UCT/CUDA: Runtime CUDA >= 12.3 to enable VMM #10396
Open
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
tvegas1
force-pushed
the
cuda_ctx_set_flags_runtime
branch
from
1ce967f to
68a5f51
Compare
rakhmets
reviewed
Dec 20, 2024
rakhmets
reviewed
Dec 20, 2024
Comment on lines
891
to
896
| status = uct_cuda_copy_md_check_is_ctx_set_flags_supported(); | ||
| if ((status != UCS_OK) && (md->config.enable_fabric != UCS_NO)) { | ||
| ucs_warn("disabled fabric memory allocations as cuda driver " | ||
| "library does not support cuCtxSetFlags()"); | ||
| md->config.enable_fabric = UCS_NO; | ||
| } |
There was a problem hiding this comment.
Suggested change
| status = uct_cuda_copy_md_check_is_ctx_set_flags_supported(); | |
| if ((status != UCS_OK) && (md->config.enable_fabric != UCS_NO)) { | |
| ucs_warn("disabled fabric memory allocations as cuda driver " | |
| "library does not support cuCtxSetFlags()"); | |
| md->config.enable_fabric = UCS_NO; | |
| } | |
| if (md->config.enable_fabric != UCS_NO) { | |
| status = uct_cuda_copy_md_check_is_ctx_set_flags_supported(); | |
| if (status != UCS_OK) { | |
| if (md->config.enable_fabric == UCS_YES) { | |
| ucs_error("fabric memory allocation requested but cuda driver " | |
| "library does not support cuCtxSetFlags()"); | |
| goto err_free_md; | |
| } else { | |
| ucs_diag("disabled fabric memory allocations as cuda driver " | |
| "library does not support cuCtxSetFlags()"); | |
| md->config.enable_fabric = UCS_NO; | |
| } | |
| } | |
| } |
There was a problem hiding this comment.
this would not work as we will try to use set ctx even if fabric is not enabled.
There was a problem hiding this comment.
Maybe we can check md->config.enable_fabric instead of ctx_set_flags_func in uct_cuda_copy_sync_memops.
| CUresult cu_err; | ||
|
|
||
| if (status == UCS_ERR_INVALID_ADDR) { | ||
| pthread_mutex_lock(&lock); |
There was a problem hiding this comment.
Why do you need a mutex here?
There was a problem hiding this comment.
theoretical: multiple workers writing func pointer
|
we have tests for different cuda versions, which include cuda memory hooks (for example, Test Cuda Docker ubuntu18_cuda_12_0). can we add a test that would have caught the new api usage? |
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?
Do not use
cuCtxSetFlags()if CUDA driver does not support it.Why?
Unresolved symbol for
cuCtxSetFlagson CUDA driver < 12.1 causes crash.How?
Assumptions:
cuCtxSetFlagsis only needed for VMM, which has UCX support starting from CUDA driver >= 12.3cuCtxSetFlagsis not strictly needed for malloc asyncTesting
Locally tested, needs final testing on platform with actual older drivers.