> AFAIK, resampling uses the original event timestamps which in android-input
> are accurate because they come directly from the kernel. So if those
> timestamps are being faked in libinput then that might explain the
> irregularity and sensitivity to scheduling. Better check that libinput (and
> Mir itself) is not inserting new timestamps. Our input reports also rely on
> those timestamps being the original event time from the kernel.
>
> It's also worth noting my mouse setting is 1000Hz (kernel param
> usbhid.mousepoll=1). You might need that to notice some issues.
Well whatever you observe I cannot observe that. I see negative effects of resampling with and without libinput involved.
> AFAIK, resampling uses the original event timestamps which in android-input mousepoll= 1). You might need that to notice some issues.
> are accurate because they come directly from the kernel. So if those
> timestamps are being faked in libinput then that might explain the
> irregularity and sensitivity to scheduling. Better check that libinput (and
> Mir itself) is not inserting new timestamps. Our input reports also rely on
> those timestamps being the original event time from the kernel.
>
> It's also worth noting my mouse setting is 1000Hz (kernel param
> usbhid.
Well whatever you observe I cannot observe that. I see negative effects of resampling with and without libinput involved.