[BUG-231941] CPU Frequency is miscalculated on Apple Silicon (M1) under Rosetta

[sl-service-account]

How would you like the feature to work?

The M1 chip that we see appearing in SL has a nominal Frequency of 3.2 GHz,, has 8 cores, split 50/50 performance and economy. The actual frequency of these cores is not fixed. When running under Rosetta it would appear that the frequency is reported as 2.4GHz. this may lead to spurious timing related data if the CPU frequency is being used as a clock cycle divisor.

Why is this feature important to you? How would it benefit the community?

Noticed on Firestorm where the frequency is used when converting raw cycle counts to clock units. Overall, the impact of this is low, but it puts a large question mark over any timing data that has not been reviewed to verify the clock source/divisors. In theory, the timer issues can be worked around, as the use of lower resolution timers such as clock_gettime() should work.

Original Jira Fields

| Field | Value | | ------------- | ------------- | | Issue | BUG-231941 | | Summary | CPU Frequency is miscalculated on Apple Silicon (M1) under Rosetta | | Type | Bug | | Priority | Unset | | Status | Accepted | | Resolution | Triaged | | Labels | rosetta | | Reporter | Beq Janus (beq.janus) | | Created at | 2022-03-18T19:48:46Z | | Updated at | 2022-04-26T21:22:34Z | ``` { 'Build Id': 'unset', 'Business Unit': ['Platform'], 'Date of First Response': '2022-03-21T12:55:10.882-0500', 'How would you like the feature to work?': 'The M1 chip that we see appearing in SL has a nominal Frequency of 3.2 GHz,, has 8 cores, split 50/50 performance and economy. The actual frequency of these cores is not fixed. \r\nWhen running under Rosetta it would appear that the frequency is reported as 2.4GHz. this may lead to spurious timing related data if the CPU frequency is being used as a clock cycle divisor.', 'ReOpened Count': 0.0, 'Severity': 'Unset', 'System': 'SL Viewer', 'Target Viewer Version': 'viewer-development', 'What just happened?': '.', 'What were you doing when it happened?': '.', 'What were you expecting to happen instead?': '.', 'Why is this feature important to you? How would it benefit the community?': 'Noticed on Firestorm where the frequency is used when converting raw cycle counts to clock units. Overall, the impact of this is low, but it puts a large question mark over any timing data that has not been reviewed to verify the clock source/divisors. \r\nIn theory, the timer issues can be worked around, as the use of lower resolution timers such as clock_gettime() should work. ', } ```

[sl-service-account]

Dan Linden commented at 2022-03-21T17:55:11Z

Thank you for the report, Beq.