std/jpeg: tweak decode_mcu z variable assignment

name                                             old speed     new speed     delta

wuffs_jpeg_decode_19k_8bpp/clang11               227MB/s ± 0%  229MB/s ± 0%  +0.93%  (p=0.008 n=5+5)
wuffs_jpeg_decode_30k_24bpp_progressive/clang11  213MB/s ± 0%  212MB/s ± 0%  -0.50%  (p=0.008 n=5+5)
wuffs_jpeg_decode_30k_24bpp_sequential/clang11   380MB/s ± 0%  380MB/s ± 0%    ~     (p=1.000 n=5+5)
wuffs_jpeg_decode_77k_24bpp/clang11              329MB/s ± 0%  331MB/s ± 0%  +0.73%  (p=0.008 n=5+5)
wuffs_jpeg_decode_552k_24bpp_420/clang11         342MB/s ± 0%  344MB/s ± 0%  +0.72%  (p=0.008 n=5+5)
wuffs_jpeg_decode_552k_24bpp_444/clang11         469MB/s ± 0%  476MB/s ± 0%  +1.46%  (p=0.008 n=5+5)
wuffs_jpeg_decode_4002k_24bpp/clang11            348MB/s ± 0%  350MB/s ± 0%  +0.51%  (p=0.008 n=5+5)

wuffs_jpeg_decode_19k_8bpp/gcc10                 213MB/s ± 0%  215MB/s ± 0%  +0.70%  (p=0.008 n=5+5)
wuffs_jpeg_decode_30k_24bpp_progressive/gcc10    246MB/s ± 0%  246MB/s ± 0%    ~     (p=0.095 n=5+5)
wuffs_jpeg_decode_30k_24bpp_sequential/gcc10     482MB/s ± 0%  485MB/s ± 0%  +0.61%  (p=0.008 n=5+5)
wuffs_jpeg_decode_77k_24bpp/gcc10                392MB/s ± 0%  394MB/s ± 0%  +0.53%  (p=0.008 n=5+5)
wuffs_jpeg_decode_552k_24bpp_420/gcc10           412MB/s ± 0%  414MB/s ± 0%  +0.45%  (p=0.008 n=5+5)
wuffs_jpeg_decode_552k_24bpp_444/gcc10           460MB/s ± 0%  462MB/s ± 0%  +0.43%  (p=0.016 n=5+4)
wuffs_jpeg_decode_4002k_24bpp/gcc10              419MB/s ± 0%  420MB/s ± 0%  +0.32%  (p=0.008 n=5+5)
This commit is contained in:
Nigel Tao committed 2023-06-19 20:13:01 +10:00
1 parent 48fa7182f4
commit 6eb92544bd
3 files changed
+4 -12

No files matched your search

+1 -3
View File
@@ -204,6 +204,7 @@ pri func decoder.decode_mcu!(workbuf: slice base.u8, mx: base.u32[..= 0x1FFF], m
// Split the 8-bit ac_symbol into two 4-bit halves, per section
// F.2.2.2 "Decoding procedure for AC coefficients".
ac_rrrr = ac_symbol >> 4
z += ac_rrrr + 1
ac_ssss = ac_symbol & 15
ac_extend = EXTEND[ac_ssss]
@@ -214,13 +215,10 @@ pri func decoder.decode_mcu!(workbuf: slice base.u8, mx: base.u32[..= 0x1FFF], m
(((this.util.sign_extend_rshift_u64(a: bits, n: 63) & 0xFFFF) as base.u16) ^ 0xFFFF)
bits ~mod<<= ac_ssss
n_bits ~mod-= ac_ssss
z += ac_rrrr + 1
this.mcu_blocks[0][UNZIG[z]] =
ac_value
} else if ac_rrrr < 15 {
break.ac_components
} else {
z += 16
}
} endwhile.ac_components
@@ -120,6 +120,7 @@ pri func decoder.decode_mcu_progressive_ac_high_bits!(workbuf: slice base.u8, mx
// Split the 8-bit ac_symbol into two 4-bit halves, per section
// F.2.2.2 "Decoding procedure for AC coefficients".
ac_rrrr = ac_symbol >> 4
z += ac_rrrr + 1
ac_ssss = ac_symbol & 15
ac_extend = EXTEND[ac_ssss]
@@ -131,7 +132,6 @@ pri func decoder.decode_mcu_progressive_ac_high_bits!(workbuf: slice base.u8, mx
(((this.util.sign_extend_rshift_u64(a: bits, n: 63) & 0xFFFF) as base.u16) ^ 0xFFFF)
bits ~mod<<= ac_ssss
n_bits ~mod-= ac_ssss
z += ac_rrrr + 1
this.mcu_blocks[0][UNZIG[z]] =
((ac_value ~mod<< this.scan_al) & 0xFFFF) as base.u16
} else if ac_rrrr < 15 {
@@ -142,8 +142,6 @@ pri func decoder.decode_mcu_progressive_ac_high_bits!(workbuf: slice base.u8, mx
n_bits ~mod-= ac_rrrr
}
break.ac_components
} else {
z += 16
}
} endwhile.ac_components