Skip to content

[AUTOGENERATED] [release/2.5] Change gfx110x BLAS preferred backend #2169

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
Merged
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
31 changes: 28 additions & 3 deletions aten/src/ATen/Context.h
Original file line number Diff line number Diff line change
Expand Up @@ -425,12 +425,37 @@ class TORCH_API Context {
c10::utils::check_env("TORCH_LINALG_PREFER_CUSOLVER") == true
? at::LinalgBackend::Cusolver
: at::LinalgBackend::Default;
at::BlasBackend blas_preferred_backend =
#ifdef USE_ROCM
(c10::utils::check_env("TORCH_BLAS_PREFER_HIPBLASLT") != false)
// AMD Instinct targets prefer hipblaslt
const bool _hipblaslt_preferred_default = []() {
const std::vector<std::string> archs = {
"gfx90a", "gfx942"
};
for (auto index: c10::irange(detail::getCUDAHooks().deviceCount())) {
if (!detail::getCUDAHooks().isGPUArch(index, archs)) {
return false;
}
}
return true;
}();
#else
(c10::utils::check_env("TORCH_BLAS_PREFER_CUBLASLT") == true)
const bool _hipblaslt_preferred_default = false;
#endif
const bool _blaslt_preferred = [&]() {
auto env = c10::utils::check_env("TORCH_BLAS_PREFER_CUBLASLT");
if (env.has_value()) {
return env.value();
}
env = c10::utils::check_env("TORCH_BLAS_PREFER_HIPBLASLT");
if (env.has_value()) {
return env.value();
}
#ifdef USE_ROCM
return _hipblaslt_preferred_default;
#endif
return false;
}();
at::BlasBackend blas_preferred_backend = _blaslt_preferred
? at::BlasBackend::Cublaslt
: at::BlasBackend::Cublas;
at::ROCmFABackend rocm_fa_preferred_backend =
Expand Down