R: Cygwin 1.7.7 fork/exec performance MUCH slower than 1.5.25
Thu Sep 30 13:53:00 GMT 2010
--- Gio 30/9/10, SZABÓ Gergely ha scritto:
> we have a Cygwin-based build system for an embedded
> Build performance has deteriorated terribly since we
> upgraded to 1.7.x from 1.5.25.
> I've created 3 shell-scripts to benchmark 1.5.25 and 1.7.7
> against each other.
> - null : read file "inp" line by line, write
> to /dev/null.
> - builtin : write each line to file "out", only
> bash-builtins used (no fork)
> - fork : do some substitution for each line
> with sed (huge number of forks)
> - null : 1.7.7 is about 5 times faster
> - builtin : 1.7.7 is slightly faster
> - fork : 1.7.7 is 5 times SLOWER !!!
> All measurements were made with the "time" command.
> The attached results.txt lists the "real" times.
> I've also attached the 3 shell-scripts (null, builtin,
> Input file "inp" is also attached.
> Also see the cygcheck_xxx.out files for both Cygwin
> Some notes:
> both cygwin versions use the same /home, mounted from
> "D:\Documents and Settings",
I bet is a network/acl issues.
I suggest to create a real home on "/home"
> so all user-level config-files are shared between the
> Cygwin versions.
> Cygwin 1.7.7 is basically unusable at the moment.
> Any ideas why it's so much slower at spawning processes? Or
> is it the pipes? Am I doing something wrong?
> And more importantly: is there an easy way to fix this?
> Thanks in advance for any help!
> Best regards
> Gergely Szabó
testing on /tmp/benchmark with XPS P2
cygwin 1.7.8s 20100924
$ time ./fork
$ time ./fork
So no difference
Problem reports: http://cygwin.com/problems.html
Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple
More information about the Cygwin