Commit Graph
17866 Commits
Author SHA1 Message Date
Behdad Esfahbod 69057e6ec0 [Ligature] Cache coverage the same way as PairPos1
Another 7% down on Roboto-Regular.
2025-02-05 19:50:51 +00:00
Behdad Esfahbod 3975e220ce Merge branch 'pairpos-cache' 2025-02-05 19:44:08 +00:00
Behdad Esfahbod e535c6835b [Context] Simplify cache costing 2025-02-05 19:43:43 +00:00
Behdad Esfahbod 7a746c3295 [Coverage] Fix cache cost reporting 2025-02-05 19:38:25 +00:00
Behdad Esfahbod 9a15e0e69a Merge pull request #5043 from harfbuzz/pairpos-cache
[OT] PairPos cache
2025-02-05 19:37:27 +00:00
Behdad Esfahbod 5be62a9f66 [PairPosFormat1] Use a coverage cache as well
Speeds up Arial shaping by 3%.
2025-02-05 19:30:01 +00:00
Behdad Esfahbod f22943a2a2 [PairPos] Cache coverage as well
Another 3% down in Roboto-Regular.
2025-02-05 19:18:25 +00:00
Behdad Esfahbod 726af2e4e5 [PairPos] Use a class cache
10% speedup on Roboto-Regular. 5% on SF Pro.
2025-02-05 19:18:25 +00:00
Behdad Esfahbod acd122d030 [gsubgpos] Prepare for per-lookup allocated caches 2025-02-05 19:18:25 +00:00
Behdad Esfahbod 3f40edbfa0 [ClassDef] Add ability to use a cache 2025-02-05 19:18:25 +00:00
Behdad Esfahbod 7392f32e47 [face] Tweak stack again
Let's go with one loop iteration. It's cheap.
2025-02-05 19:16:08 +00:00
Behdad Esfahbod 515b6a58d9 [face] Minor stack var size change 2025-02-05 19:13:19 +00:00
Behdad Esfahbod 601dd1a748 [face-builder] Fix an OOM issue
Fixes https://oss-fuzz.com/testcase-detail/5721476158521344
2025-02-05 19:10:41 +00:00
Behdad Esfahbod c05002afb8 [set] Add may_have() as alias for get() 2025-02-05 16:46:13 +00:00
Behdad Esfahbod ad7f428be4 [set-digest] Comment 2025-02-05 16:44:04 +00:00
Behdad Esfahbod 900db7b1bd [bit-set] Add a couple of alias methods 2025-02-05 16:41:49 +00:00
Behdad Esfahbod 942a30b7df [aat] Remove now-unused code 2025-02-05 14:21:49 +00:00
Behdad Esfahbod 44281b2856 [aat] Back to simplified heuristic
Performs overall better.
2025-02-05 14:19:01 +00:00
Behdad Esfahbod dc8da6129f Merge pull request #5042 from harfbuzz/aat-regression
[aat] Fix performance regression
2025-02-05 14:12:55 +00:00
Behdad Esfahbod b857b21a45 [morx] Try using buffer-glyph-set adaptive to number of chains 2025-02-05 13:50:09 +00:00
Behdad Esfahbod 491c8c4e10 Revert "[aat] Change threshold from 4 to 16"
This reverts commit 624d6edfe4.

Some other benchmarks see better results for 4. Let's leave at that.
2025-02-05 13:00:09 +00:00
Behdad Esfahbod 624d6edfe4 [aat] Change threshold from 4 to 16
Still worth it it seems.
2025-02-05 12:46:25 +00:00
Behdad Esfahbod a26cf0f409 [aat] Fix regression in performance of shaping tiny strings 2025-02-05 12:41:40 +00:00
Behdad Esfahbod cec5c57130 [aat] Refactor a method 2025-02-05 12:39:22 +00:00
Behdad Esfahbod cd5dfc6dbe [kerx] Use hb_bit_set_t instead of hb_set_t 2025-02-05 12:33:00 +00:00
Behdad Esfahbod d40c079fed [morx] Use hb_bit_set_t instead of hb_set_t directly 2025-02-05 12:30:56 +00:00
Behdad Esfahbod 7b9277486f [bit-set] Rename a method 2025-02-05 12:25:54 +00:00
Behdad Esfahbod b44f6ff5fe [perf/texts] Add a list of just 2letter strings 2025-02-05 12:19:48 +00:00
Behdad Esfahbod f738534042 Merge pull request #5041 from harfbuzz/aat-machine-initial
[morx] Only collect glyphs that can initiate action in the machine
2025-02-05 12:19:30 +00:00
Behdad Esfahbod fc825168d0 [aat] Remove idempotent check
Not necessary after the recent filtering.

Another 10% speedup with LucidaGrande.
2025-02-05 11:17:34 +00:00
Behdad Esfahbod 849b17351e [aat] Divide the logic of is_actionable() into two
The two different uses require different logic. I had combined
the two logic into one in 37bfdf79c6
2025-02-05 11:15:36 +00:00
Behdad Esfahbod b4787838e4 [aat] Speed up per-machine initial-glyph-set calculation
Use a bit-page instead of a full hb-set.
2025-02-05 10:52:26 +00:00
Behdad Esfahbod 7d14eb5eb3 [aat] Minor rename 2025-02-05 10:46:28 +00:00
Behdad Esfahbod 6215d68b69 [aat] Minor tweak to logic
DELETED_GLYPH can also initiate action.
2025-02-05 10:43:24 +00:00
Behdad Esfahbod dc4ff7012c [bit-page] Add a couple of operators 2025-02-05 10:42:33 +00:00
Behdad Esfahbod 37bfdf79c6 [morx] Only collect glyphs that can initiate action in the machine
And match them against the buffer glyph_set.

4% speedup in LucidaGrande.
2025-02-05 10:38:24 +00:00
Behdad Esfahbod d9e99baa9a [aat] Move is_actionable outside of the driver_context 2025-02-05 02:06:59 +00:00
Behdad Esfahbod 5ade5bab01 [aat] Move Flags out of driver_context_t 2025-02-05 01:53:09 +00:00
Behdad Esfahbod 9f38690e8e [aat] Remove another unused parameter
One logic change, but should be fine.
2025-02-05 01:39:28 +00:00
Behdad Esfahbod d9058c64e0 [aat] Remove an unused parameter 2025-02-05 01:38:06 +00:00
Behdad Esfahbod 1a2780735c [array] Micro-optimize 2025-02-05 00:35:40 +00:00
Behdad Esfahbod c87bfe410b [morx] Reuse a variable 2025-02-05 00:22:27 +00:00
Behdad Esfahbod e99e6538d9 [aat-map] Micro-optimize a malloc 2025-02-05 00:08:29 +00:00
Behdad Esfahbod 3e27038fca [aat-map] Micro-optimize 2025-02-04 23:39:42 +00:00
Behdad Esfahbod 7f5fac7c87 [test-buffer] Fix unintionally merged change 2025-02-04 23:35:34 +00:00
Behdad Esfahbod f0ead85283 [aat] Implement set filtering for Noncontextual chains as well 2025-02-04 23:30:30 +00:00
Behdad Esfahbod e5bdba0b3c [aat] Rename a variable 2025-02-04 23:03:23 +00:00
Behdad Esfahbod cb09fe7995 [bit-set] Remove a likely() 2025-02-04 22:54:10 +00:00
Behdad Esfahbod 749dc1eee0 [bit-set] Fix may_intersect() logic 2025-02-04 22:53:17 +00:00
Behdad Esfahbod b30c65e271 Merge pull request #5040 from harfbuzz/aat-set
[aat] Speed up subchain skipping using exact sets
2025-02-04 22:48:58 +00:00