xiaoxiang781216 commented on PR #17276: URL: https://github.com/apache/nuttx/pull/17276#issuecomment-3511699532
> > > Please hold this PR and do not merge it easily. @Fix-Point Could you split this patch into multiple PRs: > > > > > > 1. Split the ONESHOT_COUNT core implementation > > > 2. Split the division optimization > > > 3. Split the implementation for each architecture > > > 4. Split the defconfig changes > > > > > > Before the defconfig update in phase 4, the new clock framework should not be enabled in the system. The ONESHOT_COUNT framework needs time to prove itself; each developer should decide for themselves whether to enable this feature. > > > > > > But we had to revert the tick-based interfaces, which would cause performance degradation on these platforms and incur the overhead of re-verification. Besides, I think the new ClockDevice drivers are more reliable than the original, since they have fixed many timer driver bugs. E.g, the tricore systimer will miss interrupt if the next count to be set has already expired. > > The ticks loss issue can be fixed by the counter register, and it shouldn't be a serious problem unless more than 2 toggle cycles are involved. The new framework needs further evaluation; although it's more accurate, I'm concerned it might increase CPU load, especially on 32-bit systems tick unit impact up_?delay reported here: https://github.com/apache/nuttx/pull/17221 -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
