Skip to content

fix #161: XQA cubin generation ignores resolved output dir under spawn/forkserver #177

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.

Sign up for GitHub

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

Open
Hi5808 wants to merge 1 commit into
base: main
Choose a base branch
from
Open

fix #161: XQA cubin generation ignores resolved output dir under spawn/forkserver #177

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
17 changes: 16 additions & 1 deletion kernelSrcs/xqa/gen_cubins.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -454,6 +454,18 @@ def convert_cubin_cpp_np(cubin_file_name: str):
return cpp_array, cubin_size


def init_cubin_gen_worker(worker_cubin_dir: str, worker_nvcc_bin: str,
worker_clean_cubin: bool):
# Workers under a start method that re-imports this module (spawn/forkserver)
# never execute the `if __name__ == "__main__":` block, so the module-scope
# defaults for these globals would otherwise be used instead of the values
# resolved from CLI args. Set them explicitly in each worker process.
global cubin_dir, nvcc_bin, clean_cubin
cubin_dir = worker_cubin_dir
nvcc_bin = worker_nvcc_bin
clean_cubin = worker_clean_cubin


def run_cubin_gen(arch_micro_file_list: CompileArchMacrosAndFile):
nvcc_command, xxd_command, cubin_file_name = build_commands(
build_func_name_prefix, arch_micro_file_list.arch,
Expand Down Expand Up @@ -833,7 +845,10 @@ def get_generated_output_paths(

cpu_count = os.cpu_count()
thread_count = cpu_count // 2 if cpu_count >= 2 else cpu_count
with multiprocessing.Pool(processes=thread_count) as pool:
with multiprocessing.Pool(
processes=thread_count,
initializer=init_cubin_gen_worker,
initargs=(cubin_dir, nvcc_bin, clean_cubin)) as pool:
name_size_list = pool.map(run_cubin_gen, arch_macro_lists)
header_file_contents = generate_header_file_contents(
arch_macro_lists, name_size_list)
Expand Down