From owner-chemistry@ccl.net Thu Nov 10 15:48:00 2005 From: "Perry E. Metzger perry*piermont.com" To: CCL Subject: CCL: Intel Fortran compiler Message-Id: <-29929-051110135108-23129-nM9HDi7z4eLaYBu6Rd7UMw:server.ccl.net> X-Original-From: "Perry E. Metzger" Content-Type: text/plain; charset=us-ascii Date: Thu, 10 Nov 2005 13:50:59 -0500 MIME-Version: 1.0 Sent to CCL by: "Perry E. Metzger" [perry::piermont.com] "David F. Green dfgreen,+,ams.sunysb.edu" writes: > I should note, I'm generally a big fan of Free Software, and the Gnu > compilers are impressive in the range of achitectures they support, > but they simply do not produce particularly fast code, and sometimes > that needs to be considered (this applies to C/C++ as well as to > Fortran). That depends. For some architectures, languages, target programs, and versions of GCC, GCC works best. For others, it does not. It is very dependent, as I said, on the architecture you're running on, the language you're compiling, what code you're compiling, and (very very important!) which version of gcc and which optimization flags you set. I recommend benchmarking the code you're actually going to run on the machines you'll run it on -- nothing beats an actual experiment in cases like this. Especially if you're going to do a lot of computation, taking a little bit of time and trying a few things (including different settings of the optimization flags in your compilers) pays off handsomely in the end. Oh, and repeating that last bit one more time -- modern compilers have LOTS of optimizations you can turn on and off. Reading the documentation and experimenting with the flags is key. Never assume that the same optimizations will produce the best results for all programs you compile. Perry