From: Dave Airlie <airlied@gmail.com>
To: "Christian König" <christian.koenig@amd.com>
Cc: Ben Skeggs <bskeggs@redhat.com>,
dri-devel <dri-devel@lists.freedesktop.org>
Subject: Re: [PATCH 37/45] drm/ttm: add a helper to allocate a temp tt for copies.
Date: Fri, 25 Sep 2020 09:14:34 +1000 [thread overview]
Message-ID: <CAPM=9tw6a6E+KZV7LceHUj=SHFaTe9M+KjjYgouhdX0gkDsuzw@mail.gmail.com> (raw)
In-Reply-To: <39ca897a-e7f2-93c8-e844-449ccbfe6041@amd.com>
On Thu, 24 Sep 2020 at 22:42, Christian König <christian.koenig@amd.com> wrote:
>
> Am 24.09.20 um 07:18 schrieb Dave Airlie:
> > From: Dave Airlie <airlied@redhat.com>
> >
> > All the accel moves do the same pattern here, provide a helper
>
> And exactly that pattern I want to get away from.
Currently this is just refactoring out the helper code in each driver, but I see
since it calls bo_mem_space we are probably moving a bit in the wrong direction.
> See what happens if we (for example) have a VRAM -> SYSTEM move is the
> following:
>
> 1. TTM allocates a new ttm_resource object in the SYSTEM domain.
> 2. We call the driver to move from VRAM to SYSTEM.
> 3. Driver finds that it can't do this and calls TTM to allocate GTT.
> 4. Since we are maybe out of GTT TTM evicts a different BO from GTT to
> SYSTEM and call driver again.
>
> This is a horrible ping/pong between driver/TTM/driver/TTM/driver and we
> should stop that immediately.
>
> My suggestion is that we rewrite how drivers call the ttm_bo_validate()
> function so that we can guarantee that this never happens.
>
> What do you think?
I think that is likely the next step I'd like to take after this
refactor, it's a lot bigger, and I'm not sure how it will look yet.
Do we envision the driver calling validate in a loop but when it can't
find space it tells the driver and the driver does eviction and
recalls validate?
Dave.
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel
next prev parent reply other threads:[~2020-09-24 23:14 UTC|newest]
Thread overview: 79+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-09-24 5:18 [PATCH 00/45] TTM move refactoring Dave Airlie
2020-09-24 5:18 ` [PATCH 01/45] drm/qxl: drop unused code Dave Airlie
2020-09-24 10:23 ` Christian König
2020-09-24 5:18 ` [PATCH 02/45] drm/ttm: handle the SYSTEM->TT path in same place as others Dave Airlie
2020-09-24 10:24 ` Christian König
2020-09-24 5:18 ` [PATCH 03/45] drm/amdgpu/ttm: handle tt moves properly Dave Airlie
2020-09-24 11:10 ` Christian König
2020-09-24 5:18 ` [PATCH 04/45] drm/radeon/ttm: handle ttm " Dave Airlie
2020-09-24 11:10 ` Christian König
2020-09-24 5:18 ` [PATCH 05/45] drm/nouveau/ttm: " Dave Airlie
2020-09-24 11:11 ` Christian König
2020-09-24 5:18 ` [PATCH 06/45] drm/vmwgfx: move null mem checks outside move notifies Dave Airlie
2020-09-24 11:12 ` Christian König
2020-09-24 5:18 ` [PATCH 07/45] drm/vmwgfx: add a move callback Dave Airlie
2020-09-24 11:13 ` Christian König
2020-09-24 5:18 ` [PATCH 08/45] drm/vram_helper: implement a ttm " Dave Airlie
2020-09-24 11:14 ` Christian König
2020-09-24 5:18 ` [PATCH 09/45] drm/ttm: make move callback compulstory Dave Airlie
2020-09-24 11:15 ` Christian König
2020-09-24 5:18 ` [PATCH 10/45] drm/ttm: refactor out common code to setup a new tt backed resource Dave Airlie
2020-09-24 11:17 ` Christian König
2020-09-24 5:18 ` [PATCH 11/45] drm/ttm: split out the move to system from move ttm code Dave Airlie
2020-09-24 11:22 ` Christian König
2020-09-24 5:18 ` [PATCH 12/45] drm/ttm: drop free old node wrapper Dave Airlie
2020-09-24 11:23 ` Christian König
2020-09-24 5:18 ` [PATCH 13/45] drm/ttm: use new move interface for known system->ttm moves Dave Airlie
2020-09-24 11:26 ` Christian König
2020-09-24 5:18 ` [PATCH 14/45] drm/ttm: add move old to system to drivers Dave Airlie
2020-09-24 12:02 ` Christian König
2020-09-24 5:18 ` [PATCH 15/45] drm/ttm: push copy unbind into drivers Dave Airlie
2020-09-24 12:17 ` Christian König
2020-09-24 5:18 ` [PATCH 16/45] drm/radeon/ttm: do move notify actions inside move Dave Airlie
2020-09-24 5:18 ` [PATCH 17/45] drm/amdgpu/ttm: handle invalidation in move callback Dave Airlie
2020-09-24 5:18 ` [PATCH 18/45] drm/nouveau: handle move notify inside " Dave Airlie
2020-09-30 5:53 ` Ben Skeggs
2020-09-24 5:18 ` [PATCH 19/45] drm/qxl/ttm: " Dave Airlie
2020-09-24 5:18 ` [PATCH 20/45] drm/vmwgfx/ttm: handle move notify inside move Dave Airlie
2020-09-24 5:18 ` [PATCH 21/45] drm/vram_helper: call move notify from the move callback Dave Airlie
2020-09-24 5:18 ` [PATCH 22/45] drm/ttm: don't call move notify around move Dave Airlie
2020-09-24 5:18 ` [PATCH 23/45] drm/ttm: move bind out of move_to_new_tt_mem Dave Airlie
2020-09-24 5:18 ` [PATCH 24/45] drm/ttm: handle binding in move callback Dave Airlie
2020-09-24 5:18 ` [PATCH 25/45] drm/ttm: don't call ttm_bo_move_ttm from drivers Dave Airlie
2020-09-24 5:18 ` [PATCH 26/45] drm/ttm: remove bind/unbind callbacks Dave Airlie
2020-09-24 13:32 ` Christian König
2020-09-24 5:18 ` [PATCH 27/45] drm/radeon/ttm: cleanup move exit paths Dave Airlie
2020-09-24 5:18 ` [PATCH 28/45] drm/nouveau/ttm: memcpy waits for bo already Dave Airlie
2020-09-24 5:18 ` [PATCH 29/45] drm/ttm: don't expose move old to system helper to drivers Dave Airlie
2020-09-24 5:18 ` [PATCH 30/45] drm/ttm: add a new invalidate notify callback Dave Airlie
2020-09-24 9:33 ` Daniel Vetter
2020-09-24 12:25 ` Christian König
2020-09-29 3:23 ` Dave Airlie
2020-09-29 7:05 ` Christian König
2020-09-24 5:18 ` [PATCH 31/45] drm/radeon: switch to " Dave Airlie
2020-09-24 5:18 ` [PATCH 32/45] drm/amdgpu/ttm: " Dave Airlie
2020-09-24 5:18 ` [PATCH 33/45] drm/nouveau/ttm: " Dave Airlie
2020-09-24 5:18 ` [PATCH 34/45] drm/qxl/ttm: move " Dave Airlie
2020-09-24 5:18 ` [PATCH 35/45] drm/vram-helper: move to invalidate callback Dave Airlie
2020-09-24 7:10 ` Thomas Zimmermann
2020-09-24 5:18 ` [PATCH 36/45] drm/ttm: drop move_notify callback Dave Airlie
2020-09-24 5:18 ` [PATCH 37/45] drm/ttm: add a helper to allocate a temp tt for copies Dave Airlie
2020-09-24 12:42 ` Christian König
2020-09-24 23:14 ` Dave Airlie [this message]
2020-09-25 7:39 ` Christian König
2020-09-25 8:16 ` Daniel Vetter
2020-09-25 8:18 ` Daniel Vetter
2020-09-25 9:34 ` Christian König
2020-09-25 13:17 ` Daniel Vetter
2020-09-25 13:40 ` Christian König
2020-09-25 13:56 ` Daniel Vetter
2020-09-24 5:18 ` [PATCH 38/45] drm/nouveau/ttm: use helper to allocate tt temp Dave Airlie
2020-09-24 5:18 ` [PATCH 39/45] drm/radeon/ttm: use new helper to create tmp tt Dave Airlie
2020-09-24 5:18 ` [PATCH 40/45] drm/amdgpu/ttm: use new ttm helper to create temp tt Dave Airlie
2020-09-24 5:18 ` [PATCH 41/45] drm/amdgpu/ttm: use helper function for caching/populate Dave Airlie
2020-09-24 5:18 ` [PATCH 42/45] drm/radeon/ttm: " Dave Airlie
2020-09-24 5:18 ` [PATCH 43/45] drm/nouveau/ttm: use helper for placement + populaate Dave Airlie
2020-09-24 5:18 ` [PATCH 44/45] drm/ttm: move more functionality into helper function Dave Airlie
2020-09-24 5:18 ` [PATCH 45/45] drm/ttm: add a new helper for a cleaning up after a ram move Dave Airlie
2020-09-24 6:40 ` [PATCH 00/45] TTM move refactoring Dave Airlie
2020-09-30 5:51 ` Ben Skeggs
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='CAPM=9tw6a6E+KZV7LceHUj=SHFaTe9M+KjjYgouhdX0gkDsuzw@mail.gmail.com' \
--to=airlied@gmail.com \
--cc=bskeggs@redhat.com \
--cc=christian.koenig@amd.com \
--cc=dri-devel@lists.freedesktop.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.