MCPcopy Create free account
hub / github.com/DeepRec-AI/DeepRec / PtxCompilationOptions

Class PtxCompilationOptions

tensorflow/stream_executor/cuda/ptxas_utils.h:33–54  ·  view source on GitHub ↗

Compilation options for compiling ptxas.

Source from the content-addressed store, hash-verified

31
32// Compilation options for compiling ptxas.
33struct PtxCompilationOptions {
34 bool disable_ptxas_optimizations;
35
36 // CUDA directory which would be searched first.
37 std::string preferred_cuda_dir;
38
39 std::vector<std::string> extra_flags;
40
41 explicit PtxCompilationOptions(bool disable_ptxas_optimizations = false,
42 absl::string_view preferred_cuda_dir = "",
43 absl::Span<const std::string> extra_flags = {})
44 : disable_ptxas_optimizations(disable_ptxas_optimizations),
45 preferred_cuda_dir(preferred_cuda_dir),
46 extra_flags(extra_flags.begin(), extra_flags.end()) {}
47
48 using PtxOptionsTuple =
49 std::tuple<bool, std::string, std::vector<std::string>>;
50
51 PtxOptionsTuple ToTuple() {
52 return std::make_tuple(disable_ptxas_optimizations, preferred_cuda_dir, extra_flags);
53 }
54};
55
56// Compiles the given PTX string using ptxas and returns the resulting machine
57// code (i.e. a cubin) as a byte array.

Callers 8

PtxOptsFromConfigFunction · 0.85
operator()Method · 0.85
ComputeMethod · 0.85
operator()Method · 0.85
launchMethod · 0.85
operator()Method · 0.85

Calls

no outgoing calls

Tested by

no test coverage detected