GH-17211: [C++] Add hash_64 scalar compute function (old) #13487
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
Member
For a hashing function I would expect:
|
|
|
Member
|
@save-buffer would you mind reviewing this when you get a chance? |
Contributor
Author
|
I renamed this PR for ARROW-16945, so that I can track just a 32-bit version for now. Not sure if this will confuse some of the automated issue tracking/associations |
Member
I'd add that non-random but likely selections should also show a nice hash distribution. Including:
Also, ideally, the bottom K bits of the hash should also keep those nice statistical properties (so that a hash table implementation can simply use a power-of-two modulo). The implementations from |
Member
|
Why 32-bit rather than 64-bit? |
Contributor
Author
I think Weston mentioned 32-bit in passing and I just made a semi-arbitrary decision to start there. I plan on following this up with a 64-bit version, though not sure if other things should be addressed before I do so (memory management, etc.). |
Member
It would be great, by the way, to have benchmarks of the two hashing utilities we have. I believe the utilities we have in Also, can we expand in the function doc that these hashes are not suitable for cryptographic purposes? |
Member
Contributor
Author
|
Thanks @pitrou , I'll start with that as a base and add benchmarks that use the utilities from key_hash.h (I'll see if these also exist elsewhere). I recently was exposed to |
6af47cc to
7787bca
Compare
Contributor
Author
|
rebased |
Contributor
Author
|
I wanted to update with initial changes to Barring code style/organization improvements, I was hoping to get some feedback on whether these benchmarks are fundamentally what we would like to see. Some questions I have:
|
Contributor
Author
|
Also, I see now that I need to experiment a bit, but I suspect that for int primitives, Hashing32 and Hashing64 just wrap xxHash and add logic for iterating over rows and columns. Specifically, key_hash.cc#L745 looks a lot like hashing.h#L88. hashing_benchmark.cc#L70 benchmarks using 2 multipliers, and I think if it uses just the 1 multiplier then the implementation looks a lot like Hashing64. All this to say that I think it should be accommodated in the benchmark, but I also wonder if some of the hashing code can be simplified/better organized in case they aren't as independent as they seem at first glance. cc @michalursa |
Member
|
Here is my understanding, which may not be complete:
In theory, |
Contributor
|
Weston is correct, the I can leave a review of the current code, but I'd suggest moving away from the |
save-buffer
suggested changes
Jul 20, 2022
Contributor
save-buffer
left a comment
There was a problem hiding this comment.
Biggest feedback is to not allocate a large temporary std::vector and instead write directly into the buffer you'll be returning. Also you made your TempVectorStack way too big.
Contributor
Author
thanks! I had troubles with smaller sizes and tried to not size it too large, but I'll revisit. For writing directly into the buffer to return, I assume that means I need to create an ArrayData? is there a better interface via Builders? |
Contributor
Author
|
never mind, you actually mentioned how to do it already. the mobile interface hid it |
Member
That's how it ended up working in practice for non-tiny strings, but the basic objective is to have some fast though statistically good hash functions. |
Member
|
@drin could you please add entries for these functions in the Compute Functions docs ( |
Contributor
Author
|
@ursabot please benchmark lang=C++ |
|
Benchmark runs are scheduled for baseline = 4d931ff and contender = 3c1fd3b03fa2b143d582244ca1bb93fbc0c84bcf. Results will be available as each benchmark for each run completes. |
Contributor
Author
|
before figuring out the conbench stuff, I put some of the local benchmark results in a google slide: Just wanted to put this in as a sneak peek even though the test size is small (10,000 items) and this is only for a couple primitives. |
Contributor
Author
|
@ursabot please benchmark lang=C++ |
|
Commit 3c1fd3b03fa2b143d582244ca1bb93fbc0c84bcf already has scheduled benchmark runs. |
…ark (apache#39794) ### Rationale for this change `ScanOnlyBench` use 2 kind of options to scan, but ignore the function `MakeScanNode` will return 4 extra columns. ### What changes are included in this PR? Change the expected result of bench `ScanOnlyBench` ### Are these changes tested? They covered by existing tests. * Closes: apache#39765 Authored-by: xiansen.chen <xiansen.chen@openpie.com> Signed-off-by: Antoine Pitrou <antoine@python.org>
### Rationale for this change Fix a conversion warning that fails compiling arrow-dataset-file-benchmark on MSVC. ### Are these changes tested? Yes, by CI. ### Are there any user-facing changes? No. Authored-by: Antoine Pitrou <antoine@python.org> Signed-off-by: Sutou Kouhei <kou@clear-code.com>
…ey hash avx2 (apache#39800) ### Rationale for this change Issue apache#39778 seems caused by a careless (but hard to spot) bug in key hash avx2. ### What changes are included in this PR? Fix the careless bug. ### Are these changes tested? UT included. ### Are there any user-facing changes? No. * Closes: apache#39778 Authored-by: Ruoxi Sun <zanmato1984@gmail.com> Signed-off-by: Antoine Pitrou <antoine@python.org>
…to int32 (apache#39528) Be defensive instead of writing invalid data. ### Rationale for this change Users can provide this API pages that are large to validly write and we silently truncate lengths before writing. ### What changes are included in this PR? Add validations and throw an exception if sizes are too large (this was previously checked only if page indexes are being build). ### Are these changes tested? Unit tested ### Are there any user-facing changes? This might start raising exceptions instead of writing out invalid parquet files. Closes apache#39527 **This PR contains a "Critical Fix".** * Closes: apache#39527 Lead-authored-by: emkornfield <emkornfield@gmail.com> Co-authored-by: Micah Kornfield <micahk@google.com> Co-authored-by: mwish <maplewish117@gmail.com> Co-authored-by: Antoine Pitrou <pitrou@free.fr> Co-authored-by: Gang Wu <ustcwg@gmail.com> Signed-off-by: mwish <maplewish117@gmail.com>
…roto for building arrays (apache#39721) This PR updates the ArrowReaderHelper to use an ArrowField object for building an Array instead of a protobuf field obj. This removes leveraging protobuf from building out the Arrays and makes the code easier to reuse (like for the C Data Interface) * Closes: apache#39720 Authored-by: Alva Bandy <abandy@live.com> Signed-off-by: Sutou Kouhei <kou@clear-code.com>
…lve() (apache#39817) ### Rationale for this change There has been interest in improving operations on chunked-arrays and even though `ChunkResolver::Resolve()` is not a big contributor in most kernels, the fact that it can be used from tight loops warrants careful attention to branch prediction and memory effects of its implementation. ### What changes are included in this PR? - Documentation of invariants and behavior of functions - Multiple optimizations justified by microbenchmarks - Addition of a variation of `Resolve` that takes a hint as parameter - Fix of an out-of-bounds memory access that doesn't affect correctness (it can only reduce effectiveness of cache in very rare situations, but is nevertheless an issue) ### Are these changes tested? Yes, by existing tests. ### Are there any user-facing changes? - The `arrow::internal::ChunkResolver::Bisect()` function was `protected` and is now `private` with a different signature * Closes: apache#39815 Authored-by: Felipe Oliveira Carvalho <felipekde@gmail.com> Signed-off-by: Felipe Oliveira Carvalho <felipekde@gmail.com>
…ntervals (apache#39795) ### Rationale for this change The filter and take functions were not correctly supported on month_day_nano intervals. ### What changes are included in this PR? * Expand the primitive filter implementation to handle all possible fixed-width primitive types (including fixed-size binary) * Expand the take filter implementation to handle all well-known fixed-width primitive types (including month_day_nano, decimal128 and decimal256) * Add benchmarks for taking and filtering fixed-size binary These changes allow for very significant performance improvements filtering and taking fixed-size binary data: ``` ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Non-regressions: (90) ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- benchmark baseline contender change % counters FilterFixedSizeBinaryFilterNoNulls/524288/0/8 1.716 GiB/sec 33.814 GiB/sec 1870.862 {'family_index': 0, 'per_family_instance_index': 0, 'run_name': 'FilterFixedSizeBinaryFilterNoNulls/524288/0/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 2462, 'byte_width': 8.0, 'data null%': 0.0, 'mask null%': 0.0, 'select%': 99.9} TakeFixedSizeBinaryRandomIndicesWithNulls/524288/1/8 380.056M items/sec 7.098G items/sec 1767.491 {'family_index': 3, 'per_family_instance_index': 6, 'run_name': 'TakeFixedSizeBinaryRandomIndicesWithNulls/524288/1/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 505, 'byte_width': 8.0, 'null_percent': 100.0} FilterFixedSizeBinaryFilterNoNulls/524288/0/9 1.916 GiB/sec 33.721 GiB/sec 1659.766 {'family_index': 0, 'per_family_instance_index': 1, 'run_name': 'FilterFixedSizeBinaryFilterNoNulls/524288/0/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 2750, 'byte_width': 9.0, 'data null%': 0.0, 'mask null%': 0.0, 'select%': 99.9} FilterFixedSizeBinaryFilterNoNulls/524288/9/8 917.713 MiB/sec 9.193 GiB/sec 925.719 {'family_index': 0, 'per_family_instance_index': 18, 'run_name': 'FilterFixedSizeBinaryFilterNoNulls/524288/9/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1271, 'byte_width': 8.0, 'data null%': 10.0, 'mask null%': 0.0, 'select%': 99.9} FilterFixedSizeBinaryFilterNoNulls/524288/12/8 1.004 GiB/sec 9.374 GiB/sec 833.673 {'family_index': 0, 'per_family_instance_index': 24, 'run_name': 'FilterFixedSizeBinaryFilterNoNulls/524288/12/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1440, 'byte_width': 8.0, 'data null%': 90.0, 'mask null%': 0.0, 'select%': 99.9} FilterFixedSizeBinaryFilterNoNulls/524288/3/8 1.625 GiB/sec 15.009 GiB/sec 823.442 {'family_index': 0, 'per_family_instance_index': 6, 'run_name': 'FilterFixedSizeBinaryFilterNoNulls/524288/3/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 2328, 'byte_width': 8.0, 'data null%': 0.1, 'mask null%': 0.0, 'select%': 99.9} FilterFixedSizeBinaryFilterNoNulls/524288/9/9 1021.638 MiB/sec 9.126 GiB/sec 814.670 {'family_index': 0, 'per_family_instance_index': 19, 'run_name': 'FilterFixedSizeBinaryFilterNoNulls/524288/9/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1428, 'byte_width': 9.0, 'data null%': 10.0, 'mask null%': 0.0, 'select%': 99.9} FilterFixedSizeBinaryFilterNoNulls/524288/6/8 1.235 GiB/sec 10.814 GiB/sec 775.869 {'family_index': 0, 'per_family_instance_index': 12, 'run_name': 'FilterFixedSizeBinaryFilterNoNulls/524288/6/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1762, 'byte_width': 8.0, 'data null%': 1.0, 'mask null%': 0.0, 'select%': 99.9} FilterFixedSizeBinaryFilterNoNulls/524288/12/9 1.123 GiB/sec 9.120 GiB/sec 712.196 {'family_index': 0, 'per_family_instance_index': 25, 'run_name': 'FilterFixedSizeBinaryFilterNoNulls/524288/12/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1598, 'byte_width': 9.0, 'data null%': 90.0, 'mask null%': 0.0, 'select%': 99.9} FilterFixedSizeBinaryFilterNoNulls/524288/6/9 1.370 GiB/sec 10.499 GiB/sec 666.348 {'family_index': 0, 'per_family_instance_index': 13, 'run_name': 'FilterFixedSizeBinaryFilterNoNulls/524288/6/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1958, 'byte_width': 9.0, 'data null%': 1.0, 'mask null%': 0.0, 'select%': 99.9} FilterFixedSizeBinaryFilterNoNulls/524288/3/9 1.814 GiB/sec 13.394 GiB/sec 638.343 {'family_index': 0, 'per_family_instance_index': 7, 'run_name': 'FilterFixedSizeBinaryFilterNoNulls/524288/3/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 2600, 'byte_width': 9.0, 'data null%': 0.1, 'mask null%': 0.0, 'select%': 99.9} FilterFixedSizeBinaryFilterNoNulls/524288/2/8 12.155 GiB/sec 77.799 GiB/sec 540.051 {'family_index': 0, 'per_family_instance_index': 4, 'run_name': 'FilterFixedSizeBinaryFilterNoNulls/524288/2/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 17222, 'byte_width': 8.0, 'data null%': 0.0, 'mask null%': 0.0, 'select%': 1.0} FilterFixedSizeBinaryFilterNoNulls/524288/2/9 13.507 GiB/sec 84.361 GiB/sec 524.592 {'family_index': 0, 'per_family_instance_index': 5, 'run_name': 'FilterFixedSizeBinaryFilterNoNulls/524288/2/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 19469, 'byte_width': 9.0, 'data null%': 0.0, 'mask null%': 0.0, 'select%': 1.0} TakeFixedSizeBinaryMonotonicIndices/524288/1/8 194.493M items/sec 732.378M items/sec 276.557 {'family_index': 4, 'per_family_instance_index': 6, 'run_name': 'TakeFixedSizeBinaryMonotonicIndices/524288/1/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 259, 'byte_width': 8.0, 'null_percent': 100.0} TakeFixedSizeBinaryRandomIndicesNoNulls/524288/1/8 200.981M items/sec 747.628M items/sec 271.989 {'family_index': 2, 'per_family_instance_index': 6, 'run_name': 'TakeFixedSizeBinaryRandomIndicesNoNulls/524288/1/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 268, 'byte_width': 8.0, 'null_percent': 100.0} FilterFixedSizeBinaryFilterWithNulls/524288/0/8 947.631 MiB/sec 3.318 GiB/sec 258.565 {'family_index': 1, 'per_family_instance_index': 0, 'run_name': 'FilterFixedSizeBinaryFilterWithNulls/524288/0/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1329, 'byte_width': 8.0, 'data null%': 0.0, 'mask null%': 5.0, 'select%': 99.9} FilterFixedSizeBinaryFilterWithNulls/524288/3/8 911.406 MiB/sec 3.121 GiB/sec 250.677 {'family_index': 1, 'per_family_instance_index': 6, 'run_name': 'FilterFixedSizeBinaryFilterWithNulls/524288/3/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1275, 'byte_width': 8.0, 'data null%': 0.1, 'mask null%': 5.0, 'select%': 99.9} FilterFixedSizeBinaryFilterNoNulls/524288/1/8 1.045 GiB/sec 3.535 GiB/sec 238.406 {'family_index': 0, 'per_family_instance_index': 2, 'run_name': 'FilterFixedSizeBinaryFilterNoNulls/524288/1/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1496, 'byte_width': 8.0, 'data null%': 0.0, 'mask null%': 0.0, 'select%': 50.0} FilterFixedSizeBinaryFilterWithNulls/524288/6/8 899.161 MiB/sec 2.915 GiB/sec 232.029 {'family_index': 1, 'per_family_instance_index': 12, 'run_name': 'FilterFixedSizeBinaryFilterWithNulls/524288/6/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1260, 'byte_width': 8.0, 'data null%': 1.0, 'mask null%': 5.0, 'select%': 99.9} FilterFixedSizeBinaryFilterWithNulls/524288/9/8 829.852 MiB/sec 2.617 GiB/sec 222.914 {'family_index': 1, 'per_family_instance_index': 18, 'run_name': 'FilterFixedSizeBinaryFilterWithNulls/524288/9/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1157, 'byte_width': 8.0, 'data null%': 10.0, 'mask null%': 5.0, 'select%': 99.9} TakeFixedSizeBinaryRandomIndicesNoNulls/524288/0/8 234.268M items/sec 752.809M items/sec 221.345 {'family_index': 2, 'per_family_instance_index': 8, 'run_name': 'TakeFixedSizeBinaryRandomIndicesNoNulls/524288/0/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 312, 'byte_width': 8.0, 'null_percent': 0.0} FilterFixedSizeBinaryFilterNoNulls/524288/1/9 1.171 GiB/sec 3.711 GiB/sec 216.957 {'family_index': 0, 'per_family_instance_index': 3, 'run_name': 'FilterFixedSizeBinaryFilterNoNulls/524288/1/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1674, 'byte_width': 9.0, 'data null%': 0.0, 'mask null%': 0.0, 'select%': 50.0} TakeFixedSizeBinaryMonotonicIndices/524288/0/8 249.393M items/sec 787.274M items/sec 215.676 {'family_index': 4, 'per_family_instance_index': 8, 'run_name': 'TakeFixedSizeBinaryMonotonicIndices/524288/0/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 333, 'byte_width': 8.0, 'null_percent': 0.0} TakeFixedSizeBinaryRandomIndicesWithNulls/524288/0/8 234.268M items/sec 736.727M items/sec 214.481 {'family_index': 3, 'per_family_instance_index': 8, 'run_name': 'TakeFixedSizeBinaryRandomIndicesWithNulls/524288/0/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 313, 'byte_width': 8.0, 'null_percent': 0.0} TakeFixedSizeBinaryMonotonicIndices/524288/1000/8 134.852M items/sec 423.748M items/sec 214.231 {'family_index': 4, 'per_family_instance_index': 0, 'run_name': 'TakeFixedSizeBinaryMonotonicIndices/524288/1000/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 202, 'byte_width': 8.0, 'null_percent': 0.1} FilterFixedSizeBinaryFilterWithNulls/524288/12/8 913.734 MiB/sec 2.599 GiB/sec 191.245 {'family_index': 1, 'per_family_instance_index': 24, 'run_name': 'FilterFixedSizeBinaryFilterWithNulls/524288/12/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1292, 'byte_width': 8.0, 'data null%': 90.0, 'mask null%': 5.0, 'select%': 99.9} TakeFixedSizeBinaryRandomIndicesNoNulls/524288/1000/8 138.218M items/sec 309.307M items/sec 123.783 {'family_index': 2, 'per_family_instance_index': 0, 'run_name': 'TakeFixedSizeBinaryRandomIndicesNoNulls/524288/1000/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 184, 'byte_width': 8.0, 'null_percent': 0.1} TakeFixedSizeBinaryRandomIndicesWithNulls/524288/1000/8 132.755M items/sec 293.027M items/sec 120.727 {'family_index': 3, 'per_family_instance_index': 0, 'run_name': 'TakeFixedSizeBinaryRandomIndicesWithNulls/524288/1000/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 179, 'byte_width': 8.0, 'null_percent': 0.1} TakeFixedSizeBinaryRandomIndicesNoNulls/524288/10/8 125.492M items/sec 272.996M items/sec 117.540 {'family_index': 2, 'per_family_instance_index': 2, 'run_name': 'TakeFixedSizeBinaryRandomIndicesNoNulls/524288/10/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 174, 'byte_width': 8.0, 'null_percent': 10.0} FilterFixedSizeBinaryFilterWithNulls/524288/9/9 926.938 MiB/sec 1.904 GiB/sec 110.379 {'family_index': 1, 'per_family_instance_index': 19, 'run_name': 'FilterFixedSizeBinaryFilterWithNulls/524288/9/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1295, 'byte_width': 9.0, 'data null%': 10.0, 'mask null%': 5.0, 'select%': 99.9} TakeFixedSizeBinaryMonotonicIndices/524288/10/8 158.754M items/sec 331.106M items/sec 108.565 {'family_index': 4, 'per_family_instance_index': 2, 'run_name': 'TakeFixedSizeBinaryMonotonicIndices/524288/10/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 167, 'byte_width': 8.0, 'null_percent': 10.0} FilterFixedSizeBinaryFilterWithNulls/524288/0/9 1.031 GiB/sec 2.129 GiB/sec 106.621 {'family_index': 1, 'per_family_instance_index': 1, 'run_name': 'FilterFixedSizeBinaryFilterWithNulls/524288/0/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1477, 'byte_width': 9.0, 'data null%': 0.0, 'mask null%': 5.0, 'select%': 99.9} FilterFixedSizeBinaryFilterWithNulls/524288/3/9 1020.776 MiB/sec 2.056 GiB/sec 106.293 {'family_index': 1, 'per_family_instance_index': 7, 'run_name': 'FilterFixedSizeBinaryFilterWithNulls/524288/3/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1430, 'byte_width': 9.0, 'data null%': 0.1, 'mask null%': 5.0, 'select%': 99.9} FilterFixedSizeBinaryFilterWithNulls/524288/4/8 890.785 MiB/sec 1.768 GiB/sec 103.293 {'family_index': 1, 'per_family_instance_index': 8, 'run_name': 'FilterFixedSizeBinaryFilterWithNulls/524288/4/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1242, 'byte_width': 8.0, 'data null%': 0.1, 'mask null%': 5.0, 'select%': 50.0} FilterFixedSizeBinaryFilterWithNulls/524288/6/9 1005.839 MiB/sec 1.984 GiB/sec 102.023 {'family_index': 1, 'per_family_instance_index': 13, 'run_name': 'FilterFixedSizeBinaryFilterWithNulls/524288/6/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1407, 'byte_width': 9.0, 'data null%': 1.0, 'mask null%': 5.0, 'select%': 99.9} FilterFixedSizeBinaryFilterWithNulls/524288/1/8 916.810 MiB/sec 1.762 GiB/sec 96.757 {'family_index': 1, 'per_family_instance_index': 2, 'run_name': 'FilterFixedSizeBinaryFilterWithNulls/524288/1/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1270, 'byte_width': 8.0, 'data null%': 0.0, 'mask null%': 5.0, 'select%': 50.0} FilterFixedSizeBinaryFilterWithNulls/524288/7/8 890.211 MiB/sec 1.694 GiB/sec 94.853 {'family_index': 1, 'per_family_instance_index': 14, 'run_name': 'FilterFixedSizeBinaryFilterWithNulls/524288/7/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1235, 'byte_width': 8.0, 'data null%': 1.0, 'mask null%': 5.0, 'select%': 50.0} TakeFixedSizeBinaryRandomIndicesNoNulls/524288/2/8 95.788M items/sec 184.004M items/sec 92.095 {'family_index': 2, 'per_family_instance_index': 4, 'run_name': 'TakeFixedSizeBinaryRandomIndicesNoNulls/524288/2/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 124, 'byte_width': 8.0, 'null_percent': 50.0} FilterFixedSizeBinaryFilterWithNulls/524288/10/8 862.497 MiB/sec 1.616 GiB/sec 91.823 {'family_index': 1, 'per_family_instance_index': 20, 'run_name': 'FilterFixedSizeBinaryFilterWithNulls/524288/10/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1200, 'byte_width': 8.0, 'data null%': 10.0, 'mask null%': 5.0, 'select%': 50.0} FilterFixedSizeBinaryFilterWithNulls/524288/12/9 1.005 GiB/sec 1.904 GiB/sec 89.431 {'family_index': 1, 'per_family_instance_index': 25, 'run_name': 'FilterFixedSizeBinaryFilterWithNulls/524288/12/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1442, 'byte_width': 9.0, 'data null%': 90.0, 'mask null%': 5.0, 'select%': 99.9} TakeFixedSizeBinaryMonotonicIndices/524288/2/8 123.065M items/sec 228.755M items/sec 85.881 {'family_index': 4, 'per_family_instance_index': 4, 'run_name': 'TakeFixedSizeBinaryMonotonicIndices/524288/2/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 164, 'byte_width': 8.0, 'null_percent': 50.0} FilterFixedSizeBinaryFilterNoNulls/524288/10/8 930.637 MiB/sec 1.669 GiB/sec 83.659 {'family_index': 0, 'per_family_instance_index': 20, 'run_name': 'FilterFixedSizeBinaryFilterNoNulls/524288/10/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1293, 'byte_width': 8.0, 'data null%': 10.0, 'mask null%': 0.0, 'select%': 50.0} FilterFixedSizeBinaryFilterNoNulls/524288/4/8 1.034 GiB/sec 1.871 GiB/sec 81.019 {'family_index': 0, 'per_family_instance_index': 8, 'run_name': 'FilterFixedSizeBinaryFilterNoNulls/524288/4/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1482, 'byte_width': 8.0, 'data null%': 0.1, 'mask null%': 0.0, 'select%': 50.0} FilterFixedSizeBinaryFilterNoNulls/524288/7/8 1004.789 MiB/sec 1.772 GiB/sec 80.538 {'family_index': 0, 'per_family_instance_index': 14, 'run_name': 'FilterFixedSizeBinaryFilterNoNulls/524288/7/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1404, 'byte_width': 8.0, 'data null%': 1.0, 'mask null%': 0.0, 'select%': 50.0} FilterFixedSizeBinaryFilterWithNulls/524288/13/8 920.819 MiB/sec 1.616 GiB/sec 79.686 {'family_index': 1, 'per_family_instance_index': 26, 'run_name': 'FilterFixedSizeBinaryFilterWithNulls/524288/13/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1285, 'byte_width': 8.0, 'data null%': 90.0, 'mask null%': 5.0, 'select%': 50.0} FilterFixedSizeBinaryFilterNoNulls/524288/13/8 974.713 MiB/sec 1.669 GiB/sec 75.388 {'family_index': 0, 'per_family_instance_index': 26, 'run_name': 'FilterFixedSizeBinaryFilterNoNulls/524288/13/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1363, 'byte_width': 8.0, 'data null%': 90.0, 'mask null%': 0.0, 'select%': 50.0} TakeFixedSizeBinaryRandomIndicesWithNulls/524288/10/8 107.165M items/sec 187.372M items/sec 74.845 {'family_index': 3, 'per_family_instance_index': 2, 'run_name': 'TakeFixedSizeBinaryRandomIndicesWithNulls/524288/10/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 143, 'byte_width': 8.0, 'null_percent': 10.0} TakeFixedSizeBinaryRandomIndicesWithNulls/524288/2/8 72.662M items/sec 114.781M items/sec 57.965 {'family_index': 3, 'per_family_instance_index': 4, 'run_name': 'TakeFixedSizeBinaryRandomIndicesWithNulls/524288/2/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 96, 'byte_width': 8.0, 'null_percent': 50.0} FilterFixedSizeBinaryFilterWithNulls/524288/10/9 976.180 MiB/sec 1.480 GiB/sec 55.260 {'family_index': 1, 'per_family_instance_index': 21, 'run_name': 'FilterFixedSizeBinaryFilterWithNulls/524288/10/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1358, 'byte_width': 9.0, 'data null%': 10.0, 'mask null%': 5.0, 'select%': 50.0} FilterFixedSizeBinaryFilterNoNulls/524288/10/9 1.023 GiB/sec 1.581 GiB/sec 54.502 {'family_index': 0, 'per_family_instance_index': 21, 'run_name': 'FilterFixedSizeBinaryFilterNoNulls/524288/10/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1466, 'byte_width': 9.0, 'data null%': 10.0, 'mask null%': 0.0, 'select%': 50.0} FilterFixedSizeBinaryFilterWithNulls/524288/4/9 992.477 MiB/sec 1.453 GiB/sec 49.957 {'family_index': 1, 'per_family_instance_index': 9, 'run_name': 'FilterFixedSizeBinaryFilterWithNulls/524288/4/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1400, 'byte_width': 9.0, 'data null%': 0.1, 'mask null%': 5.0, 'select%': 50.0} FilterFixedSizeBinaryFilterWithNulls/524288/7/9 997.679 MiB/sec 1.450 GiB/sec 48.846 {'family_index': 1, 'per_family_instance_index': 15, 'run_name': 'FilterFixedSizeBinaryFilterWithNulls/524288/7/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1389, 'byte_width': 9.0, 'data null%': 1.0, 'mask null%': 5.0, 'select%': 50.0} FilterFixedSizeBinaryFilterNoNulls/524288/13/9 1.071 GiB/sec 1.581 GiB/sec 47.526 {'family_index': 0, 'per_family_instance_index': 27, 'run_name': 'FilterFixedSizeBinaryFilterNoNulls/524288/13/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1538, 'byte_width': 9.0, 'data null%': 90.0, 'mask null%': 0.0, 'select%': 50.0} FilterFixedSizeBinaryFilterWithNulls/524288/13/9 1.008 GiB/sec 1.485 GiB/sec 47.328 {'family_index': 1, 'per_family_instance_index': 27, 'run_name': 'FilterFixedSizeBinaryFilterWithNulls/524288/13/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1446, 'byte_width': 9.0, 'data null%': 90.0, 'mask null%': 5.0, 'select%': 50.0} FilterFixedSizeBinaryFilterWithNulls/524288/1/9 1.003 GiB/sec 1.452 GiB/sec 44.708 {'family_index': 1, 'per_family_instance_index': 3, 'run_name': 'FilterFixedSizeBinaryFilterWithNulls/524288/1/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1437, 'byte_width': 9.0, 'data null%': 0.0, 'mask null%': 5.0, 'select%': 50.0} FilterFixedSizeBinaryFilterNoNulls/524288/7/9 1.105 GiB/sec 1.568 GiB/sec 41.954 {'family_index': 0, 'per_family_instance_index': 15, 'run_name': 'FilterFixedSizeBinaryFilterNoNulls/524288/7/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1587, 'byte_width': 9.0, 'data null%': 1.0, 'mask null%': 0.0, 'select%': 50.0} FilterFixedSizeBinaryFilterNoNulls/524288/4/9 1.163 GiB/sec 1.613 GiB/sec 38.639 {'family_index': 0, 'per_family_instance_index': 9, 'run_name': 'FilterFixedSizeBinaryFilterNoNulls/524288/4/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 1662, 'byte_width': 9.0, 'data null%': 0.1, 'mask null%': 0.0, 'select%': 50.0} FilterFixedSizeBinaryFilterWithNulls/524288/14/9 8.884 GiB/sec 12.117 GiB/sec 36.381 {'family_index': 1, 'per_family_instance_index': 29, 'run_name': 'FilterFixedSizeBinaryFilterWithNulls/524288/14/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 12508, 'byte_width': 9.0, 'data null%': 90.0, 'mask null%': 5.0, 'select%': 1.0} FilterFixedSizeBinaryFilterWithNulls/524288/11/9 8.886 GiB/sec 12.075 GiB/sec 35.892 {'family_index': 1, 'per_family_instance_index': 23, 'run_name': 'FilterFixedSizeBinaryFilterWithNulls/524288/11/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 12716, 'byte_width': 9.0, 'data null%': 10.0, 'mask null%': 5.0, 'select%': 1.0} TakeFixedSizeBinaryMonotonicIndices/524288/1000/9 134.765M items/sec 182.868M items/sec 35.694 {'family_index': 4, 'per_family_instance_index': 1, 'run_name': 'TakeFixedSizeBinaryMonotonicIndices/524288/1000/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 206, 'byte_width': 9.0, 'null_percent': 0.1} FilterFixedSizeBinaryFilterNoNulls/524288/5/8 11.393 GiB/sec 15.091 GiB/sec 32.453 {'family_index': 0, 'per_family_instance_index': 10, 'run_name': 'FilterFixedSizeBinaryFilterNoNulls/524288/5/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 16510, 'byte_width': 8.0, 'data null%': 0.1, 'mask null%': 0.0, 'select%': 1.0} FilterFixedSizeBinaryFilterNoNulls/524288/8/8 11.573 GiB/sec 15.102 GiB/sec 30.496 {'family_index': 0, 'per_family_instance_index': 16, 'run_name': 'FilterFixedSizeBinaryFilterNoNulls/524288/8/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 16684, 'byte_width': 8.0, 'data null%': 1.0, 'mask null%': 0.0, 'select%': 1.0} FilterFixedSizeBinaryFilterWithNulls/524288/11/8 7.740 GiB/sec 10.059 GiB/sec 29.956 {'family_index': 1, 'per_family_instance_index': 22, 'run_name': 'FilterFixedSizeBinaryFilterWithNulls/524288/11/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 10972, 'byte_width': 8.0, 'data null%': 10.0, 'mask null%': 5.0, 'select%': 1.0} FilterFixedSizeBinaryFilterWithNulls/524288/14/8 7.733 GiB/sec 9.915 GiB/sec 28.213 {'family_index': 1, 'per_family_instance_index': 28, 'run_name': 'FilterFixedSizeBinaryFilterWithNulls/524288/14/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 10991, 'byte_width': 8.0, 'data null%': 90.0, 'mask null%': 5.0, 'select%': 1.0} FilterFixedSizeBinaryFilterWithNulls/524288/5/8 7.682 GiB/sec 9.765 GiB/sec 27.109 {'family_index': 1, 'per_family_instance_index': 10, 'run_name': 'FilterFixedSizeBinaryFilterWithNulls/524288/5/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 10991, 'byte_width': 8.0, 'data null%': 0.1, 'mask null%': 5.0, 'select%': 1.0} FilterFixedSizeBinaryFilterWithNulls/524288/8/9 8.856 GiB/sec 11.180 GiB/sec 26.241 {'family_index': 1, 'per_family_instance_index': 17, 'run_name': 'FilterFixedSizeBinaryFilterWithNulls/524288/8/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 12571, 'byte_width': 9.0, 'data null%': 1.0, 'mask null%': 5.0, 'select%': 1.0} FilterFixedSizeBinaryFilterWithNulls/524288/8/8 7.735 GiB/sec 9.710 GiB/sec 25.530 {'family_index': 1, 'per_family_instance_index': 16, 'run_name': 'FilterFixedSizeBinaryFilterWithNulls/524288/8/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 11069, 'byte_width': 8.0, 'data null%': 1.0, 'mask null%': 5.0, 'select%': 1.0} TakeFixedSizeBinaryMonotonicIndices/524288/10/9 128.606M items/sec 160.249M items/sec 24.604 {'family_index': 4, 'per_family_instance_index': 3, 'run_name': 'TakeFixedSizeBinaryMonotonicIndices/524288/10/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 209, 'byte_width': 9.0, 'null_percent': 10.0} FilterFixedSizeBinaryFilterNoNulls/524288/11/8 12.033 GiB/sec 14.737 GiB/sec 22.478 {'family_index': 0, 'per_family_instance_index': 22, 'run_name': 'FilterFixedSizeBinaryFilterNoNulls/524288/11/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 17220, 'byte_width': 8.0, 'data null%': 10.0, 'mask null%': 0.0, 'select%': 1.0} FilterFixedSizeBinaryFilterNoNulls/524288/14/8 12.141 GiB/sec 14.761 GiB/sec 21.579 {'family_index': 0, 'per_family_instance_index': 28, 'run_name': 'FilterFixedSizeBinaryFilterNoNulls/524288/14/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 17343, 'byte_width': 8.0, 'data null%': 90.0, 'mask null%': 0.0, 'select%': 1.0} FilterFixedSizeBinaryFilterWithNulls/524288/5/9 8.825 GiB/sec 10.633 GiB/sec 20.489 {'family_index': 1, 'per_family_instance_index': 11, 'run_name': 'FilterFixedSizeBinaryFilterWithNulls/524288/5/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 12543, 'byte_width': 9.0, 'data null%': 0.1, 'mask null%': 5.0, 'select%': 1.0} FilterFixedSizeBinaryFilterWithNulls/524288/2/8 8.300 GiB/sec 9.969 GiB/sec 20.117 {'family_index': 1, 'per_family_instance_index': 4, 'run_name': 'FilterFixedSizeBinaryFilterWithNulls/524288/2/8', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 11819, 'byte_width': 8.0, 'data null%': 0.0, 'mask null%': 5.0, 'select%': 1.0} FilterFixedSizeBinaryFilterNoNulls/524288/5/9 12.954 GiB/sec 15.192 GiB/sec 17.273 {'family_index': 0, 'per_family_instance_index': 11, 'run_name': 'FilterFixedSizeBinaryFilterNoNulls/524288/5/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 18572, 'byte_width': 9.0, 'data null%': 0.1, 'mask null%': 0.0, 'select%': 1.0} FilterFixedSizeBinaryFilterNoNulls/524288/8/9 13.181 GiB/sec 15.222 GiB/sec 15.490 {'family_index': 0, 'per_family_instance_index': 17, 'run_name': 'FilterFixedSizeBinaryFilterNoNulls/524288/8/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 18904, 'byte_width': 9.0, 'data null%': 1.0, 'mask null%': 0.0, 'select%': 1.0} FilterFixedSizeBinaryFilterWithNulls/524288/2/9 9.344 GiB/sec 10.632 GiB/sec 13.784 {'family_index': 1, 'per_family_instance_index': 5, 'run_name': 'FilterFixedSizeBinaryFilterWithNulls/524288/2/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 13291, 'byte_width': 9.0, 'data null%': 0.0, 'mask null%': 5.0, 'select%': 1.0} FilterFixedSizeBinaryFilterNoNulls/524288/11/9 13.566 GiB/sec 14.894 GiB/sec 9.789 {'family_index': 0, 'per_family_instance_index': 23, 'run_name': 'FilterFixedSizeBinaryFilterNoNulls/524288/11/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 19349, 'byte_width': 9.0, 'data null%': 10.0, 'mask null%': 0.0, 'select%': 1.0} FilterFixedSizeBinaryFilterNoNulls/524288/14/9 13.603 GiB/sec 14.863 GiB/sec 9.265 {'family_index': 0, 'per_family_instance_index': 29, 'run_name': 'FilterFixedSizeBinaryFilterNoNulls/524288/14/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 19490, 'byte_width': 9.0, 'data null%': 90.0, 'mask null%': 0.0, 'select%': 1.0} TakeFixedSizeBinaryRandomIndicesNoNulls/524288/10/9 124.390M items/sec 133.566M items/sec 7.377 {'family_index': 2, 'per_family_instance_index': 3, 'run_name': 'TakeFixedSizeBinaryRandomIndicesNoNulls/524288/10/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 164, 'byte_width': 9.0, 'null_percent': 10.0} TakeFixedSizeBinaryMonotonicIndices/524288/2/9 116.792M items/sec 124.182M items/sec 6.328 {'family_index': 4, 'per_family_instance_index': 5, 'run_name': 'TakeFixedSizeBinaryMonotonicIndices/524288/2/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 161, 'byte_width': 9.0, 'null_percent': 50.0} TakeFixedSizeBinaryRandomIndicesNoNulls/524288/1000/9 135.860M items/sec 142.524M items/sec 4.905 {'family_index': 2, 'per_family_instance_index': 1, 'run_name': 'TakeFixedSizeBinaryRandomIndicesNoNulls/524288/1000/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 180, 'byte_width': 9.0, 'null_percent': 0.1} TakeFixedSizeBinaryRandomIndicesWithNulls/524288/1000/9 131.123M items/sec 137.400M items/sec 4.788 {'family_index': 3, 'per_family_instance_index': 1, 'run_name': 'TakeFixedSizeBinaryRandomIndicesWithNulls/524288/1000/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 176, 'byte_width': 9.0, 'null_percent': 0.1} TakeFixedSizeBinaryRandomIndicesNoNulls/524288/0/9 220.634M items/sec 230.872M items/sec 4.640 {'family_index': 2, 'per_family_instance_index': 9, 'run_name': 'TakeFixedSizeBinaryRandomIndicesNoNulls/524288/0/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 295, 'byte_width': 9.0, 'null_percent': 0.0} TakeFixedSizeBinaryRandomIndicesNoNulls/524288/2/9 97.425M items/sec 101.477M items/sec 4.159 {'family_index': 2, 'per_family_instance_index': 5, 'run_name': 'TakeFixedSizeBinaryRandomIndicesNoNulls/524288/2/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 130, 'byte_width': 9.0, 'null_percent': 50.0} TakeFixedSizeBinaryRandomIndicesWithNulls/524288/10/9 104.830M items/sec 108.346M items/sec 3.354 {'family_index': 3, 'per_family_instance_index': 3, 'run_name': 'TakeFixedSizeBinaryRandomIndicesWithNulls/524288/10/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 100, 'byte_width': 9.0, 'null_percent': 10.0} TakeFixedSizeBinaryRandomIndicesWithNulls/524288/1/9 378.858M items/sec 387.322M items/sec 2.234 {'family_index': 3, 'per_family_instance_index': 7, 'run_name': 'TakeFixedSizeBinaryRandomIndicesWithNulls/524288/1/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 506, 'byte_width': 9.0, 'null_percent': 100.0} TakeFixedSizeBinaryRandomIndicesWithNulls/524288/0/9 221.900M items/sec 226.450M items/sec 2.050 {'family_index': 3, 'per_family_instance_index': 9, 'run_name': 'TakeFixedSizeBinaryRandomIndicesWithNulls/524288/0/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 295, 'byte_width': 9.0, 'null_percent': 0.0} TakeFixedSizeBinaryMonotonicIndices/524288/0/9 248.664M items/sec 253.037M items/sec 1.758 {'family_index': 4, 'per_family_instance_index': 9, 'run_name': 'TakeFixedSizeBinaryMonotonicIndices/524288/0/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 332, 'byte_width': 9.0, 'null_percent': 0.0} TakeFixedSizeBinaryRandomIndicesNoNulls/524288/1/9 197.730M items/sec 201.173M items/sec 1.741 {'family_index': 2, 'per_family_instance_index': 7, 'run_name': 'TakeFixedSizeBinaryRandomIndicesNoNulls/524288/1/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 264, 'byte_width': 9.0, 'null_percent': 100.0} TakeFixedSizeBinaryRandomIndicesWithNulls/524288/2/9 73.196M items/sec 74.167M items/sec 1.327 {'family_index': 3, 'per_family_instance_index': 5, 'run_name': 'TakeFixedSizeBinaryRandomIndicesWithNulls/524288/2/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 96, 'byte_width': 9.0, 'null_percent': 50.0} TakeFixedSizeBinaryMonotonicIndices/524288/1/9 192.545M items/sec 188.138M items/sec -2.289 {'family_index': 4, 'per_family_instance_index': 7, 'run_name': 'TakeFixedSizeBinaryMonotonicIndices/524288/1/9', 'repetitions': 1, 'repetition_index': 0, 'threads': 1, 'iterations': 257, 'byte_width': 9.0, 'null_percent': 100.0} ``` ### Are these changes tested? Yes. ### Are there any user-facing changes? No. * Closes: apache#39740 Authored-by: Antoine Pitrou <antoine@python.org> Signed-off-by: Antoine Pitrou <antoine@python.org>
This commit ports the latest state of scalar_hash kernels without pulling a long development history. This kernel is an element-wise function that uses an xxHash-like algorithm, prioritizing speed and not suitable for cryptographic purposes. The function is implemented by the `FastHashScalar` struct which is templated by the output type (which is assumed to be either UInt32 or UInt64, but there is no validation of that at the moment). The benchmarks in scalar_hash_benchmark.cc uses the hashing_benchmark.cc file as a reference (in cpp/src/arrow/util/), but only covers various input types and the key hashing functions (from key_hash.h). The tests in scalar_hash_test.cc use a simplified version of hashing based on what is implemented in the key_hash.cc. The idea being that the high-level entry points for high-level types should eventually reach an expected application of the low-level hash functions on simple data types; the tests do this exact comparison. At the moment, the tests pass for simple cases, but they do not work for nested types with non-trivial row layouts (e.g. ListTypes). Issue: ARROW-8991 Issue: apacheGH-17211
This commit pulls the latest changes to key_hash.h and implementations in light_array without the burden of a long development history. The only change in key_hash.h is the addition of a friend function which is used in scalar_hash_test.cc. Changes in light_array.[h,cc] are to accommodate two scenarios: (1) the use of ArraySpan, which was introduced after light_array was written; and (2) the need for a KeyColumnArray to allocate data for the purposes of interpreting (or decoding) the structure of a nested type. The main reason for the 2nd scenario is that a ListArray may have many values represented in a single row which should be hashed together; however, if the ListArray has a nested ListArray or other type, the row may have further structure. In the simplest interpretation, only the highest-level structure (the "outer" ListArray) needs to be preserved, and any further nested structures must be explicitly handled by custom kernels (or any future options, etc. that are upstreamed into Arrow). In trying to efficiently interpret complex nested types, ArraySpan can be useful because it is non-owning, thus the main reason for the 1st aforementioned scenario. Although unfinished, any tests added to light_array_test.cc should accommodate the 2 scenarios above.
This commit includes changes to register a new compute function without the burden of a long development history. The change to cpp/src/arrow/CMakeLists.txt includes scalar_hash.cc in compilation as it is used by the new Hash64 function defined in api_scalar.[h,cc]. The change to cpp/src/arrow/compute/kernels/CMakeLists.txt includes scalar_hash_test.cc in compilation for tests and it also adds a new benchmark binary that is implemented by scalar_hash_benchmark.cc. The registry files are updated to register the kernel implementations in scalar_hash.cc with the function definitions in api_scalar.[h,cc]. Finally, docs/source/cpp/compute.rst adds documentation for the Hash64 function. Issue: apacheGH-17211 Issue: ARROW-8991
This commit includes additions to the general hashing benchmarks that cover the use of hashing functions in key_hash.h without carrying the burden of a long dev history. Some existing benchmark names were changed to distinguish between the use of Int32 and Int64 types, new benchmarks were added that use the functions declared in key_hash.h. The reason the new benchmarks are added is because it is claimed they prioritize speed over cryptography as they're primarily used for join algorithms and other processing tasks, which the hashing benchmark can now provide observability for. Issue: apacheGH-17211 Issue: ARROW-8991
… ARROW-8991-newfn-scalar-hash-fresh using a merge commit to overwrite commit history while maintaining current state of code
Contributor
Author
|
alright, I attempted to force change of commit history by using While this seems to be the "correct way", clearly it makes the PR look gratuitous. I'm going to close this one, and refer to it from a new PR. I'll try to update everything as appropriate. |
|
|
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
19 participants
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.
We would like to expose hashing functions via new compute functions. This PR would add a scalar compute function,
FastHash32, which usesHashing32::HashMultiColumnto compute a 32-bit hash for each row of the input array (or scalar).This PR focuses on 32-bit hashing, but I would like to use this PR to help decide if a different design is preferable before adding 64-bit hashing. I also am not sure what should be in the unit test except for code coverage.
Potential design changes:
Optionsargument.Confusion about unit tests:
Edit (8/9/22):
This PR has been open awhile but seems to have hit a good level of functionality. This PR does not:
I think these can be addressed in future improvements: