Skip to content

Commit b502930

Browse files
authored
geotiff: share parsed VRTDataset across chunk tasks via dask.delayed (#1923) (#1927)
* geotiff: share parsed VRTDataset across chunk tasks via dask.delayed (#1923) _read_vrt_chunked in xrspatial/geotiff/_backends/vrt.py passed parsed_vrt=vrt as a plain kwarg to each per-chunk dask.delayed call, so dask embedded the full VRTDataset (filenames, src/dst rects, per-source nodata) into every task's kwargs. For an N-source VRT split into M chunks the graph held N*M copies of the source list; under distributed/process schedulers each task pickle shipped the full payload independently. Structurally verified pre-fix: 64 of 64 _vrt_chunk_read tasks held an inline VRTDataset in kwargs['parsed_vrt']. A 1000-source VRT split into 1000 chunks built a ~57 MB driver graph (60 KB pickled VRTDataset x 1000 task copies). Fix wraps the dataset in dask.delayed(vrt, pure=True) once before the per-chunk loop and threads that single shared reference through parsed_vrt=, matching the http_meta_key pattern in _backends/dask.py:147 and the meta_key pattern in _backends/gpu.py:1035. 3 new tests in test_vrt_chunked_shared_dataset_1923.py: a structural check that no task's kwargs['parsed_vrt'] is an inline VRTDataset, a decode-equivalence check vs the eager read, and a band-kwarg validation check to confirm the wrap does not change error semantics. State CSV: Pass 9 entry added; SAFE/IO-bound verdict holds. * geotiff: address copilot review on #1927 test module Remove unused pickle/tempfile imports flagged by flake8 F401 and reword the in-process scheduler note in the docstring so it no longer implies pickling happens locally.
1 parent 837ec08 commit b502930

