Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Implement gesv_batch via gesv call #1877

Merged
merged 50 commits into from
Aug 8, 2024
Merged

Conversation

vlad-perevezentsev
Copy link
Collaborator

This PR suggests adding gesv_batch implementation as a separate function and provide iteration of the batch array within it to avoid allocation of each 2D array before calling gesv to improve performance and avoid serialization of all host tasks in the queue in MKL call;

Performance of dpnp.solve on Iris Xe
image

  • Have you provided a meaningful PR description?
  • Have you added a test, reproducer or referred to issue with a reproducer?
  • Have you tested your changes locally for CPU and GPU devices?
  • Have you made sure that new changes do not introduce compiler warnings?
  • Have you checked performance impact of proposed changes?
  • If this PR is a work in progress, are you filing the PR as a draft?

Copy link
Contributor

github-actions bot commented Jun 12, 2024

View rendered docs @ https://intelpython.github.io/dpnp/index.html

dpnp/backend/extensions/lapack/gesv.cpp Outdated Show resolved Hide resolved
dpnp/backend/extensions/lapack/gesv.hpp Show resolved Hide resolved
dpnp/backend/extensions/lapack/gesv.cpp Outdated Show resolved Hide resolved
dpnp/backend/extensions/lapack/gesv.cpp Outdated Show resolved Hide resolved
dpnp/backend/extensions/lapack/gesv.cpp Outdated Show resolved Hide resolved
dpnp/linalg/dpnp_utils_linalg.py Show resolved Hide resolved
dpnp/backend/extensions/lapack/gesv.cpp Outdated Show resolved Hide resolved
dpnp/backend/extensions/lapack/gesv.cpp Outdated Show resolved Hide resolved
dpnp/backend/extensions/lapack/gesv.cpp Outdated Show resolved Hide resolved
dpnp/backend/extensions/lapack/gesv_batch.cpp Outdated Show resolved Hide resolved
dpnp/backend/extensions/lapack/gesv_batch.cpp Outdated Show resolved Hide resolved
dpnp/backend/extensions/lapack/gesv_batch.cpp Outdated Show resolved Hide resolved
dpnp/backend/extensions/lapack/gesv_batch.cpp Outdated Show resolved Hide resolved
dpnp/linalg/dpnp_utils_linalg.py Outdated Show resolved Hide resolved
dpnp/linalg/dpnp_utils_linalg.py Outdated Show resolved Hide resolved
Copy link
Contributor

@antonwolfy antonwolfy left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It's great improvement of dpnp.solve. Thank you @vlad-perevezentsev, LGTM!

@vlad-perevezentsev vlad-perevezentsev merged commit 1e5ba88 into master Aug 8, 2024
42 of 45 checks passed
@vlad-perevezentsev vlad-perevezentsev deleted the impl_gesv_batch_via_gesv branch August 8, 2024 08:36
github-actions bot added a commit that referenced this pull request Aug 8, 2024
* First working version with transpose and C contig

* Second working version with moveaxis, transpose and F contig

* Add more shape checks

* Pass sycl::queue by reference for gesv/gesv_batch

* Update _batched_solve implementation

* Remove old impl in _batched_solve

* Use py::gil_scoped_release before gesv call

* Move gesv_batch to gesv_batch.cpp

* Improve gesv_batch with independent linear streams

* Extend checks for gesv/gesv_batch

* Add common_gesv_checks

* Release GIL in gesv_batch_impl

* Remove host_task_events from gesv

* Use check_zeros_shape in gesv and gesv_batch

* Add additional checks for gesv_impl

* Move alloc_scratchpad to common_helpers.hpp

* Use helper::alloc_scratchpad in gesv_batch_impl

* Remove current_scratch_gesv check

* Remove lda, ldb pass to gesv_batch_impl, gesv_impl

* Use const and constexpr in gesv/gesv_batch

* Use dpnp.reshape in _batched_solve

* Implement alloc_ipiv in common_helpers.hpp

* Add gesv_common_utils.hpp

* Implement handle_lapack_exc function

* Use try/catch for scratchpad/ipiv allocation

* Update alloc_scratchpad/alloc_ipiv

* gesv_scratchpad_size can be 0

* Implement help functions alloc_ipiv/alloc_scratchpad

* Reuse alloc_scratchpad/ipiv in batch versions 1e5ba88
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants