archive/zip: make benchmarks more representative

Currently zip benchmarks spend 60% in the rleBuffer code,
which is used only to test zip archive/zip itself:
    17.48s 37.02% 37.02%     18.12s 38.37%  archive/zip.(*rleBuffer).ReadAt
     9.51s 20.14% 57.16%     10.43s 22.09%  archive/zip.(*rleBuffer).Write
     9.15s 19.38% 76.54%     10.85s 22.98%  compress/flate.(*compressor).deflate

This means that benchmarks currently test performance of test helper.
Updating ReadAt/Write methods to be more performant makes benchmarks closer to real world.

name                       old time/op    new time/op    delta
CompressedZipGarbage-8       2.34ms ± 0%    2.34ms ± 1%     ~     (p=0.684 n=10+10)
Zip64Test-8                  58.1ms ± 2%    10.7ms ± 1%  -81.54%  (p=0.000 n=10+10)
Zip64TestSizes/4096-8        4.05µs ± 2%    3.65µs ± 5%   -9.96%  (p=0.000 n=9+10)
Zip64TestSizes/1048576-8      238µs ± 0%      43µs ± 0%  -82.06%  (p=0.000 n=10+10)
Zip64TestSizes/67108864-8    15.3ms ± 1%     2.6ms ± 0%  -83.12%  (p=0.000 n=10+9)

name                       old alloc/op   new alloc/op   delta
CompressedZipGarbage-8       17.9kB ±14%    16.0kB ±24%  -10.48%  (p=0.026 n=9+10)

name                       old allocs/op  new allocs/op  delta
CompressedZipGarbage-8         44.0 ± 0%      44.0 ± 0%     ~     (all equal)

Change-Id: Idfd920d0e4bed4aec2f5be84dc7e3919d9f1dd2d
Reviewed-on: https://go-review.googlesource.com/83857
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

src/archive/zip/zip_test.go

@@ -140,14 +140,7 @@ func (r *rleBuffer) Write(p []byte) (n int, err error) {
 		rp = &r.buf[len(r.buf)-1]
 		// Fast path, if p is entirely the same byte repeated.
 		if lastByte := rp.b; len(p) > 0 && p[0] == lastByte {
-			all := true
+			if bytes.Count(p, []byte{lastByte}) == len(p) {
-			for _, b := range p {
-				if b != lastByte {
-					all = false
-					break
-				}
-			}
-			if all {
 				rp.n += int64(len(p))
 				return len(p), nil
 			}
@@ -165,6 +158,25 @@ func (r *rleBuffer) Write(p []byte) (n int, err error) {
 	return len(p), nil
 }
 
+func min(x, y int) int {
+	if x < y {
+		return x
+	}
+	return y
+}
+
+func memset(a []byte, b byte) {
+	if len(a) == 0 {
+		return
+	}
+	// Double, until we reach power of 2 >= len(a), same as bytes.Repeat,
+	// but without allocation.
+	a[0] = b
+	for i, l := 1, len(a); i < l; i *= 2 {
+		copy(a[i:], a[:i])
+	}
+}
+
 func (r *rleBuffer) ReadAt(p []byte, off int64) (n int, err error) {
 	if len(p) == 0 {
 		return
@@ -176,16 +188,13 @@ func (r *rleBuffer) ReadAt(p []byte, off int64) (n int, err error) {
 	parts := r.buf[skipParts:]
 	if len(parts) > 0 {
 		skipBytes := off - parts[0].off
-		for len(parts) > 0 {
+		for _, part := range parts {
-			part := parts[0]
+			repeat := min(int(part.n-skipBytes), len(p)-n)
-			for i := skipBytes; i < part.n; i++ {
+			memset(p[n:n+repeat], part.b)
-				if n == len(p) {
+			n += repeat
-					return
+			if n == len(p) {
-				}
+				return
-				p[n] = part.b
-				n++
 			}
-			parts = parts[1:]
 			skipBytes = 0
 		}
 	}