[antonwolfy]

dpnp.scipy.sparse.csr_matrix.sort_indices sorted the stored entries by (row, column) with two stable argsorts: one on the column indices, then one on the permuted row ids. This PR sorts once on the int64 key row * ncols + col. CSR rows are already in order, so the key is nearly sorted, and one merge sort of it is much cheaper than two sorts of fully permuted data.

• The per-row offsets come from repeat(arange(0, nrows * ncols, ncols), row_lengths), so only one elementwise add runs before the sort. Computing row_ids * ncols + indices instead adds kernels that made small matrices slower than before on GPU (0.71-0.95x of master below about 3·10⁴ stored values).
• The offsets are int64 even for int32 indices, so the key can't overflow. has_sorted_indices runs check_format first, which bounds the column indices to [0, ncols).
• When nrows * ncols >= 2**63 the key doesn't fit int64, and the previous two-pass sort is kept. Its row ids are now int64 too.
• kind="stable" (merge sort for int64) stays: radix sort on the int64 key was slower on both devices measured.

Two new tests: an int32-index matrix whose key overflows int32 (shape 3001 x 10⁶), and shapes just below, at and above the int64 key limit, the last of which takes the fallback. They fail when the offsets use the index dtype or when the fallback is removed. Both allocate only a few elements whatever the shape, so they aren't marked slow.

Performance

sort_indices alone (construction excluded), master vs this PR loaded in the same process, median time, on a Panther Lake H machine: the Xe3 iGPU on Level Zero and the Core Ultra 7 366H on OpenCL. Each matrix has 8 or about 1024 stored values per row, with columns shuffled within each row; results are identical to master in every case.

nnz	shape	index	GPU master, ms	GPU PR, ms	GPU speedup	CPU master, ms	CPU PR, ms	CPU speedup
10^4	1250 x 1250	int32	3.8	3.7	1.04x	3.6	3.1	1.16x
10^4	1250 x 1250	int64	5.4	5.1	1.07x	3.7	3.1	1.20x
10^4	9 x 4096	int32	12.5	12.1	1.03x	3.5	3.1	1.14x
10^4	9 x 4096	int64	13.4	12.6	1.07x	3.8	3.0	1.25x
10^5	12500 x 12500	int32	19.4	15.2	1.28x	8.5	5.2	1.62x
10^5	12500 x 12500	int64	18.1	14.3	1.26x	8.4	5.1	1.63x
10^5	97 x 4096	int32	15.5	14.3	1.08x	7.9	4.9	1.63x
10^5	97 x 4096	int64	16.9	13.9	1.22x	7.8	4.7	1.66x
10^6	125000 x 125000	int32	51.7	23.0	2.25x	56.8	18.9	3.01x
10^6	125000 x 125000	int64	178	24.3	7.35x	71.6	23.6	3.03x
10^6	976 x 4096	int32	40.1	27.4	1.46x	70.4	23.8	2.96x
10^6	976 x 4096	int64	60.6	30.1	2.01x	68.5	24.0	2.85x
10^7	1250000 x 1250000	int32	1118	166	6.75x	855	160	5.33x
10^7	1250000 x 1250000	int64	1125	174	6.48x	916	166	5.52x
10^7	9765 x 9765	int32	325	167	1.95x	708	195	3.63x
10^7	9765 x 9765	int64	366	174	2.10x	719	193	3.73x

dpnp/tests/test_scipy_sparse_linalg.py passes (406 tests) on both of those devices.

• 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?

[Copilot]

🟢 Approval recommended

No active review issues are recorded for the CSR index-sorting optimization.

No open issues 🎉

What changed in this PR

This PR optimizes CSR index sorting in dpnp.scipy.sparse by replacing two stable sorts with one stable sort on a combined row-and-column key when it fits in int64.

Changes:

• Uses int64 row offsets and preserves the two-pass fallback for shapes at or above the int64 key limit.
• Adds regression tests for int32 key overflow and shapes around the int64 limit.
• Documents the performance improvement in the changelog.

File	Description
dpnp/​tests/​test_scipy_sparse_linalg.py	Adds overflow and large-shape sorting regression tests.
dpnp/​scipy/​sparse/​_csr.py	Implements combined-key sorting with an overflow-safe fallback.
CHANGELOG.md	Records the CSR index-sorting performance improvement.

_{Review effort: Balanced}

[github-actions]

View rendered docs @ https://intelpython.github.io/dpnp/pull/3134/index.html

[github-actions]

Array API standard conformance tests for dpnp=0.22.0dev0=np2py314h8d9cdd5_4 ran successfully.
Passed: 1491
Failed: 0
Skipped: 6

[coveralls]

coverage: 82.638% (+0.002%) from 82.636% — sort-indices-composite-key into master