Replace internally developed CSR GEMM with a call to MKL #2959
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
Pull changes from oneDAL main branch
|
/intelci: run |
|
/intelci: run |
Pull the changes from main branch
…nto dev/kmeans_csr_upd
|
/intelci: run |
Vika-F
requested review from
Alexsandruss,
samir-nasibli and
Alexandr-Solovev
as code owners
Vika-F
requested review from
david-cortes-intel and
ethanglaser
and removed request for
samir-nasibli and
Alexsandruss
ahuber21
reviewed
Nov 13, 2024
|
|
||
| const auto distances_ptr = distances.get_data(); | ||
| // Workaround. Sparse gemm cannot accept transposed dense inputs in oneMKL 2025.0. |
|
|
||
| const auto finalize_range = | ||
| bk::make_multiple_nd_range_2d({ num_clusters, local_size }, { 1, local_size }); | ||
|
|
||
| // Compute the array of centroids by dividing the respective sums of observations | ||
| // by the number of observations in each centroid |
There was a problem hiding this comment.
Suggested change
| // by the number of observations in each centroid | |
| // by the number of observations in each centroid |
Comment on lines
45
to
46
| #define INSTANTIATE_SINGLE_NODE(F, M, T) \ | ||
| template struct ONEDAL_EXPORT infer_ops_dispatcher<dal::detail::data_parallel_policy, F, M, T>; |
There was a problem hiding this comment.
Suggested change
| #define INSTANTIATE_SINGLE_NODE(F, M, T) \ | |
| template struct ONEDAL_EXPORT infer_ops_dispatcher<dal::detail::data_parallel_policy, F, M, T>; | |
| #define INSTANTIATE_NON_DISTRIBUTED(F, M, T) \ | |
| template struct ONEDAL_EXPORT infer_ops_dispatcher<dal::detail::data_parallel_policy, F, M, T>; |
Maybe that's better? With multi-tile GPUs or single-node multi-GPU setups we would be using spdm on a single node?
|
Thanks for adding extensive & helpful comments! |
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.
handle_empty_clusterskernel from sparse implementation. Now both dense and sparse implementations use the same kernel for empty clusters handling and give the same results on the same input data passed in sparse and dense layout respectively.PR completeness and readability
Testing
All the failures in CI are not related to sparse K-means algorithm.
No new functionality was introduced.
Performance