File: llvm_codegen.h

package info (click to toggle)
pytorch 1.7.1-7
  • links: PTS, VCS
  • area: main
  • in suites: bullseye
  • size: 80,340 kB
  • sloc: cpp: 670,830; python: 343,991; ansic: 67,845; asm: 5,503; sh: 2,924; java: 2,888; xml: 266; makefile: 244; ruby: 148; yacc: 144; objc: 51; lex: 44
file content (60 lines) | stat: -rw-r--r-- 1,244 bytes parent folder | download
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
#pragma once

#ifdef TORCH_ENABLE_LLVM
#include <torch/csrc/WindowsTorchApiMacro.h>

#include <torch/csrc/jit/tensorexpr/codegen.h>
#include <torch/csrc/jit/tensorexpr/ir.h>
#include <torch/csrc/jit/tensorexpr/ir_visitor.h>

#include <unordered_map>
#include <vector>

namespace torch {
namespace jit {
namespace tensorexpr {

class LLVMCodeGenImpl;

class TORCH_API LLVMCodeGen : public CodeGen {
 public:
  explicit LLVMCodeGen(
      Stmt* stmt,
      const std::vector<BufferArg>& args,
      at::Device device = at::kCPU,
      Dtype dtype = kInt);
  explicit LLVMCodeGen(Stmt* stmt);

  LLVMCodeGen() = delete;
  ~LLVMCodeGen() override;

  TORCH_API void call(const std::vector<CallArg>& args) override;

  template <typename T>
  T value() {
    return value<T>(nullptr);
  }

  template <typename T>
  T value(std::vector<void*>& args) {
    return value<T>(args.data());
  }

  template <typename T>
  T value(void** args) {
    T (*fp)(void**) = (T(*)(void**))getKernelAddress(impl_.get());
    T rv = fp(args);
    return rv;
  }

 private:
  void* getKernelAddress(LLVMCodeGenImpl* impl);

  std::unique_ptr<LLVMCodeGenImpl> impl_;
};

} // namespace tensorexpr
} // namespace jit
} // namespace torch

#endif // TORCH_ENABLE_LLVM