Hi Jack,
the scheduler changes become to invasive.
You should split that patch up into two. The first one to make the
scheduler changes and the second one using them in amdgpu.
Christian.
Am 11.03.21 um 16:37 schrieb Jack Zhang:
[Why]
Previous tdr design treats the first job in job_timeout
I would use a more meaningful name then
amdgpu_device_correct_karma and also put some comment
above describing the algorithm (just as in the patch description).
Other then that looks good to me -
Reviewed-by: Andrey Grodzovsky andrey.grodzov...@amd.com
Andrey
On 2021-03-11 10:37 a.m., Jack
[Why]
Previous tdr design treats the first job in job_timeout as the bad job.
But sometimes a later bad compute job can block a good gfx job and
cause an unexpected gfx job timeout because gfx and compute ring share
internal GC HW mutually.
[How]
This patch implements an advanced tdr mode.It
[Why]
Previous tdr design treats the first job in job_timeout as the bad job.
But sometimes a later bad compute job can block a good gfx job and
cause an unexpected gfx job timeout because gfx and compute ring share
internal GC HW mutually.
[How]
This patch implements an advanced tdr mode.It