llvm-project/clang/lib/CodeGen/CGOpenCLRuntime.cpp

187 lines
7.1 KiB
C++
Raw Normal View History

//===----- CGOpenCLRuntime.cpp - Interface to OpenCL Runtimes -------------===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
//
// This provides an abstract class for OpenCL code generation. Concrete
// subclasses of this implement code generation for specific OpenCL
// runtime libraries.
//
//===----------------------------------------------------------------------===//
#include "CGOpenCLRuntime.h"
#include "CodeGenFunction.h"
#include "TargetInfo.h"
#include "clang/CodeGen/ConstantInitBuilder.h"
#include "llvm/IR/DerivedTypes.h"
#include "llvm/IR/GlobalValue.h"
#include <assert.h>
using namespace clang;
using namespace CodeGen;
CGOpenCLRuntime::~CGOpenCLRuntime() {}
void CGOpenCLRuntime::EmitWorkGroupLocalVarDecl(CodeGenFunction &CGF,
const VarDecl &D) {
return CGF.EmitStaticVarDecl(D, llvm::GlobalValue::InternalLinkage);
}
llvm::Type *CGOpenCLRuntime::convertOpenCLSpecificType(const Type *T) {
assert(T->isOpenCLSpecificType() &&
"Not an OpenCL specific type!");
llvm::LLVMContext& Ctx = CGM.getLLVMContext();
uint32_t AddrSpc = CGM.getContext().getTargetAddressSpace(
CGM.getContext().getOpenCLTypeAddrSpace(T));
switch (cast<BuiltinType>(T)->getKind()) {
default:
llvm_unreachable("Unexpected opencl builtin type!");
return nullptr;
[OpenCL] Complete image types support. I. Current implementation of images is not conformant to spec in the following points: 1. It makes no distinction with respect to access qualifiers and therefore allows to use images with different access type interchangeably. The following code would compile just fine: void write_image(write_only image2d_t img); kernel void foo(read_only image2d_t img) { write_image(img); } // Accepted code which is disallowed according to s6.13.14. 2. It discards access qualifier on generated code, which leads to generated code for the above example: call void @write_image(%opencl.image2d_t* %img); In OpenCL2.0 however we can have different calls into write_image with read_only and wite_only images. Also generally following compiler steps have no easy way to take different path depending on the image access: linking to the right implementation of image types, performing IR opts and backend codegen differently. 3. Image types are language keywords and can't be redeclared s6.1.9, which can happen currently as they are just typedef names. 4. Default access qualifier read_only is to be added if not provided explicitly. II. This patch corrects the above points as follows: 1. All images are encapsulated into a separate .def file that is inserted in different points where image handling is required. This avoid a lot of code repetition as all images are handled the same way in the code with no distinction of their exact type. 2. The Cartesian product of image types and image access qualifiers is added to the builtin types. This simplifies a lot handling of access type mismatch as no operations are allowed by default on distinct Builtin types. Also spec intended access qualifier as special type qualifier that are combined with an image type to form a distinct type (see statement above - images can't be created w/o access qualifiers). 3. Improves testing of images in Clang. Author: Anastasia Stulova Reviewers: bader, mgrang. Subscribers: pxli168, pekka.jaaskelainen, yaxunl. Differential Revision: http://reviews.llvm.org/D17821 llvm-svn: 265783
2016-04-08 21:40:33 +08:00
#define IMAGE_TYPE(ImgType, Id, SingletonId, Access, Suffix) \
case BuiltinType::Id: \
return llvm::PointerType::get( \
llvm::StructType::create(Ctx, "opencl." #ImgType "_" #Suffix "_t"), \
AddrSpc);
#include "clang/Basic/OpenCLImageTypes.def"
case BuiltinType::OCLSampler:
return getSamplerType(T);
case BuiltinType::OCLEvent:
return llvm::PointerType::get(
llvm::StructType::create(Ctx, "opencl.event_t"), AddrSpc);
case BuiltinType::OCLClkEvent:
return llvm::PointerType::get(
llvm::StructType::create(Ctx, "opencl.clk_event_t"), AddrSpc);
case BuiltinType::OCLQueue:
return llvm::PointerType::get(
llvm::StructType::create(Ctx, "opencl.queue_t"), AddrSpc);
case BuiltinType::OCLReserveID:
return llvm::PointerType::get(
llvm::StructType::create(Ctx, "opencl.reserve_id_t"), AddrSpc);
#define EXT_OPAQUE_TYPE(ExtType, Id, Ext) \
case BuiltinType::Id: \
return llvm::PointerType::get( \
llvm::StructType::create(Ctx, "opencl." #ExtType), AddrSpc);
#include "clang/Basic/OpenCLExtensionTypes.def"
}
}
llvm::Type *CGOpenCLRuntime::getPipeType(const PipeType *T) {
if (T->isReadOnly())
return getPipeType(T, "opencl.pipe_ro_t", PipeROTy);
else
return getPipeType(T, "opencl.pipe_wo_t", PipeWOTy);
}
llvm::Type *CGOpenCLRuntime::getPipeType(const PipeType *T, StringRef Name,
llvm::Type *&PipeTy) {
if (!PipeTy)
PipeTy = llvm::PointerType::get(llvm::StructType::create(
CGM.getLLVMContext(), Name),
CGM.getContext().getTargetAddressSpace(
CGM.getContext().getOpenCLTypeAddrSpace(T)));
return PipeTy;
}
2016-07-29 03:26:30 +08:00
llvm::PointerType *CGOpenCLRuntime::getSamplerType(const Type *T) {
2016-07-29 03:26:30 +08:00
if (!SamplerTy)
SamplerTy = llvm::PointerType::get(llvm::StructType::create(
CGM.getLLVMContext(), "opencl.sampler_t"),
CGM.getContext().getTargetAddressSpace(
CGM.getContext().getOpenCLTypeAddrSpace(T)));
2016-07-29 03:26:30 +08:00
return SamplerTy;
}
llvm::Value *CGOpenCLRuntime::getPipeElemSize(const Expr *PipeArg) {
const PipeType *PipeTy = PipeArg->getType()->getAs<PipeType>();
// The type of the last (implicit) argument to be passed.
