From trevor@grserv.wustl.edu  Wed Jul 21 09:46:36 1993
Date: Wed, 21 Jul 93 15:46:36 -0600
From: trevor@grserv.wustl.edu (Trevor Creamer)
Message-Id: <9307212146.AA22282@giovanni>
To: chemistry@ccl.net
Subject: SGI compiler options



Bruce Luxon wrote :
>
>Dan Severance recently wrote regarding SGI machines -
>
>>If you have an R4000 or R4400 machine, you need to add the "-mips2" 
>>option, since "-mips1" is the default (compatibility mode, will run on
>>both R3000 and R4000 machines).  (The fast square-root is incorporated
>>when you specify -mips2) ...
>>
>>   Another compiler option (fairly new) to try is the -sopt, and is
>>documented in the f77 man page, and under the "fopt" manpage (-sopt
>>runs the program fopt).  One user got a 40% speedup, we don't; it
>>depends on your code.
>
>I and several other users attempted this and were not able to get either 
>of these options to work. I have an Indigo XS24 R4000 with the F77 compiler
>3.4.1, Development option 4.0.1, and Irix 4.0.5 so I believe I have most
>(but apparently not all) the goodies.
>
>When trying to make using the -mips2 option I received the following note -
>
>	f77: Error: malformed or unknown option -mips2
>	*** Error code 2

This worked fine for me with the same set-up. Are you using IRIX 4.0.5F ?
Or just plain old 4.0.5 ?

>
>I checked my cdroms to see if I had all the options installed and it 'appears'
>I do.  What am I doing wrong or what isn't installed? 
> 
>Regarding -sopt, when this option was included in the makefile the compiler
>seemed to go into hibernation after it encountered the first routine to 
>compile.  After awhile (ca. 10 minutes) I lost patience and aborted the make.
>Again, any suggestions what went wrong or I didn't do correctly?

The -sopt option seems to require LOTS of memory - when I did this my
machine seemed to spend a lot of time swapping stuff in and out of the
swap space. Try watching gr_osview while compiling - it can be educational.

>
>Thanks in advance to Dan and any others with responses,
>
>Bruce

I tried these options on my Monte Carlo program. Normally I use the -O4
optimization flag with f77. When I added -mips2 I got a 1.15x speed-up.
With -sopt as well I got no appreciable difference (other than a much
longer compilation time; ~15min.s with -sopt vs. ~1.5 min.s without).

Cheers,
       Trevor

Trevor P. Creamer                          Dept. of Biochemistry
Email  - trevor@grserv.wustl.edu   ,-_|\   & Molecular Biophysics
Phone  - (314) 362 4123           /     \  Washington University
Fax    - (314) 362 4115           *_,-._/  St. Louis, Missouri 63110
                                       v   U.S.A.