Commit d205438

committed

unicode/utf8: make DecodeRune{,InString} inlineable

This change makes the fast path for ASCII characters inlineable in DecodeRune and DecodeRuneInString. Here are some benchmark results (no change to allocations): goos: darwin goarch: amd64 pkg: unicode/utf8 cpu: Intel(R) Core(TM) i7-6700HQ CPU @ 2.60GHz │ old │ new │ │ sec/op │ sec/op vs base │ DecodeASCIIRune-8 2.4585n ± 1% 0.6187n ± 1% -74.83% (p=0.000 n=20) DecodeJapaneseRune-8 3.987n ± 1% 3.984n ± 1% ~ (p=0.941 n=20) DecodeASCIIRuneInString-8 2.2945n ± 1% 0.3123n ± 2% -86.39% (p=0.000 n=20) DecodeJapaneseRuneInString-8 3.821n ± 1% 3.983n ± 1% +4.25% (p=0.000 n=20) geomean 3.045n 1.323n -56.53% Note: when #61502 gets resolved, this change should likely be reverted. Updates #48195

1 parent d029eaa commit d205438Copy full SHA for d205438

File tree

2 files changed

+28

-0

lines changed

src
- cmd/compile/internal/test
  - inl_test.go
- unicode/utf8
  - utf8.go

2 files changed

+28

-0

lines changed

`‎src/cmd/compile/internal/test/inl_test.go`

Lines changed: 2 additions & 0 deletions

Original file line number	Diff line number	Diff line change
`@@ -125,6 +125,8 @@ func TestIntendedInlining(t *testing.T) {`
`125`	`125`	`"assemble64",`
`126`	`126`	`},`
`127`	`127`	`"unicode/utf8": {`
	`128`	`+ "DecodeRune",`
	`129`	`+ "DecodeRuneInString",`
`128`	`130`	`"FullRune",`
`129`	`131`	`"FullRuneInString",`
`130`	`132`	`"RuneLen",`

`‎src/unicode/utf8/utf8.go`

Lines changed: 26 additions & 0 deletions

Original file line number	Diff line number	Diff line change
`@@ -155,6 +155,20 @@ func FullRuneInString(s string) bool {`
`155`	`155`	`// out of range, or is not the shortest possible UTF-8 encoding for the`
`156`	`156`	`// value. No other validation is performed.`
`157`	`157`	`func DecodeRune(p []byte) (r rune, size int) {`
	`158`	`+ // Inlineable fast path for ASCII characters; see #48195.`
	`159`	`+ // This implementation is weird but effective at rendering the`
	`160`	`+ // function inlineable.`
	`161`	`+ for _, b := range p {`
	`162`	`+ if b < RuneSelf {`
	`163`	`+ return rune(b), 1`
	`164`	`+ }`
	`165`	`+ break`
	`166`	`+ }`
	`167`	`+ r, size = decodeRune(p)`
	`168`	`+ return`
	`169`	`+}`
	`170`	`+`
	`171`	`+func decodeRune(p []byte) (r rune, size int) {`
`158`	`172`	`n := len(p)`
`159`	`173`	`if n < 1 {`
`160`	`174`	`return RuneError, 0`
`@@ -203,6 +217,18 @@ func DecodeRune(p []byte) (r rune, size int) {`
`203`	`217`	`// out of range, or is not the shortest possible UTF-8 encoding for the`
`204`	`218`	`// value. No other validation is performed.`
`205`	`219`	`func DecodeRuneInString(s string) (r rune, size int) {`
	`220`	`+ // Inlineable fast path for ASCII characters; see #48195.`
	`221`	`+ // This implementation is a bit weird but effective at rendering the`
	`222`	`+ // function inlineable.`
	`223`	`+ if s != "" && s[0] < RuneSelf {`
	`224`	`+ return rune(s[0]), 1`
	`225`	`+ } else {`
	`226`	`+ r, size = decodeRuneInString(s)`
	`227`	`+ }`
	`228`	`+ return`
	`229`	`+}`
	`230`	`+`
	`231`	`+func decodeRuneInString(s string) (rune, int) {`
`206`	`232`	`n := len(s)`
`207`	`233`	`if n < 1 {`
`208`	`234`	`return RuneError, 0`

0 commit comments

Comments

(0)

Navigation Menu

Search code, repositories, users, issues, pull requests...

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Commit d205438

File tree

2 files changed

2 files changed

`‎src/cmd/compile/internal/test/inl_test.go`

`‎src/unicode/utf8/utf8.go`

0 commit comments