draw: RGBA src fast path for scaling.
benchmark old ns/op new ns/op delta
BenchmarkScaleSrcRGBA 15124800 2091946 -86.17%
Change-Id: Id8d3088793ebf1d75b929fcf6945987817e87463
Reviewed-on: https://go-review.googlesource.com/6234
Reviewed-by: Rob Pike <r@golang.org>
diff --git a/draw/gen.go b/draw/gen.go
index 7b7bfc3..8ffea52 100644
--- a/draw/gen.go
+++ b/draw/gen.go
@@ -284,10 +284,26 @@
switch d.sType {
default:
log.Fatalf("bad sType %q", d.sType)
- case "image.Image", "*image.NRGBA", "*image.RGBA", "*image.Uniform", "*image.YCbCr": // TODO: separate code for concrete types.
+ case "image.Image", "*image.NRGBA", "*image.Uniform", "*image.YCbCr": // TODO: separate code for concrete types.
fmt.Fprintf(buf, "%sr%s, %sg%s, %sb%s, %sa%s := "+
"src.At(sp.X + int(%s), sp.Y+int(%s)).RGBA()\n",
- lhs, tmp, lhs, tmp, lhs, tmp, lhs, tmp, args[0], args[1])
+ lhs, tmp, lhs, tmp, lhs, tmp, lhs, tmp,
+ args[0], args[1],
+ )
+ case "*image.RGBA":
+ // TODO: there's no need to multiply by 0x101 if the next thing
+ // we're going to do is shift right by 8.
+ fmt.Fprintf(buf, "%si := src.PixOffset(sp.X + int(%s), sp.Y+int(%s))\n"+
+ "%sr%s := uint32(src.Pix[%si+0]) * 0x101\n"+
+ "%sg%s := uint32(src.Pix[%si+1]) * 0x101\n"+
+ "%sb%s := uint32(src.Pix[%si+2]) * 0x101\n"+
+ "%sa%s := uint32(src.Pix[%si+3]) * 0x101\n",
+ lhs, args[0], args[1],
+ lhs, tmp, lhs,
+ lhs, tmp, lhs,
+ lhs, tmp, lhs,
+ lhs, tmp, lhs,
+ )
}
if dollar == "srcf" {
@@ -399,7 +415,14 @@
if dr.Empty() {
return
}
- $switch z.scale_$dTypeRN_$sTypeRN(dst, dp, dr, src, sp)
+ // sr is the source pixels. If it extends beyond the src bounds,
+ // we cannot use the type-specific fast paths, as they access
+ // the Pix fields directly without bounds checking.
+ if sr := (image.Rectangle{sp, sp.Add(image.Point{int(z.sw), int(z.sh)})}); !sr.In(src.Bounds()) {
+ z.scale_Image_Image(dst, dp, dr, src, sp)
+ } else {
+ $switch z.scale_$dTypeRN_$sTypeRN(dst, dp, dr, src, sp)
+ }
}
`
@@ -478,7 +501,16 @@
// scaleY distributes the temporary image's rows over the destination image.
// TODO: is it worth having a sync.Pool for this temporary buffer?
tmp := make([][4]float64, z.dw*z.sh)
- $switchS z.scaleX_$sTypeRN(tmp, src, sp)
+
+ // sr is the source pixels. If it extends beyond the src bounds,
+ // we cannot use the type-specific fast paths, as they access
+ // the Pix fields directly without bounds checking.
+ if sr := (image.Rectangle{sp, sp.Add(image.Point{int(z.sw), int(z.sh)})}); !sr.In(src.Bounds()) {
+ z.scaleX_Image(tmp, src, sp)
+ } else {
+ $switchS z.scaleX_$sTypeRN(tmp, src, sp)
+ }
+
$switchD z.scaleY_$dTypeRN(dst, dp, dr, tmp)
}
`
diff --git a/draw/impl.go b/draw/impl.go
index c284c4c..3c07d31 100644
--- a/draw/impl.go
+++ b/draw/impl.go
@@ -16,24 +16,31 @@
if dr.Empty() {
return
}
- switch dst := dst.(type) {
- case *image.RGBA:
- switch src := src.(type) {
- case *image.NRGBA:
- z.scale_RGBA_NRGBA(dst, dp, dr, src, sp)
+ // sr is the source pixels. If it extends beyond the src bounds,
+ // we cannot use the type-specific fast paths, as they access
+ // the Pix fields directly without bounds checking.
+ if sr := (image.Rectangle{sp, sp.Add(image.Point{int(z.sw), int(z.sh)})}); !sr.In(src.Bounds()) {
+ z.scale_Image_Image(dst, dp, dr, src, sp)
+ } else {
+ switch dst := dst.(type) {
case *image.RGBA:
- z.scale_RGBA_RGBA(dst, dp, dr, src, sp)
- case *image.Uniform:
- z.scale_RGBA_Uniform(dst, dp, dr, src, sp)
- case *image.YCbCr:
- z.scale_RGBA_YCbCr(dst, dp, dr, src, sp)
+ switch src := src.(type) {
+ case *image.NRGBA:
+ z.scale_RGBA_NRGBA(dst, dp, dr, src, sp)
+ case *image.RGBA:
+ z.scale_RGBA_RGBA(dst, dp, dr, src, sp)
+ case *image.Uniform:
+ z.scale_RGBA_Uniform(dst, dp, dr, src, sp)
+ case *image.YCbCr:
+ z.scale_RGBA_YCbCr(dst, dp, dr, src, sp)
+ default:
+ z.scale_RGBA_Image(dst, dp, dr, src, sp)
+ }
default:
- z.scale_RGBA_Image(dst, dp, dr, src, sp)
- }
- default:
- switch src := src.(type) {
- default:
- z.scale_Image_Image(dst, dp, dr, src, sp)
+ switch src := src.(type) {
+ default:
+ z.scale_Image_Image(dst, dp, dr, src, sp)
+ }
}
}
}
@@ -60,7 +67,11 @@
d := dst.PixOffset(dp.X+dr.Min.X, dp.Y+int(dy))
for dx := int32(dr.Min.X); dx < int32(dr.Max.X); dx++ {
sx := (2*uint64(dx) + 1) * uint64(z.sw) / (2 * uint64(z.dw))
- pr, pg, pb, pa := src.At(sp.X+int(sx), sp.Y+int(sy)).RGBA()
+ pi := src.PixOffset(sp.X+int(sx), sp.Y+int(sy))
+ pr := uint32(src.Pix[pi+0]) * 0x101
+ pg := uint32(src.Pix[pi+1]) * 0x101
+ pb := uint32(src.Pix[pi+2]) * 0x101
+ pa := uint32(src.Pix[pi+3]) * 0x101
dst.Pix[d+0] = uint8(uint32(pr) >> 8)
dst.Pix[d+1] = uint8(uint32(pg) >> 8)
dst.Pix[d+2] = uint8(uint32(pb) >> 8)
@@ -144,24 +155,31 @@
if dr.Empty() {
return
}
- switch dst := dst.(type) {
- case *image.RGBA:
- switch src := src.(type) {
- case *image.NRGBA:
- z.scale_RGBA_NRGBA(dst, dp, dr, src, sp)
+ // sr is the source pixels. If it extends beyond the src bounds,
+ // we cannot use the type-specific fast paths, as they access
+ // the Pix fields directly without bounds checking.
+ if sr := (image.Rectangle{sp, sp.Add(image.Point{int(z.sw), int(z.sh)})}); !sr.In(src.Bounds()) {
+ z.scale_Image_Image(dst, dp, dr, src, sp)
+ } else {
+ switch dst := dst.(type) {
case *image.RGBA:
- z.scale_RGBA_RGBA(dst, dp, dr, src, sp)
- case *image.Uniform:
- z.scale_RGBA_Uniform(dst, dp, dr, src, sp)
- case *image.YCbCr:
- z.scale_RGBA_YCbCr(dst, dp, dr, src, sp)
+ switch src := src.(type) {
+ case *image.NRGBA:
+ z.scale_RGBA_NRGBA(dst, dp, dr, src, sp)
+ case *image.RGBA:
+ z.scale_RGBA_RGBA(dst, dp, dr, src, sp)
+ case *image.Uniform:
+ z.scale_RGBA_Uniform(dst, dp, dr, src, sp)
+ case *image.YCbCr:
+ z.scale_RGBA_YCbCr(dst, dp, dr, src, sp)
+ default:
+ z.scale_RGBA_Image(dst, dp, dr, src, sp)
+ }
default:
- z.scale_RGBA_Image(dst, dp, dr, src, sp)
- }
- default:
- switch src := src.(type) {
- default:
- z.scale_Image_Image(dst, dp, dr, src, sp)
+ switch src := src.(type) {
+ default:
+ z.scale_Image_Image(dst, dp, dr, src, sp)
+ }
}
}
}
@@ -267,12 +285,20 @@
sx1 = sx0
xFrac0, xFrac1 = 1, 0
}
- s00ru, s00gu, s00bu, s00au := src.At(sp.X+int(sx0), sp.Y+int(sy0)).RGBA()
+ s00i := src.PixOffset(sp.X+int(sx0), sp.Y+int(sy0))
+ s00ru := uint32(src.Pix[s00i+0]) * 0x101
