Use int64 row_ptr in _build_row_csr_numba (#1388)#1391
Merged
Conversation
Cumulative edge counts in _build_row_csr_numba previously stored to an int32 row_ptr / diff / running scalar / offsets array. On a tall raster (~50k rows) carrying many long polygon edges, the prefix sum could pass 2**31 - 1 and wrap, after which np.empty(total, dtype=np.int32) either raised on a negative size or allocated an undersized buffer that the Pass 2 fill walked off the end of inside the numba kernel. Cast row_ptr, diff, running, and offsets to int64. col_idx values are edge indices and stay int32. Downstream consumers (the GPU scanline kernel and np.diff(row_ptr).max()) accept int64 unchanged. Tests cover the dtype contract (row_ptr is int64 on both the populated and empty-edge paths), int64 representability past int32 max, CSR correctness for overlapping edges, and an end-to-end rasterize call.
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
row_ptr,diff,running, andoffsetsin_build_row_csr_numbato int64 so the cumulative edge count cannot wrap on a tall raster with many long polygon edgescol_idxat int32 (it stores edge indices, not cumulative counts)TestBuildRowCsrInt64covering the dtype contract, the empty-edges short-circuit, int64 representability past the int32 boundary, CSR layout for overlapping edges, and an end-to-endrasterize()callFixes #1388. Audit follow-up to #1223.
Test plan
pytest xrspatial/tests/test_rasterize.pypasses (143 passed, 2 skipped on this CPU host)pytest xrspatial/tests/test_rasterize.py::TestBuildRowCsrInt64 -vpasses (5/5)