From: Felix Kuehling <felix.kuehling-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
To: "Koenig,
Christian" <Christian.Koenig-5C7GfCeVMHo@public.gmane.org>,
"Kuehling, Felix" <Felix.Kuehling-5C7GfCeVMHo@public.gmane.org>,
"amd-gfx-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org"
<amd-gfx-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org>
Cc: "Russell, Kent" <Kent.Russell-5C7GfCeVMHo@public.gmane.org>
Subject: Re: [PATCH 27/27] drm/amdgpu: Fix GTT size calculation
Date: Sat, 13 Jul 2019 16:24:57 -0400 [thread overview]
Message-ID: <44747586-8c98-6c85-dc5c-7464f59b205e@gmail.com> (raw)
In-Reply-To: <f5c698ad-2aff-b3c5-2041-05a10983438a-5C7GfCeVMHo@public.gmane.org>
[-- Attachment #1.1.1.1: Type: text/plain, Size: 1602 bytes --]
Am 2019-04-30 um 1:03 p.m. schrieb Koenig, Christian:
>>> The only real solution I can see is to be able to reliable kill shaders
>>> in an OOM situation.
>> Well, we can in fact preempt our compute shaders with low latency.
>> Killing a KFD process will do exactly that.
> I've taken a look at that thing as well and to be honest it is not even
> remotely sufficient.
>
> We need something which stops the hardware *immediately* from accessing
> system memory, and not wait for the SQ to kill all waves, flush caches
> etc...
>
> One possibility I'm playing around with for a while is to replace the
> root PD for the VMIDs in question on the fly. E.g. we just let it point
> to some dummy which redirects everything into nirvana.
>
> But implementing this is easier said than done...
Warming up this thread, since I just fixed another bug that was enabled
by artificial memory pressure due to the GTT limit.
I think disabling the PD for the VMIDs is a good idea. A problem is that
HWS firmware updates PD pointers in the background for its VMIDs. So
this would require a reliable and fast way to kill the HWS first.
An alternative I thought about is, disabling bus access at the BIF level
if that's possible somehow. Basically we would instantaneously kill all
GPU system memory access, signal all fences or just remove all fences
from all BO reservations (reservation_object_add_excl_fence(resv, NULL))
to allow memory to be freed, let the OOM killer do its thing, and when
the dust settles, reset the GPU.
Regards,
Felix
>
> Regards,
> Christian.
[-- Attachment #1.1.1.2: Type: text/html, Size: 2629 bytes --]
[-- Attachment #1.2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 228 bytes --]
[-- Attachment #2: Type: text/plain, Size: 153 bytes --]
_______________________________________________
amd-gfx mailing list
amd-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
next prev parent reply other threads:[~2019-07-13 20:24 UTC|newest]
Thread overview: 37+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-04-28 7:44 [PATCH 00/27] KFD upstreaming Kuehling, Felix
[not found] ` <20190428074331.30107-1-Felix.Kuehling-5C7GfCeVMHo@public.gmane.org>
2019-04-28 7:44 ` [PATCH 01/27] drm/amdkfd: Use 64 bit sdma_bitmap Kuehling, Felix
2019-04-28 7:44 ` [PATCH 02/27] drm/amdkfd: Add sdma allocation debug message Kuehling, Felix
2019-04-28 7:44 ` [PATCH 03/27] drm/amdkfd: Differentiate b/t sdma_id and sdma_queue_id Kuehling, Felix
2019-04-28 7:44 ` [PATCH 05/27] drm/amdkfd: Fix a potential memory leak Kuehling, Felix
2019-04-28 7:44 ` [PATCH 04/27] drm/amdkfd: Shift sdma_engine_id and sdma_queue_id in mqd Kuehling, Felix
2019-04-28 7:44 ` [PATCH 06/27] drm/amdkfd: Introduce asic-specific mqd_manager_init function Kuehling, Felix
