perf_counter() returns computers uptime

[txhx38]

BPO	43585
Nosy	@terryjreedy, @eryksun, @sweeneyde

^{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 2021-08-09.03:31:39.973>
created_at = <Date 2021-03-21.16:38:13.520>
labels = ['type-bug', 'invalid', '3.10']
title = 'perf_counter() returns computers uptime'
updated_at = <Date 2021-08-09.03:31:39.973>
user = 'https://bugs.python.org/txhx38'

bugs.python.org fields:

activity = <Date 2021-08-09.03:31:39.973>
actor = 'Dennis Sweeney'
assignee = 'none'
closed = True
closed_date = <Date 2021-08-09.03:31:39.973>
closer = 'Dennis Sweeney'
components = []
creation = <Date 2021-03-21.16:38:13.520>
creator = 'txhx38'
dependencies = []
files = []
hgrepos = []
issue_num = 43585
keywords = []
message_count = 5.0
messages = ['389248', '389249', '389262', '389561', '389563']
nosy_count = 4.0
nosy_names = ['terry.reedy', 'eryksun', 'Dennis Sweeney', 'txhx38']
pr_nums = []
priority = 'normal'
resolution = 'not a bug'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue43585'
versions = ['Python 3.10']

[txhx38]

perf_counter() does not behave as expected

[sweeneyde]

Can you explain why you think this is the wrong behavior, and what operating system and version of Python you're using?

from https://docs.python.org/3/library/time.html#time.perf_counter

Return the value (in fractional seconds) of a performance counter, i.e. a clock with the highest available resolution to measure a short duration. It does include time elapsed during sleep and is system-wide. The reference point of the returned value is undefined, so that only the difference between the results of two calls is valid.

Returning the uptime seems like acceptable behavior to me, since "the reference point ... is undefined".

[eryksun]

For Windows, starting with Python 3.10, the value of time.perf_counter_ns is system wide instead of relative to the first time it's accessed in the current process. This allows comparing the counter value across processes. See bpo-37205.

[terryjreedy]

Eryk, if I understand correctly, this should be closed as 'not a bug'. Correct?

[eryksun]

I think so. But I'll mark it pending for now in case Tobi provides more information to explain why this is a problem. Making time.perf_counter() system-wide in Windows was fixing it to match the documented behavior. That this provides the system uptime is a natural consequence of how the system performance counter is implemented consistently for all processes for the entire uptime of the system.