llvm::Type *Int32Ty = llvm::IntegerType::getInt32Ty(CGM.getLLVMContext());
unsigned TypeSize = CGM.getContext()
.getTypeSizeInChars(PipeTy->getElementType())
.getQuantity();
return llvm::ConstantInt::get(Int32Ty, TypeSize, false);
}
llvm::Value *CGOpenCLRuntime::getPipeElemAlign(const Expr *PipeArg) {
const PipeType *PipeTy = PipeArg->getType()->getAs<PipeType>();
// The type of the last (implicit) argument to be passed.
llvm::Type *Int32Ty = llvm::IntegerType::getInt32Ty(CGM.getLLVMContext());
unsigned TypeSize = CGM.getContext()
.getTypeAlignInChars(PipeTy->getElementType())
.getQuantity();
return llvm::ConstantInt::get(Int32Ty, TypeSize, false);
}
llvm::PointerType *CGOpenCLRuntime::getGenericVoidPointerType() {
assert(CGM.getLangOpts().OpenCL);
return llvm::IntegerType::getInt8PtrTy(
CGM.getLLVMContext(),
CGM.getContext().getTargetAddressSpace(LangAS::opencl_generic));
}
// Get the block literal from an expression derived from the block expression.
// OpenCL v2.0 s6.12.5:
// Block variable declarations are implicitly qualified with const. Therefore
// all block variables must be initialized at declaration time and may not be
// reassigned.
static const BlockExpr *getBlockExpr(const Expr *E) {
const Expr *Prev = nullptr; // to make sure we do not stuck in infinite loop.
while(!isa<BlockExpr>(E) && E != Prev) {
Prev = E;
E = E->IgnoreCasts();
if (auto DR = dyn_cast<DeclRefExpr>(E)) {
E = cast<VarDecl>(DR->getDecl())->getInit();
}
}
return cast<BlockExpr>(E);
}
/// Record emitted llvm invoke function and llvm block literal for the
/// corresponding block expression.
void CGOpenCLRuntime::recordBlockInfo(const BlockExpr *E,
llvm::Function *InvokeF,
llvm::Value *Block) {
assert(EnqueuedBlockMap.find(E) == EnqueuedBlockMap.end() &&
"Block expression emitted twice");
assert(isa<llvm::Function>(InvokeF) && "Invalid invoke function");
assert(Block->getType()->isPointerTy() && "Invalid block literal type");
EnqueuedBlockMap[E].InvokeFunc = InvokeF;
EnqueuedBlockMap[E].BlockArg = Block;
EnqueuedBlockMap[E].Kernel = nullptr;
}
llvm::Function *CGOpenCLRuntime::getInvokeFunction(const Expr *E) {
return EnqueuedBlockMap[getBlockExpr(E)].InvokeFunc;
}
CGOpenCLRuntime::EnqueuedBlockInfo
CGOpenCLRuntime::emitOpenCLEnqueuedBlock(CodeGenFunction &CGF, const Expr *E) {
CGF.EmitScalarExpr(E);
// The block literal may be assigned to a const variable. Chasing down
// to get the block literal.
const BlockExpr *Block = getBlockExpr(E);
assert(EnqueuedBlockMap.find(Block) != EnqueuedBlockMap.end() &&
"Block expression not emitted");
// Do not emit the block wrapper again if it has been emitted.
if (EnqueuedBlockMap[Block].Kernel) {
return EnqueuedBlockMap[Block];
}
auto *F = CGF.getTargetHooks().createEnqueuedBlockKernel(
CGF, EnqueuedBlockMap[Block].InvokeFunc,
EnqueuedBlockMap[Block].BlockArg->stripPointerCasts());
// The common part of the post-processing of the kernel goes here.
F->addFnAttr(llvm::Attribute::NoUnwind);
F->setCallingConv(
CGF.getTypes().ClangCallConvToLLVMCallConv(CallingConv::CC_OpenCLKernel));
EnqueuedBlockMap[Block].Kernel = F;
return EnqueuedBlockMap[Block];
}