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Outdated suggestions cannot be applied.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.
Description
This PR adds
?gemm_batch
routines and newxerblai
error handler for batched routines.?gemm_batch
routines can be extremely useful in case of large set of small matrixes in the case of optimal implementation for multicore machines.See more about performance improvement by the following link: https://www.intel.com/content/www/us/en/developer/articles/technical/introducing-batch-gemm-operations.html
In Intel MKL and cublas, there are corresponding routines.
A good article about batched API for BLAS/LAPACK can be found here: https://eprints.maths.manchester.ac.uk/2464/1/batched_api.pdf
I took an API from Intel MKL since it can take both options for
?gemm_batch
for fixed size matrixes, and for different size matrixes. While GROUP_COUNT specifies the number of matrixes' shapes, GROUP_SIZE specifies the number of matrixes with given shapes.Note, in current implementation, some features from Fortran 2003 are used. Another option is to use
LOC
operation which is an extension for Fortran standard, so I would prefer not to use it.Checklist