Update BLAS, LAPACK and linear algebra routines #103
Conversation
This file contains hidden or 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
- Allow general nda::MemoryArray types of rank 1 or 2
- do not allow conjugate expressions - update docs and tests
- Remove outer_product from nda::blas namespace - Update docs and tests
- move outer_product from blas/ger.hpp to linalg/outer_product.hpp
- move the generic dot routines from the blas to the linalg namespace - blas::dot and blas::dotc - thin wrappers around the BLAS routines - restrict to nda::MemoryVectors - linalg::dot linalg::dotc - allow scalar-scalar and vector-vector dot products - only call blas::dot and blas::dotc if it is possible without making a copy
- move matvecmul into linalg namespace and into its own linalg/matvecmul.hpp header - move blas::gemv_generic into linalg::detail namespace in linalg/matvecmul.hpp - linalg::matvecmul - allow non-contiguous matrices - move various helper lambdas into linalg::detail namespace - avoid unnecessary vector copies
- move blas::gemm_generic into linalg::detail namespace in linalg/matmul.hpp - linalg::matvecmul - allow non-contiguous matrices - move various helper lambdas into linalg::detail namespace
- remove explicit throw statements --> always return the LAPACK info - consistent checks and function layout
- remove explicit throw statement --> always return the LAPACK info - allow C-layout matrices
- require F-layout for B matrices
- lapack::getrf - allow C-layout matrices - lapack::getri - restrict to host (there is not getri for cuda) - lapack::getrs - allow conjugate lazy expressions - allow B to be either a vector or matrix - restrict B to F-layout
- move norm into linalg namespace
- allow vectors with different value types - require vectors to have a host compatible address space
…rse.hpp to matrix_functions.hpp
- rename inverse to inv, inverse_in_place to inv_in_place, inverse1_in_place to inv_in_place_1d, inverse2_in_place to inv_place_2d, inverse3_in_place to inv_in_place_3d - allow in place functions for host compatible address spaces - use getrf + getrs to allow device matrices in inv
- rename determinant to det, determinant_in_place to det_in_place - introduce det_1d, det_2d and det_3d - only allow matrices on host compatible address spaces and with double or complex value types - call optimized routines for matrices with less than 4 rows/columns
- replace throw statements with std::terminate()
- wrap the BLAS gerc routine - restrict to Fortran layout matrices
- rename linalg::eigenelements to linalg::eigh - rename linalg::eigenvalues to linalg::eigvalsh - rename linalg/eigenelements.hpp to linalg/eigh.hpp - call the LAPACK wrappers lapack::syev and lapack::heev
- increase overall test coverage - add tests for unified memory
- cusolverDn?gesvd only supports matrices with m >= n
- increase overall test coverage - add (some) tests for unified memory
- try to get compile time errors instead of runtime errors whenever possible
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.
nda::blasandnda::lapacknamespaces are only very thin wrappers around the BLAS/LAPACK callsnda::linalgnamespace and use numpy function names whenever it makes sense