lp:~reviczky/texlive/+git/harfbuzz

Author: BehdadEsfahbod
Author Date: 2023-11-15 21:55:11 UTC

Merge pull request #4473 from googlefonts/help

[instancer] add the option to leave blanks in the min:def:max syntax.

Author: BehdadEsfahbod
Author Date: 2023-11-15 21:00:52 UTC

[ChainRule] Speed up sanitize

Limit backtrack-len and lookahead-len to HB_MAX_CONTEXT_LENGTH.
Add a max_size for the struct, and bail early if max_size is
within sanitize range.

Speeds up Gulzar sanitization by over 20%.

For consistency, put the same limit on backtrack/lookahead of
ReverseChainContextSubstFormat1.

Author: BehdadEsfahbod
Author Date: 2023-11-11 02:10:51 UTC

[set] Speed up next_range()

Author: BehdadEsfahbod
Author Date: 2023-11-03 23:51:50 UTC

[sanitize] Yet another hack to "fix" barrier issue

Author: BehdadEsfahbod
Author Date: 2023-08-22 22:59:27 UTC

WIP [COLR] Add PaintLocation

Doesn't work...

Author: BehdadEsfahbod
Author Date: 2023-08-07 17:27:29 UTC

Revert "Fix bot"

This reverts commit 483d4a8741890ac72fa22d167ed53d4e39e9bd1d.

Author: BehdadEsfahbod
Author Date: 2023-08-06 02:10:07 UTC

[COLRv2] Implement PaintGlyphDelta

Author: BehdadEsfahbod
Author Date: 2023-08-04 22:53:00 UTC

Revert "[COLR] Make template arguments sharable"

This reverts commit 7e1c71cfb9ea6d63932af7d8b6af298be7fa6bcd.

Discarding idea. Didn't save any.

Author: BehdadEsfahbod
Author Date: 2023-07-27 21:47:43 UTC

[benchmark] Switch -font and -shape to monster-fonts submodule

Author: Khaled Hosny
Author Date: 2023-06-06 13:08:39 UTC

[shape] Add “ot” and “aat” shapers

They currently fail if relevant tables are missing, but don’t actually
enforce the relevant shaping model.

Author: BehdadEsfahbod
Author Date: 2023-06-02 02:52:36 UTC

[subset/glyf] Use a hashmap instead of vector for SubsetGlyph

Doens't seem to provide any speedup though.

Author: BehdadEsfahbod
Author Date: 2023-06-01 19:28:54 UTC

[map] Keep one byte of the hash in a separate array

Still no speedup. Just for the record.

Author: BehdadEsfahbod
Author Date: 2023-05-04 18:24:13 UTC

Merge branch 'main' into simd

Author: BehdadEsfahbod
Author Date: 2023-05-04 17:58:39 UTC

[layout] Use NEON SIMD instructions for apply_forward()

To skip four glyphs at a time.

Unfortunately slows down benchmark-shape by up to 25% for Amiri, instead of
speeding anything up. This is due to the overhead of switching back and forth
between the ARM and NEON processors.

To be experimented with: Same architecture but with AVX on x86_64 to skip 8
glyphs at a time.

https://github.com/harfbuzz/harfbuzz/issues/566

Author: BehdadEsfahbod
Author Date: 2022-06-15 22:08:28 UTC

[layout] Use AVX2 SIMD instructions for apply_forward()

To skip eight glyphs at a time.

Like the NEON instructions, this slows down shaping by up to 35%,
for Amiri.

https://github.com/harfbuzz/harfbuzz/issues/566

Author: BehdadEsfahbod
Author Date: 2023-03-09 21:27:25 UTC

[cairo] Add some GI annotations

Author: BehdadEsfahbod
Author Date: 2023-01-30 19:19:17 UTC

[coverage] Add a static cache

Speeds up Roboto by around 5%; costs 2kb globally.

Author: Matthias Clasen
Author Date: 2023-01-16 02:47:50 UTC

hb-shape: Print table version where relevant

When listing tables with --list-tables, show
the version for CPAL and COLR tables.

Author: Matthias Clasen
Author Date: 2023-01-02 22:48:06 UTC

Add hb_language_get_scripts

This is a copy of the Pango implementation, minus caching.

Tests included.

Author: Matthias Clasen
Author Date: 2022-12-29 03:24:45 UTC

Some tweaks

Author: Garret Rieger
Author Date: 2022-10-17 20:14:02 UTC

[subset] only preprocess in benchmark when experimental api is enabled.

Author: BehdadEsfahbod
Author Date: 2022-09-21 18:22:41 UTC

[gpos] in PairPos, always advance next-glyph by one only

This goes against the spec.

https://github.com/harfbuzz/harfbuzz/issues/3824

Author: Garret Rieger
Author Date: 2022-07-25 21:55:03 UTC

[repacker] for ext promotion choose lookups from highest subtables per byte to lowest.

Attempts to roughly maximize the number of subtables that are non-ext.

Author: BehdadEsfahbod
Author Date: 2022-06-08 12:17:27 UTC

[inplace] Adjust

Author: BehdadEsfahbod
Author Date: 2021-02-12 21:13:11 UTC

WIP

Author: BehdadEsfahbod
Author Date: 2018-01-10 12:59:36 UTC

[WIP][test] Speed up test runner

Keep one hb-shape process open and stream text to it.

Author: BehdadEsfahbod
Author Date: 2013-04-30 01:16:15 UTC

[OTLayout] Accelerate lookups by batching

If we need to apply many many lookups, we can fasten that up by applying
them in batches. For each batch we keep the union of the coverage of
the lookups participating. We can then skip glyph ranges that do NOT
participate in any lookup in the batch. The batch partition is
determined optimally by a mathematical probability model on the glyphs
and a dynamic-program to optimize the partition.

The net effect is 30% speedup on Amiri. the downside is more memory
consuption as each batch will keep an hb_set_t of its coverage.

I'm not yet convinced that the tradeoff is worth pursuing. I'm trying
to find out ways to optimized this more, with less memory overhead.

This work also ignores the number of subtables per lookup. That may
prove to be very important for the performance numbers from here on.