Implement dpnp.linalg.lu_solve() 2D inputs
#2575
Open
+426
−0
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
vlad-perevezentsev
requested review from
antonwolfy,
AlexanderKalistratov,
vtavana and
ndgrigorian
as code owners
antonwolfy
changed the title
dpnp.linalg.solve() 2D inputs dpnp.linalg.lu_solve() 2D inputs
|
Array API standard conformance tests for dpnp=0.19.0dev3=py313h509198e_27 ran successfully. |
|
View rendered docs @ https://intelpython.github.io/dpnp/pull/2575/index.html |
antonwolfy
reviewed
Sep 5, 2025
| @@ -966,6 +968,75 @@ def lu_factor(a, overwrite_a=False, check_finite=True): | |||
| return dpnp_lu_factor(a, overwrite_a=overwrite_a, check_finite=check_finite) | |||
|
|
|||
|
|
|||
| def lu_solve(lu_and_piv, b, trans=0, overwrite_b=False, check_finite=True): | |||
| """ | |||
| Solve an equation system, a x = b, given the LU factorization of `a` | |||
There was a problem hiding this comment.
Suggested change
| Solve an equation system, a x = b, given the LU factorization of `a` | |
| Solve an equation system, a x = b, given the LU factorization of `a`. |
|
|
||
| Parameters | ||
| ---------- | ||
| (lu, piv) : {tuple of dpnp.ndarrays or usm_ndarrays} |
There was a problem hiding this comment.
Now it is rendered in a wrong way in the documentation
Suggested change
| (lu, piv) : {tuple of dpnp.ndarrays or usm_ndarrays} | |
| lu, piv : {tuple of dpnp.ndarrays or usm_ndarrays} |
| Parameters | ||
| ---------- | ||
| (lu, piv) : {tuple of dpnp.ndarrays or usm_ndarrays} | ||
| LU factorization of matrix `a` ((M, N)) together with pivot indices. |
There was a problem hiding this comment.
Why do we need double brackets here?
Suggested change
| LU factorization of matrix `a` ((M, N)) together with pivot indices. | |
| LU factorization of matrix `a` (M, N) together with pivot indices. |
| A[i, i] = A.dtype.type(off + 1.0) | ||
| return A | ||
|
|
||
| @pytest.mark.parametrize("shape", [(1, 1), (2, 2), (3, 3), (5, 5)]) |
There was a problem hiding this comment.
I failed to find a test with (M, N) shapes, only with (M, M)
| _manager = dpu.SequentialOrderManager[exec_q] | ||
| dep_evs = _manager.submitted_events | ||
|
|
||
| # oneMKL LAPACK getrf overwrites `a`. |
There was a problem hiding this comment.
Suggested change
| # oneMKL LAPACK getrf overwrites `a`. | |
| # oneMKL LAPACK getrf overwrites `lu`. |
Comment on lines
+2585
to
+2592
| ht_ev, dep_ev = ti._copy_usm_ndarray_into_usm_ndarray( | ||
| src=b_usm_arr, | ||
| dst=b_h.get_array(), | ||
| sycl_queue=b.sycl_queue, | ||
| depends=_manager.submitted_events, | ||
| ) | ||
| _manager.add_event_pair(ht_ev, dep_ev) | ||
| dep_ev = [dep_ev] |
There was a problem hiding this comment.
Isn't we need to override dep_evs with list of copy events?
Suggested change
| ht_ev, dep_ev = ti._copy_usm_ndarray_into_usm_ndarray( | |
| src=b_usm_arr, | |
| dst=b_h.get_array(), | |
| sycl_queue=b.sycl_queue, | |
| depends=_manager.submitted_events, | |
| ) | |
| _manager.add_event_pair(ht_ev, dep_ev) | |
| dep_ev = [dep_ev] | |
| ht_ev, b_copy_ev = ti._copy_usm_ndarray_into_usm_ndarray( | |
| src=b_usm_arr, | |
| dst=b_h.get_array(), | |
| sycl_queue=b.sycl_queue, | |
| depends=dep_evs, | |
| ) | |
| _manager.add_event_pair(ht_ev, b_copy_ev) | |
| dep_evs = [lu_copy_ev, b_copy_ev] |
| else: | ||
| # input is suitable for in-place modification | ||
| b_h = b | ||
| dep_ev = _manager.submitted_events |
There was a problem hiding this comment.
and then here:
Suggested change
| dep_ev = _manager.submitted_events | |
| dep_evs = [lu_copy_ev] |
| b_h = b | ||
| dep_ev = _manager.submitted_events | ||
|
|
||
| # oneMKL LAPACK getrf overwrites `a`. |
There was a problem hiding this comment.
Suggested change
| # oneMKL LAPACK getrf overwrites `a`. | |
| # oneMKL LAPACK getrf overwrites `piv`. |
Comment on lines
+2598
to
+2611
| # oneMKL LAPACK getrf overwrites `a`. | ||
| piv_h = dpnp.empty_like(piv, order="F", usm_type=res_usm_type) | ||
|
|
||
| # use DPCTL tensor function to fill the сopy of the pivot array | ||
| # from the pivot array | ||
| ht_ev, piv_copy_ev = ti._copy_usm_ndarray_into_usm_ndarray( | ||
| src=piv_usm_arr, | ||
| dst=piv_h.get_array(), | ||
| sycl_queue=piv.sycl_queue, | ||
| depends=dep_evs, | ||
| ) | ||
| _manager.add_event_pair(ht_ev, piv_copy_ev) | ||
| # MKL lapack uses 1-origin while SciPy uses 0-origin | ||
| piv_h += 1 |
There was a problem hiding this comment.
it'd make sense to do before line#2581, to build a list of copy events dep_evs = [lu_copy_ev, piv_copy_ev, b_copy_ev]
| depends=dep_evs, | ||
| ) | ||
| _manager.add_event_pair(ht_ev, piv_copy_ev) | ||
| # MKL lapack uses 1-origin while SciPy uses 0-origin |
There was a problem hiding this comment.
I guess SciPy also uses MKL to call getrs, so it seems unclear for me.
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.
This PR suggests adding
dpnp.linalg.lu_solve()for 2D arrays similar to scipy.linalg.lu_solveSupport for ND inputs will be added in the next phase.