[PATCH i-g-t] tests/intel/xe_drm_fdinfo: Extend mercy to the upper end

[Lucas De Marchi]

On Sat, Aug 24, 2024 at 02:24:37PM GMT, Lucas De Marchi wrote:
>When we are processing the fdinfo of each client, the gpu time is read
>first, and then later all the exec queues are accumulated. It's thus
>possible that the total gpu time is smaller than the time reported in
>the exec queues. A preemption in the middle of second sample would
>exaggerate the problem:
>				  total_cycles	      cycles
>	s1: read gpu time		|
>	s1: read exec queues times	|		*
>					|		*
>	..				|		*
>					|		*
>	s2: read gpu time		|		*
>	-> prempted					*


on a second thought, this reasoning seems wrong... because we do
call spin_sync_end() before s2. So delta for exec queue time
shouldn't be larger than the the one for gpu time (since in s1 we
got the gpu timestamp before).... need to check if the sync end is
actually doing its job, or if there's something else going on.
I couldn't actually reproduce this > 100% on my end.

Lucas De Marchi