[tisonkun]

XOR construction repeats fingerprint allocation and assignment for each width, while the benchmark matrix emphasizes adjacent cardinalities and a specially selected retry input. Share fingerprint allocation with assignment, centralize the validated 32-bit range conversion, and remove the builder forwarding method used only to expose an attempt limit to tests.

Keep retry, order-independence, empty-filter, and compatibility coverage at their existing behavior boundaries while removing overlapping assertions and tests. Replace the matched-retry performance matrix with small, shard-sized, and large builds; retain independent-filter batches and imported-filter queries. Deduplicating comparisons now use repeated values and include the sorting and deduplication required by xorf's caller, so both sides perform equivalent work.

Validation: cargo x check, cargo x test (815 tests), cargo x lint (identical clean source copy), and cargo x bench -- --test (246 implementation/workload combinations). Optimized before/after spot checks show similar construction and query timings; this change does not claim a new broad speedup.

The complete benchmark results include ordinary bench and fat LTO, all raw samples, and the remaining slower cases.

[tisonkun]

Focused XOR benchmark results

This refactor keeps construction and query performance near the preceding revision; it does not claim a new broad speedup. The matrix now covers 45 scenarios / 93 implementation rows per profile, replacing the forced-retry and adjacent-cardinality cases with representative build sizes. Deduplicating comparisons include the same collect/sort/deduplicate work on both sides and use repeated values.

Counts below mean lower / equal / higher elapsed-time medians versus xorf. They are not statistical wins or an aggregate score.

Workload	Ordinary bench	Fat LTO, one codegen unit
Distinct builds	6 / 0 / 3	6 / 0 / 3
Deduplicating builds	9 / 0 / 0	9 / 0 / 0
Independent-filter batches	6 / 0 / 0	6 / 0 / 0
Single-filter queries	11 / 1 / 0	10 / 0 / 2
Imported-filter queries	12 / 0 / 0	7 / 0 / 5

The default-seed 16K distinct builds still need an extra attempt and remain slower. LTO also has seven query medians 1.0%–2.2% above xorf; six of those seven sample ranges overlap. Exact remaining gaps are listed below alongside all results. The changed build definitions establish a new baseline: improvements relative to xorf in the deduplicating rows must not be interpreted as before/after speedups caused by the refactor.

Both profiles use optimized Cargo bench builds with five repeats, fixed rounds, and setup/cleanup included. Apple M4 Max, macOS arm64, rustc 1.101.0-nightly (8006091eb, 2026-10-17), xorf 0.13.0. Measured production and benchmark code: 5cb4807; e200080 only simplifies a unit test. The ScopeDB end-to-end report separately measures real ingestion and query paths with ThinLTO.

Reproduction and before/after checks

cargo x bench -- xor --repeat 5 --raw

CARGO_PROFILE_BENCH_LTO=fat CARGO_PROFILE_BENCH_CODEGEN_UNITS=1 cargo x bench -- xor --repeat 5 --raw

Before/after checks used identical retained benchmark definitions on 541f87e and this change, with three repetitions. Independent-filter batch medians changed from 265/149/173 ms to 267/148/170 ms for 128/16K/50K inputs. Single-filter build changes ranged from -2.7% to +2.2%, and query changes from -4.1% to +2.5%. These small mixed differences do not establish a new general performance improvement.

Every remaining above-xorf median

Profile	Scenario	DS entry	DS ms	xorf ms	More elapsed time	Ranges overlap
Ordinary bench	xor::build::8_distinct/16384	DataSketches	361	279	+29.4%	no
Ordinary bench	xor::build::16_distinct/16384	DataSketches	361	280	+28.9%	no
Ordinary bench	xor::build::32_distinct/16384	DataSketches	366	284	+28.9%	no
Fat LTO, one codegen unit	xor::build::8_distinct/16384	DataSketches	357	279	+28.0%	no
Fat LTO, one codegen unit	xor::build::16_distinct/16384	DataSketches	365	288	+26.7%	no
Fat LTO, one codegen unit	xor::build::32_distinct/16384	DataSketches	366	284	+28.9%	no
Fat LTO, one codegen unit	xor::query::16_present/50000	DataSketches	238	235	+1.3%	no
Fat LTO, one codegen unit	xor::query::16_absent/50000	DataSketches	239	235	+1.7%	yes
Fat LTO, one codegen unit	xor::query_shards::8_absent/16384	DataSketches	358	351	+2.0%	yes
Fat LTO, one codegen unit	xor::query_shards::16_present/128	DataSketches	289	286	+1.0%	yes
Fat LTO, one codegen unit	xor::query_shards::16_present/16384	DataSketches	362	358	+1.1%	yes
Fat LTO, one codegen unit	xor::query_shards::16_absent/16384	DataSketches	370	364	+1.6%	yes
Fat LTO, one codegen unit	xor::query_shards::32_absent/16384	DataSketches	370	362	+2.2%	yes

All measurements and raw samples (ms)

Delta = (DataSketches median / xorf median − 1) × 100%; negative means less elapsed time. Each scenario keeps identical work and rounds across implementations.

