Nigel Tao
f5c68a13cb
Fix std/png for depth-16 grayscale interlaced
2021-11-20 22:54:21 +11:00
Nigel Tao
5e8b448c12
Drop std/deflate's "Re-purpose the hdist variable"
2021-11-20 22:03:29 +11:00
Nigel Tao
49a431f98d
Fix std/deflate decode_huffman_slow suspends
...
Unlike the other decode_huffman_etc! methods, decode_huffman_slow? (with
a ?) can suspend. The (hlen, hdist) and the remainder (length, distance)
pairs might not be stable across a suspension, as the args.dst io_writer
might change its history length. We have to re-calculate those pairs
after a short write.
The test/3pdata/nia-checksums-of-pngsuite.txt for interlaced and
non-interlaced flavors of the same image now match.
2021-11-20 21:31:40 +11:00
Nigel Tao
ede0590ead
Add while label to std/deflate decode_huffman_slow
2021-11-20 20:14:37 +11:00
Nigel Tao
bb8623b472
Have deflate slow-Huffman use extended ringbuffer
...
This is the same optimization as commit e65c8612 "Extend std/deflate
ringbuffer to simplify copies", applied to decoder.decode_huffman_slow.
2021-11-20 20:10:10 +11:00
Nigel Tao
a684928240
Let std/png allow PLTE for RGB, RGBA color types
2021-11-20 15:39:07 +11:00
Nigel Tao
e44920d382
Let gif "ignore too much" quirk skip lzw errors
...
The modified test/data/*/*.gif files differ by only one byte:
00000000 47 49 46 38 39 61 02 00 02 00 81 00 00 00 00 ff |GIF89a..........|
00000010 11 00 ff 22 00 ff 33 00 ff 2c 00 00 00 00 02 00 |..."..3..,......|
-00000020 02 00 00 02 02 84 f1 00 3b |........;|
+00000020 02 00 00 02 02 84 5d 00 3b |......].;|
The etc-good-lzw.gif bits have changed, even though the
etc-good-lzw.etc.txt file only had comments added, due to a Go standard
library change: "compress/lzw: output a Clear code first, per GIF spec"
https://github.com/golang/go/commit/9c1dbdf60edbffeff10f58af21fa055eb0fdd29f
Updates PR #57
Updates https://crbug.com/1270631
2021-11-19 13:30:23 +11:00
Nigel Tao
29560a336b
Convert std/png text from Latin-1 to UTF-8
...
Fixes #55
2021-11-17 23:04:49 +11:00
Nigel Tao
e0cdaeb54f
Fix std/png tell_me_more minfo with suspensions
2021-11-17 13:04:04 +11:00
Nigel Tao
e7ed1dd43e
Add std/png tell_me_more while.loop name
2021-11-17 13:01:05 +11:00
Nigel Tao
417b03a0ad
Add std/png "#unsupported PNG compression method"
2021-11-17 12:07:44 +11:00
Nigel Tao
7dd0dd264b
Delete superfluous "// --------" comment
2021-11-17 12:03:43 +11:00
Nigel Tao
0d37acd6f6
Factor out std/png/common_consts.wuffs
2021-11-17 12:01:37 +11:00
Nigel Tao
6737565c20
Let std/png decode iTXt, tEXt and zTXt chunks
...
Updates #55
2021-11-17 11:56:49 +11:00
Nigel Tao
f0c2a40d66
Resize std/png chunk_length field from u64 to u32
2021-11-16 13:06:30 +11:00
Nigel Tao
ce9ab13ecd
Let std/png decode first fcTL after the last IDAT
2021-11-16 12:55:43 +11:00
Nigel Tao
0045ce49f3
Rename table.row as table.row_u32
2021-11-15 10:49:42 +11:00
Nigel Tao
fa56d92410
Clip std/png frame_dirty_rect to frame rect
2021-11-15 10:46:28 +11:00
Nigel Tao
8f1adea523
Hoist subslice out of std/png filter_and_swizzle
...
On an Intel NUC (Comet Lake, BXNUC10i5FNKPA):
name old speed new speed delta
wuffs_png_decode_image_19k_8bpp/clang9 274MB/s ± 0% 273MB/s ± 0% -0.25% (p=0.008 n=5+5)
wuffs_png_decode_image_40k_24bpp/clang9 309MB/s ± 0% 308MB/s ± 0% -0.22% (p=0.008 n=5+5)
wuffs_png_decode_image_77k_8bpp/clang9 985MB/s ± 0% 982MB/s ± 0% -0.29% (p=0.008 n=5+5)
wuffs_png_decode_image_552k_32bpp_ignore_checksum/clang9 846MB/s ± 0% 841MB/s ± 0% -0.56% (p=0.008 n=5+5)
wuffs_png_decode_image_552k_32bpp_verify_checksum/clang9 812MB/s ± 0% 808MB/s ± 0% -0.55% (p=0.008 n=5+5)
wuffs_png_decode_image_4002k_24bpp/clang9 320MB/s ± 0% 320MB/s ± 0% -0.15% (p=0.008 n=5+5)
wuffs_png_decode_image_19k_8bpp/gcc10 263MB/s ± 0% 276MB/s ± 0% +5.00% (p=0.008 n=5+5)
wuffs_png_decode_image_40k_24bpp/gcc10 328MB/s ± 0% 333MB/s ± 0% +1.49% (p=0.016 n=5+4)
wuffs_png_decode_image_77k_8bpp/gcc10 950MB/s ± 0% 976MB/s ± 0% +2.78% (p=0.008 n=5+5)
wuffs_png_decode_image_552k_32bpp_ignore_checksum/gcc10 865MB/s ± 0% 912MB/s ± 0% +5.45% (p=0.008 n=5+5)
wuffs_png_decode_image_552k_32bpp_verify_checksum/gcc10 825MB/s ± 0% 868MB/s ± 0% +5.19% (p=0.008 n=5+5)
wuffs_png_decode_image_4002k_24bpp/gcc10 336MB/s ± 0% 341MB/s ± 0% +1.37% (p=0.008 n=5+5)
2021-11-15 10:43:55 +11:00
Nigel Tao
4d64aff554
Rename std/png/decode_swizzle_fallback.wuffs
2021-11-14 19:58:23 +11:00
Nigel Tao
a4fa6a2554
Let std/png decode (animated) APNG
...
The test/nia-checksums-of-data.txt checksums for the *.apng files now
match the (multiple frame) *.gif files. Previously, they matched the
(single frame) *.000000.nie files.
Fixes #41
2021-11-14 18:20:15 +11:00
Nigel Tao
9af2666aec
Move std/png first_frame_io_position earlier
2021-11-10 16:52:59 +11:00
Nigel Tao
9c3d27fb43
Fix std/png decode_other_chunk skipping chunks
2021-11-09 22:10:00 +11:00
Nigel Tao
2373340f83
Add base.ANIMATION_DISPOSAL__ETC constants
2021-11-09 09:53:29 +11:00
Nigel Tao
e847766b02
Tweak std/png decode_other_chunk if-else chain
2021-11-09 09:05:11 +11:00
Nigel Tao
74c1dfce96
Rename std/gif num_loops field
2021-11-09 08:55:15 +11:00
Nigel Tao
ca76a01ac2
Have std/png support remapped transparency
2021-11-07 09:51:04 +11:00
Nigel Tao
00b329872a
Tighten png.INTERLACING type refinement
2021-11-06 09:54:56 +11:00
Nigel Tao
744fbdd732
Have std/png decode iCCP chunks
...
Updates #39
2021-11-05 16:29:11 +11:00
Nigel Tao
8f9a4f40c2
Rename METADATA_RAW as METADATA_RAW_PASSTHROUGH
2021-10-31 15:17:46 +11:00
Nigel Tao
db5a407bab
Have std/png decode sRGB chunks
...
Updates #39
2021-10-16 09:27:30 +11:00
Nigel Tao
911049e9f9
Have std/png decode cHRM and gAMA chunks
...
Updates #39
2021-10-15 15:08:47 +11:00
Nigel Tao
d8ad363ef5
Tweak png.decoder.decode_image_config indentation
2021-10-15 13:35:15 +11:00
Nigel Tao
ad1d983698
Add ETC__MORE_INFORMATION__FLAVOR__METADATA_PARSED
2021-10-14 17:17:11 +11:00
Nigel Tao
f9c029c788
Fix std/png history_position for interlacing
2021-10-07 13:52:40 +11:00
Nigel Tao
592bc14ed1
Have std/deflate allow non-compact dst history
...
Fixes #52
On a mid-range x86_64 laptop (2016, Skylake):
name old speed new speed delta
wuffs_deflate_decode_1k_full_init/clang9 101MB/s ± 1% 100MB/s ± 1% -1.22% (p=0.000 n=9+9)
wuffs_deflate_decode_1k_part_init/clang9 119MB/s ± 1% 119MB/s ± 1% ~ (p=0.139 n=8+9)
wuffs_deflate_decode_10k_full_init/clang9 221MB/s ± 1% 210MB/s ± 1% -4.94% (p=0.000 n=9+9)
wuffs_deflate_decode_10k_part_init/clang9 225MB/s ± 3% 215MB/s ± 1% -4.33% (p=0.000 n=10+8)
wuffs_deflate_decode_100k_just_one_read/clang9 290MB/s ± 0% 284MB/s ± 1% -2.26% (p=0.000 n=8+9)
wuffs_deflate_decode_100k_many_big_reads/clang9 188MB/s ± 3% 177MB/s ± 1% -5.78% (p=0.000 n=9+10)
wuffs_deflate_decode_1k_full_init/gcc10 101MB/s ± 0% 97MB/s ± 1% -3.65% (p=0.000 n=8+9)
wuffs_deflate_decode_1k_part_init/gcc10 119MB/s ± 1% 113MB/s ± 2% -4.65% (p=0.000 n=9+9)
wuffs_deflate_decode_10k_full_init/gcc10 214MB/s ± 3% 224MB/s ± 1% +4.83% (p=0.000 n=10+10)
wuffs_deflate_decode_10k_part_init/gcc10 222MB/s ± 1% 231MB/s ± 1% +4.03% (p=0.000 n=9+10)
wuffs_deflate_decode_100k_just_one_read/gcc10 281MB/s ± 2% 297MB/s ± 0% +5.78% (p=0.000 n=9+9)
wuffs_deflate_decode_100k_many_big_reads/gcc10 184MB/s ± 0% 194MB/s ± 0% +5.85% (p=0.000 n=10+8)
mimic_deflate_decode_1k_full_init 138MB/s ± 1% 138MB/s ± 0% ~ (p=0.546 n=9+9)
mimic_deflate_decode_10k_full_init 156MB/s ± 2% 156MB/s ± 0% ~ (p=0.387 n=9+9)
mimic_deflate_decode_100k_just_one_read 191MB/s ± 1% 190MB/s ± 2% -0.94% (p=0.004 n=9+10)
mimic_deflate_decode_100k_many_big_reads 147MB/s ± 1% 147MB/s ± 1% ~ (p=0.356 n=9+10)
wuffs_gzip_decode_10k/clang9 220MB/s ± 1% 209MB/s ± 1% -5.29% (p=0.000 n=10+9)
wuffs_gzip_decode_100k/clang9 283MB/s ± 2% 279MB/s ± 1% -1.47% (p=0.002 n=10+10)
wuffs_gzip_decode_10k/gcc10 218MB/s ± 1% 214MB/s ± 2% -1.76% (p=0.000 n=9+10)
wuffs_gzip_decode_100k/gcc10 274MB/s ± 0% 264MB/s ± 2% -3.61% (p=0.000 n=7+10)
mimic_gzip_decode_10k 131MB/s ± 2% 132MB/s ± 0% +0.81% (p=0.000 n=10+8)
mimic_gzip_decode_100k 156MB/s ± 0% 156MB/s ± 1% ~ (p=0.423 n=8+9)
wuffs_zlib_decode_10k/clang9 223MB/s ± 2% 209MB/s ± 2% -6.41% (p=0.000 n=10+10)
wuffs_zlib_decode_100k/clang9 285MB/s ± 0% 280MB/s ± 0% -1.87% (p=0.000 n=8+9)
wuffs_zlib_decode_10k/gcc10 235MB/s ± 0% 215MB/s ± 1% -8.22% (p=0.000 n=10+9)
wuffs_zlib_decode_100k/gcc10 293MB/s ± 0% 268MB/s ± 1% -8.38% (p=0.000 n=8+10)
mimic_zlib_decode_10k 145MB/s ± 1% 145MB/s ± 1% -0.45% (p=0.004 n=9+9)
mimic_zlib_decode_100k 175MB/s ± 1% 174MB/s ± 0% ~ (p=0.055 n=10+8)
wuffs_png_decode_image_19k_8bpp/clang9 142MB/s ± 1% 152MB/s ± 1% +6.98% (p=0.000 n=9+9)
wuffs_png_decode_image_40k_24bpp/clang9 172MB/s ± 0% 170MB/s ± 0% -1.63% (p=0.000 n=8+9)
wuffs_png_decode_image_77k_8bpp/clang9 504MB/s ± 2% 540MB/s ± 1% +7.28% (p=0.000 n=9+10)
wuffs_png_decode_image_552k_32bpp_ignore_checksum/clang9 462MB/s ± 1% 470MB/s ± 0% +1.57% (p=0.000 n=9+10)
wuffs_png_decode_image_552k_32bpp_verify_checksum/clang9 445MB/s ± 1% 451MB/s ± 0% +1.49% (p=0.000 n=8+10)
wuffs_png_decode_image_4002k_24bpp/clang9 175MB/s ± 3% 175MB/s ± 1% ~ (p=0.356 n=9+10)
wuffs_png_decode_image_19k_8bpp/gcc10 150MB/s ± 0% 153MB/s ± 1% +2.03% (p=0.000 n=10+10)
wuffs_png_decode_image_40k_24bpp/gcc10 186MB/s ± 0% 186MB/s ± 0% ~ (p=0.408 n=10+8)
wuffs_png_decode_image_77k_8bpp/gcc10 543MB/s ± 0% 542MB/s ± 0% -0.23% (p=0.019 n=9+9)
wuffs_png_decode_image_552k_32bpp_ignore_checksum/gcc10 478MB/s ± 1% 498MB/s ± 0% +4.12% (p=0.000 n=10+9)
wuffs_png_decode_image_552k_32bpp_verify_checksum/gcc10 451MB/s ± 4% 476MB/s ± 0% +5.37% (p=0.000 n=10+9)
wuffs_png_decode_image_4002k_24bpp/gcc10 185MB/s ± 2% 190MB/s ± 1% +2.78% (p=0.000 n=10+9)
mimic_png_decode_image_19k_8bpp 63.0MB/s ± 0% 62.9MB/s ± 1% ~ (p=0.673 n=8+9)
mimic_png_decode_image_40k_24bpp 78.7MB/s ± 0% 78.9MB/s ± 0% ~ (p=0.059 n=9+8)
mimic_png_decode_image_77k_8bpp 192MB/s ± 1% 193MB/s ± 3% +0.44% (p=0.028 n=10+9)
mimic_png_decode_image_552k_32bpp_verify_checksum 157MB/s ± 0% 158MB/s ± 1% ~ (p=0.053 n=10+9)
mimic_png_decode_image_4002k_24bpp 113MB/s ± 1% 113MB/s ± 1% ~ (p=0.053 n=10+9)
2021-08-15 10:50:41 +10:00
Nigel Tao
28a54baa19
Add io_bind history_position arg
2021-08-15 08:36:14 +10:00
Nigel Tao
77f855206d
Have JSON line-comment tokens exclude the "\n"
...
Before this commit, line-comment token chains included the final "\n"
the way block-comments included the final "*/". The "\n" was mandatory.
However, when parsing with both QUIRK_ALLOW_COMMENT_LINE and
QUIRK_ALLOW_TRAILING_FILLER, input like "123//xyz" (without a "\n"
before the EOF) was rejected before and accepted after this commit.
So that line-comments are consistent, in that they either always or
never include the final "\n", they now never include the final "\n".
An input fragment (parsed with QUIRK_ALLOW_COMMENT_LINE) like
"//abc\n\t\t" was, before this commit, parsed as a 6 byte comment token
and a 2 byte other-filler token. That split is now 5+3.
2021-07-20 14:17:09 +10:00
Nigel Tao
3358730f6f
Have std/png reject zero width or height
2021-05-11 12:34:19 +10:00
Nigel Tao
e77d73e3ce
Add std/adler32 bench comment for the roll back
2021-04-28 22:38:10 +10:00
Nigel Tao
0a6cce5e0e
Tweak some adler32 hasher.up_x86_sse42 var names
2021-04-28 00:32:32 +10:00
Nigel Tao
abd37fa3f8
Roll back 'Add std/adler32 hasher.up_x86_avx2'
...
See the previous commit for the rationale.
2021-04-28 00:26:54 +10:00
Nigel Tao
baec831fc8
Add std/adler32 hasher.up_x86_avx2
...
While the std/adler32 benchmarks got better, the std/zlib and std/png
benchmarks got worse.
This commit will soon be followed by a rollback, but it is committed
anyway so that we can refer to these numbers in the git log.
Binary size, before:
4056 gen/lib/c/clang-9-dynamic/wuffs-std-adler32.lo
4104 gen/lib/c/clang-9-static/wuffs-std-adler32.o
4624 gen/lib/c/gcc-dynamic/wuffs-std-adler32.lo
4680 gen/lib/c/gcc-static/wuffs-std-adler32.o
After:
4952 gen/lib/c/clang-9-dynamic/wuffs-std-adler32.lo
5016 gen/lib/c/clang-9-static/wuffs-std-adler32.o
5512 gen/lib/c/gcc-dynamic/wuffs-std-adler32.lo
5576 gen/lib/c/gcc-static/wuffs-std-adler32.o
name old speed new speed delta
wuffs_adler32_10k/clang9 11.5GB/s ± 1% 15.0GB/s ± 5% +30.73% (p=0.000 n=9+10)
wuffs_adler32_100k/clang9 13.5GB/s ± 1% 15.6GB/s ±14% +16.06% (p=0.001 n=9+10)
wuffs_adler32_10k/gcc10 11.3GB/s ± 2% 15.6GB/s ± 1% +38.27% (p=0.000 n=9+9)
wuffs_adler32_100k/gcc10 11.3GB/s ± 1% 19.4GB/s ± 1% +71.20% (p=0.000 n=7+9)
wuffs_zlib_decode_10k/clang9 223MB/s ± 1% 216MB/s ± 2% -3.45% (p=0.000 n=10+10)
wuffs_zlib_decode_100k/clang9 285MB/s ± 1% 284MB/s ± 1% ~ (p=0.243 n=9+10)
wuffs_zlib_decode_10k/gcc10 232MB/s ± 1% 216MB/s ± 1% -6.70% (p=0.000 n=10+9)
wuffs_zlib_decode_100k/gcc10 291MB/s ± 1% 280MB/s ± 1% -3.70% (p=0.000 n=9+8)
wuffs_png_decode_image_19k_8bpp/clang9 152MB/s ± 1% 143MB/s ± 3% -5.89% (p=0.000 n=8+9)
wuffs_png_decode_image_40k_24bpp/clang9 171MB/s ± 1% 173MB/s ± 1% ~ (p=0.094 n=9+9)
wuffs_png_decode_image_77k_8bpp/clang9 541MB/s ± 1% 509MB/s ± 1% -5.77% (p=0.000 n=8+9)
wuffs_png_decode_image_552k_32bpp_ignore_checksum/clang9 479MB/s ± 1% 464MB/s ± 3% -3.20% (p=0.000 n=8+10)
wuffs_png_decode_image_552k_32bpp_verify_checksum/clang9 461MB/s ± 1% 452MB/s ± 1% -2.02% (p=0.000 n=8+10)
wuffs_png_decode_image_4002k_24bpp/clang9 177MB/s ± 1% 175MB/s ± 2% -1.14% (p=0.004 n=10+10)
wuffs_png_decode_image_19k_8bpp/gcc10 159MB/s ± 1% 153MB/s ± 1% -3.54% (p=0.000 n=8+10)
wuffs_png_decode_image_40k_24bpp/gcc10 188MB/s ± 1% 187MB/s ± 1% -0.47% (p=0.021 n=10+9)
wuffs_png_decode_image_77k_8bpp/gcc10 557MB/s ± 0% 546MB/s ± 4% -1.94% (p=0.000 n=8+10)
wuffs_png_decode_image_552k_32bpp_ignore_checksum/gcc10 510MB/s ± 0% 493MB/s ± 0% -3.33% (p=0.000 n=10+9)
wuffs_png_decode_image_552k_32bpp_verify_checksum/gcc10 486MB/s ± 1% 478MB/s ± 0% -1.77% (p=0.000 n=8+9)
wuffs_png_decode_image_4002k_24bpp/gcc10 189MB/s ± 1% 187MB/s ± 1% -1.26% (p=0.000 n=10+10)
2021-04-28 00:18:20 +10:00
Nigel Tao
5764f77913
Have std/png produce PIXEL_FORMAT__Y_16LE
...
Updates #46
2021-04-16 14:34:48 +10:00
Nigel Tao
2dddd388bc
Add crc32.ieee_hasher.up_x86_avx2
...
name old speed new speed delta
wuffs_crc32_ieee_10k/clang9 8.42GB/s ± 1% 8.12GB/s ± 1% -3.55% (p=0.000 n=9+8)
wuffs_crc32_ieee_100k/clang9 11.0GB/s ± 1% 10.9GB/s ± 1% ~ (p=0.931 n=9+9)
wuffs_crc32_ieee_10k/gcc10 9.20GB/s ± 0% 9.71GB/s ± 0% +5.54% (p=0.000 n=8+10)
wuffs_crc32_ieee_100k/gcc10 11.9GB/s ± 2% 13.1GB/s ± 1% +10.22% (p=0.000 n=10+8)
wuffs_gzip_decode_10k/clang9 219MB/s ± 1% 218MB/s ± 2% ~ (p=0.497 n=9+10)
wuffs_gzip_decode_100k/clang9 283MB/s ± 2% 281MB/s ± 1% ~ (p=0.063 n=10+10)
wuffs_gzip_decode_10k/gcc10 224MB/s ± 1% 223MB/s ± 1% ~ (p=0.489 n=9+9)
wuffs_gzip_decode_100k/gcc10 285MB/s ± 4% 290MB/s ± 0% +1.68% (p=0.004 n=10+8)
wuffs_png_decode_image_19k_8bpp/clang9 138MB/s ± 2% 138MB/s ± 1% ~ (p=0.156 n=10+9)
wuffs_png_decode_image_40k_24bpp/clang9 171MB/s ± 2% 172MB/s ± 1% ~ (p=0.971 n=10+10)
wuffs_png_decode_image_77k_8bpp/clang9 495MB/s ± 4% 504MB/s ± 1% +1.79% (p=0.000 n=9+9)
wuffs_png_decode_image_552k_32bpp_ignore_checksum/clang9 457MB/s ± 4% 463MB/s ± 0% +1.35% (p=0.002 n=9+9)
wuffs_png_decode_image_552k_32bpp_verify_checksum/clang9 435MB/s ± 5% 445MB/s ± 1% +2.31% (p=0.000 n=10+9)
wuffs_png_decode_image_4002k_24bpp/clang9 172MB/s ± 2% 174MB/s ± 1% ~ (p=0.053 n=10+9)
wuffs_png_decode_image_19k_8bpp/gcc10 157MB/s ± 1% 157MB/s ± 1% ~ (p=0.549 n=9+10)
wuffs_png_decode_image_40k_24bpp/gcc10 187MB/s ± 1% 186MB/s ± 1% ~ (p=0.182 n=10+9)
wuffs_png_decode_image_77k_8bpp/gcc10 554MB/s ± 1% 559MB/s ± 0% +0.84% (p=0.000 n=9+9)
wuffs_png_decode_image_552k_32bpp_ignore_checksum/gcc10 502MB/s ± 1% 500MB/s ± 0% -0.38% (p=0.011 n=10+10)
wuffs_png_decode_image_552k_32bpp_verify_checksum/gcc10 481MB/s ± 1% 478MB/s ± 0% -0.46% (p=0.002 n=8+9)
wuffs_png_decode_image_4002k_24bpp/gcc10 189MB/s ± 2% 190MB/s ± 0% +0.62% (p=0.008 n=10+9)
2021-04-09 23:12:32 +10:00
Nigel Tao
1dd7ad4413
Optimize std/deflate decoding for RLE sections
...
name old speed new speed delta
wuffs_deflate_decode_1k_full_init/clang9 104MB/s ± 1% 104MB/s ± 1% ~ (p=0.161 n=9+9)
wuffs_deflate_decode_1k_part_init/clang9 121MB/s ± 1% 119MB/s ± 1% -0.92% (p=0.000 n=9+9)
wuffs_deflate_decode_10k_full_init/clang9 223MB/s ± 1% 223MB/s ± 1% ~ (p=0.796 n=9+9)
wuffs_deflate_decode_10k_part_init/clang9 230MB/s ± 1% 228MB/s ± 2% ~ (p=0.065 n=10+9)
wuffs_deflate_decode_100k_just_one_read/clang9 279MB/s ± 1% 292MB/s ± 0% +4.78% (p=0.000 n=9+8)
wuffs_deflate_decode_100k_many_big_reads/clang9 178MB/s ± 5% 189MB/s ± 2% +6.41% (p=0.000 n=10+10)
wuffs_deflate_decode_1k_full_init/gcc10 103MB/s ± 1% 102MB/s ± 1% -0.57% (p=0.022 n=10+9)
wuffs_deflate_decode_1k_part_init/gcc10 122MB/s ± 1% 121MB/s ± 3% -1.46% (p=0.000 n=10+9)
wuffs_deflate_decode_10k_full_init/gcc10 235MB/s ± 1% 231MB/s ± 1% -1.34% (p=0.000 n=10+8)
wuffs_deflate_decode_10k_part_init/gcc10 242MB/s ± 1% 239MB/s ± 0% -1.40% (p=0.001 n=9+8)
wuffs_deflate_decode_100k_just_one_read/gcc10 300MB/s ± 4% 297MB/s ± 2% ~ (p=0.156 n=10+9)
wuffs_deflate_decode_100k_many_big_reads/gcc10 193MB/s ± 2% 197MB/s ± 1% +2.29% (p=0.000 n=10+9)
wuffs_gzip_decode_10k/clang9 220MB/s ± 1% 219MB/s ± 1% ~ (p=0.605 n=9+9)
wuffs_gzip_decode_100k/clang9 269MB/s ± 2% 283MB/s ± 2% +4.98% (p=0.000 n=9+10)
wuffs_gzip_decode_10k/gcc10 227MB/s ± 1% 224MB/s ± 1% -1.42% (p=0.000 n=10+9)
wuffs_gzip_decode_100k/gcc10 283MB/s ± 1% 285MB/s ± 4% ~ (p=0.182 n=9+10)
wuffs_zlib_decode_10k/clang9 226MB/s ± 1% 220MB/s ± 4% -2.66% (p=0.000 n=9+10)
wuffs_zlib_decode_100k/clang9 272MB/s ± 2% 285MB/s ± 1% +4.77% (p=0.000 n=10+8)
wuffs_zlib_decode_10k/gcc10 227MB/s ± 1% 212MB/s ± 1% -6.80% (p=0.000 n=10+9)
wuffs_zlib_decode_100k/gcc10 282MB/s ± 1% 273MB/s ± 0% -3.32% (p=0.000 n=9+8)
wuffs_png_decode_image_19k_8bpp/clang9 148MB/s ± 1% 138MB/s ± 2% -6.79% (p=0.000 n=9+10)
wuffs_png_decode_image_40k_24bpp/clang9 170MB/s ± 1% 171MB/s ± 2% ~ (p=0.133 n=9+10)
wuffs_png_decode_image_77k_8bpp/clang9 525MB/s ± 1% 495MB/s ± 4% -5.81% (p=0.000 n=8+9)
wuffs_png_decode_image_552k_32bpp_ignore_checksum/clang9 400MB/s ± 2% 457MB/s ± 4% +14.16% (p=0.000 n=9+9)
wuffs_png_decode_image_552k_32bpp_verify_checksum/clang9 388MB/s ± 1% 435MB/s ± 5% +12.17% (p=0.000 n=9+10)
wuffs_png_decode_image_4002k_24bpp/clang9 175MB/s ± 1% 172MB/s ± 2% -1.78% (p=0.005 n=10+10)
wuffs_png_decode_image_19k_8bpp/gcc10 154MB/s ± 3% 157MB/s ± 1% +1.72% (p=0.001 n=9+9)
wuffs_png_decode_image_40k_24bpp/gcc10 182MB/s ± 1% 187MB/s ± 1% +2.56% (p=0.000 n=9+10)
wuffs_png_decode_image_77k_8bpp/gcc10 552MB/s ± 3% 554MB/s ± 1% ~ (p=0.863 n=9+9)
wuffs_png_decode_image_552k_32bpp_ignore_checksum/gcc10 426MB/s ± 2% 502MB/s ± 1% +17.85% (p=0.000 n=10+10)
wuffs_png_decode_image_552k_32bpp_verify_checksum/gcc10 410MB/s ± 2% 481MB/s ± 1% +17.17% (p=0.000 n=9+8)
wuffs_png_decode_image_4002k_24bpp/gcc10 186MB/s ± 1% 189MB/s ± 2% +1.33% (p=0.001 n=9+10)
2021-04-09 14:31:52 +10:00
Nigel Tao
a7ef82f7db
Add deflate.decoder.decode_huffman_bmi2
...
Binary size, before:
21864 gen/lib/c/clang-9-dynamic/wuffs-std-deflate.lo
21488 gen/lib/c/clang-9-static/wuffs-std-deflate.o
21136 gen/lib/c/gcc-dynamic/wuffs-std-deflate.lo
21536 gen/lib/c/gcc-static/wuffs-std-deflate.o
After:
24592 gen/lib/c/clang-9-dynamic/wuffs-std-deflate.lo
24368 gen/lib/c/clang-9-static/wuffs-std-deflate.o
23320 gen/lib/c/gcc-dynamic/wuffs-std-deflate.lo
23928 gen/lib/c/gcc-static/wuffs-std-deflate.o
I've upgraded my kernel and compilers recently, so the absolute Wuffs
benchmark numbers below aren't directly comparable to those reported in
previous commits, but the relative changes still seems valid.
On a mid-range x86_64 laptop:
name old speed new speed delta
wuffs_deflate_decode_1k_full_init/clang9 107MB/s ± 1% 104MB/s ± 1% -2.36% (p=0.000 n=8+9)
wuffs_deflate_decode_1k_part_init/clang9 123MB/s ± 1% 121MB/s ± 1% -1.78% (p=0.000 n=10+9)
wuffs_deflate_decode_10k_full_init/clang9 221MB/s ± 2% 223MB/s ± 1% +0.86% (p=0.035 n=10+9)
wuffs_deflate_decode_10k_part_init/clang9 228MB/s ± 0% 230MB/s ± 1% +0.59% (p=0.003 n=8+10)
wuffs_deflate_decode_100k_just_one_read/clang9 276MB/s ± 0% 279MB/s ± 1% +1.12% (p=0.001 n=8+9)
wuffs_deflate_decode_100k_many_big_reads/clang9 174MB/s ± 1% 178MB/s ± 5% +2.38% (p=0.028 n=9+10)
wuffs_deflate_decode_1k_full_init/gcc10 103MB/s ± 3% 103MB/s ± 1% ~ (p=0.105 n=10+10)
wuffs_deflate_decode_1k_part_init/gcc10 123MB/s ± 1% 122MB/s ± 1% -0.94% (p=0.001 n=9+10)
wuffs_deflate_decode_10k_full_init/gcc10 216MB/s ± 2% 235MB/s ± 1% +8.83% (p=0.000 n=10+10)
wuffs_deflate_decode_10k_part_init/gcc10 223MB/s ± 1% 242MB/s ± 1% +8.45% (p=0.000 n=9+9)
wuffs_deflate_decode_100k_just_one_read/gcc10 270MB/s ± 4% 300MB/s ± 4% +11.08% (p=0.000 n=9+10)
wuffs_deflate_decode_100k_many_big_reads/gcc10 179MB/s ± 1% 193MB/s ± 2% +7.97% (p=0.000 n=10+10)
wuffs_gzip_decode_10k/clang9 218MB/s ± 3% 220MB/s ± 1% ~ (p=0.549 n=10+9)
wuffs_gzip_decode_100k/clang9 270MB/s ± 0% 269MB/s ± 2% ~ (p=0.730 n=9+9)
wuffs_gzip_decode_10k/gcc10 208MB/s ± 1% 227MB/s ± 1% +8.87% (p=0.000 n=9+10)
wuffs_gzip_decode_100k/gcc10 267MB/s ± 1% 283MB/s ± 1% +6.05% (p=0.000 n=10+9)
wuffs_zlib_decode_10k/clang9 214MB/s ± 2% 226MB/s ± 1% +5.69% (p=0.000 n=10+9)
wuffs_zlib_decode_100k/clang9 259MB/s ± 6% 272MB/s ± 2% +5.05% (p=0.000 n=9+10)
wuffs_zlib_decode_10k/gcc10 211MB/s ± 1% 227MB/s ± 1% +7.53% (p=0.000 n=10+10)
wuffs_zlib_decode_100k/gcc10 262MB/s ± 1% 282MB/s ± 1% +7.69% (p=0.000 n=10+9)
wuffs_png_decode_image_19k_8bpp/clang9 144MB/s ± 0% 148MB/s ± 1% +2.68% (p=0.000 n=9+9)
wuffs_png_decode_image_40k_24bpp/clang9 166MB/s ± 1% 170MB/s ± 1% +2.38% (p=0.000 n=9+9)
wuffs_png_decode_image_77k_8bpp/clang9 515MB/s ± 1% 525MB/s ± 1% +1.91% (p=0.000 n=9+8)
wuffs_png_decode_image_552k_32bpp_ignore_checksum/clang9 401MB/s ± 1% 400MB/s ± 2% ~ (p=0.913 n=9+9)
wuffs_png_decode_image_552k_32bpp_verify_checksum/clang9 387MB/s ± 1% 388MB/s ± 1% ~ (p=0.546 n=9+9)
wuffs_png_decode_image_4002k_24bpp/clang9 170MB/s ± 3% 175MB/s ± 1% +3.02% (p=0.000 n=9+10)
wuffs_png_decode_image_19k_8bpp/gcc10 143MB/s ± 2% 154MB/s ± 3% +8.32% (p=0.000 n=9+9)
wuffs_png_decode_image_40k_24bpp/gcc10 174MB/s ± 2% 182MB/s ± 1% +4.42% (p=0.000 n=10+9)
wuffs_png_decode_image_77k_8bpp/gcc10 513MB/s ± 2% 552MB/s ± 3% +7.66% (p=0.000 n=10+9)
wuffs_png_decode_image_552k_32bpp_ignore_checksum/gcc10 408MB/s ± 1% 426MB/s ± 2% +4.43% (p=0.000 n=9+10)
wuffs_png_decode_image_552k_32bpp_verify_checksum/gcc10 390MB/s ± 4% 410MB/s ± 2% +5.05% (p=0.000 n=10+9)
wuffs_png_decode_image_4002k_24bpp/gcc10 177MB/s ± 1% 186MB/s ± 1% +5.13% (p=0.000 n=9+9)
2021-04-09 00:06:33 +10:00
Nigel Tao
9b228fab2a
Make deflate.decoder.decode_huffman_fast64 choosy
2021-04-09 00:06:32 +10:00
Nigel Tao
b24e046670
Replace SIMD CRC-32 k6' magic constant with Px'
...
This is just a change of name (and a change of comment), not a change of
value.
From my colleague, Eric Biggers:
The number is not k6' = (x^32 mod P(x) << 32)' << 1 = 0x1DB710640, but
rather P(x)' = 0x1DB710641. This value represents the generator
polynomial of CRC-32 as a 33-bit value, using the CRC-32 bit ordering
convention (lowest order bit = coefficient of highest order polynomial
term). The lowest order bit is the coefficient of x^32, which by
definition is 1 for CRC-32. P(x) is used as a multiplicand in Barrett
reduction, which is the step which reduces a 64-bit value to the final
32-bit remainder. See "Algorithm 1" in the Gopal paper.
The reason that last bit doesn't actually matter is because the output
of the Barrett reduction is supposed to be fully reduced mod P(x), so
only 32 bits (the bits that represent the coefficients of x^0..x^31) are
supposed to be able to be nonzero. Therefore, in practice
implementations will just truncate the result to 32 bits. Incorrectly
using 0x1DB710640 rather than P(x) = 0x1DB710641 only causes the result
to be wrong by a factor of x^32, so it doesn't affect those 32 bits.
2021-03-18 11:21:51 +11:00
Nigel Tao
16b4b7554b
Link to SIMD CRC-32 k6' magic constant blog post
2021-03-18 11:06:21 +11:00