[daip] Compiling AIPS with Intel Compilers

[Sally Laurent-Muehleisen]

Hi Eric and Pat,

	Well, we've completed our work on installing AIPS on linux using
Intel's (v. 5.0) compilers.  Our AIPSMark is 89.1 (7m 29s runtime for the
large y2k).  Unless I'm mistaken, that's the winner of "Highest AIPSMark"!
The machine is a 2 GHz dual processor Pentium 4 with 2 GB of memory.  We have
2 70 GB data disks as well as the 9 GB system disk (on which the AIPS
executables lie).  When I ran the y2k's, I had the data on one of the 70 GB
disks, so I might have gotten a higher AIPSMark if I'd put the data on the
same disk as the executables, but we'll usually be running by accessing data
on the 70 GB disks, so I figured it was just as well that I benchmarked it
this way.

	Here are the things we did in the course of installing AIPS (I *think*
I have everything here...)


We put the INTEL compilers in /opt/intel/


Set the following environmental variables:
setenv INTEL_FLEXLM_LICENSE /opt/intel/licenses
setenv IA32ROOT /opt/intel/compiler50/ia32


Added the following to our LD_LIBRARY_PATH:
/opt/intel/compiler50/ia32/lib:


Added the following to our $path:
/opt/intel/compiler50/ia32/bin /opt/intel/compiler50/ia32/include
/opt/intel/compiler50/ia32/lib /opt/intel/compiler50/ia32/substitute_headers



We had to compile PP.EXE oursevles...we simply could not get the install
wizard to do it.  The format was:
ifc -o PP.EXE PP.f ZTRLOP.o -Vaxlib

We *think* the option '-lPEPCF90' will do same thing as -Vaxlib, but we used
-Vaxlib, so I can't say for sure that '-lPEPCF90' will work.



The C COMPILER should be set to /opt/intel/compiler50/ia32/bin/icc


The Fortran COMPILER should be set to /opt/intel/compiler50/ia32/bin/ifc

with the (fortran) compiler options of: -unroll -w -tpp7 -save -c -O2

(For your information, we did try -O3 optimization, but it did not work.
Everything *seemed* to compile okay, but when I ran the y2k, we got into some
sort of infinite loop (?) were CALIB (I think...) just ran and ran and ran
until I had to just kill it.)

By the way, I don't think we've by any means taken full advantage of
optimization capabilities here, but we're happy with our results and don't
think it's worth it to proceed further at this point.  Using exactly the same
machine/setup and using the default egcs compiler options, we had an AIPSMark
of 72.  Like I said, we're now at 89, so the improvement we did manage to get
was still significant (~24% faster), despite the fact that we haven't
optimized as much as we probably could.  If you are curious, I could send
you the full output of the Y2K, but here are the numbers from the end of
the testing that I think you might want to see:

Task   #   User       Start time         Real      Cpu        Iocnt  Ver  Abort
Sequential list                          secs      secs

APCLN  1     1  24-JAN-2002 15:10:16       55     54.24           0  TST
CALIB  1     1  24-JAN-2002 15:11:16       19     19.24           0  TST
IMAGR  1     1  24-JAN-2002 15:11:36      294    294.72           0  TST
VTESS  1     1  24-JAN-2002 15:16:37       59     58.88           0  TST

  Tot real         7.5
  Tot cpu        7.441

and an avg cp/re ratio of 0.842...by comparison, we had a cp/re ratio of 0.735
on the same machine when we used the egcs compilers.

I guess the only other thing to note is that we need to get this version
registered.  I ran the registration script this morning so you should have
all the stuff you need to make up the keys.

Lastly, let me extend my thanks to both you and Pat for the guidance you gave
us (me and Gabe Prochter) along the way.  We never could have gotten this to
work without your help, so, of course, if there's any assistance we can give
you, should you guys decide to try to investigate compiling with the Intel
compilers yourself, let us know.  (By the way, using the Intel compilers is
claimed to make a significant different for Pentium 4's, but not so much for
Pentium III's.)


					-Sally
				      (slauren at igpp.ucllnl.org)