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
gh-100344: Provide C implementation for asyncio.current_task #100345
gh-100344: Provide C implementation for asyncio.current_task #100345
Conversation
Implementing it in C makes it about 4x-6x faster Microbenchmark: ``` # bench.py import time import timeit import asyncio ITERS: int = 10**6 NANO: int = 10**9 NANO_PER_ITER: float = NANO / ITERS async def main(): # avoid load attr noise py_current_task = asyncio.tasks._py_current_task c_current_task = asyncio.tasks._c_current_task asyncio.current_task() # warmup py_current_task() # warmup c_current_task() # warmup print( "current_task: {}ns".format(timeit.timeit(py_current_task, number=ITERS, timer=time.process_time) * NANO_PER_ITER) ) print( "current_task: {}ns".format(timeit.timeit(c_current_task, number=ITERS, timer=time.process_time) * NANO_PER_ITER) ) asyncio.run(main()) ``` a few runs on MacBook Pro 2.4 GHz 8-Core Intel Core i9 64 GB 2667 MHz DDR4: debug build: ``` ~/work/pyexe/main-dbg⌚ 9:57:34 $ ./python.exe bench.py [py] current_task: 606.234ns [c] current_task: 104.47699999999993ns ~/work/pyexe/main-dbg⌚ 9:57:59 $ ./python.exe bench.py [py] current_task: 631.856ns [c] current_task: 110.22500000000002ns ~/work/pyexe/main-dbg⌚ 9:58:08 $ ./python.exe bench.py [py] current_task: 637.746ns [c] current_task: 105.03899999999999ns ~/work/pyexe/main-dbg⌚ 9:58:16 $ ./python.exe bench.py [py] current_task: 621.3169999999999ns [c] current_task: 103.01300000000002ns ``` opt build: ``` ~/work/pyexe/main-opt⌚ 10:33:17 $ ./python.exe bench.py [py] current_task: 128.743ns [c] current_task: 31.997999999999998ns ~/work/pyexe/main-opt⌚ 10:33:24 $ ./python.exe bench.py [py] current_task: 126.388ns [c] current_task: 32.64599999999998ns ~/work/pyexe/main-opt⌚ 10:33:26 $ ./python.exe bench.py [py] current_task: 137.053ns [c] current_task: 32.066999999999986ns ~/work/pyexe/main-opt⌚ 10:33:28 $ ./python.exe bench.py [py] current_task: 131.17800000000003ns [c] current_task: 32.06600000000001ns ```
If you want to schedule another build, you need to add the |
Congrats @itamaro! Your PR has been selected to try the new refleak buildbot label! |
Such an honor, thank you @pablogsal Checking out a few of the additional refleak test runs, I saw (across different buildbots): timeout, failure of
I think the issue is not decrefing the |
I'm waiting for the benchmark test suggested by Mark in the issue. |
replied on the issue and pushed the python optimization @markshannon suggested for ~40% speedup. |
A Python core developer has requested some changes be made to your pull request before we can consider merging it. If you could please address their requests along with any other requests in other reviews from core developers that would be appreciated. Once you have made the requested changes, please leave a comment on this pull request containing the phrase |
I have made the requested changes; please review again |
Thanks for making the requested changes! @kumaraditya303: please review the changes made to this pull request. |
Is it still ok to push another change to revert the Python implementation as discussed on the issue, after approval? |
That's fine, I'll review anyways before merging so go ahead! |
Merged, thanks for working on this! |
Implementing it in C makes it about 4x-6x faster
Microbenchmark:
a few runs on MacBook Pro
2.4 GHz 8-Core Intel Core i9
64 GB 2667 MHz DDR4:
debug build:
opt build:
Co-authored-by: @pranavtbhat