image/draw: remove some bounds checks from DrawYCbCr

It’d be nicer to write just

_ = dpix[x+3]

but the compiler isn’t able to reason about offsets
from symbols (yet).

image/draw benchmark:

YCbCr-8   722µs ± 3%   682µs ± 3%  -5.54%  (p=0.000 n=50+50)

Change-Id: Ia1e399496ed87c282bf0f9ca56c0b2d4948a0df9
Reviewed-on: https://go-review.googlesource.com/22146
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

src/image/internal/imageutil/gen.go

@@ -133,10 +133,13 @@ const sratioCase = `
 					b = ^(b >> 31)
 				}
 
-				dpix[x+0] = uint8(r)
-				dpix[x+1] = uint8(g)
-				dpix[x+2] = uint8(b)
-				dpix[x+3] = 255
+
+				// use a temp slice to hint to the compiler that a single bounds check suffices
+				rgba := dpix[x : x+4 : len(dpix)]
+				rgba[0] = uint8(r)
+				rgba[1] = uint8(g)
+				rgba[2] = uint8(b)
+				rgba[3] = 255
 			}
 		}
 `

src/image/internal/imageutil/impl.go

@@ -82,10 +82,12 @@ func DrawYCbCr(dst *image.RGBA, r image.Rectangle, src *image.YCbCr, sp image.Po
 					b = ^(b >> 31)
 				}
 
-				dpix[x+0] = uint8(r)
-				dpix[x+1] = uint8(g)
-				dpix[x+2] = uint8(b)
-				dpix[x+3] = 255
+				// use a temp slice to hint to the compiler that a single bounds check suffices
+				rgba := dpix[x : x+4 : len(dpix)]
+				rgba[0] = uint8(r)
+				rgba[1] = uint8(g)
+				rgba[2] = uint8(b)
+				rgba[3] = 255
 			}
 		}
 
@@ -137,10 +139,12 @@ func DrawYCbCr(dst *image.RGBA, r image.Rectangle, src *image.YCbCr, sp image.Po
 					b = ^(b >> 31)
 				}
 
-				dpix[x+0] = uint8(r)
-				dpix[x+1] = uint8(g)
-				dpix[x+2] = uint8(b)
-				dpix[x+3] = 255
+				// use a temp slice to hint to the compiler that a single bounds check suffices
+				rgba := dpix[x : x+4 : len(dpix)]
+				rgba[0] = uint8(r)
+				rgba[1] = uint8(g)
+				rgba[2] = uint8(b)
+				rgba[3] = 255
 			}
 		}
 
@@ -192,10 +196,12 @@ func DrawYCbCr(dst *image.RGBA, r image.Rectangle, src *image.YCbCr, sp image.Po
 					b = ^(b >> 31)
 				}
 
-				dpix[x+0] = uint8(r)
-				dpix[x+1] = uint8(g)
-				dpix[x+2] = uint8(b)
-				dpix[x+3] = 255
+				// use a temp slice to hint to the compiler that a single bounds check suffices
+				rgba := dpix[x : x+4 : len(dpix)]
+				rgba[0] = uint8(r)
+				rgba[1] = uint8(g)
+				rgba[2] = uint8(b)
+				rgba[3] = 255
 			}
 		}
 
@@ -246,10 +252,12 @@ func DrawYCbCr(dst *image.RGBA, r image.Rectangle, src *image.YCbCr, sp image.Po
 					b = ^(b >> 31)
 				}
 
-				dpix[x+0] = uint8(r)
-				dpix[x+1] = uint8(g)
-				dpix[x+2] = uint8(b)
-				dpix[x+3] = 255
+				// use a temp slice to hint to the compiler that a single bounds check suffices
+				rgba := dpix[x : x+4 : len(dpix)]
+				rgba[0] = uint8(r)
+				rgba[1] = uint8(g)
+				rgba[2] = uint8(b)
+				rgba[3] = 255
 			}
 		}