Runtime for "make check-fortran" on Cygwin? What's typical?
Jerry DeLisle
jvdelisle@verizon.net
Sat Nov 4 21:10:00 GMT 2006
Janne Blomqvist wrote:
> Brooks Moses wrote:
>> So, I finally got dejagnu worked out to run properly on my Cygwin
>> installation. (The relevant incantation seemed to be "setenv
>> DEJAGNULIBS /usr/share/dejagnu", to fix an error of "couldn't find
>> runtext.exp") And so I did a run of "make check-fortran", to make
>> sure it was working, and get a baseline before making changes.
>>
>> Eight hours later, I think it's almost done. Most of the time, it
>> hasn't been using the CPU much at all, but the disk is working away.
>>
>> Is this normal for Cygwin? How long does a "make check-fortran" run
>> usually take? What about on non-Cygwin machines?
>
> On i686-pc-linux-gnu, with my 1.8 GHz Athlon 64, make check-fortran
> takes about 20 minutes. With the cpu pegged, moderate io load.
>
> I've heard that cygwin is slow, but 8 hours vs. 20 minutes, wow.
>
Yes, I have seen this too and was very surprised being use to checking in a few
minutes. It is my understanding that in the shell, windows does not handle
forking well, so an entire copy of the shell environment has to be made for each
fork. There ought to be some sort of magic bullet to fix this. I am sure
others have given it a lot of thought already.
More information about the Fortran
mailing list