Commit Graph

268 Commits

Author SHA1 Message Date
Behdad Esfahbod f13a79548f [subset] Convert another use of hashmap to unique_ptr 2022-06-03 01:17:20 -06:00
Behdad Esfahbod 997d9cc466 [map] Make unique_ptr hashable 2022-06-02 18:47:42 -06:00
Behdad Esfahbod 5af5a56593 [VarStore] Implement writing 32bit var-store
Untested.

Finishes fixing https://github.com/harfbuzz/harfbuzz/issues/2965
2022-05-31 03:58:35 -06:00
Behdad Esfahbod 75112098ac [VarStore] Implement reading 32bit var-store
Untested.

Part of https://github.com/harfbuzz/harfbuzz/issues/2965

Serializing is incomplete.
2022-05-31 03:58:29 -06:00
Behdad Esfahbod e5d6da7905 [varStore] rename shortCount to wordCount 2022-05-30 07:48:58 -06:00
Behdad Esfahbod 371e14d99c Combine uses of map has() then get() with has(.., &..) 2022-05-28 13:40:30 -06:00
Behdad Esfahbod 6010feeeb5 [varStore] Rename variable as per review
https://github.com/harfbuzz/harfbuzz/pull/3605
2022-05-24 09:01:02 -06:00
Behdad Esfahbod d9acc045f1 [VarStore] Sprinkle cache_t type around
It's available so no need to use void*.
2022-05-21 15:12:14 -06:00
Behdad Esfahbod da38312e42 [VarStore] Pepper cache with likely() 2022-05-21 15:12:14 -06:00
Behdad Esfahbod cf8f00e354 [VarStore] Don't use NAN
Is faster.

With this, I'm seeing 25 to 28% speedup in glyph_h_advances benchmark
of benchmark-font for var/hb tests.
2022-05-21 15:12:14 -06:00
Behdad Esfahbod 880f50f7e4 Refactor varstore cache 2022-05-21 15:12:14 -06:00
Behdad Esfahbod f2a2fb91a3 [GPOS] Cache VarStore region scalars 2022-05-21 15:12:14 -06:00
Behdad Esfahbod 5fbc70c59b [VarStore] Add cache API 2022-05-21 15:12:14 -06:00
Garret Rieger 3ab2c7935f [subset-perf] Signficiantly speed up ClassDef*::subset.
Eliminates the usage of a glyph -> klass hash map and replaces it with a vector storing the mapping. This allows us to use the vector directly as the iterator driving the serialize. Approximately 1% speedup for Noto Nastaliq.
2022-05-19 12:37:28 -06:00
Garret Rieger 0b1c2ff96a [subset-perf] Remove extra map lookup in ClassDef subset methods. 2022-05-18 18:25:36 -06:00
Garret Rieger 13ace77f1d [subset-perf] Use glyph_map instead of set in ClassDefFormat. 2022-05-18 18:25:36 -06:00
Garret Rieger adae2f2272 [subset-perf] Cache a glyph map for gsub.
This allows us in some cases to avoid using glyph_set_gsub as a filter.
2022-05-18 18:25:36 -06:00
Behdad Esfahbod cedf739646 Add some commented-out code 2022-05-18 16:53:26 -06:00
Garret Rieger 48dfbd54a3 [subset] minor cleanup. 2022-05-18 15:24:54 -06:00
Garret Rieger 482c6e5dc4 [subset-perf] Speed up Coverage::serialize by caching iterator. 2022-05-18 15:24:54 -06:00
Behdad Esfahbod 14b18725f0 In Coverage::iter_t, assume iterators are from same Coverage object
No need to support otherwise.
2022-05-18 15:14:32 -06:00
Behdad Esfahbod 27141735c3 [subset] Add Coverage::__end__ implementation 2022-05-18 15:12:49 -06:00
Behdad Esfahbod c476f58adb [subset] Write CoverageFormat2::intersects_coverage() as bsearch() 2022-05-18 14:20:23 -06:00
Behdad Esfahbod 63c6695108 [ot-layout] Cosmetic
The implementation of HBUINT16 operator == is slower than just
comparing to ints.
2022-05-18 13:53:52 -06:00
Behdad Esfahbod 777debd748 [subset] Rewrite CoverageFormat2::intersects as dagger 2022-05-18 13:46:06 -06:00
Behdad Esfahbod cf5001fac7 [subset] Optimize CoverageFormat2::intersected_coverage_glyphs 2022-05-18 13:38:29 -06:00
Behdad Esfahbod 6f37c20798 [subset] Minor rewrite in CoverageFormat2::serialize() 2022-05-18 13:25:42 -06:00
Behdad Esfahbod 175319cd89 [gsubgpos] Clean up OT::ClassDefFormat2::intersected_class_glyphs 0 case 2022-05-11 13:47:17 -06:00
Behdad Esfahbod 137af3612b [gsubgpos] Simplify OT::ClassDefFormat2::intersected_class_glyphs() 2022-05-11 13:39:30 -06:00
Behdad Esfahbod 3261e05bdb [subset] Optimize ClassDef1::intersected_class_glyphs() for class0 2022-05-11 13:16:31 -06:00
Behdad Esfahbod a792e16e48 [ot-layout] Change max nesting level of lookups from 6 to 64 2022-03-24 13:08:51 -06:00
Qunxin Liu 7cb002cb58 [subset] bug fix in prune_langsys
we should not cache visited langsys cause 2 different Record<Langsys>
could have different Tag while pointing to the same Langsys, a langsys
is redundant in Record<Langsys> A does not mean it's redundant in Record
B. Same thing for visited_script.
Also adding the number of features in the LangSys's feature list to the
visited langsys count so it's more accurate.
Plus some improvement in langsys compare()
2022-03-16 13:08:13 -06:00
Behdad Esfahbod de2cef42b4 [bimap] Remove init/fini 2022-01-18 09:21:31 -07:00
luz paz e2e305066a Fix various typos
Found via `codespell -q 3 -S ./perf/texts -L actualy,ba,beng,fo,gir,inout,nd,ot,pres,ro,te,teh,timne`
2022-01-16 05:39:03 -08:00
Garret Rieger 067f90a820 [subset] Fix for fuzzer timeout.
Fixes https://oss-fuzz.com/testcase-detail/5549945449480192

In prune_langsys: move LangSys visited check up before any work is done for a LangSys. In this particular case the compare() method is responsible for the majority of the time spent and wasn't being guarded with a visisted check.
2021-12-14 17:34:21 -07:00
Qunxin Liu 706014f69d [subset] (Chain)ContextSubst glyph_closure fix
- When pos_glyphs is empty, use current full glyphs set as input for
subsequent recursive closure process
- Also increase max_lookup_visit_count to 35000 cause a real font file hit
previous limit 20000 and some lookups are dropped unexpectedly
2021-12-10 11:08:36 -07:00
Behdad Esfahbod 394f772937 [map] Allow storing classes in the hashmap
Fixes https://github.com/harfbuzz/harfbuzz/issues/3293

The trick was to change the type of the invalid key/value to be non-class.
2021-11-19 11:49:23 -07:00
Garret Rieger a75b96f7e5 [sanitize] In lookup sanitize don't try to access subtable 0 if it isn't there. 2021-11-16 15:00:50 -08:00
Qunxin Liu 540f19b6fe [subset] fix bug in (Chain)ContextFormat2
Only keep rulesets for glyphs class numbers that survived in coverage
2021-11-02 16:05:55 -07:00
Behdad Esfahbod 6d555ce82e [meta] Use std::forward instead of hb_forward() 2021-11-02 00:18:22 -06:00
Qunxin Liu ca7b9daef0 [subset] subset MATH table 2021-09-22 18:08:30 -04:00
Garret Rieger c2cc566c9d [subset] Fix subset_offset_array adding unused space to serializer.
ArrayOf.serialize_append allocates space for the new item, but ArrayOf.pop() does not recover the allocated space. So in the case where the revert path was entered the extra space added by serialize_append gets left in the serialization buffer. This moves the snapshot to before ArrayOf.serialize_append is called so that revert cleans up the buffer extend.
2021-09-22 16:07:18 -06:00
Behdad Esfahbod 14a2df7bc2 Add stub HBUINT15 2021-09-19 23:06:09 -04:00
Behdad Esfahbod c852b86841 Rename HBGlyphID to HBGlyphID16 2021-09-19 16:30:12 -04:00
Garret Rieger b63ac57174 [subset] bail if collection region indices is in error. 2021-07-29 18:30:37 -07:00
Behdad Esfahbod 30579f5a37 [set] Document & use open-ended del_range() 2021-07-29 11:19:13 -06:00
Behdad Esfahbod 2d54c34680 [subset/VarStore] Fix OOM in fuzzer test
The test in question is the one added in c68a00b92e.
Culprit is that it's allocating lots of memory because of region_indices that
are out-of-range anyway. So, try to filter those out first.
2021-07-29 00:49:36 -06:00
Behdad Esfahbod ff1fe25a5d [VarStore] Rename internal get_scalars() to get_region_scalars() 2021-07-29 00:49:36 -06:00
Behdad Esfahbod f0a1892ff9 [serialize] Remove unnecessary pointer indirection 2021-07-28 17:36:22 -06:00
Behdad Esfahbod 1c511ed6f6 [subset] Simplify VarStore serialization 2021-07-28 17:33:26 -06:00