Re: performance issue with cgywin make

[Jon Grant]

Hi Bill,

Bill Hoffman elucidated on 07/12/06 22:11:
[...]
> Can someone give me a pointer to where
> I could put different calls into make just so I can try it and see if it
> helps.  Thanks.

I took a quick look, didn't have a project to hand so I ran Make without
a Makefile.

dir_contents_file_exists_p () is the main call can see running Make
without a makefile, it got 287 calls. See (3) below for the line you
could try implementing an alternative w32 version for.

1) dirent.c:40
2) read.c:2901
3) dir.c:649  stat(dir->path_key, &st) == 0 && st.st_mtime > dir->mtime)
4) remake.c:1361  EINTRLOOP (e, stat (name, &st));

__stat64i32 is the resulting Clib function, it was called 287 times,
taking up around 11% of overall running time.

Would be interesting to see the results of your comparison between
stat() and the Win32 call. You might already have a profiler, but if you
don't you might like to try run the build through GlowCode (has a free
evaluation), gives % of processing and other useful stats like a
callstack etc. I'll see if I've got a Makefile around on my Windows
machine I can run it through GlowCode with.

Cheers
Jon