I pushed my patch series that simplifies eviction fence handling in KFD.
If you rebase this, it should be OK now.
Regards,
Felix
On 2019-02-04 7:42 a.m., Christian König wrote:
> Let's start to allocate VM PDs/PTs on demand instead of pre-allocating
> them during mapping.
>
> Signed-off-by:
t;> I saw a backtrace from the GPU scheduler when I was debugging this
>>>> yesterday, but those backtraces never tell us where the command submission
>>>> came from. It could be memory initialization, or some migration at
>>>> buffer-validation. Basically,
e the command submission
>>> came from. It could be memory initialization, or some migration at
>>> buffer-validation. Basically, any command submission triggered by page
>>> table allocation that synchronizes with the PD reservation object is a
>>> pro
gt;>
>> -Original Message-
>> From: Christian König
>> Sent: Tuesday, February 05, 2019 10:40 AM
>> To: Kuehling, Felix ; Koenig, Christian
>> ; amd-gfx@lists.freedesktop.org
>> Subject: Re: [PATCH 6/8] drm/amdgpu: allocate VM PDs/PTs on demand
&
e-
> From: Christian König
> Sent: Tuesday, February 05, 2019 10:40 AM
> To: Kuehling, Felix ; Koenig, Christian
> ; amd-gfx@lists.freedesktop.org
> Subject: Re: [PATCH 6/8] drm/amdgpu: allocate VM PDs/PTs on demand
>
> Am 05.02.19 um 16:20 schrieb Kuehling, Felix:
>>
allocation that
synchronizes with the PD reservation object is a problem.
Regards,
Felix
-Original Message-
From: Christian König
Sent: Tuesday, February 05, 2019 10:40 AM
To: Kuehling, Felix ; Koenig, Christian
; amd-gfx@lists.freedesktop.org
Subject: Re: [PATCH 6/8] drm/amdgpu: allocate VM
a call chain?
Regards,
Christian.
Regards,
Felix
-Original Message-
From: Christian König
Sent: Tuesday, February 05, 2019 6:37 AM
To: Kuehling, Felix ; amd-gfx@lists.freedesktop.org
Subject: Re: [PATCH 6/8] drm/amdgpu: allocate VM PDs/PTs on demand
This may cause regressions
edesktop.org
Subject: Re: [PATCH 6/8] drm/amdgpu: allocate VM PDs/PTs on demand
> This may cause regressions in KFD if PT BO allocation can trigger
> eviction fences.
I don't think that can happen, but need to double check as well.
Otherwise allocating page tables could try to evict ot
Am 05.02.19 um 01:33 schrieb Kuehling, Felix:
On 2019-02-04 3:17 p.m., Kuehling, Felix wrote:
This may cause regressions in KFD if PT BO allocation can trigger
eviction fences. The previous call to amdgpu_vm_alloc_pts was in a
context where we had temporarily removed the eviction fence. PT BO
This may cause regressions in KFD if PT BO allocation can trigger
eviction fences.
I don't think that can happen, but need to double check as well.
Otherwise allocating page tables could try to evict other page tables
from the same process and that seriously doesn't make much sense.
Do you
On 2019-02-04 3:17 p.m., Kuehling, Felix wrote:
> This may cause regressions in KFD if PT BO allocation can trigger
> eviction fences. The previous call to amdgpu_vm_alloc_pts was in a
> context where we had temporarily removed the eviction fence. PT BO
> allocation in amdgpu_vm_bo_update is not
This may cause regressions in KFD if PT BO allocation can trigger
eviction fences. The previous call to amdgpu_vm_alloc_pts was in a
context where we had temporarily removed the eviction fence. PT BO
allocation in amdgpu_vm_bo_update is not protected like that.
I vaguely remember looking into
Let's start to allocate VM PDs/PTs on demand instead of pre-allocating
them during mapping.
Signed-off-by: Christian König
---
.../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 6 -
drivers/gpu/drm/amd/amdgpu/amdgpu_csa.c | 9 --
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c | 10 --
13 matches
Mail list logo