e630a65e60
[gvar] Micro-optize vector extend
2022-11-22 11:29:13 -07:00
49d4f62135
[gvar] Micro-optimize
2022-11-22 11:14:56 -07:00
1758ee6646
[glyf] Minor write loop more idiomatically
2022-11-22 10:55:16 -07:00
16ec9dcc1b
[gvar] Whitespace
2022-11-22 10:55:16 -07:00
b567ce51d3
[subset] Don't trim glyf's again if preprocessed
...
Speeds up M1/10000 benchmark by 30%!
2022-11-22 10:55:08 -07:00
72059a4789
[gvar] Optimize IUP alg
2022-11-22 10:41:37 -07:00
ee9873b5ed
[gvar] Disable initializing vectors when not necessary
2022-11-22 10:23:17 -07:00
b0d2641186
[vector] Add "initialize" argument to resize()
2022-11-22 10:20:11 -07:00
a2059f8f55
[gvar] Optimize unpack_points
2022-11-22 10:16:21 -07:00
6d7206b68b
[gvar] Optimize unpack_deltas
2022-11-22 10:13:14 -07:00
bca569ae53
[array] Speed up hash() for byte arrays
2022-11-21 23:19:42 -07:00
d7b492e3f5
Revert "[array] Remove hash specializations for bytes"
...
This reverts commit 213117317c
2022-11-21 23:08:51 -07:00
1572ba281a
[subset-cff] Return in subr closure if already seen subr
...
Not sure why this was not done before.
2022-11-21 22:26:44 -07:00
a29ca6efbc
[subset-cff] Comment
2022-11-21 22:02:17 -07:00
28e767ddea
[subset-cff] Really optimize op_str_t / parsed_cs_op_t layout
...
Now parsed_cs_op_t and op_str_t are both 16 bytes.
Saves another 7% in SourceHanSans/10000 benchmark.
2022-11-21 21:59:51 -07:00
2d5ee23731
[subset-cff] Readjust parsed_cs_op_t
...
Now it doesn't matter anymore since op_str_t is adjusted and
is 16 bytes with 8byte alignment.
2022-11-21 21:55:21 -07:00
4f056b923a
[subset-cff] Optimize op_str_t layout
2022-11-21 21:37:57 -07:00
a750cb0128
Simplify rvalue creation
2022-11-21 21:03:32 -07:00
86a763c651
[map] Make keys moveable
2022-11-21 20:53:44 -07:00
cf20d2ec5d
[map] Take const key
2022-11-21 20:47:17 -07:00
3d1c76f713
[serializer] Don't hash objects twice
2022-11-21 19:40:32 -07:00
35878df215
[algs] Implement swap() for pair_t
...
Helps priority_queue::pop_minimum and friends, which help subsetter
repacker. Shows a few percentage improvement on NotoNastaliq benchmark.
2022-11-21 19:14:03 -07:00
a2984a2932
[cff] Remove unnecessary namespacing
2022-11-21 18:40:52 -07:00
dc3bb5e0ed
[subset-cff] Pre-allocate values array for subroutines as well
2022-11-21 18:18:48 -07:00
c6279224db
[cff] Adjust pre-allocation
...
This better matches actual usage, given that ops are one or two
bytes, and vector also allocates 50% extra.
2022-11-21 18:01:50 -07:00
bab8ec58b0
[subset-cff] Disable sharing when packing charstring INDEXes
...
Saves another 8%ish.
2022-11-21 17:46:32 -07:00
2cadacad6c
[cff] Simplify str_encoder_t error handling
2022-11-21 17:17:15 -07:00
f263e3fe2e
[cff] Manually copy short strings instead of memcpy()
2022-11-21 17:04:55 -07:00
38efd1862f
[cff] Add a likely()
2022-11-21 17:02:11 -07:00
191025cc96
[cff] Adjust buffer pre-allocation
...
Most ops take one or two bytes, so allocate count*2, not count*3.
Shows minor speedup in subsetting benchmark (around 2%).
2022-11-21 16:58:19 -07:00
4b2caafea2
[subset-cff] Optimize parsed_cs_op_t size
...
Shows 5% speedup on SourceHanSans-Regular/10000 benchmark.
2022-11-21 16:46:20 -07:00
e0b06bd1b1
[subset] Cache has_seac in accelerator
...
Speeds up SourceHanSans-Regular/10000 benchmark by %25.
2022-11-21 16:30:34 -07:00
dd1ba328a8
[repacker] fix fuzzer timeout.
...
For https://oss-fuzz.com/testcase-detail/5845846876356608 . Only process the set of unique overflows.
2022-11-21 16:24:48 -07:00
59451502e9
[cff] Optimize env error checking
2022-11-21 15:23:16 -07:00
b238578a9c
[cff] Optimize INDEX operator[]
2022-11-21 14:36:57 -07:00
d9de515a38
[cff] Optimize byte_str_ref_t array access
2022-11-21 14:23:07 -07:00
a81ec9b2b6
[cff] Optimize byte_str_ref_t inc()
...
Shows a couple percent speedup.
2022-11-21 14:03:28 -07:00
3ff75411bd
[algs] Fix bot fail
2022-11-21 13:08:55 -07:00
b81e3989f8
Try fixing arm-eabi build after a10cfe3f32
2022-11-21 13:02:40 -07:00
4c14043b06
[algs] Add output argument to hb_unsigned_mul_overflows()
2022-11-21 12:56:33 -07:00
25adbb3825
[algs] Use __builtin_mul_overflow
...
Compiles to smaller binary.
2022-11-21 12:47:53 -07:00
a10cfe3f32
[algs] Write hb_in_ranges() recursively
2022-11-21 12:37:59 -07:00
2e86700e30
[gvar] Add memory-barrier as in ArrayOf
2022-11-21 12:28:10 -07:00
b00a911fa7
[sorted-array] Add faster iterator implementation here as well
2022-11-21 12:11:30 -07:00
7cc79a8a86
[vector] Adjust comment
2022-11-21 12:09:24 -07:00
e82a3c69dd
[array/vector] Optimize range-based for-loop
...
Avoid bounds-checking.
2022-11-21 12:00:10 -07:00
0387182c2a
[ot-map] Minor refactor features[i] access
2022-11-21 11:05:47 -07:00
5ee6d5d77e
[cff] Add memory-barrier to INDEX
...
Like we do for ArrayOf.
2022-11-21 11:00:29 -07:00
6905d36d73
[cff] Fix fetch_op() bounds-checking
2022-11-21 10:51:33 -07:00
b51ab1a9e5
[cff] Optimize byte_str_ref_t
...
Make it 16 bytes instead of 24. This struct is used in the subroutine
call stack heavily.
This change makes the HB AdobeVFPrototype benchmark to become faster
than FT one, with about 6% speedup as a result of this change.
2022-11-21 10:27:07 -07:00
Behdad Esfahbod
Garret Rieger