Add dummy nvidia headers for offline builds

[stephen-huan]

Fixes #202. Based on my read of download_and_copy, specifically the usage of shutil.copytree and shutil.copy, this should be safely replaced by the actual nvidia headers once they are downloaded, but I have not actually tested this.

triton-cpu/python/setup.py

Lines 297 to 330 in daa7eb0

	def download_and_copy(name, src_path, dst_path, variable, version, url_func):
	if is_offline_build():
	return
	triton_cache_path = get_triton_cache_path()
	if variable in os.environ:
	return
	base_dir = os.path.dirname(__file__)
	system = platform.system()
	try:
	arch = {"x86_64": "64", "arm64": "aarch64", "aarch64": "aarch64"}[platform.machine()]
	except KeyError:
	arch = platform.machine()
	supported = {"Linux": "linux", "Darwin": "linux"}
	url = url_func(supported[system], arch, version)
	tmp_path = os.path.join(triton_cache_path, "nvidia", name) # path to cache the download
	dst_path = os.path.join(base_dir, os.pardir, "third_party", "nvidia", "backend", dst_path) # final binary path
	platform_name = "sbsa-linux" if arch == "aarch64" else "x86_64-linux"
	src_path = src_path(platform_name, version) if callable(src_path) else src_path
	src_path = os.path.join(tmp_path, src_path)
	download = not os.path.exists(src_path)
	if os.path.exists(dst_path) and system == "Linux" and shutil.which(dst_path) is not None:
	curr_version = subprocess.check_output([dst_path, "--version"]).decode("utf-8").strip()
	curr_version = re.search(r"V([.|\d]+)", curr_version).group(1)
	download = download or curr_version != version
	if download:
	print(f'downloading and extracting {url} ...')
	file = tarfile.open(fileobj=open_url(url), mode="r|*")
	file.extractall(path=tmp_path)
	os.makedirs(os.path.split(dst_path)[0], exist_ok=True)
	print(f'copy {src_path} to {dst_path} ...')
	if os.path.isdir(src_path):
	shutil.copytree(src_path, dst_path, dirs_exist_ok=True)
	else:
	shutil.copy(src_path, dst_path)

triton-cpu/python/setup.py

Lines 327 to 330 in daa7eb0

	if os.path.isdir(src_path):
	shutil.copytree(src_path, dst_path, dirs_exist_ok=True)
	else:
	shutil.copy(src_path, dst_path)

shutil.copytree

If dirs_exist_ok is true, the copying operation will continue if it encounters existing directories, and files within the dst tree will be overwritten by corresponding files from the src tree.

shutil.copy

If dst specifies a file that already exists, it will be replaced. Returns the path to the newly created file.

Opening here instead of triton upstream because it's most relevant to the cpu backend and there probably won't be any merge conflicts (as third_party/nvidia/backend/include is ignored upstream).

cc @pawelszczerbuk

New contributor declaration

• I am not making a trivial change, such as fixing a typo in a comment.

• I have written a PR description following these
  rules.

• I have run pre-commit run --from-ref origin/main --to-ref HEAD.

• Select one of the following.

  • I have added tests.
    • /test for lit tests
    • /unittest for C++ tests
    • /python/test for end-to-end tests
  • This PR does not need a test because this is a build change.

• Select one of the following.

  • I have not added any lit tests.
  • The lit tests I have added follow these best practices,
    including the "tests should be minimal" section. (Usually running Python code
    and using the instructions it generates is not minimal.)

[minjang]

Like the previous PR, this needs to go to the upstream. You are updating the nvidia backend.

[stephen-huan]

Opened triton-lang#5492 upstream.

But in this specific circumstance, I doubt there will ever be a merge conflict as (1) these files are not actual Triton source code, but intended to be replaced by the downloaded cuda libraries by setup.py in the course of an ordinary build and (2) accordingly, they are .gitignore'd both here and upstream.

triton-cpu/.gitignore

Lines 59 to 61 in daa7eb0

	# Third-party include
	third_party/nvidia/backend/include
	third_party/nvidia/backend/lib/cupti

[minjang]

Once your PR is merged in the upstream, after our periodic rebasing, we will eventually get this PR. You don't need to make duplicate PRs. So, let me close it as this PR doesn't seem like an urgent fix.

[stephen-huan]

Understood, thanks for your patience!