Add time.thread_time()

[pitrou]

BPO	32025
Nosy	@abalkin, @pitrou, @vstinner, @koobs
PRs	bpo-32025: Add time.thread_time() #4410

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = <Date 2017-11-20.17:21:29.014>
created_at = <Date 2017-11-14.13:50:20.131>
labels = ['type-feature', '3.7']
title = 'Add time.thread_time()'
updated_at = <Date 2017-11-20.17:21:29.004>
user = 'https://github.com/pitrou'

bugs.python.org fields:

activity = <Date 2017-11-20.17:21:29.004>
actor = 'vstinner'
assignee = 'none'
closed = True
closed_date = <Date 2017-11-20.17:21:29.014>
closer = 'vstinner'
components = ['FreeBSD']
creation = <Date 2017-11-14.13:50:20.131>
creator = 'pitrou'
dependencies = []
files = []
hgrepos = []
issue_num = 32025
keywords = ['patch']
message_count = 9.0
messages = ['306210', '306247', '306249', '306265', '306308', '306309', '306313', '306398', '306570']
nosy_count = 4.0
nosy_names = ['belopolsky', 'pitrou', 'vstinner', 'koobs']
pr_nums = ['4410']
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'enhancement'
url = 'https://bugs.python.org/issue32025'
versions = ['Python 3.7']

[pitrou]

Currently, the time module has time.process_time(), a cross-platform function for getting per-process elapsed CPU time.

Similarly, we could expose time.thread_time(), to get per-thread elapsed CPU time.
On a modern POSIX platform, it can use clock_gettime(CLOCK_THREAD_CPUTIME_ID).
On Windows, it can use GetThreadTimes(): https://msdn.microsoft.com/en-us/library/ms683237%28VS.85%29.aspx
On other platforms, it can simply raise NotImplementedError.

Currently, you would need ctypes hacks to call GetThreadTimes(), which is not very nice.

[vstinner]

On other platforms, it can simply raise NotImplementedError.

Would it be possible to not provide the function if it's not supported by a platform?

Do you know platforms which don't support time.thread_time()?

Do you want to add time.thread_time_ns() with nanosecond resolution (PEP-564) ? I suggest to add it to be consistent with time.process_time().

[vstinner]

It seems like macOS doesn't implement CLOCK_THREAD_CPUTIME_ID but provides thread_info() which can be used by suming user and system times.

[pitrou]

Le 15/11/2017 à 02:48, STINNER Victor a écrit :

Would it be possible to not provide the function if it's not supported by a platform?

Probably, yes.

Do you know platforms which don't support time.thread_time()?

Any POSIX platform which doesn't have CLOCK_THREAD_CPUTIME_ID? It's an
optional feature in the POSIX spec.

Do you want to add time.thread_time_ns() with nanosecond resolution (PEP-564) ?

Yes, I think so.

It seems like macOS doesn't implement CLOCK_THREAD_CPUTIME_ID but
provides thread_info() which can be used by suming user and system times.

I don't have a macOS machine to test on, though :-)

[vstinner]

I can work on the macOS implementation, once your PR will be merged.

[pitrou]

Le 15/11/2017 à 22:06, STINNER Victor a écrit :

I can work on the macOS implementation, once your PR will be merged.

Great!

[pitrou]

New changeset 4bd41c9 by Antoine Pitrou in branch 'master':
bpo-32025: Add time.thread_time() (bpo-4410)
4bd41c9

[pitrou]

Should I leave this open for you to work on the macOS implementation, Victor?

[vstinner]

I opened bpo-32093 for macOS. I close this one.

Nice enhancement, I already like this new clock ;-)