+ s00gu := uint32(src.Pix[s00i+1]) * 0x101
+ s00bu := uint32(src.Pix[s00i+2]) * 0x101
+ s00au := uint32(src.Pix[s00i+3]) * 0x101
s00r := float64(s00ru)
s00g := float64(s00gu)
s00b := float64(s00bu)
s00a := float64(s00au)
- s10ru, s10gu, s10bu, s10au := src.At(sp.X+int(sx1), sp.Y+int(sy0)).RGBA()
+ s10i := src.PixOffset(sp.X+int(sx1), sp.Y+int(sy0))
+ s10ru := uint32(src.Pix[s10i+0]) * 0x101
+ s10gu := uint32(src.Pix[s10i+1]) * 0x101
+ s10bu := uint32(src.Pix[s10i+2]) * 0x101
+ s10au := uint32(src.Pix[s10i+3]) * 0x101
s10r := float64(s10ru)
s10g := float64(s10gu)
s10b := float64(s10bu)
@@ -281,12 +307,20 @@
s10g = xFrac1*s00g + xFrac0*s10g
s10b = xFrac1*s00b + xFrac0*s10b
s10a = xFrac1*s00a + xFrac0*s10a
- s01ru, s01gu, s01bu, s01au := src.At(sp.X+int(sx0), sp.Y+int(sy1)).RGBA()
+ s01i := src.PixOffset(sp.X+int(sx0), sp.Y+int(sy1))
+ s01ru := uint32(src.Pix[s01i+0]) * 0x101
+ s01gu := uint32(src.Pix[s01i+1]) * 0x101
+ s01bu := uint32(src.Pix[s01i+2]) * 0x101
+ s01au := uint32(src.Pix[s01i+3]) * 0x101
s01r := float64(s01ru)
s01g := float64(s01gu)
s01b := float64(s01bu)
s01a := float64(s01au)
- s11ru, s11gu, s11bu, s11au := src.At(sp.X+int(sx1), sp.Y+int(sy1)).RGBA()
+ s11i := src.PixOffset(sp.X+int(sx1), sp.Y+int(sy1))
+ s11ru := uint32(src.Pix[s11i+0]) * 0x101
+ s11gu := uint32(src.Pix[s11i+1]) * 0x101
+ s11bu := uint32(src.Pix[s11i+2]) * 0x101
+ s11au := uint32(src.Pix[s11i+3]) * 0x101
s11r := float64(s11ru)
s11g := float64(s11gu)
s11b := float64(s11bu)
@@ -607,18 +641,27 @@
// scaleY distributes the temporary image's rows over the destination image.
// TODO: is it worth having a sync.Pool for this temporary buffer?
tmp := make([][4]float64, z.dw*z.sh)
- switch src := src.(type) {
- case *image.NRGBA:
- z.scaleX_NRGBA(tmp, src, sp)
- case *image.RGBA:
- z.scaleX_RGBA(tmp, src, sp)
- case *image.Uniform:
- z.scaleX_Uniform(tmp, src, sp)
- case *image.YCbCr:
- z.scaleX_YCbCr(tmp, src, sp)
- default:
+
+ // sr is the source pixels. If it extends beyond the src bounds,
+ // we cannot use the type-specific fast paths, as they access
+ // the Pix fields directly without bounds checking.
+ if sr := (image.Rectangle{sp, sp.Add(image.Point{int(z.sw), int(z.sh)})}); !sr.In(src.Bounds()) {
z.scaleX_Image(tmp, src, sp)
+ } else {
+ switch src := src.(type) {
+ case *image.NRGBA:
+ z.scaleX_NRGBA(tmp, src, sp)
+ case *image.RGBA:
+ z.scaleX_RGBA(tmp, src, sp)
+ case *image.Uniform:
+ z.scaleX_Uniform(tmp, src, sp)
+ case *image.YCbCr:
+ z.scaleX_YCbCr(tmp, src, sp)
+ default:
+ z.scaleX_Image(tmp, src, sp)
+ }
}
+
switch dst := dst.(type) {
case *image.RGBA:
z.scaleY_RGBA(dst, dp, dr, tmp)
@@ -656,7 +699,11 @@
for _, s := range z.horizontal.sources {
var pr, pg, pb, pa float64
for _, c := range z.horizontal.contribs[s.i:s.j] {
- pru, pgu, pbu, pau := src.At(sp.X+int(c.coord), sp.Y+int(y)).RGBA()
+ pi := src.PixOffset(sp.X+int(c.coord), sp.Y+int(y))
+ pru := uint32(src.Pix[pi+0]) * 0x101
+ pgu := uint32(src.Pix[pi+1]) * 0x101
+ pbu := uint32(src.Pix[pi+2]) * 0x101
+ pau := uint32(src.Pix[pi+3]) * 0x101
pr += float64(pru) * c.weight
pg += float64(pgu) * c.weight
pb += float64(pbu) * c.weight
