Date: Mon, 31 Mar 1997 20:18:07 -0600 From: Fred Streitz To: meflan1@engr.uky.edu Subject: Re: Your Message Sent on Mon, 31 Mar 1997 17:13:00 -0500 (EST) Hi Matt, It seems like I keep falling further and further behind... I was able to get pretty decent optimization on the SGI's - my optimization flag was: OPTIMIZER = -O2 -mips2 -col120 \ -sopt,-inll=5,-ind=5,-inline,-inff=mkcijkl.f:blaslike.f:calvij.f:util.f If I recall correctly, O3 was not faster, but I don't think it actually failed. (I could be wrong.. I know O3 failed on other programs - it was more sensitive to common block alignment and whatnot...) the -sopt flag (with the attendent options, seperated by commas) was very productive. (It performs "scalar optimization" ). Hope this helps. Dr. Streitz --------------------------------------------------------------------- Frederick H. Streitz Phone: (334) 844-2943 Physics Department FAX: (334) 844-4613 Auburn University Auburn, AL 36849-5311 email: streitz@physics.auburn.edu