UCT/ROCM/COPY: Use faster memcpy for device to host copies #4532
+89
−5
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
Can one of the admins verify this patch? |
|
ok to test |
shamisp
requested changes
Dec 2, 2019
Mellanox CI: FAILED on 3 of 25 workers (click for details)Note: the logs will be deleted after 10-Dec-2019
|
souravzzz
force-pushed
the
topic/sourav/rocm-memcpy-nt
branch
2 times, most recently
from
ce7608e
to
2bf407f
Compare
|
@souravzzz Do you happen to have performance numbers that can help support this optimization ? |
Mellanox CI: PASSED on 25 workers (click for details)Note: the logs will be deleted after 10-Dec-2019
|
|
Hi @shamisp here are some intra-node D2D numbers with rocm_copy transport that shows the improvements from this change.
|
Mellanox CI: PASSED on 25 workers (click for details)Note: the logs will be deleted after 10-Dec-2019
|
|
Looks impressive. What about d2h ? thanks |
|
@shamisp We see good improvement for D2H transfers as well.
|
|
Looks good to me. @yosefe ? |
yosefe
reviewed
Dec 4, 2019
souravzzz
force-pushed
the
topic/sourav/rocm-memcpy-nt
branch
from
2bf407f
to
80bbc67
Compare
|
Thanks for the feedback @yosefe. I have incorporated the suggested changes. |
yosefe
approved these changes
Dec 4, 2019
shamisp
approved these changes
Dec 4, 2019
Mellanox CI: PASSED on 25 workers (click for details)Note: the logs will be deleted after 11-Dec-2019
|
|
@souravzzz Curious to know if this test has the GPU touch the buffers transferred before each transfer. |
|
@Akshay-Venkatesh No these are results from the standard osu_latency benchmark |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
What
Use faster memcpy for device to host copies
Why ?
Native memcpy is slow for device to host copies
How
Use non-temporal load (MOVNTDQA) to implement memcpy