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
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
|
// slang-emit-torch.cpp
#include "slang-emit-torch.h"
#include "../core/slang-writer.h"
#include "slang-emit-source-writer.h"
#include "slang-mangled-lexer.h"
#include <assert.h>
namespace Slang
{
void emitTorchScalarTypeName(SourceWriter* m_writer, IRInst* type)
{
m_writer->emit("torch::");
// Get the element type of the tensor.
auto instType = as<IRTorchTensorType>(type)->getOperand(0);
// If instType is a vector type, then we need to get the element type.
if (auto vectorType = as<IRVectorType>(instType))
{
instType = vectorType->getElementType();
}
switch (instType->getOp())
{
case kIROp_FloatType: m_writer->emit("kFloat32"); break;
case kIROp_HalfType: m_writer->emit("kFloat16"); break;
case kIROp_DoubleType: m_writer->emit("kFloat64"); break;
case kIROp_UInt8Type: m_writer->emit("kUInt8"); break;
case kIROp_Int8Type: m_writer->emit("kInt8"); break;
case kIROp_Int16Type: m_writer->emit("kInt16"); break;
case kIROp_IntType: m_writer->emit("kInt32"); break;
case kIROp_Int64Type: m_writer->emit("kInt64"); break;
case kIROp_BoolType: m_writer->emit("kBool"); break;
default:
SLANG_UNEXPECTED((std::string("unknown scalar type in allocTorchTensor: ") +
std::string(getIROpInfo(type->getOp()).name))
.c_str());
break;
}
}
bool TorchCppSourceEmitter::tryEmitInstStmtImpl(IRInst* inst)
{
switch (inst->getOp())
{
default: return false;
case kIROp_CudaKernelLaunch:
{
m_writer->emit("AT_CUDA_CHECK(cudaLaunchKernel(");
// func
m_writer->emit("(const void*)(");
emitOperand(inst->getOperand(0), getInfo(EmitOp::General));
m_writer->emit("), ");
// gridDim
m_writer->emit("slang_bit_cast<dim3>(");
emitOperand(inst->getOperand(1), getInfo(EmitOp::General));
m_writer->emit("), ");
// blockDim
m_writer->emit("slang_bit_cast<dim3>(");
emitOperand(inst->getOperand(2), getInfo(EmitOp::General));
m_writer->emit("), ");
// args
emitOperand(inst->getOperand(3), getInfo(EmitOp::General));
m_writer->emit(", ");
// shared mem
m_writer->emit("0, ");
// stream
m_writer->emit("((cudaStream_t)");
emitOperand(inst->getOperand(4), getInfo(EmitOp::General));
m_writer->emit(")));\n");
return true;
}
}
}
bool TorchCppSourceEmitter::tryEmitInstExprImpl(IRInst* inst, const EmitOpInfo& inOuterPrec)
{
switch (inst->getOp())
{
default:
{
return Super::tryEmitInstExprImpl(inst, inOuterPrec);
}
case kIROp_MakeTensorView:
{
m_writer->emit("make_tensor_view(");
emitOperand(inst->getOperand(0), getInfo(EmitOp::General));
m_writer->emit(", ");
emitStringLiteral(getUnmangledName(inst->getOperand(0)));
m_writer->emit(", ");
emitTorchScalarTypeName(m_writer, inst->getOperand(0)->getDataType());
m_writer->emit(", ");
auto tensorViewType = as<IRTensorViewType>(inst->getDataType());
if (as<IRVectorType>(tensorViewType->getElementType()))
m_writer->emit("true");
else
m_writer->emit("false");
m_writer->emit(")");
return true;
}
case kIROp_TorchGetCudaStream:
{
m_writer->emit("at::cuda::getCurrentCUDAStream()");
return true;
}
case kIROp_AllocateTorchTensor:
{
if (as<IRTorchTensorType>(inst->getOperand(0)->getDataType()))
{
/*
Emit something like:
```
torch::Tensor out = torch::empty_like(other);
```
*/
m_writer->emit("torch::empty_like(");
emitOperand(inst->getOperand(0), getInfo(EmitOp::General));
m_writer->emit(", torch::TensorOptions().device(torch::kCUDA).dtype(");
emitTorchScalarTypeName(m_writer, inst->getDataType());
m_writer->emit("))");
}
else
{
/*
Emit something like:
```
torch::Tensor out = torch::empty({ dimX, dimY, dimZ, ... },
torch::TensorOptions().device(torch::kCUDA).dtype(torch::kFloat32));
```
*/
m_writer->emit("torch::empty({ ");
for (UInt i = 0; i < inst->getOperandCount(); i++)
{
if (i > 0)
m_writer->emit(", ");
auto arg = inst->getOperand(i);
emitOperand(arg, getInfo(EmitOp::General));
}
if (as<IRTorchTensorType>(inst->getDataType()))
{
if (auto vectorType = as<IRVectorType>(inst->getDataType()->getOperand(0)))
{
// If the element type of the tensor is a vector, we need to add the vector
// size to the shape.
m_writer->emit(", ");
emitOperand(vectorType->getElementCount(), getInfo(EmitOp::General));
}
}
m_writer->emit("}, torch::TensorOptions().device(torch::kCUDA).dtype(");
emitTorchScalarTypeName(m_writer, inst->getDataType());
m_writer->emit("))");
}
return true;
}
}
}
SlangResult TorchCppSourceEmitter::calcTypeName(
IRType* type,
CodeGenTarget target,
StringBuilder& out)
{
switch (type->getOp())
{
default: return Super::calcTypeName(type, target, out);
case kIROp_TensorViewType:
{
out << "TensorView";
return SLANG_OK;
}
case kIROp_TorchTensorType:
{
out << "torch::Tensor";
return SLANG_OK;
}
}
}
void TorchCppSourceEmitter::emitModuleImpl(IRModule* module, DiagnosticSink* sink)
{
Super::emitModuleImpl(module, sink);
// Emit PyBind declarations.
m_writer->emit("PYBIND11_MODULE(TORCH_EXTENSION_NAME, m) {\n");
m_writer->indent();
for (auto inst : module->getGlobalInsts())
{
auto func = as<IRFunc>(inst);
if (!func)
continue;
auto decor = func->findDecoration<IRTorchEntryPointDecoration>();
if (!decor)
continue;
m_writer->emit("m.def(");
emitStringLiteral(decor->getFunctionName());
m_writer->emit(", &");
m_writer->emit(decor->getFunctionName());
m_writer->emit(", ");
emitStringLiteral(decor->getFunctionName());
m_writer->emit(");\n");
}
m_writer->dedent();
m_writer->emit("}\n");
}
} // namespace Slang
|
