From: Maarten Lankhorst <dev@lankhorst.se>
To: intel-xe@lists.freedesktop.org
Cc: Maarten Lankhorst <dev@lankhorst.se>
Subject: [PATCH 0/4] drm/xe: Update migration tests.
Date: Fri, 2 May 2025 11:35:09 +0200 [thread overview]
Message-ID: <20250502093514.510621-1-dev@lankhorst.se> (raw)
Now that we have a reliable migration implementation, it makes sense to update
and clean up the tests.
I'm also adding a testcase that copies 1/3rd of VRAM to sysmem or another buffer,
to measure the performance impact of our implementation of migration.
This makes it easier to see if our migration is performant, or how changes to optimize
it affect the speed, compared to a direct copy of the same buffer, using the same commands,
but without our VM mapping done in migration.
For completeness, all 4 paths are tested:
- VRAM -> VRAM
- VRAM -> sysmem
- sysmem -> VRAM
- sysmem -> sysmem (same buffer, don't care about contents)
During my first testing, copying a buffer through migration is about 7% slower than performing the
same copy directly. This is likely the result from "<flush>update PT<flush>copy 8MB<flush> next job to repeat" compared
to "copy the entire BO in 8MB chunks in a single job" repeatedly.
Maarten Lankhorst (4):
drm/xe: Simplify migration kunit tests
drm/xe: Silence migrate debug message.
drm/xe: Fix small leak in migration error path.
drm/xe: Add migrate performance kunit test
drivers/gpu/drm/xe/tests/xe_migrate.c | 401 +++++++++++++++++++-------
drivers/gpu/drm/xe/xe_migrate.c | 64 ++--
2 files changed, 338 insertions(+), 127 deletions(-)
--
2.45.2
next reply other threads:[~2025-05-02 9:35 UTC|newest]
Thread overview: 17+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-05-02 9:35 Maarten Lankhorst [this message]
2025-05-02 9:35 ` [PATCH 1/4] drm/xe: Simplify migration kunit tests Maarten Lankhorst
2025-05-02 16:54 ` Matthew Brost
2025-05-02 9:35 ` [PATCH 2/4] drm/xe: Silence migrate debug message Maarten Lankhorst
2025-05-02 16:55 ` Matthew Brost
2025-05-02 9:35 ` [PATCH 3/4] drm/xe: Fix small leak in migration error path Maarten Lankhorst
2025-05-02 16:59 ` Matthew Brost
2025-05-02 9:35 ` [PATCH 4/4] drm/xe: Add migrate performance kunit test Maarten Lankhorst
2025-05-08 6:49 ` Matthew Brost
2025-05-02 11:02 ` ✓ CI.Patch_applied: success for drm/xe: Update migration tests Patchwork
2025-05-02 11:02 ` ✗ CI.checkpatch: warning " Patchwork
2025-05-02 11:03 ` ✓ CI.KUnit: success " Patchwork
2025-05-02 11:12 ` ✓ CI.Build: " Patchwork
2025-05-02 11:14 ` ✓ CI.Hooks: " Patchwork
2025-05-02 11:15 ` ✓ CI.checksparse: " Patchwork
2025-05-02 15:21 ` ✗ Xe.CI.Full: failure " Patchwork
2025-05-06 7:59 ` ✗ Xe.CI.BAT: " Patchwork
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=20250502093514.510621-1-dev@lankhorst.se \
--to=dev@lankhorst.se \
--cc=intel-xe@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.