[OpenCL] Emit enqueued block as kernel
In OpenCL the kernel function and non-kernel function has different calling conventions.
For certain targets they have different argument ABIs. Also kernels have special function
attributes and metadata for runtime to launch them.
The blocks passed to enqueue_kernel is supposed to be executed as kernels. As such,
the block invoke function should be emitted as kernel with proper calling convention and
argument ABI.
This patch emits enqueued block as kernel. If a block is both called directly and passed
to enqueue_kernel, separate functions will be generated.
Differential Revision: https://reviews.llvm.org/D38134
llvm-svn: 315804
diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index e8d2c59..1de7c6e 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -2779,12 +2779,17 @@
// The most basic form of the call with parameters:
// queue_t, kernel_enqueue_flags_t, ndrange_t, block(void)
Name = "__enqueue_kernel_basic";
- llvm::Type *ArgTys[] = {QueueTy, Int32Ty, RangeTy, GenericVoidPtrTy};
+ llvm::Type *ArgTys[] = {QueueTy, Int32Ty, RangeTy, GenericVoidPtrTy,
+ GenericVoidPtrTy};
llvm::FunctionType *FTy = llvm::FunctionType::get(
- Int32Ty, llvm::ArrayRef<llvm::Type *>(ArgTys, 4), false);
+ Int32Ty, llvm::ArrayRef<llvm::Type *>(ArgTys), false);
- llvm::Value *Block = Builder.CreatePointerCast(
- EmitScalarExpr(E->getArg(3)), GenericVoidPtrTy);
+ auto Info =
+ CGM.getOpenCLRuntime().emitOpenCLEnqueuedBlock(*this, E->getArg(3));
+ llvm::Value *Kernel =
+ Builder.CreatePointerCast(Info.Kernel, GenericVoidPtrTy);
+ llvm::Value *Block =
+ Builder.CreatePointerCast(Info.BlockArg, GenericVoidPtrTy);
AttrBuilder B;
B.addAttribute(Attribute::ByVal);
@@ -2793,7 +2798,7 @@
auto RTCall =
Builder.CreateCall(CGM.CreateRuntimeFunction(FTy, Name, ByValAttrSet),
- {Queue, Flags, Range, Block});
+ {Queue, Flags, Range, Kernel, Block});
RTCall->setAttributes(ByValAttrSet);
return RValue::get(RTCall);
}
@@ -2825,21 +2830,23 @@
if (E->getArg(3)->getType()->isBlockPointerType()) {
// No events passed, but has variadic arguments.
Name = "__enqueue_kernel_vaargs";
- auto *Block = Builder.CreatePointerCast(EmitScalarExpr(E->getArg(3)),
- GenericVoidPtrTy);
+ auto Info =
+ CGM.getOpenCLRuntime().emitOpenCLEnqueuedBlock(*this, E->getArg(3));
+ llvm::Value *Kernel =
+ Builder.CreatePointerCast(Info.Kernel, GenericVoidPtrTy);
+ auto *Block = Builder.CreatePointerCast(Info.BlockArg, GenericVoidPtrTy);
auto *PtrToSizeArray = CreateArrayForSizeVar(4);
// Create a vector of the arguments, as well as a constant value to
// express to the runtime the number of variadic arguments.
- std::vector<llvm::Value *> Args = {Queue,
- Flags,
- Range,
- Block,
- ConstantInt::get(IntTy, NumArgs - 4),
- PtrToSizeArray};
- std::vector<llvm::Type *> ArgTys = {QueueTy, IntTy,
- RangeTy, GenericVoidPtrTy,
- IntTy, PtrToSizeArray->getType()};
+ std::vector<llvm::Value *> Args = {
+ Queue, Flags, Range,
+ Kernel, Block, ConstantInt::get(IntTy, NumArgs - 4),
+ PtrToSizeArray};
+ std::vector<llvm::Type *> ArgTys = {
+ QueueTy, IntTy, RangeTy,
+ GenericVoidPtrTy, GenericVoidPtrTy, IntTy,
+ PtrToSizeArray->getType()};
llvm::FunctionType *FTy = llvm::FunctionType::get(
Int32Ty, llvm::ArrayRef<llvm::Type *>(ArgTys), false);
@@ -2863,15 +2870,19 @@
// Convert to generic address space.
EventList = Builder.CreatePointerCast(EventList, EventPtrTy);
ClkEvent = Builder.CreatePointerCast(ClkEvent, EventPtrTy);
- llvm::Value *Block = Builder.CreatePointerCast(
- EmitScalarExpr(E->getArg(6)), GenericVoidPtrTy);
+ auto Info =
+ CGM.getOpenCLRuntime().emitOpenCLEnqueuedBlock(*this, E->getArg(6));
+ llvm::Value *Kernel =
+ Builder.CreatePointerCast(Info.Kernel, GenericVoidPtrTy);
+ llvm::Value *Block =
+ Builder.CreatePointerCast(Info.BlockArg, GenericVoidPtrTy);
std::vector<llvm::Type *> ArgTys = {
- QueueTy, Int32Ty, RangeTy, Int32Ty,
- EventPtrTy, EventPtrTy, GenericVoidPtrTy};
+ QueueTy, Int32Ty, RangeTy, Int32Ty,
+ EventPtrTy, EventPtrTy, GenericVoidPtrTy, GenericVoidPtrTy};
- std::vector<llvm::Value *> Args = {Queue, Flags, Range, NumEvents,
- EventList, ClkEvent, Block};
+ std::vector<llvm::Value *> Args = {Queue, Flags, Range, NumEvents,
+ EventList, ClkEvent, Kernel, Block};
if (NumArgs == 7) {
// Has events but no variadics.
@@ -2905,24 +2916,30 @@
case Builtin::BIget_kernel_work_group_size: {
llvm::Type *GenericVoidPtrTy = Builder.getInt8PtrTy(
getContext().getTargetAddressSpace(LangAS::opencl_generic));
- Value *Arg = EmitScalarExpr(E->getArg(0));
- Arg = Builder.CreatePointerCast(Arg, GenericVoidPtrTy);
+ auto Info =
+ CGM.getOpenCLRuntime().emitOpenCLEnqueuedBlock(*this, E->getArg(0));
+ Value *Kernel = Builder.CreatePointerCast(Info.Kernel, GenericVoidPtrTy);
+ Value *Arg = Builder.CreatePointerCast(Info.BlockArg, GenericVoidPtrTy);
return RValue::get(Builder.CreateCall(
CGM.CreateRuntimeFunction(
- llvm::FunctionType::get(IntTy, GenericVoidPtrTy, false),
+ llvm::FunctionType::get(IntTy, {GenericVoidPtrTy, GenericVoidPtrTy},
+ false),
"__get_kernel_work_group_size_impl"),
- Arg));
+ {Kernel, Arg}));
}
case Builtin::BIget_kernel_preferred_work_group_size_multiple: {
llvm::Type *GenericVoidPtrTy = Builder.getInt8PtrTy(
getContext().getTargetAddressSpace(LangAS::opencl_generic));
- Value *Arg = EmitScalarExpr(E->getArg(0));
- Arg = Builder.CreatePointerCast(Arg, GenericVoidPtrTy);
+ auto Info =
+ CGM.getOpenCLRuntime().emitOpenCLEnqueuedBlock(*this, E->getArg(0));
+ Value *Kernel = Builder.CreatePointerCast(Info.Kernel, GenericVoidPtrTy);
+ Value *Arg = Builder.CreatePointerCast(Info.BlockArg, GenericVoidPtrTy);
return RValue::get(Builder.CreateCall(
CGM.CreateRuntimeFunction(
- llvm::FunctionType::get(IntTy, GenericVoidPtrTy, false),
+ llvm::FunctionType::get(IntTy, {GenericVoidPtrTy, GenericVoidPtrTy},
+ false),
"__get_kernel_preferred_work_group_multiple_impl"),
- Arg));
+ {Kernel, Arg}));
}
case Builtin::BIget_kernel_max_sub_group_size_for_ndrange:
case Builtin::BIget_kernel_sub_group_count_for_ndrange: {
@@ -2930,8 +2947,10 @@
getContext().getTargetAddressSpace(LangAS::opencl_generic));
LValue NDRangeL = EmitAggExprToLValue(E->getArg(0));
llvm::Value *NDRange = NDRangeL.getAddress().getPointer();
- Value *Block = EmitScalarExpr(E->getArg(1));
- Block = Builder.CreatePointerCast(Block, GenericVoidPtrTy);
+ auto Info =
+ CGM.getOpenCLRuntime().emitOpenCLEnqueuedBlock(*this, E->getArg(1));
+ Value *Kernel = Builder.CreatePointerCast(Info.Kernel, GenericVoidPtrTy);
+ Value *Block = Builder.CreatePointerCast(Info.BlockArg, GenericVoidPtrTy);
const char *Name =
BuiltinID == Builtin::BIget_kernel_max_sub_group_size_for_ndrange
? "__get_kernel_max_sub_group_size_for_ndrange_impl"
@@ -2939,9 +2958,10 @@
return RValue::get(Builder.CreateCall(
CGM.CreateRuntimeFunction(
llvm::FunctionType::get(
- IntTy, {NDRange->getType(), GenericVoidPtrTy}, false),
+ IntTy, {NDRange->getType(), GenericVoidPtrTy, GenericVoidPtrTy},
+ false),
Name),
- {NDRange, Block}));
+ {NDRange, Kernel, Block}));
}
case Builtin::BI__builtin_store_half: