Sapphire Rapids: Confirm ASCII root matches without trie replay
perfloop-oss/casei · DEPENDENT LOAD CHAIN
https://perfloop.ai/t/oss/case_zww66a1p23
Verdict
VERIFIED · settled 2026-10-04 · merged as tsenart/casei#25
What happened: The paired measurements met the required improvement.
Hypothesis
Speed up the required English five-name count on Sapphire Rapids with complete ASCII word confirmation instead of serialized trie lookup. Nominations and match priority stay unchanged; this is a confirmation prerequisite, not proved field leadership.
The retained899232-byte fixture reaches matchAtStart. Census recorded895 replays,725 matches and12484 edges;876 windows pass even4*maxUnits ASCII coverage. A complete maxUnits-byte ASCII prefix excludes wider spellings for every terminal; unknown prefixes keep decoding.
Replay costs1.75/3.54 sampled seconds beside1.70 scanning. It is material here; Tom is a regression control, not benefit coverage. Certificate exploration changed about180us to111-124us,31-40%, below the59.7% removal needed at provisional2.480. Even free replay has only about49% sampled headroom. No extra filter, matcher or public API; report descriptor setup/memory.
Exact ASCII token classes admit bounded word certificates cheaper than the observed dependent replay.
Prove complete ASCII certificate equivalence and fallback for any unknown lower-ID spelling. Pair the required complete five-name count on model143 against pristine baseline; no repeatable ns/op saving, any result/order/ID/width/folding mismatch, overread or guard slowdown rejects it. Guard Tom/Russian counts, existing BenchmarkBar wins and exercised fallback; report descriptor setup/memory.
Recovered flag-selected experiments are not pristine-baseline official pairs. Safe partial loads, tests and18-row preflight are retained; initial checks lacked fixture environment before correction. Broader filter/confirmation Cases have no Candidates. This replaces complete confirmation, not nomination selectivity.
Change to test: On Sapphire Rapids, replace eligible four/five-pattern replay with complete word certificates from the plan's ASCII token classes. Prove the prefix ASCII for all terminals, test IDs in order and retain trie replay for unknown spellings.
Where it lives
perfloop-oss/casei · audit/rebar/runner/main.go
Evidence
Rebar curated/02 English five-name retained-Matcher Each count on the full 899,232-byte fixture (725 matches), Sapphire Rapids model 143 · 10 sample pairs
| metric | baseline | candidate | paired median change | confidence range | required | result |
|---|---|---|---|---|---|---|
ns/op |
181630 |
105390 |
−41.6% (−75562) |
−82430 to −74446 |
< −9081 |
PASSED |
MB/s |
4951 |
8532 |
+71.7% (+3551) |
+3518 to +3769 |
≥ −247.6 |
PASSED |
Rebar Leipzig Tom/Sawyer/Huckleberry/Finn retained-Matcher Each count (4,152 matches), Sapphire Rapids model 143 regression guard · 10 sample pairs
| metric | baseline | candidate | paired median change | confidence range | required | result |
|---|---|---|---|---|---|---|
ns/op |
1717584 |
1655469 |
−2.4% (−41472) |
−61807 to −27853 |
≤ 0 |
PASSED |
MB/s |
9324 |
9673 |
+2.5% (+231.2) |
+163.7 to +344.8 |
≥ 0 |
PASSED |
Rebar curated/02 Russian five-name retained-Matcher Each count (971 matches), Sapphire Rapids model 143 regression guard · 10 sample pairs
| metric | baseline | candidate | paired median change | confidence range | required | result |
|---|---|---|---|---|---|---|
ns/op |
514106 |
509617 |
−0.9% (−4474) |
−6193 to −3349 |
≤ 0 |
PASSED |
MB/s |
3055 |
3082 |
+0.9% (+26.82) |
+20.03 to +37.26 |
≥ 0 |
PASSED |
Preserve each baseline-winning row in the complete 38-row BenchmarkBar on Sapphire Rapids model 143 · 10 sample pairs
| metric | baseline | candidate | paired median change | confidence range | required | result |
|---|---|---|---|---|---|---|
bar_multi_multi_N1_unicode_pair_miss_1_5mb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_multi_multi_N2_miss_log_1mb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_multi_multi_N512_miss_hazard_64kb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_multi_multi_N512_miss_log_64kb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_multi_multi_N5_raw_transition_late_hit_5mb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_multi_multi_N5_raw_transition_miss_5mb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_multi_multi_N64_miss_log_64kb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_multi_multi_N64_miss_ru_64kb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_multi_multi_N8_hazard_hit_1mb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_multi_multi_N8_hit_log_1mb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_multi_multi_N8_miss_hazard_1mb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_multi_multi_N8_miss_log_1mb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_multi_multi_N8_miss_ru_1mb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_multi_multi_forms4_complete_triple_miss_1mb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_multi_multi_forms8_complete_triple_near_miss_64kb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_single_code_hit_brackets_256kb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_single_code_miss_256kb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_single_kelvin_hazard_1mb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_single_latency_match_end_1kb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_single_latency_match_mid_1kb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_single_latency_match_start_1kb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_single_latency_miss_1kb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_single_log_hit_sparse_1mb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_single_log_miss_1kb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_single_log_miss_1mb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_single_log_miss_64kb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_single_log_needle16_64kb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_single_log_needle32_64kb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_single_log_needle3_64kb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_single_log_needle8_64kb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_single_periodic_miss_64kb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_single_prose_hit_dense_1mb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_single_prose_miss_1mb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_single_ru_hit_sparse_1mb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_single_ru_latency_miss_1kb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_single_ru_miss_1mb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_single_samechar_miss_64kb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
bar_single_torture_miss_64kb_all_samples_win |
1 |
1 |
0 |
0 to 0 |
≥ 0 |
PASSED |
Checks: 5 of 5 passed. Verification: no defect found.