Commit Graph
54 Commits
Author SHA1 Message Date
Nigel Tao 65c2f4971f Rename some initialize methods to prepare or set 2019-01-26 15:46:42 +11:00
Nigel Tao 796c8ccc30 Simplify the gen.writeBuiltinQuestionCall caller 2018-12-27 17:06:59 +11:00
Nigel Tao 29c884dfcc Remove temporary coroSuspPoint numbering padding 2018-12-27 16:48:43 +11:00
Nigel Tao 28366f744f Drop duplicate CoroSuspPoint for io_reader.skip 2018-12-27 16:47:37 +11:00
Nigel Tao 9d11334e2b Gen explicit read_uxx_as_uyy width conversions 2018-12-23 09:41:07 +11:00
Nigel Tao 1349c7900c Add a yy arg to writeReadUxxAsUyy 2018-12-23 09:30:58 +11:00
Nigel Tao 5891e1a163 Make writeReadUXX more table-driven 2018-12-23 09:25:41 +11:00
Nigel Tao 155f544c6b Move writeReadUXX to the file where it's called 2018-12-23 09:14:50 +11:00
Nigel Tao 86a96969cc Tweak cgen of peek_uxx_as_uyy 2018-12-23 09:08:57 +11:00
Nigel Tao 33a4557d98 Collect GIF's LZW data into larger batches
name                                    old speed     new speed     delta

wuffs_gif_decode_1k_bw/clang            224MB/s ± 1%  227MB/s ± 0%  +0.98%  (p=0.000 n=10+9)
wuffs_gif_decode_1k_color/clang         116MB/s ± 1%  118MB/s ± 1%  +2.26%  (p=0.000 n=10+10)
wuffs_gif_decode_10k_bgra/clang         541MB/s ± 0%  568MB/s ± 1%  +4.97%  (p=0.000 n=9+10)
wuffs_gif_decode_10k_indexed/clang      144MB/s ± 0%  152MB/s ± 0%  +5.61%  (p=0.000 n=9+10)
wuffs_gif_decode_100k_artificial/clang  438MB/s ± 1%  454MB/s ± 1%  +3.56%  (p=0.000 n=9+10)
wuffs_gif_decode_100k_realistic/clang   171MB/s ± 0%  179MB/s ± 0%  +4.64%  (p=0.000 n=8+10)
wuffs_gif_decode_1000k/clang            174MB/s ± 0%  180MB/s ± 0%  +3.90%  (p=0.000 n=9+10)
wuffs_gif_decode_anim_screencap/clang   897MB/s ± 1%  943MB/s ± 0%  +5.10%  (p=0.000 n=9+9)

wuffs_gif_decode_1k_bw/gcc              250MB/s ± 1%  241MB/s ± 1%  -3.26%  (p=0.000 n=10+10)
wuffs_gif_decode_1k_color/gcc           132MB/s ± 1%  130MB/s ± 1%  -2.15%  (p=0.000 n=10+10)
wuffs_gif_decode_10k_bgra/gcc           534MB/s ± 1%  535MB/s ± 1%    ~     (p=0.211 n=9+10)
wuffs_gif_decode_10k_indexed/gcc        162MB/s ± 1%  163MB/s ± 0%    ~     (p=0.105 n=10+10)
wuffs_gif_decode_100k_artificial/gcc    460MB/s ± 0%  456MB/s ± 0%  -0.80%  (p=0.000 n=10+10)
wuffs_gif_decode_100k_realistic/gcc     186MB/s ± 0%  186MB/s ± 1%  -0.33%  (p=0.001 n=9+10)
wuffs_gif_decode_1000k/gcc              189MB/s ± 1%  188MB/s ± 1%    ~     (p=0.053 n=9+10)
wuffs_gif_decode_anim_screencap/gcc     981MB/s ± 1%  971MB/s ± 1%  -0.95%  (p=0.000 n=10+10)
2018-12-15 14:10:10 +11:00
Nigel Tao fad2ddce68 Generate less code for foo.since_mark().length() 2018-12-09 11:15:10 +11:00
Nigel Tao f9249d4089 Tweak how std/gif's uncompressed_wi is updated 2018-12-09 10:35:24 +11:00
Nigel Tao d256df0fe3 Refactor cgen of question-calls 2018-12-02 10:20:43 +11:00
Nigel Tao b162747934 Remove cgen's obsolete replacementPolicy concept 2018-12-01 22:41:41 +11:00
Nigel Tao c986536230 Rename some read_uxx methods as read_uxx_as_uyy 2018-11-25 12:15:05 +11:00
Nigel Tao e3486eec17 Remove unused gif.decoder.skip_frame variable 2018-11-17 10:58:39 +11:00
Nigel Tao 51a0b347ed Add WUFFS_BASE__LOW_BITS_MASK__UXX macros 2018-10-21 21:53:44 +11:00
Nigel Tao f60723f624 Simplify cgen output for copy_from_slice 2018-10-21 18:03:54 +11:00
Nigel Tao 3338d6d1bc Write LZW output 8 bytes at a time, not 1
The benchmarks are mixed, but I think it's still a net win.

----
name                                    old speed     new speed     delta

wuffs_gif_decode_1k_bw/clang            258MB/s ± 0%  253MB/s ± 1%    -2.04%  (p=0.008 n=5+5)
wuffs_gif_decode_1k_color/clang         159MB/s ± 0%  116MB/s ± 0%   -26.76%  (p=0.008 n=5+5)
wuffs_gif_decode_10k/clang              167MB/s ± 0%  138MB/s ± 0%   -17.42%  (p=0.008 n=5+5)
wuffs_gif_decode_100k_artificial/clang  205MB/s ± 0%  439MB/s ± 0%  +114.53%  (p=0.008 n=5+5)
wuffs_gif_decode_100k_realistic/clang   151MB/s ± 0%  164MB/s ± 0%    +8.41%  (p=0.008 n=5+5)
wuffs_gif_decode_1000k/clang            155MB/s ± 0%  165MB/s ± 0%    +6.50%  (p=0.008 n=5+5)
wuffs_gif_decode_anim_screencap/clang   255MB/s ± 0%  921MB/s ± 0%  +261.50%  (p=0.008 n=5+5)

wuffs_gif_decode_1k_bw/gcc              316MB/s ± 2%  267MB/s ± 1%   -15.78%  (p=0.008 n=5+5)
wuffs_gif_decode_1k_color/gcc           184MB/s ± 1%  134MB/s ± 0%   -27.57%  (p=0.008 n=5+5)
wuffs_gif_decode_10k/gcc                196MB/s ± 0%  161MB/s ± 0%   -17.50%  (p=0.008 n=5+5)
wuffs_gif_decode_100k_artificial/gcc    287MB/s ± 0%  449MB/s ± 1%   +56.22%  (p=0.008 n=5+5)
wuffs_gif_decode_100k_realistic/gcc     182MB/s ± 0%  182MB/s ± 1%      ~     (p=0.690 n=5+5)
wuffs_gif_decode_1000k/gcc              187MB/s ± 1%  185MB/s ± 1%      ~     (p=0.056 n=5+5)
wuffs_gif_decode_anim_screencap/gcc     397MB/s ± 0%  973MB/s ± 0%  +145.42%  (p=0.008 n=5+5)

wuffs_lzw_decode_20k/clang              170MB/s ± 0%  191MB/s ± 1%   +12.52%  (p=0.008 n=5+5)
wuffs_lzw_decode_100k/clang             242MB/s ± 0%  312MB/s ± 2%   +29.00%  (p=0.008 n=5+5)

wuffs_lzw_decode_20k/gcc                220MB/s ± 0%  227MB/s ± 0%    +3.31%  (p=0.008 n=5+5)
wuffs_lzw_decode_100k/gcc               317MB/s ± 0%  369MB/s ± 1%   +16.53%  (p=0.008 n=5+5)
----

LZW is the compression method used by GIF images. Roughly speaking, LZW
is a sequence of codes, and each code emits either 1 byte (a literal
code) or multiple bytes (a copy code). The speed gain in this commit
comes from, when emitting an N-byte copy code, this now requires N/8
iterations instead of N iterations, as each iteration now writes 8 bytes
instead of 1. Literal codes, both before and after this commit, requires
a single iteration.

Decoding should therefore get faster or slower depending on whether it's
copy-code-heavy or literal-code-heavy respectively. For what it's worth,
copy-code-heavy GIF images also usually compress better.

Running script/extract-giflzw.go (or examining the cached output at
test/data/analysis/giflzw-emission-lengths.txt) on the anim_screencap
(which got faster) and 1k_color (which got slower) cases (also known as
gifplayer-muybridge.gif and hippopotamus.regular.gif) shows that the
proportion of literal codes agrees with that explanation.

----
gifplayer-muybridge.gif (all frames)
 Percent   Count  Emission_Length_Bucket
   12.78   34658           1 (Literal Code)
   14.05   38106           2
   13.25   35937           3
   10.59   28707           4
    7.85   21286           5
    6.08   16481           6
    4.85   13156           7
    3.60    9764           8
    9.34   25318    9 ..  16
    4.89   13257   17 ..  32
    4.78   12954   33 ..  64
    4.72   12806   65 .. 128
    3.02    8190  127 .. 256
    0.20     550  256+

hippopotamus.regular.gif (all frames)
 Percent   Count  Emission_Length_Bucket
   82.66     672           1 (Literal Code)
   13.04     106           2
    2.83      23           3
    0.86       7           4
    0.37       3           5
    0.25       2           6
    0.00       0           7
    0.00       0           8
    0.00       0    9 ..  16
    0.00       0   17 ..  32
    0.00       0   33 ..  64
    0.00       0   65 .. 128
    0.00       0  127 .. 256
    0.00       0  256+
----

The new algorithm also adds more CPU and RAM overhead, regardless of
whether the input GIF image's LZW data is copy-code-heavy or
literal-code-heavy. For example, the suffix table grows by 8x, from 4KiB
to 32 KiB.

The gifplayer-muybridge.gif image is also larger (472 x 298) and the
hippopotamus.regular.gif image is also smaller (36 x 28). Higher fixed
costs may have larger (negative) relative effects on decoding smaller
images.

To repeat, I think it's still a net win.

An upcoming commit will update the std/lzw/README.md file to provide
more detail on the algorithm.

Thanks to Tushar Khandelwal for the idea.
2018-10-21 14:48:40 +11:00
Nigel Tao 97f2570977 Fix codegen for skip_fast
args[0] is actual, args[1] is worst_case.
2018-10-14 12:52:52 +11:00
Nigel Tao 6fed251b40 Split an io_buffer's fields into data and meta 2018-09-15 08:45:50 +10:00
Nigel Tao 7798beaa8c Add I/O positions 2018-08-26 16:31:19 +10:00
Nigel Tao 4e7ef5fb88 Rename I/O bounds fields to mark and limit 2018-08-18 09:38:02 +10:00
Nigel Tao b840c80040 Make the status type a const char*, not a struct 2018-08-12 16:05:14 +10:00
Nigel Tao 1e8438f43b Have cgen code call wuffs_base__status methods 2018-08-12 09:44:58 +10:00
Nigel Tao 2a32a86a6c Make the status type a struct, not an int32_t 2018-08-11 13:27:29 +10:00
Nigel Tao 9c0b2d2f5c Rename ioptr_etc cgen variables 2018-08-11 08:32:49 +10:00
Nigel Tao f2ee78aa2a Remove short_read_src cgen code path 2018-08-04 09:45:35 +10:00
Nigel Tao ca06083f8a Rename the implicit in variable to args 2018-07-29 10:26:22 +10:00
Nigel Tao 29eb64a0a4 Add a nptr T type 2018-07-15 17:10:03 +10:00
Nigel Tao e49d464432 Make _fast explicit in copy_n_from_history_fast 2018-07-08 21:39:18 +10:00
Nigel Tao 8a449fa92e Rename __bco to _fast 2018-07-08 21:10:40 +10:00
Nigel Tao 9e300fc4f8 Rename unread_u8? to undo_byte!; add can_undo_byte 2018-07-08 21:02:29 +10:00
Nigel Tao ce3d99a6d5 Remove no longer used optimizeSuspendible code 2018-07-08 17:10:34 +10:00
Nigel Tao a32683e5ea Drop the 32 in the skip32 and skip32_fast names 2018-07-08 14:02:32 +10:00
Nigel Tao 26eba56f57 Rename some copy_from_xxx methods 2018-07-08 13:56:49 +10:00
Nigel Tao 732365b80d Add write_fast_uxx methods 2018-07-08 13:26:28 +10:00
Nigel Tao ccb78f5db7 Rename ast.Node methods from Foo to AsFoo 2018-07-01 20:41:21 +10:00
Nigel Tao 2599476d2c Add skip32_fast method 2018-07-01 11:14:25 +10:00
Nigel Tao bb4477d91f Add peek_uxx methods 2018-07-01 10:26:17 +10:00
Nigel Tao 008b83e4b4 Fix some more C++ (not C) compile errors 2018-06-30 13:41:19 +10:00
Nigel Tao 50a955317a Remove the WUFFS_FOO__STATUS_ETC names 2018-06-10 22:05:06 +10:00
Nigel Tao 20e459cfb2 Add read_u24be etc methods
Fixes #12
2018-06-02 13:03:24 +10:00
Nigel Tao 47cf248e57 Remove the (table T).linearize method 2018-05-29 08:23:10 +10:00
Nigel Tao f43afd8236 Add a (table T).row method 2018-05-26 11:32:54 +10:00
Nigel Tao 85256c7a45 Add a (table T).linearize method 2018-05-18 13:54:07 +10:00
Nigel Tao e4320c3c6d Add built-in table T methods 2018-05-18 12:50:53 +10:00
Nigel Tao 7452ec8035 Rename b_rptr_foo as ioptr_foo 2018-05-17 09:59:04 +10:00
Nigel Tao 2448b2ae7a Factor out writeBuiltinCallSuspendibles 2018-05-16 12:12:47 +10:00
Nigel Tao 8b4a82ab83 Remove hard-coded output in writeCallSuspendibles
Updates #8
2018-05-16 11:41:48 +10:00