IfCond)
IfCondVal = Bld.CreateIntCast(CGF.EvaluateExprAsBool(IfCond), CGF.Int32Ty,		IfCondVal = Bld.CreateIntCast(CGF.EvaluateExprAsBool(IfCond), CGF.Int32Ty,
/* isSigned */ false);		/* isSigned */ false);
else		else
IfCondVal = llvm::ConstantInt::get(CGF.Int32Ty, 1);		IfCondVal = llvm::ConstantInt::get(CGF.Int32Ty, 1);

		if (!NumThreadsVal)
		NumThreadsVal = llvm::ConstantInt::get(CGF.Int32Ty, -1);
		else
		NumThreadsVal = Bld.CreateZExtOrTrunc(NumThreadsVal, CGF.Int32Ty),

assert(IfCondVal && "Expected a value");		assert(IfCondVal && "Expected a value");
llvm::Value *RTLoc = emitUpdateLocation(CGF, Loc);		llvm::Value *RTLoc = emitUpdateLocation(CGF, Loc);
llvm::Value *Args[] = {		llvm::Value *Args[] = {
RTLoc,		RTLoc,
getThreadID(CGF, Loc),		getThreadID(CGF, Loc),
IfCondVal,		IfCondVal,
llvm::ConstantInt::get(CGF.Int32Ty, -1),		NumThreadsVal,
llvm::ConstantInt::get(CGF.Int32Ty, -1),		llvm::ConstantInt::get(CGF.Int32Ty, -1),
FnPtr,		FnPtr,
ID,		ID,
Bld.CreateBitOrPointerCast(CapturedVarsAddrs.getPointer(),		Bld.CreateBitOrPointerCast(CapturedVarsAddrs.getPointer(),
CGF.VoidPtrPtrTy),		CGF.VoidPtrPtrTy),
llvm::ConstantInt::get(CGM.SizeTy, CapturedVars.size())};		llvm::ConstantInt::get(CGM.SizeTy, CapturedVars.size())};
CGF.EmitRuntimeCall(OMPBuilder.getOrCreateRuntimeFunction(		CGF.EmitRuntimeCall(OMPBuilder.getOrCreateRuntimeFunction(
CGM.getModule(), OMPRTL___kmpc_parallel_51),		CGM.getModule(), OMPRTL___kmpc_parallel_51),
▲ Show 20 Lines • Show All 2,402 Lines • Show Last 20 Lines