Repository navigation
What's Changed
Other changes
- Seed compile-time hash tables per process by @dain in #116
- Count class runs without branching on membership by @dain in #144
- Stop rescanning shared roots in dominator pass by @dain in #119
- Let C2 inline the class-run span search again by @dain in #121
- Compare candidate bytes once on 128-bit SVE by @dain in #132
- Drop always-true guards from the RE2 set stop by @dain in #134
- Read short literals without slice bounds checks by @dain in #153
- Keep model checks out of timed bulk loops by @dain in #137
- Speed up line grep with captures by @dain in #138
- Build capture instruction words at compile time by @dain in #156
- Give CI build legs thirty minutes by @dain in #157
- Scan byte sets with one compare on 128-bit SVE by @dain in #135
- Register the inner literal DFA pair once by @dain in #149
- Speed up the literal prefix scan by @dain in #140
- Reuse byte-scan samples within a DFA count by @dain in #148
- End short literal scans with an overlapping word by @dain in #155
- Keep the full-match OnePass byte loop call-free by @dain in #150
- Scan ASCII word-boundary runs without the DFA by @dain in #139
- Use the literal filter for rare start sets by @dain in #142
- Pair a multibyte literal's continuation byte by @dain in #141
- Cut per-attempt work in unanchored scan plans by @dain in #158
- Replace DFA reader slots with copy-on-reset by @dain in #151
- Read inlining trees per compiler thread by @dain in #161
- Count literal alternation routes only in tests by @dain in #130
- Measure the routes one engine path moves by @dain in #117
- Start capture searches with sixteen jobs by @dain in #159
- Search short contained literals in every language by @dain in #118
- Use vector scans for long inner literal searches by @dain in #143
- Search large dictionaries the DFA cannot hold by @dain in #120
- Find one-byte matches on the table in matchers by @dain in #122
- Speed up candidate checks on ReDoS shapes by @dain in #163
- Reject short windows without required bytes by @dain in #147
- Find literal and fixed-width spans in matchers by @dain in #123
- Let the candidate fallback test reach the cursor by @dain in #164
- Keep the pointer table from forcing NFA fallback by @dain in #165
- Share the short fixed-width matcher by @dain in #125
- Keep the one-of fact for prefix-scan patterns by @dain in #154
- Fix Java word boundaries near supplementary text by @dain in #167
- Fold Perl classes before negating in brackets by @dain in #171
- Step over encoded surrogates in replaceAll by @dain in #172
- Keep deep patterns from overflowing FilteredRe2 by @dain in #174
- Skip DFA self-loop runs with exit-byte scans by @dain in #176
- Find longest matches of RE2 lazy repeats by @dain in #177
- Fold Latin-1 classes by range by @dain in #179
- Add differential fuzzers for every language by @dain in #181
- Reject oversized repetition before expanding it by @dain in #178
- Cap unanchored BitState at upstream's bitmap by @dain in #173
- Keep the crate's priority on Rust nullable loops by @dain in #180
- Add a heap pointer table for safe-mode DFAs by @dain in #162
- Skip inner literal hits on impossible neighbors by @dain in #152
- Accept Rust repetition counts above 1,000 by @dain in #169
- Step self-loop runs on the grown DFA tables by @dain in #186
- Stop compiled patterns keeping their parse trees by @dain in #184
- Bound literal extraction by the memory budget by @dain in #187
- Fit flattening scratch within its charge by @dain in #185
- Skip inner literals in patterns that never match by @dain in #166
- Find dot-star literal spans in every language by @dain in #126
- Find RE2 and Rust ordered literals over byte gaps by @dain in #127
- End Java loops on an empty iteration by @dain in #175
- Run Java and Trino end anchors on OnePass by @dain in #128
- Start each lazy loop iteration afresh by @dain in #183
- Scan dictionary trie edges without a bound test by @dain in #160
- Reuse nullable-loop prefixes instead of cloning by @dain in #189
- Cut Java zero-width repetitions to one copy by @dain in #191
- Wait out C1 code in allocation tests by @dain in #190
- Pin JDK 26 sharp S matches on every JVM by @dain in #192
- Scan fixed-width patterns for a rare class by @dain in #194
- Explain where Regulator's speed comes from by @dain in #168
