RE: [PATCH 2/3] drm/amdgpu: drop the sched_sync

2018-11-05 Thread Liu, Monk
eah, that’s true and more simple /Monk -Original Message- From: Koenig, Christian Sent: Monday, November 5, 2018 9:59 PM To: Liu, Monk ; amd-gfx@lists.freedesktop.org; Zhou, David(ChunMing) Subject: Re: [PATCH 2/3] drm/amdgpu: drop the sched_sync > and later its VMID's "curre

RE: [PATCH 2/3] drm/amdgpu: drop the sched_sync

2018-11-05 Thread Liu, Monk
istian Sent: Monday, November 5, 2018 9:59 PM To: Liu, Monk ; amd-gfx@lists.freedesktop.org; Zhou, David(ChunMing) Subject: Re: [PATCH 2/3] drm/amdgpu: drop the sched_sync > and later its VMID's "current_gpu_reset_count" is updated to > "adev->gpu_reset_count"

Re: [PATCH 2/3] drm/amdgpu: drop the sched_sync

2018-11-05 Thread Koenig, Christian
it dep on Job A, this explicit dep would > get lost and there will be no pipeline sync inserted prior to Job B ... > > Do you think that's a possible corner case ? > > /Monk > > -Original Message----- > From: Koenig, Christian > Sent: Monday, November 5, 2018 3:

RE: [PATCH 2/3] drm/amdgpu: drop the sched_sync

2018-11-05 Thread Liu, Monk
:12 PM To: Liu, Monk ; amd-gfx@lists.freedesktop.org; Zhou, David(ChunMing) Subject: Re: [PATCH 2/3] drm/amdgpu: drop the sched_sync > BTW: could we let the Job remember the hw fence seq that it need to > sync up to ? e.g. in "drm_sched_entity_clear_dep" we not only wake up > scheduler but also

RE: [PATCH 2/3] drm/amdgpu: drop the sched_sync

2018-11-05 Thread Liu, Monk
and should only surface when a specific test is run many many times. Regards, Christian. > > /Monk > -Original Message----- > From: Koenig, Christian > Sent: Monday, November 5, 2018 3:02 AM > To: Liu, Monk ; amd-gfx@lists.freedesktop.org > Subject: Re: [PATCH 2/3] drm/amdg

Re: [PATCH 2/3] drm/amdgpu: drop the sched_sync

2018-11-05 Thread Koenig, Christian
s parameter, and > we use ">=" operation instead of "==") > > /Monk > -----Original Message- > From: Koenig, Christian > Sent: Monday, November 5, 2018 3:48 PM > To: Liu, Monk ; amd-gfx@lists.freedesktop.org; Zhou, > David(ChunMing) > Subject

RE: [PATCH 2/3] drm/amdgpu: drop the sched_sync

2018-11-05 Thread Liu, Monk
tian Sent: Monday, November 5, 2018 3:48 PM To: Liu, Monk ; amd-gfx@lists.freedesktop.org; Zhou, David(ChunMing) Subject: Re: [PATCH 2/3] drm/amdgpu: drop the sched_sync Am 05.11.18 um 08:24 schrieb Liu, Monk: >> David Zhou had an use case which saw a >10% performance drop the last time >>

RE: [PATCH 2/3] drm/amdgpu: drop the sched_sync

2018-11-05 Thread Liu, Monk
--Original Message- From: Zhou, David(ChunMing) Sent: Monday, November 5, 2018 3:51 PM To: Koenig, Christian ; Liu, Monk ; amd-gfx@lists.freedesktop.org Subject: RE: [PATCH 2/3] drm/amdgpu: drop the sched_sync > -Original Message- > From: Koenig, Christian > Sent: Monday, Novemb

RE: [PATCH 2/3] drm/amdgpu: drop the sched_sync

2018-11-04 Thread Zhou, David(ChunMing)
> -Original Message- > From: Koenig, Christian > Sent: Monday, November 05, 2018 3:48 PM > To: Liu, Monk ; amd-gfx@lists.freedesktop.org; Zhou, > David(ChunMing) > Subject: Re: [PATCH 2/3] drm/amdgpu: drop the sched_sync > > Am 05.11.18 um 08:24 schrieb Liu, M

Re: [PATCH 2/3] drm/amdgpu: drop the sched_sync

2018-11-04 Thread Koenig, Christian
n. > > /Monk > -Original Message- > From: Koenig, Christian > Sent: Monday, November 5, 2018 3:02 AM > To: Liu, Monk ; amd-gfx@lists.freedesktop.org > Subject: Re: [PATCH 2/3] drm/amdgpu: drop the sched_sync > >> Can you tell me which game/benchmark will have

RE: [PATCH 2/3] drm/amdgpu: drop the sched_sync

2018-11-04 Thread Liu, Monk
-next, maybe there is gap in this part /Monk -Original Message- From: Koenig, Christian Sent: Monday, November 5, 2018 3:02 AM To: Liu, Monk ; amd-gfx@lists.freedesktop.org Subject: Re: [PATCH 2/3] drm/amdgpu: drop the sched_sync > Can you tell me which game/benchmark will have pe

Re: [PATCH 2/3] drm/amdgpu: drop the sched_sync

2018-11-04 Thread Koenig, Christian
, > but there is a JobB between A and C and the pipeline sync of before JobC will > wrongly wait on the JobB ... > > while it is not a big issue but obviously not necessary: C have no relation > with B > > /Monk > > > > -----Original Message- > From: Chris

RE: [PATCH 2/3] drm/amdgpu: drop the sched_sync

2018-11-03 Thread Liu, Monk
day, November 4, 2018 3:50 AM To: Liu, Monk ; amd-gfx@lists.freedesktop.org Subject: Re: [PATCH 2/3] drm/amdgpu: drop the sched_sync Am 03.11.18 um 06:33 schrieb Monk Liu: > Reasons to drop it: > > 1) simplify the code: just introduce field member "need_pipe_sync" > for

RE: [PATCH 2/3] drm/amdgpu: drop the sched_sync

2018-11-03 Thread Liu, Monk
, Monk ; amd-gfx@lists.freedesktop.org Subject: Re: [PATCH 2/3] drm/amdgpu: drop the sched_sync Am 03.11.18 um 06:33 schrieb Monk Liu: > Reasons to drop it: > > 1) simplify the code: just introduce field member "need_pipe_sync" > for job is good enough to tell if the explicit

Re: [PATCH 2/3] drm/amdgpu: drop the sched_sync

2018-11-03 Thread Christian König
Am 03.11.18 um 06:33 schrieb Monk Liu: Reasons to drop it: 1) simplify the code: just introduce field member "need_pipe_sync" for job is good enough to tell if the explicit dependency fence need followed by a pipeline sync. 2) after GPU_recover the explicit fence from sched_syn will not come ba