2019-04-28 7:44 ` [PATCH 07/27] drm/amdkfd: Introduce DIQ type mqd manager Kuehling, Felix
2019-04-28 7:44 ` [PATCH 08/27] drm/amdkfd: Init mqd managers in device queue manager init Kuehling, Felix
2019-04-28 7:44 ` [PATCH 09/27] drm/amdkfd: Add mqd size in mqd manager struct Kuehling, Felix
2019-04-28 7:44 ` [PATCH 10/27] drm/amdkfd: Allocate MQD trunk for HIQ and SDMA Kuehling, Felix
2019-04-28 7:44 ` [PATCH 11/27] drm/amdkfd: Move non-sdma mqd allocation out of init_mqd Kuehling, Felix
2019-04-28 7:44 ` [PATCH 12/27] drm/amdkfd: Allocate hiq and sdma mqd from mqd trunk Kuehling, Felix
2019-04-28 7:44 ` [PATCH 13/27] drm/amdkfd: Move sdma_queue_id calculation into allocate_sdma_queue() Kuehling, Felix
2019-04-28 7:44 ` [PATCH 14/27] drm/amdkfd: Fix compute profile switching Kuehling, Felix
2019-04-28 7:44 ` [PATCH 15/27] drm/amdkfd: Fix sdma queue map issue Kuehling, Felix
2019-04-28 7:44 ` [PATCH 16/27] drm/amdkfd: Introduce XGMI SDMA queue type Kuehling, Felix
2019-04-28 7:44 ` [PATCH 17/27] drm/amdkfd: Expose sdma engine numbers to topology Kuehling, Felix
2019-04-28 7:44 ` [PATCH 18/27] drm/amdkfd: Delete alloc_format field from map_queue struct Kuehling, Felix
2019-04-28 7:44 ` [PATCH 19/27] drm/amdkfd: Fix a circular lock dependency Kuehling, Felix
2019-04-28 7:44 ` [PATCH 20/27] drm/amdkfd: Fix gfx8 MEM_VIOL exception handler Kuehling, Felix
2019-04-28 7:44 ` [PATCH 21/27] drm/amdkfd: Preserve wave state after instruction fetch MEM_VIOL Kuehling, Felix
2019-04-28 7:44 ` [PATCH 22/27] drm/amdkfd: Fix gfx9 XNACK state save/restore Kuehling, Felix
2019-04-28 7:44 ` [PATCH 23/27] drm/amdkfd: Preserve ttmp[4:5] instead of ttmp[14:15] Kuehling, Felix
2019-04-28 7:44 ` [PATCH 25/27] drm/amdkfd: Add domain number into gpu_id Kuehling, Felix
2019-04-28 7:44 ` [PATCH 24/27] drm/amdkfd: Add VegaM support Kuehling, Felix
2019-04-28 7:44 ` [PATCH 26/27] drm/amdgpu: Use heavy weight for tlb invalidation on xgmi configuration Kuehling, Felix
2019-04-28 7:44 ` [PATCH 27/27] drm/amdgpu: Fix GTT size calculation Kuehling, Felix
[not found] ` <20190428074331.30107-28-Felix.Kuehling-5C7GfCeVMHo@public.gmane.org>
2019-04-29 12:34 ` Christian König
[not found] ` <86fa9fc3-7a8f-9855-ae1d-5c7ccf2b5260-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
2019-04-29 23:16 ` Kuehling, Felix
[not found] ` <1b1ec993-1c4b-8661-9b3f-ac0ad8ae64c7-5C7GfCeVMHo@public.gmane.org>
2019-04-30 9:32 ` Christian König
[not found] ` <134a4999-776f-44c6-99a2-42e8b9366a73-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
2019-04-30 15:36 ` Kuehling, Felix
[not found] ` <9f882acd-c48f-3bbd-2d90-659c2edead39-5C7GfCeVMHo@public.gmane.org>
2019-04-30 17:03 ` Koenig, Christian
[not found] ` <f5c698ad-2aff-b3c5-2041-05a10983438a-5C7GfCeVMHo@public.gmane.org>
2019-04-30 17:25 ` Kuehling, Felix
[not found] ` <8ba952ab-4836-4ca3-cd80-99f7367a7979-5C7GfCeVMHo@public.gmane.org>
2019-05-02 13:06 ` Koenig, Christian
2019-07-13 20:24 ` Felix Kuehling [this message]
2019-04-29 23:23 ` [PATCH 00/27] KFD upstreaming Kuehling, Felix
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=44747586-8c98-6c85-dc5c-7464f59b205e@gmail.com \
--to=felix.kuehling-re5jqeeqqe8avxtiumwx3w@public.gmane.org \
--cc=Christian.Koenig-5C7GfCeVMHo@public.gmane.org \
--cc=Felix.Kuehling-5C7GfCeVMHo@public.gmane.org \
--cc=Kent.Russell-5C7GfCeVMHo@public.gmane.org \
--cc=amd-gfx-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox