RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-25 04:02:41 +01:00
Code Issues Projects Releases Wiki Activity

[openmp] Apply code change from D109500

Browse Source 
(cherry picked from commit 71052ea1e3c63b7209731fdc1726d10640d97480)
This commit is contained in:
Jon Chesterfield 2021-09-13 18:00:15 +01:00 committed by Tom Stellard
parent fca643847d
commit 47e63cb60e
1 changed files with 19 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

23
lib/Transforms/IPO/OpenMPOpt.cpp
View File

@ -1996,7 +1996,8 @@ bool OpenMPOpt::rewriteDeviceCodeStateMachine() {
UndefValue::get(Int8Ty), F->getName() + ".ID"); UndefValue::get(Int8Ty), F->getName() + ".ID");
for (Use *U : ToBeReplacedStateMachineUses) for (Use *U : ToBeReplacedStateMachineUses)
U->set(ConstantExpr::getBitCast(ID, U->get()->getType())); U->set(ConstantExpr::getPointerBitCastOrAddrSpaceCast(
ID, U->get()->getType()));
++NumOpenMPParallelRegionsReplacedInGPUStateMachine; ++NumOpenMPParallelRegionsReplacedInGPUStateMachine;
@ -3183,10 +3184,14 @@ struct AAKernelInfoFunction : AAKernelInfo {
IsWorker->setDebugLoc(DLoc); IsWorker->setDebugLoc(DLoc);
BranchInst::Create(StateMachineBeginBB, UserCodeEntryBB, IsWorker, InitBB); BranchInst::Create(StateMachineBeginBB, UserCodeEntryBB, IsWorker, InitBB);
Module &M = *Kernel->getParent();
// Create local storage for the work function pointer. // Create local storage for the work function pointer.
const DataLayout &DL = M.getDataLayout();
Type *VoidPtrTy = Type::getInt8PtrTy(Ctx); Type *VoidPtrTy = Type::getInt8PtrTy(Ctx);
AllocaInst *WorkFnAI = new AllocaInst(VoidPtrTy, 0, "worker.work_fn.addr", Instruction *WorkFnAI =
&Kernel->getEntryBlock().front()); new AllocaInst(VoidPtrTy, DL.getAllocaAddrSpace(), nullptr,
"worker.work_fn.addr", &Kernel->getEntryBlock().front());
WorkFnAI->setDebugLoc(DLoc); WorkFnAI->setDebugLoc(DLoc);
auto &OMPInfoCache = static_cast<OMPInformationCache &>(A.getInfoCache()); auto &OMPInfoCache = static_cast<OMPInformationCache &>(A.getInfoCache());
@ -3199,13 +3204,23 @@ struct AAKernelInfoFunction : AAKernelInfo {
Value *Ident = KernelInitCB->getArgOperand(0); Value *Ident = KernelInitCB->getArgOperand(0);
Value *GTid = KernelInitCB; Value *GTid = KernelInitCB;
Module &M = *Kernel->getParent();
FunctionCallee BarrierFn = FunctionCallee BarrierFn =
OMPInfoCache.OMPBuilder.getOrCreateRuntimeFunction( OMPInfoCache.OMPBuilder.getOrCreateRuntimeFunction(
M, OMPRTL___kmpc_barrier_simple_spmd); M, OMPRTL___kmpc_barrier_simple_spmd);
CallInst::Create(BarrierFn, {Ident, GTid}, "", StateMachineBeginBB) CallInst::Create(BarrierFn, {Ident, GTid}, "", StateMachineBeginBB)
->setDebugLoc(DLoc); ->setDebugLoc(DLoc);
if (WorkFnAI->getType()->getPointerAddressSpace() !=
(unsigned int)AddressSpace::Generic) {
WorkFnAI = new AddrSpaceCastInst(
WorkFnAI,
PointerType::getWithSamePointeeType(
cast<PointerType>(WorkFnAI->getType()),
(unsigned int)AddressSpace::Generic),
WorkFnAI->getName() + ".generic", StateMachineBeginBB);
WorkFnAI->setDebugLoc(DLoc);
}
FunctionCallee KernelParallelFn = FunctionCallee KernelParallelFn =
OMPInfoCache.OMPBuilder.getOrCreateRuntimeFunction( OMPInfoCache.OMPBuilder.getOrCreateRuntimeFunction(
M, OMPRTL___kmpc_kernel_parallel); M, OMPRTL___kmpc_kernel_parallel);