On 4/6/20 1:28 PM, Dmitry Osipenko wrote:
External email: Use caution opening links or attachments
06.04.2020 23:05, Sowjanya Komatineni пишет:
..
Wait time for HW syncpt to reach threshold is tuned to work for all
frame rates. So if increment doesn't happen by then, its definitely
missed event.
This is questionable. Technically, speculating about whether the tuned
value is good for all possible cases is incorrect thing to do.
Although, I guess in practice it should be good enough for the starter
and could be improved later on, once the host1x driver will be improved.
By tuned value I meant about 200ms wait timeout for frame event to
happen is what we have been using in downstream and with BSP release
images which works good for all sensors and bridges we supported so far.
I don't know anything about the state of today's downstream, but
downstream of older Tegra SoCs was pretty awful in regards to the host1x
syncing, unfortunately it was borrowed into the upstream host1x years
ago and nothing was done about it so far. I'd suggest to be careful
about it.
200ms timeout we wait for event to happen is the case even with
T186/T194 as well and internally it was tuned from lots of testing with
various sensors and frame rate computations which is known to work good.
