Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

strconv: ParseInt, ParseFloat use two loops instead of one #33330

Open
sarnold64 opened this issue Jul 28, 2019 · 1 comment

Comments

@sarnold64
Copy link

commented Jul 28, 2019

What version of Go are you using (go version)?

$ go version
go version devel +919594830f Tue Jul 23 23:42:43 2019 +0000 darwin/amd64

Does this issue reproduce with the latest release?

Yes, see version

What operating system and processor architecture are you using (go env)?

go env Output
$ go env
GO111MODULE=""
GOARCH="amd64"
GOBIN=""
GOCACHE="/Users/sarnold64/Library/Caches/go-build"
GOENV="/Users/sarnold64/Library/Application Support/go/env"
GOEXE=""
GOFLAGS=""
GOHOSTARCH="amd64"
GOHOSTOS="darwin"
GONOPROXY=""
GONOSUMDB=""
GOOS="darwin"
GOPATH="/Users/sarnold64/go"
GOPRIVATE=""
GOPROXY="https://proxy.golang.org,direct"
GOROOT="/Users/sarnold64/golang/go"
GOSUMDB="sum.golang.org"
GOTMPDIR=""
GOTOOLDIR="/Users/sarnold64/golang/go/pkg/tool/darwin_amd64"
GCCGO="gccgo"
AR="ar"
CC="clang"
CXX="clang++"
CGO_ENABLED="1"
GOMOD="/Users/sarnold64/golang/go/src/go.mod"
CGO_CFLAGS="-g -O2"
CGO_CPPFLAGS=""
CGO_CXXFLAGS="-g -O2"
CGO_FFLAGS="-g -O2"
CGO_LDFLAGS="-g -O2"
PKG_CONFIG="pkg-config"
GOGCCFLAGS="-fPIC -m64 -pthread -fno-caret-diagnostics -Qunused-arguments -fmessage-length=0 -fdebug-prefix-map=/var/folders/5s/zkkc95rs08731f9t6df3fvsc0000gr/T/go-build178433439=/tmp/go-build -gno-record-gcc-switches -fno-common"

What did you do?

I noticed that strconv takes more time than it needs to for ParseInt and ParseFloat because it loops over the input twice, once in underscoreOK and once to do the parsing.

What did you expect to see?

One loop

What did you see instead?

Two loops

I was playing around with pprof after the gophercon talks and saw this looking suspicious.

Some performance results on my laptop after fixing:

name                   old time/op  new time/op  delta
ParseInt/Pos/7bit-12   19.0ns ± 3%  12.1ns ± 0%  -36.58%  (p=0.000 n=6+5)
ParseInt/Pos/26bit-12  28.5ns ± 0%  18.4ns ± 2%  -35.58%  (p=0.002 n=6+6)
ParseInt/Pos/31bit-12  32.4ns ± 1%  20.9ns ± 1%  -35.71%  (p=0.002 n=6+6)
ParseInt/Pos/56bit-12  47.0ns ± 2%  29.3ns ± 1%  -37.70%  (p=0.002 n=6+6)
ParseInt/Pos/63bit-12  50.9ns ± 2%  32.0ns ± 1%  -37.24%  (p=0.002 n=6+6)
ParseInt/Neg/7bit-12   18.3ns ± 6%  12.2ns ± 1%  -33.18%  (p=0.002 n=6+6)
ParseInt/Neg/26bit-12  28.2ns ± 3%  18.2ns ± 2%  -35.38%  (p=0.002 n=6+6)
ParseInt/Neg/31bit-12  31.4ns ± 1%  20.7ns ± 1%  -34.04%  (p=0.004 n=6+5)
ParseInt/Neg/56bit-12  45.4ns ± 0%  29.4ns ± 1%  -35.17%  (p=0.004 n=5+6)
ParseInt/Neg/63bit-12  49.4ns ± 1%  32.0ns ± 1%  -35.30%  (p=0.002 n=6+6)

name                   old time/op  new time/op  delta
Atof64Decimal-12       30.1ns ± 1%  25.4ns ± 1%  -15.75%  (p=0.002 n=6+6)
Atof64Float-12         38.3ns ± 1%  31.5ns ± 0%  -17.84%  (p=0.000 n=6+5)
Atof64FloatExp-12      63.1ns ± 0%  53.9ns ± 0%  -14.53%  (p=0.002 n=6+6)
Atof64Big-12            113ns ± 0%    96ns ± 2%  -14.85%  (p=0.010 n=4+6)
Atof64RandomBits-12     180ns ±10%   160ns ±16%  -11.36%  (p=0.028 n=6+6)
Atof64RandomFloats-12   123ns ± 2%   102ns ± 1%  -17.03%  (p=0.002 n=6+6)
Atof32Decimal-12       31.4ns ± 0%  25.4ns ± 1%  -18.86%  (p=0.002 n=6+6)
Atof32Float-12         36.5ns ± 0%  29.4ns ± 0%  -19.40%  (p=0.008 n=5+5)
Atof32FloatExp-12      67.0ns ± 0%  56.6ns ± 0%  -15.58%  (p=0.008 n=5+5)
Atof32Random-12        90.6ns ± 1%  75.1ns ± 0%  -17.15%  (p=0.004 n=6+5)

@ALTree ALTree added the Performance label Jul 28, 2019

@ALTree ALTree added this to the Go1.14 milestone Jul 28, 2019

@gopherbot

This comment has been minimized.

Copy link

commented Jul 29, 2019

Change https://golang.org/cl/187957 mentions this issue: strconv: Speed improvement to number parsing

@golang golang deleted a comment from yangzhaoyu520418 Jul 29, 2019

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
3 participants
You can’t perform that action at this time.