Optimize zero reads

Use optimized range loop[1], optimized by the compiler to single memclr
call. This dramatically speeds up zero reads.

| format | compression | utilization | speedup |
|--------|-------------|-------------|---------|
| qcow2  | -           |          0% |   28.72 |
| qcow2  | zlib        |          0% |   28.04 |
| qcow2  | -           |         50% |    4.54 |
| qcow2  | zlib        |         50% |    1.03 |
| qcow2  | -           |        100% |    1.01 |
| qcow2  | zlib        |        100% |    1.00 |

Before:

    % go test -bench Read
    BenchmarkRead0p/qcow2-12           14      77515735 ns/op     3462.98 MB/s      1050518 B/op        39 allocs/op
    BenchmarkRead0p/qcow2_zlib-12      14      77823402 ns/op     3449.29 MB/s      1050504 B/op        39 allocs/op
    BenchmarkRead50p/qcow2-12          24      48812158 ns/op     5499.36 MB/s      1181856 B/op        45 allocs/op
    BenchmarkRead50p/qcow2_zlib-12      2     899659187 ns/op      298.37 MB/s    184996316 B/op     43247 allocs/op
    BenchmarkRead100p/qcow2-12         61      19306020 ns/op    13904.24 MB/s      1181854 B/op        45 allocs/op
    BenchmarkRead100p/qcow2_zlib-12     1    1732168542 ns/op      154.97 MB/s    368850952 B/op     86460 allocs/op

After:

    % go test -bench Read
    BenchmarkRead0p/qcow2-12          471       2698377 ns/op    99480.34 MB/s      1050514 B/op        39 allocs/op
    BenchmarkRead0p/qcow2_zlib-12     468       2774952 ns/op    96735.15 MB/s      1050511 B/op        39 allocs/op
    BenchmarkRead50p/qcow2-12         100      10735870 ns/op    25003.61 MB/s      1181854 B/op        45 allocs/op
    BenchmarkRead50p/qcow2_zlib-12      2     868310583 ns/op      309.15 MB/s    185038456 B/op     43263 allocs/op
    BenchmarkRead100p/qcow2-12         63      18977718 ns/op    14144.77 MB/s      1181851 B/op        45 allocs/op
    BenchmarkRead100p/qcow2_zlib-12     1    1727832917 ns/op      155.36 MB/s    368886656 B/op     86471 allocs/op

Comparing with qemu-img show that we match qemu-img performance for
uncompressed version of the lima default image:

    % time ./go-qcow2reader-example /tmp/test.qcow2 > /tmp/tmp.img
    ./go-qcow2reader-example /tmp/test.qcow2 > /tmp/tmp.img  0.06s user 0.73s system 93% cpu 0.854 total

    % time qemu-img convert -O raw /tmp/test.qcow2 /tmp/tmp.img
    qemu-img convert -O raw /tmp/test.qcow2 /tmp/tmp.img  0.04s user 0.70s system 98% cpu 0.756 total

[1] https://go-review.googlesource.com/c/go/+/2520

Signed-off-by: Nir Soffer <[email protected]>

image/qcow2/qcow2.go

@@ -902,10 +902,15 @@ func readZero(p []byte, off int64, sz uint64) (int, error) {
 			l = 0
 		}
 		err = io.EOF
+		p = p[:l]
 	}
-	for i := 0; i < l; i++ {
+
+	// Optimized by the compiler to memclr call.
+	// https://go-review.googlesource.com/c/go/+/2520
+	for i := range p {
 		p[i] = 0
 	}
+
 	return l, err
 }