home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!watson.ibm.com
- From: jbs@watson.ibm.com
- Message-ID: <19921230.110650.57@almaden.ibm.com>
- Date: Wed, 30 Dec 92 13:56:48 EST
- Newsgroups: comp.benchmarks
- Subject: lorenz.c (was Alpha, Risc/6000, HP benchmarks)
- Lines: 15
-
- Steve Thompson posted:
- >Having just received a DEC 3000/400-AXP, I have had time to run one
- >benchmark. This is written in C, and solves the lorenz equations
- >using a Runge-Kutta 4 algorithm. Here are the CPU times, using the
- >highest optimization level in each case, on some machines that we
- >have:
- and
- >The compiler for the RS6000 should have been XLC 1.2, not XLF 2.2. I
- >can also add a time for the RS6000-350: 10.74 seconds.
- You are not compiling this optimally for the RISC/6000. Adding
- -Q (to enable inlining) improves the 550 time from 10.74 seconds to 1.60
- seconds (real time). Adding -qhsflt as well (changes /3. to *(1./3.))
- further improves the time to 1.31 seconds (but changes the answers
- slightly).
- James B. Shearer
-