Ordinary bench

Scenario	Implementation	Rounds	Median	Min–max	Raw samples	Delta vs xorf
xor::build::8_distinct/1000	DataSketches	60000	305	302–322	322, 303, 302, 305, 305	-33.4%
xor::build::8_distinct/1000	xorf	60000	458	453–459	458, 459, 453, 456, 458	reference
xor::build::8_deduplicate/1000	DataSketches	60000	356	355–359	355, 359, 357, 356, 356	-11.7%
xor::build::8_deduplicate/1000	xorf	60000	403	400–411	400, 411, 401, 403, 403	reference
xor::build::16_distinct/1000	DataSketches	60000	300	299–301	300, 301, 301, 300, 299	-35.5%
xor::build::16_distinct/1000	xorf	60000	465	461–466	461, 465, 466, 466, 464	reference
xor::build::16_deduplicate/1000	DataSketches	60000	354	352–359	359, 354, 353, 357, 352	-11.1%
xor::build::16_deduplicate/1000	xorf	60000	398	395–406	396, 398, 395, 400, 406	reference
xor::build::32_distinct/1000	DataSketches	60000	308	308–311	311, 308, 309, 308, 308	-34.6%
xor::build::32_distinct/1000	xorf	60000	471	470–474	473, 474, 470, 471, 470	reference
xor::build::32_deduplicate/1000	DataSketches	60000	356	354–361	359, 356, 361, 356, 354	-10.8%
xor::build::32_deduplicate/1000	xorf	60000	399	398–402	402, 399, 400, 398, 398	reference
xor::build::8_distinct/16384	DataSketches	2000	361	359–363	359, 361, 361, 363, 360	+29.4%
xor::build::8_distinct/16384	xorf	2000	279	276–292	292, 276, 279, 277, 280	reference
xor::build::8_deduplicate/16384	DataSketches	2000	249	248–250	248, 248, 249, 250, 250	-8.8%
xor::build::8_deduplicate/16384	xorf	2000	273	271–274	273, 271, 274, 273, 273	reference
xor::build::16_distinct/16384	DataSketches	2000	361	359–364	359, 361, 362, 361, 364	+28.9%
xor::build::16_distinct/16384	xorf	2000	280	275–285	285, 277, 281, 275, 280	reference
xor::build::16_deduplicate/16384	DataSketches	2000	249	245–254	245, 254, 248, 249, 249	-7.8%
xor::build::16_deduplicate/16384	xorf	2000	270	269–270	270, 270, 269, 270, 270	reference
xor::build::32_distinct/16384	DataSketches	2000	366	364–377	364, 374, 366, 377, 365	+28.9%
xor::build::32_distinct/16384	xorf	2000	284	280–288	288, 283, 284, 280, 284	reference
xor::build::32_deduplicate/16384	DataSketches	2000	250	248–261	250, 261, 250, 248, 251	-7.7%
xor::build::32_deduplicate/16384	xorf	2000	271	270–273	270, 271, 272, 273, 271	reference
xor::build::8_distinct/1000000	DataSketches	12	159	154–161	161, 159, 159, 154, 154	-41.3%
xor::build::8_distinct/1000000	xorf	12	271	269–280	280, 273, 271, 269, 270	reference
xor::build::8_deduplicate/1000000	DataSketches	12	152	151–154	154, 151, 152, 152, 153	-13.6%
xor::build::8_deduplicate/1000000	xorf	12	176	174–180	174, 179, 175, 180, 176	reference
xor::build::16_distinct/1000000	DataSketches	12	158	156–163	163, 160, 158, 156, 156	-42.5%
xor::build::16_distinct/1000000	xorf	12	275	272–280	280, 275, 278, 274, 272	reference
xor::build::16_deduplicate/1000000	DataSketches	12	153	151–153	153, 152, 151, 153, 153	-12.1%
xor::build::16_deduplicate/1000000	xorf	12	174	173–175	174, 173, 173, 175, 174	reference
xor::build::32_distinct/1000000	DataSketches	12	159	157–163	163, 160, 159, 157, 157	-41.5%
xor::build::32_distinct/1000000	xorf	12	272	270–282	282, 272, 279, 271, 270	reference
xor::build::32_deduplicate/1000000	DataSketches	12	154	151–154	154, 151, 153, 154, 154	-11.5%
xor::build::32_deduplicate/1000000	xorf	12	174	173–180	173, 174, 173, 176, 180	reference
xor::build_batches::8/128	default_seed	10000	263	262–272	272, 262, 263, 262, 263	-34.9%
xor::build_batches::8/128	matched_seed	10000	272	270–274	274, 272, 271, 270, 273	-32.7%
xor::build_batches::8/128	xorf	10000	404	403–406	405, 403, 406, 404, 404	reference
xor::build_batches::8/16384	default_seed	40	151	151–153	152, 151, 153, 151, 151	-58.1%
xor::build_batches::8/16384	matched_seed	40	142	140–145	145, 142, 142, 140, 141	-60.6%
xor::build_batches::8/16384	xorf	40	360	359–361	360, 360, 361, 359, 361	reference
xor::build_batches::8/50000	default_seed	12	172	171–172	172, 172, 171, 172, 171	-48.3%
xor::build_batches::8/50000	matched_seed	12	167	167–170	167, 167, 167, 170, 167	-49.8%
xor::build_batches::8/50000	xorf	12	333	332–340	333, 340, 333, 335, 332	reference
xor::query::8_present/16384	DataSketches	10000	275	272–280	275, 273, 272, 280, 276	-3.2%
xor::query::8_present/16384	xorf	10000	284	282–286	282, 283, 285, 284, 286	reference
xor::query::8_present/50000	DataSketches	3000	247	245–248	247, 246, 245, 248, 247	-3.9%
xor::query::8_present/50000	xorf	3000	257	257–257	257, 257, 257, 257, 257	reference
xor::query::8_absent/16384	DataSketches	10000	267	265–269	267, 267, 269, 267, 265	-7.0%
xor::query::8_absent/16384	xorf	10000	287	283–290	286, 290, 287, 287, 283	reference
xor::query::8_absent/50000	DataSketches	3000	243	242–244	244, 242, 244, 243, 242	-5.4%
xor::query::8_absent/50000	xorf	3000	257	255–260	258, 255, 260, 257, 257	reference
xor::query::16_present/16384	DataSketches	10000	277	276–278	278, 276, 276, 277, 277	-2.1%
xor::query::16_present/16384	xorf	10000	283	282–288	283, 284, 282, 282, 288	reference
xor::query::16_present/50000	DataSketches	3000	267	267–270	267, 270, 267, 267, 267	0.0%
xor::query::16_present/50000	xorf	3000	267	267–269	267, 267, 268, 267, 269	reference
xor::query::16_absent/16384	DataSketches	10000	273	271–275	271, 274, 275, 273, 273	-3.9%
xor::query::16_absent/16384	xorf	10000	284	283–290	284, 285, 290, 283, 284	reference
xor::query::16_absent/50000	DataSketches	3000	266	265–266	265, 266, 266, 266, 265	-1.5%
xor::query::16_absent/50000	xorf	3000	270	269–274	269, 272, 270, 270, 274	reference
xor::query::32_present/16384	DataSketches	10000	268	265–276	268, 276, 275, 268, 265	-5.6%
xor::query::32_present/16384	xorf	10000	284	283–290	285, 284, 283, 290, 284	reference
xor::query::32_present/50000	DataSketches	3000	280	279–281	280, 279, 279, 281, 281	-8.5%
xor::query::32_present/50000	xorf	3000	306	304–313	306, 306, 306, 313, 304	reference
xor::query::32_absent/16384	DataSketches	10000	273	267–275	267, 269, 275, 273, 273	-8.4%
xor::query::32_absent/16384	xorf	10000	298	287–305	287, 298, 292, 305, 298	reference
xor::query::32_absent/50000	DataSketches	3000	286	285–297	297, 286, 291, 285, 285	-6.2%
xor::query::32_absent/50000	xorf	3000	305	304–322	322, 305, 305, 304, 305	reference
xor::query_shards::8_present/128	DataSketches	20000	277	276–281	277, 277, 276, 278, 281	-14.2%
xor::query_shards::8_present/128	xorf	20000	323	322–327	327, 323, 322, 323, 325	reference
xor::query_shards::8_absent/128	DataSketches	20000	278	276–283	278, 278, 276, 277, 283	-14.7%
xor::query_shards::8_absent/128	xorf	20000	326	324–333	324, 333, 326, 326, 328	reference
xor::query_shards::8_present/16384	DataSketches	20000	354	349–355	355, 350, 354, 355, 349	-15.7%
xor::query_shards::8_present/16384	xorf	20000	420	416–423	418, 421, 420, 423, 416	reference
xor::query_shards::8_absent/16384	DataSketches	20000	348	347–355	352, 347, 355, 348, 348	-17.5%
xor::query_shards::8_absent/16384	xorf	20000	422	413–425	421, 413, 422, 425, 422	reference
xor::query_shards::16_present/128	DataSketches	20000	286	283–287	286, 283, 286, 285, 287	-12.8%
xor::query_shards::16_present/128	xorf	20000	328	326–328	326, 326, 328, 328, 328	reference
xor::query_shards::16_absent/128	DataSketches	20000	285	283–293	285, 293, 284, 283, 285	-12.6%
xor::query_shards::16_absent/128	xorf	20000	326	325–327	326, 326, 325, 326, 327	reference
xor::query_shards::16_present/16384	DataSketches	20000	364	358–367	364, 358, 363, 367, 364	-13.9%
xor::query_shards::16_present/16384	xorf	20000	423	422–428	423, 428, 423, 424, 422	reference
xor::query_shards::16_absent/16384	DataSketches	20000	363	361–365	363, 361, 365, 363, 364	-15.2%
xor::query_shards::16_absent/16384	xorf	20000	428	426–433	433, 432, 428, 426, 426	reference
xor::query_shards::32_present/128	DataSketches	20000	277	275–279	279, 276, 275, 278, 277	-23.9%
xor::query_shards::32_present/128	xorf	20000	364	363–365	364, 364, 365, 363, 363	reference
xor::query_shards::32_absent/128	DataSketches	20000	278	276–291	276, 277, 279, 291, 278	-23.2%
xor::query_shards::32_absent/128	xorf	20000	362	356–364	362, 364, 356, 359, 362	reference
xor::query_shards::32_present/16384	DataSketches	20000	372	368–373	368, 373, 369, 372, 372	-16.2%
xor::query_shards::32_present/16384	xorf	20000	444	441–446	446, 444, 446, 441, 442	reference
xor::query_shards::32_absent/16384	DataSketches	20000	367	360–375	375, 360, 367, 368, 366	-17.3%
xor::query_shards::32_absent/16384	xorf	20000	444	442–450	450, 443, 448, 442, 444	reference

Fat LTO, one codegen unit

Scenario	Implementation	Rounds	Median	Min–max	Raw samples	Delta vs xorf
xor::build::8_distinct/1000	DataSketches	60000	315	315–334	334, 319, 315, 315, 315	-32.7%
xor::build::8_distinct/1000	xorf	60000	468	468–488	468, 488, 470, 468, 468	reference
xor::build::8_deduplicate/1000	DataSketches	60000	365	365–371	365, 371, 365, 365, 367	-10.3%
xor::build::8_deduplicate/1000	xorf	60000	407	407–411	407, 408, 411, 407, 407	reference
xor::build::16_distinct/1000	DataSketches	60000	318	314–382	318, 382, 318, 316, 314	-32.2%
xor::build::16_distinct/1000	xorf	60000	469	466–479	468, 477, 479, 469, 466	reference
xor::build::16_deduplicate/1000	DataSketches	60000	360	356–366	365, 366, 360, 358, 356	-10.9%
xor::build::16_deduplicate/1000	xorf	60000	404	401–410	410, 404, 405, 401, 401	reference
xor::build::32_distinct/1000	DataSketches	60000	304	303–306	303, 306, 304, 304, 303	-33.9%
xor::build::32_distinct/1000	xorf	60000	460	457–470	470, 464, 459, 457, 460	reference
xor::build::32_deduplicate/1000	DataSketches	60000	360	357–364	361, 360, 357, 364, 358	-11.1%
xor::build::32_deduplicate/1000	xorf	60000	405	401–407	405, 407, 406, 403, 401	reference
xor::build::8_distinct/16384	DataSketches	2000	357	354–361	355, 361, 360, 357, 354	+28.0%
xor::build::8_distinct/16384	xorf	2000	279	278–285	279, 278, 285, 278, 280	reference
xor::build::8_deduplicate/16384	DataSketches	2000	253	251–258	255, 252, 251, 258, 253	-9.0%
xor::build::8_deduplicate/16384	xorf	2000	278	277–303	278, 277, 281, 303, 278	reference
xor::build::16_distinct/16384	DataSketches	2000	365	363–368	365, 364, 363, 368, 368	+26.7%
xor::build::16_distinct/16384	xorf	2000	288	286–292	292, 288, 288, 286, 290	reference
xor::build::16_deduplicate/16384	DataSketches	2000	253	250–254	253, 254, 251, 250, 254	-8.7%
xor::build::16_deduplicate/16384	xorf	2000	277	274–281	276, 281, 277, 274, 278	reference
xor::build::32_distinct/16384	DataSketches	2000	366	362–374	362, 370, 374, 366, 366	+28.9%
xor::build::32_distinct/16384	xorf	2000	284	282–290	284, 288, 290, 284, 282	reference
xor::build::32_deduplicate/16384	DataSketches	2000	253	248–259	259, 248, 250, 253, 254	-8.7%
xor::build::32_deduplicate/16384	xorf	2000	277	275–278	278, 275, 275, 277, 277	reference
xor::build::8_distinct/1000000	DataSketches	12	154	153–158	158, 153, 154, 155, 153	-42.3%
xor::build::8_distinct/1000000	xorf	12	267	255–278	278, 263, 255, 270, 267	reference
xor::build::8_deduplicate/1000000	DataSketches	12	152	150–152	150, 152, 151, 152, 152	-12.6%
xor::build::8_deduplicate/1000000	xorf	12	174	172–176	172, 176, 174, 174, 174	reference
xor::build::16_distinct/1000000	DataSketches	12	160	152–167	160, 153, 152, 163, 167	-39.8%
xor::build::16_distinct/1000000	xorf	12	266	260–277	273, 260, 260, 266, 277	reference
xor::build::16_deduplicate/1000000	DataSketches	12	153	151–154	153, 151, 152, 154, 153	-12.6%
xor::build::16_deduplicate/1000000	xorf	12	175	174–176	174, 174, 175, 176, 176	reference
xor::build::32_distinct/1000000	DataSketches	12	157	155–161	160, 155, 155, 161, 157	-42.3%
xor::build::32_distinct/1000000	xorf	12	272	265–289	284, 272, 265, 272, 289	reference
xor::build::32_deduplicate/1000000	DataSketches	12	153	152–154	153, 152, 153, 154, 154	-13.1%
xor::build::32_deduplicate/1000000	xorf	12	176	175–177	175, 175, 176, 176, 177	reference
xor::build_batches::8/128	default_seed	10000	271	269–273	271, 273, 273, 271, 269	-32.8%
xor::build_batches::8/128	matched_seed	10000	267	267–273	273, 267, 271, 267, 267	-33.7%
xor::build_batches::8/128	xorf	10000	403	401–413	403, 403, 410, 401, 413	reference
xor::build_batches::8/16384	default_seed	40	152	148–152	150, 152, 152, 152, 148	-57.7%
xor::build_batches::8/16384	matched_seed	40	141	138–143	138, 140, 142, 143, 141	-60.7%
xor::build_batches::8/16384	xorf	40	359	357–362	357, 357, 362, 361, 359	reference
xor::build_batches::8/50000	default_seed	12	173	171–174	171, 171, 173, 173, 174	-47.9%
xor::build_batches::8/50000	matched_seed	12	169	168–173	170, 169, 169, 168, 173	-49.1%
xor::build_batches::8/50000	xorf	12	332	329–336	332, 329, 336, 332, 331	reference
xor::query::8_present/16384	DataSketches	10000	248	247–252	248, 247, 248, 252, 249	-1.2%
xor::query::8_present/16384	xorf	10000	251	250–254	251, 252, 250, 251, 254	reference
xor::query::8_present/50000	DataSketches	3000	224	220–225	224, 224, 225, 220, 220	-3.0%
xor::query::8_present/50000	xorf	3000	231	226–236	236, 236, 229, 226, 231	reference
xor::query::8_absent/16384	DataSketches	10000	248	244–248	248, 244, 248, 247, 248	-1.2%
xor::query::8_absent/16384	xorf	10000	251	249–253	251, 251, 253, 249, 252	reference
xor::query::8_absent/50000	DataSketches	3000	221	221–226	226, 221, 221, 223, 221	-3.1%
xor::query::8_absent/50000	xorf	3000	228	224–233	233, 226, 229, 228, 224	reference
xor::query::16_present/16384	DataSketches	10000	251	249–253	250, 251, 252, 253, 249	-2.0%
xor::query::16_present/16384	xorf	10000	256	252–257	256, 257, 256, 257, 252	reference
xor::query::16_present/50000	DataSketches	3000	238	236–241	236, 238, 240, 241, 238	+1.3%
xor::query::16_present/50000	xorf	3000	235	234–235	235, 235, 234, 235, 234	reference
xor::query::16_absent/16384	DataSketches	10000	252	250–256	256, 251, 250, 252, 253	-1.2%
xor::query::16_absent/16384	xorf	10000	255	254–285	259, 285, 254, 254, 255	reference
xor::query::16_absent/50000	DataSketches	3000	239	237–241	237, 241, 239, 239, 240	+1.7%
xor::query::16_absent/50000	xorf	3000	235	230–241	231, 241, 236, 235, 230	reference
xor::query::32_present/16384	DataSketches	10000	244	243–246	244, 246, 245, 243, 244	-3.2%
xor::query::32_present/16384	xorf	10000	252	251–255	251, 252, 253, 251, 255	reference
xor::query::32_present/50000	DataSketches	3000	254	248–258	254, 256, 258, 248, 249	-3.8%
xor::query::32_present/50000	xorf	3000	264	262–270	264, 262, 270, 265, 263	reference
xor::query::32_absent/16384	DataSketches	10000	242	241–243	241, 242, 243, 242, 243	-6.6%
xor::query::32_absent/16384	xorf	10000	259	256–261	259, 256, 257, 259, 261	reference
xor::query::32_absent/50000	DataSketches	3000	248	247–253	253, 247, 252, 248, 248	-5.3%
xor::query::32_absent/50000	xorf	3000	262	259–268	261, 259, 262, 268, 265	reference
xor::query_shards::8_present/128	DataSketches	20000	277	274–279	276, 278, 279, 274, 277	-1.4%
xor::query_shards::8_present/128	xorf	20000	281	280–282	282, 281, 281, 280, 280	reference
xor::query_shards::8_absent/128	DataSketches	20000	277	273–279	273, 277, 274, 279, 277	-0.7%
xor::query_shards::8_absent/128	xorf	20000	279	277–286	277, 279, 278, 286, 283	reference
xor::query_shards::8_present/16384	DataSketches	20000	341	341–345	341, 341, 341, 345, 345	-1.2%
xor::query_shards::8_present/16384	xorf	20000	345	340–347	340, 345, 344, 346, 347	reference
xor::query_shards::8_absent/16384	DataSketches	20000	358	352–358	358, 355, 358, 358, 352	+2.0%
xor::query_shards::8_absent/16384	xorf	20000	351	349–355	355, 350, 349, 351, 351	reference
xor::query_shards::16_present/128	DataSketches	20000	289	283–295	295, 293, 288, 289, 283	+1.0%
xor::query_shards::16_present/128	xorf	20000	286	284–297	287, 297, 286, 284, 284	reference
xor::query_shards::16_absent/128	DataSketches	20000	284	282–294	284, 282, 294, 283, 285	-0.7%
xor::query_shards::16_absent/128	xorf	20000	286	281–289	284, 281, 289, 286, 287	reference
xor::query_shards::16_present/16384	DataSketches	20000	362	359–365	362, 365, 359, 364, 360	+1.1%
xor::query_shards::16_present/16384	xorf	20000	358	354–359	358, 358, 354, 359, 357	reference
xor::query_shards::16_absent/16384	DataSketches	20000	370	368–375	368, 368, 370, 375, 370	+1.6%
xor::query_shards::16_absent/16384	xorf	20000	364	360–372	372, 363, 360, 364, 367	reference
xor::query_shards::32_present/128	DataSketches	20000	277	277–278	278, 277, 277, 277, 277	-2.5%
xor::query_shards::32_present/128	xorf	20000	284	281–299	284, 299, 282, 281, 284	reference
xor::query_shards::32_absent/128	DataSketches	20000	276	275–279	279, 276, 278, 275, 275	-2.5%
xor::query_shards::32_absent/128	xorf	20000	283	281–290	283, 290, 283, 281, 282	reference
xor::query_shards::32_present/16384	DataSketches	20000	364	359–368	364, 368, 364, 363, 359	-0.5%
xor::query_shards::32_present/16384	xorf	20000	366	359–371	367, 371, 363, 366, 359	reference
xor::query_shards::32_absent/16384	DataSketches	20000	370	364–380	364, 368, 372, 380, 370	+2.2%
xor::query_shards::32_absent/16384	xorf	20000	362	361–367	367, 362, 367, 362, 361	reference