From 125e91a174644b6cfc003ae04b8e127b0951edba Mon Sep 17 00:00:00 2001
From: Russ Cox <rsc@golang.org>
Date: Tue, 3 May 2011 01:50:43 -0400
Subject: [PATCH] [release-branch.r57] jpeg: speed up RGBA encoding ~%50
MIME-Version: 1.0
Content-Type: text/plain; charset=utf8
Content-Transfer-Encoding: 8bit

««« CL 4433088 / 099dd59d3976
jpeg: speed up RGBA encoding ~%50

Avoids image.At(), color.RGBA(), opposing 8 bit shifts,
and min function calls in a loop.  Not as pretty as before,
but the pure version is still there to revert back to
later if/when the compiler gets better.

before (best of 5)
jpeg.BenchmarkEncodeRGBOpaque   50   64781360 ns/op   18.97 MB/s

after (best of 5)
jpeg.BenchmarkEncodeRGBOpaque   50   42044300 ns/op   29.23 MB/s

(benchmarked on an HP z600; 16 core Xeon E5520 @ 2.27Ghz)

R=r, r2, nigeltao
CC=golang-dev
https://golang.org/cl/4433088
»»»

TBR=adg
CC=golang-dev
https://golang.org/cl/4431086
---
 src/pkg/image/jpeg/writer.go | 32 +++++++++++++++++++++++++++++++-
 1 file changed, 31 insertions(+), 1 deletion(-)

diff --git a/src/pkg/image/jpeg/writer.go b/src/pkg/image/jpeg/writer.go
index 505cce04fb..52b3dc4e2c 100644
--- a/src/pkg/image/jpeg/writer.go
+++ b/src/pkg/image/jpeg/writer.go
@@ -391,6 +391,31 @@ func toYCbCr(m image.Image, p image.Point, yBlock, cbBlock, crBlock *block) {
 	}
 }
 
+// rgbaToYCbCr is a specialized version of toYCbCr for image.RGBA images.
+func rgbaToYCbCr(m *image.RGBA, p image.Point, yBlock, cbBlock, crBlock *block) {
+	b := m.Bounds()
+	xmax := b.Max.X - 1
+	ymax := b.Max.Y - 1
+	for j := 0; j < 8; j++ {
+		sj := p.Y + j
+		if sj > ymax {
+			sj = ymax
+		}
+		yoff := sj * m.Stride
+		for i := 0; i < 8; i++ {
+			sx := p.X + i
+			if sx > xmax {
+				sx = xmax
+			}
+			col := &m.Pix[yoff+sx]
+			yy, cb, cr := ycbcr.RGBToYCbCr(col.R, col.G, col.B)
+			yBlock[8*j+i] = int(yy)
+			cbBlock[8*j+i] = int(cb)
+			crBlock[8*j+i] = int(cr)
+		}
+	}
+}
+
 // scale scales the 16x16 region represented by the 4 src blocks to the 8x8
 // dst block.
 func scale(dst *block, src *[4]block) {
@@ -431,13 +456,18 @@ func (e *encoder) writeSOS(m image.Image) {
 		prevDCY, prevDCCb, prevDCCr int
 	)
 	bounds := m.Bounds()
+	rgba, _ := m.(*image.RGBA)
 	for y := bounds.Min.Y; y < bounds.Max.Y; y += 16 {
 		for x := bounds.Min.X; x < bounds.Max.X; x += 16 {
 			for i := 0; i < 4; i++ {
 				xOff := (i & 1) * 8
 				yOff := (i & 2) * 4
 				p := image.Point{x + xOff, y + yOff}
-				toYCbCr(m, p, &yBlock, &cbBlock[i], &crBlock[i])
+				if rgba != nil {
+					rgbaToYCbCr(rgba, p, &yBlock, &cbBlock[i], &crBlock[i])
+				} else {
+					toYCbCr(m, p, &yBlock, &cbBlock[i], &crBlock[i])
+				}
 				prevDCY = e.writeBlock(&yBlock, 0, prevDCY)
 			}
 			scale(&cBlock, &cbBlock)
-- 
2.50.0