- Count runs of classes that end at U+10FFFF by @dain in #197
- Survive paired-row release in self-loop scans by @dain in #196
- Reject long windows lacking a required literal by @dain in #188
- Scan DFA self-loops by the bytes they stay on by @dain in #202
- Return at the earliest DFA match by @dain in #198
- Encode UTF-8 by code point to avoid a C2 bug by @dain in #200
- Require a catalog entry for each optimization by @dain in #182
- Correct stale engine claims in reference docs by @dain in #170
- Match Trino fixed-width regions on the matcher by @dain in #204
- Scan for a match's first two or three classes by @dain in #201
- Skip the DFA only after a streak of bailed counts by @dain in #203
- Hold the catalog to optimizations worth knowing by @dain in #199
- Compile large literal sets without a program by @dain in #205
- Probe the rarest folded pair of a prefix by @dain in #195
- Split Rebar grep lines with a word search by @dain in #207
- Search inner literals behind a run of one class by @dain in #193
- Accept a zero duration in Rebar runner tests by @dain in #217
- Report the memory budget a rejected pattern needs by @dain in #206
- Reject spans too short or too long to match by @dain in #209
- Admit inner literals after a folded letter by @dain in #211
- Run probe children at the build's vector width by @dain in #212
- Search folded inner literals with the pair kernel by @dain in #213
- Flatten alternations nested in x{1} once by @dain in #208
- Give shallow literal-set states dense rows by @dain in #210
- Skip ahead in literal sets of long words by @dain in #219
- Read long class runs with a vector scan by @dain in #216
- Bound long searches by the last required suffix by @dain in #218
- Vector-scan a fixed-width one-byte candidate by @dain in #221
- Gate inner literal hits on more nearby bytes by @dain in #214
- Ask the JVM for the test vector width by @dain in #222
- Compare folded literals at filter stops by @dain in #225
- Hold SWAR constants in registers in word loops by @dain in #226
- Run bounded repeats on a bit-parallel NFA by @dain in #220
- Step paired DFA rows on a flat table on aarch64 by @dain in #227
- Test every start of a 64-byte block at once by @dain in #224
- Scan for bytes alternatives share past a prefix by @dain in #223
- Settle the JVM before Rebar samples by @dain in #228
- Check long windows for one of the required bytes by @dain in #230
- Count single dense classes on the class counter by @dain in #231
- Address reverse scan blocks from their low end by @dain in #234
- Walk each mark run once for Java word boundaries by @dain in #236
- Record Joni's iota subscript fold difference by @dain in #240
- Run budgeted benchmark rows at their own budget by @dain in #238
- Lower the default memory budget to 32 MiB by @dain in #239
- Skip a thrashing DFA in boolean line searches by @dain in #237
- Let start literal filter groups span three bits by @dain in #232
- Search a rare folded run behind a start literal by @dain in #235
- Start DFA state tables small and grow them by @dain in #229
- Hand long-window literal hits to inner search by @dain in #241
- Resume the candidate cursor past a DFA quit by @dain in #242
- Search window literals where the window wins by @dain in #233
- Scan a folded one-byte prefix with vectors by @dain in #243
- Decide class-run alternation captures by count by @dain in #244
- Find one-shot dot-star spans without the DFA by @dain in #245
- Keep program instructions as packed words by @dain in #246
- Explain how threads share a pattern's cache by @dain in #247
- Compare compiled patterns by pattern and options by @dain in #248
- Use the tagged capture DFA for small programs by @dain in #249
- Add an access-log line search to everyday rows by @dain in #250
- Report the memory patterns and matchers retain by @dain in #251
- Count DFA scan routes only in tests by @dain in #252
- Let Java dot-star literals take the span route by @dain in #253
- Fall back on dense gate-rejected literal hits by @dain in #254
- Count unbounded byte runs from their starts by @dain in #255
- Search contained literals only in long windows by @dain in #257
- Load fixed-width position masks by position by @dain in #258
- Find exact-literal spans without the DFA by @dain in #256
- Publish preliminary 2.1 benchmark results by @dain in #259
Full Changelog: 2.0...2.1