Move `_tensor_reductions_impl` extension and use it for dpnp by vlad-perevezentsev · Pull Request #2794 · IntelPython/dpnp

vlad-perevezentsev · 2026-03-03T13:38:09Z

This PR completely moves _tensor_reductions_impl pybind11 extension into dpctl_ext.tensor and extends dpctl_ext.tensor Python API with the functions:

all
any
diff
argmax
argmin
count_nonzero
logsumexp
max
min
prod
reduce_hypot
sum

reusing them in dpnp

Have you provided a meaningful PR description?
Have you added a test, reproducer or referred to an 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?
Have you added documentation for your changes, if necessary?
Have you added your changes to the changelog?

…r and reuse them

github-actions · 2026-03-03T14:20:55Z

View rendered docs @ https://intelpython.github.io/dpnp/pull/2794/index.html

github-actions · 2026-03-03T14:37:04Z

Array API standard conformance tests for dpnp=0.20.0dev3=py313h509198e_153 ran successfully.
Passed: 1356
Failed: 1
Skipped: 16

antonwolfy · 2026-03-04T15:02:57Z

dpctl_ext/tensor/libtensor/include/kernels/reductions.hpp

+#include <complex>
+#include <cstddef>
+#include <cstdint>
+#include <stdexcept>


Suggested change

#include <stdexcept>

antonwolfy · 2026-03-04T17:22:10Z

dpctl_ext/tensor/libtensor/source/reductions/reduce_hypot.cpp

+#include <pybind11/pybind11.h>
+#include <pybind11/stl.h>
+
+#include "kernels/reductions.hpp"


missing include:

Suggested change

#include "kernels/reductions.hpp"

#include "kernels/reductions.hpp"

#include "utils/sycl_utils.hpp"

antonwolfy · 2026-03-04T17:27:46Z

dpctl_ext/tensor/libtensor/source/tensor_reductions.cpp

+
+#include "reductions/reduction_common.hpp"
+
+namespace py = pybind11;


seems never used

Suggested change

namespace py = pybind11;

antonwolfy · 2026-03-04T17:31:22Z

dpctl_ext/tensor/libtensor/source/reductions/reduction_over_axis.hpp

+
+#include <algorithm>
+#include <cstddef>
+#include <cstdint>


not used

Suggested change

#include <cstdint>

antonwolfy · 2026-03-04T17:34:06Z

dpctl_ext/tensor/libtensor/source/reductions/reduction_atomic_support.hpp

+//===---------------------------------------------------------------------===//
+
+#pragma once
+#include <complex>


not used

Suggested change

#include <complex>

antonwolfy · 2026-03-04T18:01:17Z

dpctl_ext/tensor/libtensor/source/reductions/argmin.cpp

+        auto argmin_pyapi = [&](const arrayT &src, int trailing_dims_to_reduce,
+                                const arrayT &dst, sycl::queue &exec_q,
+                                const event_vecT &depends = {}) {
+            using dpctl::tensor::py_internal::py_search_over_axis;


Suggested change

using dpctl::tensor::py_internal::py_search_over_axis;

antonwolfy · 2026-03-04T18:01:30Z

dpctl_ext/tensor/libtensor/source/reductions/logsumexp.cpp

+                                   int trailing_dims_to_reduce,
+                                   const arrayT &dst, sycl::queue &exec_q,
+                                   const event_vecT &depends = {}) {
+            using dpctl::tensor::py_internal::py_tree_reduction_over_axis;


Suggested change

using dpctl::tensor::py_internal::py_tree_reduction_over_axis;

antonwolfy · 2026-03-04T18:01:39Z

dpctl_ext/tensor/libtensor/source/reductions/logsumexp.cpp

+
+        auto logsumexp_dtype_supported = [&](const py::dtype &input_dtype,
+                                             const py::dtype &output_dtype) {
+            using dpctl::tensor::py_internal::py_tree_reduction_dtype_supported;


Suggested change

using dpctl::tensor::py_internal::py_tree_reduction_dtype_supported;

antonwolfy · 2026-03-04T18:01:50Z

dpctl_ext/tensor/libtensor/source/reductions/max.cpp

+        auto max_pyapi = [&](const arrayT &src, int trailing_dims_to_reduce,
+                             const arrayT &dst, sycl::queue &exec_q,
+                             const event_vecT &depends = {}) {
+            using dpctl::tensor::py_internal::py_reduction_over_axis;


Suggested change

using dpctl::tensor::py_internal::py_reduction_over_axis;

antonwolfy · 2026-03-04T18:02:03Z

dpctl_ext/tensor/libtensor/source/reductions/min.cpp

+        auto min_pyapi = [&](const arrayT &src, int trailing_dims_to_reduce,
+                             const arrayT &dst, sycl::queue &exec_q,
+                             const event_vecT &depends = {}) {
+            using dpctl::tensor::py_internal::py_reduction_over_axis;


Suggested change

using dpctl::tensor::py_internal::py_reduction_over_axis;

vlad-perevezentsev added 13 commits March 3, 2026 03:44

Initialize _tensor_reductions_impl extension and move _all

ad814fb

Add TODO with incorrect logic in reductions_over_axis.hpp

c4f2496

Move ti.all() to dpctl_ext.tensor and reuse it in dpnp

bd3add0

Move _any to _tensor_reductions_impl

b1953df

Move ti.any()/ti.diff() and reuse them in dpnp

32802d6

Move _min_over_axis/_max_over_axis to _tensor_reductions_impl

1b4e498

Move ti.min()/max() and reuse it in dpnp

aa313ff

Move _argmax/argmin_over_axis to _tensor_reductions_impl

c6d600a

Move ti.argmax()/argmin() to dpctl_ext.tensor and reuse them in dpnp

8ae933d

Move _prod/sum_over_axis to _tensor_reductions_impl

2ec3cc7

Move ti.sum()/prod() to dpctl_ext.tensor and reuse them in dpnp

6b27b1b

Move _logsumexp/hypot_over_axis to _tensor_reductions_impl

3041d7d

Move ti.count_nonzero()/logsumexp()/reduce_hypot() to dpctl_ext.tenso…

4872bb6

…r and reuse them

vlad-perevezentsev self-assigned this Mar 3, 2026

vlad-perevezentsev mentioned this pull request Mar 3, 2026

Restructure dpctl, delegating all array functionality to dpnp #2743

Open

vlad-perevezentsev marked this pull request as ready for review March 4, 2026 12:39

vlad-perevezentsev requested review from antonwolfy and ndgrigorian as code owners March 4, 2026 12:39

antonwolfy reviewed Mar 4, 2026

View reviewed changes

Merge move_tensor_sorting_impl into move_tensor_reductions_impl_ext

c9644ac

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Move `_tensor_reductions_impl` extension and use it for dpnp#2794

Move `_tensor_reductions_impl` extension and use it for dpnp#2794
vlad-perevezentsev wants to merge 14 commits intomove_tensor_sorting_implfrom
move_tensor_reductions_impl_ext

vlad-perevezentsev commented Mar 3, 2026

Uh oh!

github-actions bot commented Mar 3, 2026

Uh oh!

github-actions bot commented Mar 3, 2026

Uh oh!

antonwolfy Mar 4, 2026

Uh oh!

antonwolfy Mar 4, 2026

Uh oh!

antonwolfy Mar 4, 2026

Uh oh!

antonwolfy Mar 4, 2026

Uh oh!

antonwolfy Mar 4, 2026

Uh oh!

antonwolfy Mar 4, 2026

Uh oh!

antonwolfy Mar 4, 2026

Uh oh!

antonwolfy Mar 4, 2026

Uh oh!

antonwolfy Mar 4, 2026

Uh oh!

antonwolfy Mar 4, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

	#include "kernels/reductions.hpp"
	#include "kernels/reductions.hpp"
	#include "utils/sycl_utils.hpp"


		#include "reductions/reduction_common.hpp"

		namespace py = pybind11;

Conversation

vlad-perevezentsev commented Mar 3, 2026

Uh oh!

github-actions bot commented Mar 3, 2026

Uh oh!

github-actions bot commented Mar 3, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants