From: Nao YONASHIRO <owan.orisano@gmail.com>
Date: Wed, 24 Jul 2019 13:00:58 +0000 (+0900)
Subject: image/png: hoist repetitive pixels per byte out of loop in Encode
X-Git-Tag: go1.14beta1~1321
X-Git-Url: http://www.git.cypherpunks.su/?a=commitdiff_plain;h=f04f594a1fd1e8cf1bd01f9dc62599d5bd6e1d92;p=gostls13.git

image/png: hoist repetitive pixels per byte out of loop in Encode

The existing implementation has calculated pixels per byte for each pixel.
reduce the calculation of pixels per byte.

name                        old time/op    new time/op     delta
EncodeGray-4                  2.16ms ± 1%     2.16ms ± 1%    -0.28%  (p=0.000 n=86+84)
EncodeGrayWithBufferPool-4    1.99ms ± 0%     1.97ms ± 0%    -0.72%  (p=0.000 n=97+92)
EncodeNRGBOpaque-4            6.51ms ± 1%     6.48ms ± 1%    -0.45%  (p=0.000 n=90+85)
EncodeNRGBA-4                 7.33ms ± 1%     7.28ms ± 0%    -0.69%  (p=0.000 n=89+87)
EncodePaletted-4              5.10ms ± 1%     2.29ms ± 0%   -55.11%  (p=0.000 n=90+85)
EncodeRGBOpaque-4             6.51ms ± 1%     6.51ms ± 0%      ~     (p=0.311 n=94+88)
EncodeRGBA-4                  24.3ms ± 2%     24.1ms ± 1%    -0.87%  (p=0.000 n=91+91)

name                        old speed      new speed       delta
EncodeGray-4                 142MB/s ± 1%    143MB/s ± 1%    +0.26%  (p=0.000 n=86+85)
EncodeGrayWithBufferPool-4   154MB/s ± 0%    156MB/s ± 0%    +0.73%  (p=0.000 n=97+92)
EncodeNRGBOpaque-4           189MB/s ± 1%    190MB/s ± 1%    +0.44%  (p=0.000 n=90+86)
EncodeNRGBA-4                168MB/s ± 1%    169MB/s ± 0%    +0.69%  (p=0.000 n=89+87)
EncodePaletted-4            60.3MB/s ± 1%  134.2MB/s ± 0%  +122.74%  (p=0.000 n=90+85)
EncodeRGBOpaque-4            189MB/s ± 1%    189MB/s ± 0%      ~     (p=0.326 n=94+88)
EncodeRGBA-4                50.6MB/s ± 2%   51.1MB/s ± 1%    +0.87%  (p=0.000 n=91+91)

name                        old alloc/op   new alloc/op    delta
EncodeGray-4                   852kB ± 0%      852kB ± 0%    +0.00%  (p=0.000 n=100+100)
EncodeGrayWithBufferPool-4    1.49kB ± 2%     1.47kB ± 1%    -0.88%  (p=0.000 n=95+90)
EncodeNRGBOpaque-4             860kB ± 0%      860kB ± 0%    +0.00%  (p=0.003 n=98+58)
EncodeNRGBA-4                  864kB ± 0%      864kB ± 0%    +0.00%  (p=0.021 n=100+99)
EncodePaletted-4               849kB ± 0%      849kB ± 0%    +0.00%  (p=0.040 n=100+100)
EncodeRGBOpaque-4              860kB ± 0%      860kB ± 0%      ~     (p=0.062 n=66+98)
EncodeRGBA-4                  3.32MB ± 0%     3.32MB ± 0%    -0.00%  (p=0.044 n=99+99)

name                        old allocs/op  new allocs/op   delta
EncodeGray-4                    32.0 ± 0%       32.0 ± 0%      ~     (all equal)
EncodeGrayWithBufferPool-4      3.00 ± 0%       3.00 ± 0%      ~     (all equal)
EncodeNRGBOpaque-4              32.0 ± 0%       32.0 ± 0%      ~     (all equal)
EncodeNRGBA-4                   32.0 ± 0%       32.0 ± 0%      ~     (all equal)
EncodePaletted-4                36.0 ± 0%       36.0 ± 0%      ~     (all equal)
EncodeRGBOpaque-4               32.0 ± 0%       32.0 ± 0%      ~     (all equal)
EncodeRGBA-4                    614k ± 0%       614k ± 0%      ~     (all equal)

Change-Id: I716bde2dc80d8111d75d3d765fc09223f770d5a5
Reviewed-on: https://go-review.googlesource.com/c/go/+/187417
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
Reviewed-by: Nigel Tao <nigeltao@golang.org>
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
---

diff --git a/src/image/png/writer.go b/src/image/png/writer.go
index 2bd4fde692..53adc1633c 100644
--- a/src/image/png/writer.go
+++ b/src/image/png/writer.go
@@ -426,10 +426,11 @@ func (e *encoder) writeImage(w io.Writer, m image.Image, cb int, level int) erro
 
 			var a uint8
 			var c int
+			pixelsPerByte := 8 / bitsPerPixel
 			for x := b.Min.X; x < b.Max.X; x++ {
 				a = a<<uint(bitsPerPixel) | pi.ColorIndexAt(x, y)
 				c++
-				if c == 8/bitsPerPixel {
+				if c == pixelsPerByte {
 					cr[0][i] = a
 					i += 1
 					a = 0
@@ -437,7 +438,7 @@ func (e *encoder) writeImage(w io.Writer, m image.Image, cb int, level int) erro
 				}
 			}
 			if c != 0 {
-				for c != 8/bitsPerPixel {
+				for c != pixelsPerByte {
 					a = a << uint(bitsPerPixel)
 					c++
 				}