mirror of
https://github.com/ddidderr/wuffs.git
synced 2026-11-17 16:32:11 +01:00
The LICENSE has changed from a single license (Apache2) to a dual license (Apache2 or MIT, at your option). If you've previously used Wuffs under the Apache2 license, you are able to keep doing so. No further action required by you. The two files listed below (previous releases) are exceptions, as they are older snapshots, frozen in time. Their license has not changed: release/c/wuffs-v0.2.c release/c/wuffs-v0.3.c Every name in the AUTHORS file, other than Google Inc., has signed Google's CLA (Contributor License Agreement).
225 lines
5.7 KiB
Go
225 lines
5.7 KiB
Go
// Copyright 2021 The Wuffs Authors.
|
|
//
|
|
// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
|
|
// https://www.apache.org/licenses/LICENSE-2.0> or the MIT license
|
|
// <LICENSE-MIT or https://opensource.org/licenses/MIT>, at your
|
|
// option. This file may not be copied, modified, or distributed
|
|
// except according to those terms.
|
|
|
|
//go:build ignore
|
|
// +build ignore
|
|
|
|
package main
|
|
|
|
// benchstat-ratio.go prints the relative ratios of similarly named benchstat
|
|
// summary lines. The benchstat tool (written in Go) is not specific to Wuffs.
|
|
// It is documented at https://godoc.org/golang.org/x/perf/cmd/benchstat
|
|
//
|
|
// Out of the box, benchstat (when passed two filename arguments) compares
|
|
// *identically* named lines (i.e. before and after a code change). This
|
|
// program computes ratios for *different-but-similar* named lines (i.e.
|
|
// multiple implementations or compilers for the same algorithm or test data),
|
|
// grouped by the name fragment after the first '_' and before the first '/'.
|
|
//
|
|
// By default, the baseline case (normalized to a ratio of 1.00x) is the one
|
|
// whose name starts with "mimic_" and contains "/gcc".
|
|
//
|
|
// For example, given benchstat output like:
|
|
//
|
|
// wuffs_crc32_ieee_10k/clang9 8.32GB/s ± 4%
|
|
// wuffs_crc32_ieee_100k/clang9 10.8GB/s ± 1%
|
|
// mimic_crc32_ieee_10k/clang9 771MB/s ± 0%
|
|
// mimic_crc32_ieee_100k/clang9 772MB/s ± 0%
|
|
// wuffs_crc32_ieee_10k/gcc10 8.68GB/s ± 0%
|
|
// wuffs_crc32_ieee_100k/gcc10 10.9GB/s ± 0%
|
|
// mimic_crc32_ieee_10k/gcc10 775MB/s ± 0%
|
|
// mimic_crc32_ieee_100k/gcc10 773MB/s ± 0%
|
|
//
|
|
// Piping that through this program prints:
|
|
//
|
|
// wuffs_crc32_ieee_10k/clang9 8.32GB/s ± 4% 10.74x
|
|
// wuffs_crc32_ieee_100k/clang9 10.8GB/s ± 1% 13.97x
|
|
// mimic_crc32_ieee_10k/clang9 771MB/s ± 0% 0.99x
|
|
// mimic_crc32_ieee_100k/clang9 772MB/s ± 0% 1.00x
|
|
// wuffs_crc32_ieee_10k/gcc10 8.68GB/s ± 0% 11.20x
|
|
// wuffs_crc32_ieee_100k/gcc10 10.9GB/s ± 0% 14.10x
|
|
// mimic_crc32_ieee_10k/gcc10 775MB/s ± 0% 1.00x
|
|
// mimic_crc32_ieee_100k/gcc10 773MB/s ± 0% 1.00x
|
|
//
|
|
// There are two groups here: "crc32_ieee_10k" and "crc32_ieee_100k".
|
|
//
|
|
// Usage: benchstat etc | go run benchstat-ratio.go
|
|
|
|
import (
|
|
"bufio"
|
|
"flag"
|
|
"fmt"
|
|
"os"
|
|
"strconv"
|
|
"strings"
|
|
)
|
|
|
|
var (
|
|
baseline = flag.String("baseline", "mimic", "benchmark name prefix of the 1.00x case")
|
|
)
|
|
|
|
func main() {
|
|
if err := main1(); err != nil {
|
|
os.Stderr.WriteString(err.Error() + "\n")
|
|
os.Exit(1)
|
|
}
|
|
}
|
|
|
|
func main1() error {
|
|
flag.Parse()
|
|
baselineUnderscore = *baseline + "_"
|
|
|
|
r := bufio.NewScanner(os.Stdin)
|
|
for r.Scan() {
|
|
line := strings.TrimSpace(string(r.Bytes()))
|
|
pmIndex := strings.Index(line, plusMinus)
|
|
if pmIndex < 0 {
|
|
// Collapse multiple blank lines to a single one.
|
|
if (len(entries) > 0) && (entries[len(entries)-1].key != "") {
|
|
entries = append(entries, entry{})
|
|
}
|
|
continue
|
|
}
|
|
|
|
// Find the "/gcc10" in "mimic_etc/gcc10".
|
|
if (gccSuffix == "") && strings.HasPrefix(line, baselineUnderscore) {
|
|
if i := strings.Index(line, slashGCC); i >= 0 {
|
|
gccSuffix = line[i:]
|
|
if j := strings.IndexByte(gccSuffix, ' '); j >= 0 {
|
|
gccSuffix = gccSuffix[:j]
|
|
}
|
|
}
|
|
}
|
|
|
|
key, val := parse(line[:pmIndex])
|
|
if key == "" {
|
|
return fmt.Errorf("could not parse %q", line)
|
|
} else if _, ok := values[key]; ok {
|
|
return fmt.Errorf("duplicate key for %q", line)
|
|
}
|
|
entries = append(entries, entry{key, line})
|
|
values[key] = val
|
|
}
|
|
if err := r.Err(); err != nil {
|
|
return err
|
|
}
|
|
|
|
for _, e := range entries {
|
|
if e.key == "" {
|
|
fmt.Println()
|
|
continue
|
|
}
|
|
eVal, eOK := values[e.key]
|
|
bVal, bOK := values[baselineKeyFor(e.key)]
|
|
if eOK && bOK {
|
|
switch e.key[0] {
|
|
case '1':
|
|
fmt.Printf("%s %6.2fx\n", e.line, bVal/eVal)
|
|
continue
|
|
case '2':
|
|
fmt.Printf("%s %6.2fx\n", e.line, eVal/bVal)
|
|
continue
|
|
}
|
|
}
|
|
fmt.Printf("%s\n", e.line)
|
|
}
|
|
|
|
return nil
|
|
}
|
|
|
|
func parse(line string) (key string, val float64) {
|
|
i := strings.IndexByte(line, ' ')
|
|
if i < 0 {
|
|
return "", 0
|
|
}
|
|
k, v := line[:i], strings.TrimSpace(line[i+1:])
|
|
|
|
prefix, multiplier := "", 1e0
|
|
switch {
|
|
case strings.HasSuffix(v, suffix_GBs):
|
|
prefix, multiplier = prefix2, 1e9
|
|
v = v[:len(v)-len(suffix_GBs)]
|
|
case strings.HasSuffix(v, suffix_MBs):
|
|
prefix, multiplier = prefix2, 1e6
|
|
v = v[:len(v)-len(suffix_MBs)]
|
|
case strings.HasSuffix(v, suffix_KBs):
|
|
prefix, multiplier = prefix2, 1e3
|
|
v = v[:len(v)-len(suffix_KBs)]
|
|
case strings.HasSuffix(v, suffix__Bs):
|
|
prefix, multiplier = prefix2, 1e0
|
|
v = v[:len(v)-len(suffix__Bs)]
|
|
|
|
case strings.HasSuffix(v, suffix_ns):
|
|
prefix, multiplier = prefix1, 1e0
|
|
v = v[:len(v)-len(suffix_ns)]
|
|
case strings.HasSuffix(v, suffix_µs):
|
|
prefix, multiplier = prefix1, 1e3
|
|
v = v[:len(v)-len(suffix_µs)]
|
|
case strings.HasSuffix(v, suffix_ms):
|
|
prefix, multiplier = prefix1, 1e6
|
|
v = v[:len(v)-len(suffix_ms)]
|
|
case strings.HasSuffix(v, suffix__s):
|
|
prefix, multiplier = prefix1, 1e9
|
|
v = v[:len(v)-len(suffix__s)]
|
|
default:
|
|
return "", 0
|
|
}
|
|
|
|
x, err := strconv.ParseFloat(v, 64)
|
|
if err != nil {
|
|
return "", 0
|
|
}
|
|
return prefix + string(k), x * multiplier
|
|
}
|
|
|
|
func baselineKeyFor(key string) (bKey string) {
|
|
if key == "" {
|
|
return ""
|
|
}
|
|
i := strings.IndexByte(key, '_')
|
|
if i < 0 {
|
|
return ""
|
|
}
|
|
j := strings.IndexByte(key[i:], '/')
|
|
if j < 0 {
|
|
j = len(key) - i
|
|
}
|
|
return key[:1] + *baseline + key[i:i+j] + gccSuffix
|
|
}
|
|
|
|
type entry struct {
|
|
key string
|
|
line string
|
|
}
|
|
|
|
var (
|
|
entries []entry
|
|
values = map[string]float64{}
|
|
|
|
baselineUnderscore string
|
|
gccSuffix string
|
|
)
|
|
|
|
const (
|
|
plusMinus = " ± "
|
|
slashGCC = "/gcc"
|
|
|
|
prefix1 = "1" // val is nanoseconds.
|
|
prefix2 = "2" // val is bytes per second.
|
|
|
|
suffix_GBs = "GB/s"
|
|
suffix_MBs = "MB/s"
|
|
suffix_KBs = "KB/s"
|
|
suffix__Bs = "B/s"
|
|
|
|
suffix_ns = "ns"
|
|
suffix_µs = "µs"
|
|
suffix_ms = "ms"
|
|
suffix__s = "s"
|
|
)
|