Date: Thu, 01 Feb 1996 20:43:15 +0000 From: "/home/nbc/.mail.sig" <nbc@vulture.dmem.strath.ac.uk> To: hackers@FreeBSD.ORG Subject: Re: Pentium gcc port done Message-ID: <199602012043.UAA09021@vulture.dmem.strath.ac.uk> In-Reply-To: Your message of "Thu, 01 Feb 1996 10:19:15 PST." <9602011819.AA09036@gnu.mc.xerox.com>
next in thread | previous in thread | raw e-mail | index | archive | help
OK, I managed to get the new pentium optimizing gcc working, and for those
interested, here are a couple of results. The first program, aspsim, is
basically a ray casting program that works a bit like a simple povray; it
generates synthetic range and intensity images from an ascii description
file. As you can see, the speedup is remarkable with -mpentium. Could this
be because of the fact that the program does a great deal of FPU operations
and thus is able to take advantage of instruction ordering?
The second test is a program that analyses the files produced by aspsim, and
implements a version of the Extended Gaussian Image. This program exhibited
less of a speedup - possibly because it concentrates more on throwing data
about in memory than on floating point things? Who knows ;-)
I'm going to *have* to recompile POVRay in the light of result (1)...
Neil
----- Test 1 - aspsim -----
CC = gcc -c -O3 -m486 -fomit-frame-pointer
/usr/bin/time aspsim besl1 512 512 1024
9.08 real 7.95 user 0.32 sys
CC = pentium-unknown-freebsd-gcc -c -O3 -mpentium -fomit-frame-pointer
/usr/bin/time aspsim besl1 512 512 1024
3.76 real 2.64 user 0.42 sys
----- Test 2 - egi -----
CC = gcc -c -O3 -m486 -fomit-frame-pointer
/usr/bin/time egi besl1.xyz
12.45 real 6.57 user 0.66 sys
CC = pentium-unknown-freebsd-gcc -c -O3 -mpentium -fomit-frame-pointer
/usr/bin/time egi besl1.xyz
10.16 real 5.33 user 0.59 sys
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199602012043.UAA09021>