3 files changed

Lines changed: 182 additions & 1 deletion

File tree

.claude/sweep-performance-state.csv

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -18,6 +18,7 @@ fire,2026-03-31T18:00:00Z,SAFE,compute-bound,0,,
1818
flood,2026-03-31T18:00:00Z,SAFE,compute-bound,0,,
1919
focal,2026-03-31T18:00:00Z,SAFE,compute-bound,0,,
2020
geodesic,2026-03-31T18:00:00Z,N/A,compute-bound,0,,
21+
geotiff,2026-05-15,SAFE,IO-bound,0,1923,"Pass 9 (2026-05-15): re-audit after #1888 writers/_writers split, #1885-1888 backends extraction, #1909 chunked GPU dtype cast, #1910 band-kwarg validation, #1911-1918 various. 1 new MEDIUM found and fixed. _read_vrt_chunked in _backends/vrt.py:500 embedded the full parsed VRTDataset in every chunk task's kwargs (kwargs['parsed_vrt']=vrt) instead of routing it through a shared dask.delayed(vrt, pure=True) like the sibling http_meta_key (dask.py:147) and GDS meta_key (gpu.py:1035) patterns. Under distributed/process schedulers each task pickle ships the full source list independently; structurally verified pre-fix: 64 of 64 _vrt_chunk_read tasks held an inline VRTDataset in kwargs (1000-source VRT * 1000 chunks => ~57 MB graph). Filed #1923, fixed in this branch; 3 new tests in test_vrt_chunked_shared_dataset_1923.py (structural shared-input check, decode unchanged, band validation unchanged). 438 existing VRT tests pass; the 1 unrelated tile_size=4 GPU validator failure predates this change. Also noted but not fixed: LOW _apply_nodata_mask_gpu (_backends/_gpu_helpers.py:62-67) uses cupy.where alloc for fresh GPU buffers where in-place arr_gpu[mask] = nan would save one chunk-sized allocation; deferred since arr buffers are small (<= chunk size) and current code is already correct. SAFE/IO-bound verdict holds. | Pass 8 (2026-05-12): 1 new MEDIUM found and fixed. _assemble_standard_layout/_assemble_cog_layout returned bytes(bytearray), doubling peak memory transiently during eager writes. Filed #1756, fixed by returning the bytearray directly. Measured: 95 MB uint8 raster peak drops 202 MB -> 107 MB. _write_bytes / parse_header already accepted the buffer protocol so the change is transparent to callers. 6 new tests in test_assemble_layout_no_bytes_copy_1756.py. 2123 existing geotiff tests pass; the 10 unrelated failures (test_no_georef_windowed_coords_1710, test_predictor2_big_endian_gpu_1517) reference the now-private read_to_array attribute (commit 8adb749, issue #1708) and predate this change. SAFE/IO-bound verdict holds. | Pass 7 (2026-05-12): re-audit identified 4 MEDIUM findings, all real, all backed by microbenches. (1) unpack_bits sub-byte loops for bps=2/4/12 in _compression.py:836-878 were 100-200x slower than vectorised numpy (filed #1713, fixed in this branch: bps=4 2M pixels drops from 165ms to 3ms = 55x; bps=2/12 similar). (2) _write_vrt_tiled at __init__.py:1708 uses scheduler='synchronous' on independent tile writes; measured 33% slowdown on 256-tile zstd write vs threads scheduler (filed #1714, no fix yet). (3) _nvcomp_batch_compress at _gpu_decode.py:2522-2526 still does per-tile cupy.get().tobytes() despite #1552 / #1659 fixing the same pattern elsewhere; measured 45% reduction with concat+single get on n=1024 (filed #1712, no fix yet). (4) _nvcomp_batch_compress at _gpu_decode.py:2457 uses per-tile cupy.empty allocations; 1024 tiles 16KB drops from 4.7ms to 1.0ms with single contiguous + views (bundled into #1712). Cat 6 OOM verdict: SAFE/IO-bound holds -- read_geotiff_dask caps task count at _MAX_DASK_CHUNKS=50_000 and per-chunk memory is bounded by chunk size. _inflate_tiles_kernel resource usage on Ampere: 67 regs/thread, 2896B local/thread, 8192B shared/block (LZW kernel: 29 regs, 24576B shared) -- register pressure under control; high local memory in inflate is unavoidable (LZ77 state) but only thread 0 in each block uses it. | Pass 4 (2026-05-10): re-audit after #1559 (centralise attrs across all read backends). New _populate_attrs_from_geo_info helper at __init__.py:301 runs once per read, not per-chunk -- no perf impact. Probe: 2560x2560 deflate-tiled file opened via read_geotiff_dask yields 400 tasks (4 tasks/chunk for 100 chunks), well under 1M cap. read_geotiff_gpu(1024x1024) returns cupy.ndarray end-to-end with no host round-trip (226ms incl. write+decode). No new HIGH/MEDIUM findings. SAFE/IO-bound holds. | Pass 3 (2026-05-10): SAFE/IO-bound. Audited 4 perf commits: #1558 (in-place NaN writes on uniquely-owned buffers correct), #1556 (fp-predictor ngjit ~297us/tile for 256x256 float32), #1552 (single cupy.concatenate + one .get() for batched D2H at _gpu_decode.py:870-913), #1551 (parallel decode threshold >=65536px engages 256x256 default at _reader.py:1121). Bench: 8192x8192 f32 deflate+pred2 256-tile write 782ms; 4096x4096 f32 deflate read 83ms with parallel decode. Deferred LOW (none filed, all <10% MEDIUM threshold): _writer.py:459/1109 redundant .copy() before predictor encode (~1% per tile), _compression.py:280 lzw_decompress dst[:n].copy() (~2% per LZW tile decode), _writer.py:1419 seg_np.copy() before in-place NaN substitution (negligible, conditional path), _CloudSource.read_range opens fresh fsspec handle per range (pre-existing, predates audit scope). nvCOMP per-tile D2H batching break-even confirmed (variable sizes need staging buffer, no win). | Pass 3 (2026-05-10): audited f157746,39322c3,f23ec8f,1aac3b7. All 5 commits correct. Redundant .copy() in _writer.py:459,1109 and _compression.py:280 (1-2% overhead, LOW). _CloudSource.read_range() per-call open is pre-existing arch issue. No HIGH/MEDIUM regressions. SAFE. | re-audit 2026-05-02: 6 commits since 2026-04-16 (predictor=3 CPU encode/decode, GPU predictor stride fix, validate_tile_layout, BigTIFF LONG8 offsets, AREA_OR_POINT VRT, per-tile alloc guard). 1M dask chunk cap intact at __init__.py:948; adler32 batch transfer intact at _gpu_decode.py:1825. New code is metadata validation and dispatcher logic with no extra materialization or per-tile sync points. No HIGH/MEDIUM regressions. | Pass 5 (2026-05-12): re-audit identified MEDIUM in _gpu_decode.py:1577 _try_nvcomp_from_device_bufs: per-tile cupy.empty + trailing cupy.concatenate doubled peak VRAM and added serial concat. Filed #1659 and fixed to single-buffer + pointer offsets (matches LZW/deflate/host-buffer patterns at L1847/L1878/L1114). Microbench (alloc+concat overhead only, not full nvCOMP latency): n=256 tile_bytes=65536 drops 3.66ms->0.69ms, n=256 tile_bytes=262144 drops 8.18ms->0.13ms. Tests: 5 new tests in test_nvcomp_from_device_bufs_single_alloc_1659.py (codec short-circuit, no-lib short-circuit, memory-guard contract, real ZSTD round-trip via nvCOMP, structural single-buffer check). 1458 existing geotiff tests pass, 3 unrelated matplotlib/py3.14 failures pre-existing. SAFE/IO-bound verdict holds. | Pass 6 (2026-05-12): re-audit on top of #1659. New HIGH in _try_kvikio_read_tiles at _gpu_decode.py:941: per-tile cupy.empty() + blocking IOFuture.get() inside loop serialised GDS reads to ~1 outstanding pread, missed parallelism the kvikio worker pool was designed for, paid per-tile cupy.empty setup (matches #1659 anti-pattern in nvCOMP path), and lacked _check_gpu_memory guard. Filed #1688 and fixed to single contiguous buffer + batched submit + guard. Microbench with 8-worker pool simulation: 256 tiles@1ms latency drops 256ms->38.7ms (~6.6x); single-thread simulation 256ms->28.5ms (9x). Tests: 9 new tests in test_kvikio_batched_pread_1688.py (kvikio-absent path, single-buffer pointer arithmetic, submit-before-get ordering, memory guard, partial-read fallback, round-trip data, zero-size/all-sparse tiles). All 1577 geotiff tests pass except pre-existing matplotlib/py3.14 failures."
2122
geotiff,2026-05-15,SAFE,IO-bound,0,1756,"Rockout 2026-05-15: LOW filed #1934 -- _apply_nodata_mask_gpu used cupy.where (allocating); switched to cupy.putmask on the already-owned buffer (float path) and on the post-astype float64 buffer (int path). Saves one chunk-sized device allocation per call. 7 new tests in test_apply_nodata_mask_gpu_inplace_1934.py; 52 related nodata tests pass. | Pass 8 (2026-05-12): 1 new MEDIUM found and fixed. _assemble_standard_layout/_assemble_cog_layout returned bytes(bytearray), doubling peak memory transiently during eager writes. Filed #1756, fixed by returning the bytearray directly. Measured: 95 MB uint8 raster peak drops 202 MB -> 107 MB. _write_bytes / parse_header already accepted the buffer protocol so the change is transparent to callers. 6 new tests in test_assemble_layout_no_bytes_copy_1756.py. 2123 existing geotiff tests pass; the 10 unrelated failures (test_no_georef_windowed_coords_1710, test_predictor2_big_endian_gpu_1517) reference the now-private read_to_array attribute (commit 8adb749, issue #1708) and predate this change. SAFE/IO-bound verdict holds. | Pass 7 (2026-05-12): re-audit identified 4 MEDIUM findings, all real, all backed by microbenches. (1) unpack_bits sub-byte loops for bps=2/4/12 in _compression.py:836-878 were 100-200x slower than vectorised numpy (filed #1713, fixed in this branch: bps=4 2M pixels drops from 165ms to 3ms = 55x; bps=2/12 similar). (2) _write_vrt_tiled at __init__.py:1708 uses scheduler='synchronous' on independent tile writes; measured 33% slowdown on 256-tile zstd write vs threads scheduler (filed #1714, no fix yet). (3) _nvcomp_batch_compress at _gpu_decode.py:2522-2526 still does per-tile cupy.get().tobytes() despite #1552 / #1659 fixing the same pattern elsewhere; measured 45% reduction with concat+single get on n=1024 (filed #1712, no fix yet). (4) _nvcomp_batch_compress at _gpu_decode.py:2457 uses per-tile cupy.empty allocations; 1024 tiles 16KB drops from 4.7ms to 1.0ms with single contiguous + views (bundled into #1712). Cat 6 OOM verdict: SAFE/IO-bound holds -- read_geotiff_dask caps task count at _MAX_DASK_CHUNKS=50_000 and per-chunk memory is bounded by chunk size. _inflate_tiles_kernel resource usage on Ampere: 67 regs/thread, 2896B local/thread, 8192B shared/block (LZW kernel: 29 regs, 24576B shared) -- register pressure under control; high local memory in inflate is unavoidable (LZ77 state) but only thread 0 in each block uses it. | Pass 4 (2026-05-10): re-audit after #1559 (centralise attrs across all read backends). New _populate_attrs_from_geo_info helper at __init__.py:301 runs once per read, not per-chunk -- no perf impact. Probe: 2560x2560 deflate-tiled file opened via read_geotiff_dask yields 400 tasks (4 tasks/chunk for 100 chunks), well under 1M cap. read_geotiff_gpu(1024x1024) returns cupy.ndarray end-to-end with no host round-trip (226ms incl. write+decode). No new HIGH/MEDIUM findings. SAFE/IO-bound holds. | Pass 3 (2026-05-10): SAFE/IO-bound. Audited 4 perf commits: #1558 (in-place NaN writes on uniquely-owned buffers correct), #1556 (fp-predictor ngjit ~297us/tile for 256x256 float32), #1552 (single cupy.concatenate + one .get() for batched D2H at _gpu_decode.py:870-913), #1551 (parallel decode threshold >=65536px engages 256x256 default at _reader.py:1121). Bench: 8192x8192 f32 deflate+pred2 256-tile write 782ms; 4096x4096 f32 deflate read 83ms with parallel decode. Deferred LOW (none filed, all <10% MEDIUM threshold): _writer.py:459/1109 redundant .copy() before predictor encode (~1% per tile), _compression.py:280 lzw_decompress dst[:n].copy() (~2% per LZW tile decode), _writer.py:1419 seg_np.copy() before in-place NaN substitution (negligible, conditional path), _CloudSource.read_range opens fresh fsspec handle per range (pre-existing, predates audit scope). nvCOMP per-tile D2H batching break-even confirmed (variable sizes need staging buffer, no win). | Pass 3 (2026-05-10): audited f157746,39322c3,f23ec8f,1aac3b7. All 5 commits correct. Redundant .copy() in _writer.py:459,1109 and _compression.py:280 (1-2% overhead, LOW). _CloudSource.read_range() per-call open is pre-existing arch issue. No HIGH/MEDIUM regressions. SAFE. | re-audit 2026-05-02: 6 commits since 2026-04-16 (predictor=3 CPU encode/decode, GPU predictor stride fix, validate_tile_layout, BigTIFF LONG8 offsets, AREA_OR_POINT VRT, per-tile alloc guard). 1M dask chunk cap intact at __init__.py:948; adler32 batch transfer intact at _gpu_decode.py:1825. New code is metadata validation and dispatcher logic with no extra materialization or per-tile sync points. No HIGH/MEDIUM regressions. | Pass 5 (2026-05-12): re-audit identified MEDIUM in _gpu_decode.py:1577 _try_nvcomp_from_device_bufs: per-tile cupy.empty + trailing cupy.concatenate doubled peak VRAM and added serial concat. Filed #1659 and fixed to single-buffer + pointer offsets (matches LZW/deflate/host-buffer patterns at L1847/L1878/L1114). Microbench (alloc+concat overhead only, not full nvCOMP latency): n=256 tile_bytes=65536 drops 3.66ms->0.69ms, n=256 tile_bytes=262144 drops 8.18ms->0.13ms. Tests: 5 new tests in test_nvcomp_from_device_bufs_single_alloc_1659.py (codec short-circuit, no-lib short-circuit, memory-guard contract, real ZSTD round-trip via nvCOMP, structural single-buffer check). 1458 existing geotiff tests pass, 3 unrelated matplotlib/py3.14 failures pre-existing. SAFE/IO-bound verdict holds. | Pass 6 (2026-05-12): re-audit on top of #1659. New HIGH in _try_kvikio_read_tiles at _gpu_decode.py:941: per-tile cupy.empty() + blocking IOFuture.get() inside loop serialised GDS reads to ~1 outstanding pread, missed parallelism the kvikio worker pool was designed for, paid per-tile cupy.empty setup (matches #1659 anti-pattern in nvCOMP path), and lacked _check_gpu_memory guard. Filed #1688 and fixed to single contiguous buffer + batched submit + guard. Microbench with 8-worker pool simulation: 256 tiles@1ms latency drops 256ms->38.7ms (~6.6x); single-thread simulation 256ms->28.5ms (9x). Tests: 9 new tests in test_kvikio_batched_pread_1688.py (kvikio-absent path, single-buffer pointer arithmetic, submit-before-get ordering, memory guard, partial-read fallback, round-trip data, zero-size/all-sparse tiles). All 1577 geotiff tests pass except pre-existing matplotlib/py3.14 failures."
2223
glcm,2026-03-31T18:00:00Z,SAFE,compute-bound,0,,"Downgraded to MEDIUM. da.stack without rechunk is scheduling overhead, not OOM risk."
2324
hillshade,2026-04-16T12:00:00Z,SAFE,compute-bound,0,,"Re-audit after Horn's method rewrite (PR 1175): clean stencil, map_overlap depth=(1,1), no materialization. Zero findings."

