CGOpenCLRuntime.cpp
7.08 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
//===----- 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;
#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;
}
llvm::PointerType *CGOpenCLRuntime::getSamplerType(const Type *T) {
if (!SamplerTy)
SamplerTy = llvm::PointerType::get(llvm::StructType::create(
CGM.getLLVMContext(), "opencl.sampler_t"),
CGM.getContext().getTargetAddressSpace(
CGM.getContext().getOpenCLTypeAddrSpace(T)));
return SamplerTy;
}
llvm::Value *CGOpenCLRuntime::getPipeElemSize(const Expr *PipeArg) {
const PipeType *PipeTy = PipeArg->getType()->castAs<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()->castAs<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];
}