xrspatial/geotiff/_backends/vrt.py

Lines changed: 13 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -499,6 +499,18 @@ def _read_vrt_chunked(source, *, window, band, name, chunks, gpu, dtype,
499499

500500
delayed_read = dask.delayed(_vrt_chunk_read)
501501

502+
# Wrap the parsed VRTDataset in a single dask Delayed so every
503+
# chunk task takes it as one graph input rather than a per-task
504+
# closure copy. Without this, dask embeds the full source list
505+
# (filenames, src/dst rects, per-source nodata) into every task's
506+
# pickled graph entry; a 1000-source VRT split into 1000 chunks
507+
# would build a ~57 MB driver graph and re-serialise the same
508+
# metadata 1000 times under distributed/process schedulers. The
509+
# sibling COG-HTTP and GDS chunked paths use the same single-
510+
# delayed-input pattern (see ``http_meta_key`` in ``dask.py`` and
511+
# ``meta_key`` in ``gpu.py``). See issue #1923.
512+
parsed_vrt_key = dask.delayed(vrt, pure=True)
513+
502514
if gpu:
503515
import cupy
504516
meta = cupy.empty((0,) * (3 if out_has_band_axis else 2),
@@ -526,7 +538,7 @@ def _read_vrt_chunked(source, *, window, band, name, chunks, gpu, dtype,
526538
missing_sources=missing_sources,
527539
declared_dtype=declared_dtype,
528540
gpu=gpu,
529-
parsed_vrt=vrt,
541+
parsed_vrt=parsed_vrt_key,
530542
)
531543
block = da.from_delayed(d, shape=block_shape,
532544
dtype=declared_dtype, meta=meta)

0 commit comments

Comments
 (0)