David O'Brien wrote:
>
> On Fri, Feb 11, 2000 at 08:19:09AM -0600, Nathan Kinsman wrote:
> > I've been doing some compiler benchmarks with AMD K3-450 and CURRENT
>
> Hum... K6-3/450 I assume.
Yes, K6-3/450. Long night...
>
> > D) -s -O3 2.141 1.852 1.979
> > E) -s -O6 2.144 1.850 1.981
>
> These should be the same as -O's above 3 are ignored by GCC 2.95 (pgcc
> does offer -O's above 3). So it sounds like you are getting cacheing
> affects, or you aren't making enough runs and averaging them.
I have been using pgcc under FreeBSD 3.x for a while now. GCC does not
have the -06 flag? Yes, 4 more test reveal very little deviance with
-O3 and -O6.
One more point of interest is that after recompiling the world with -O3
-ffast-math -fexpensive-optimizations -funroll-loops
-mpreferred-stack-boundry=2 -malign-functions=4 -s -pipe and rebenching
with Nbench I get new results:
-Os -O3
MEMORY INDEX : 2.188
INTEGER INDEX : 1.911
FLOATING-POINT INDEX: 2.220
-O3 -static -ffast-math -fexpensive-optimizations -funroll-loops
-mpreferred-stack-boundry=2 -malign-functions=4 -s
MEMORY INDEX : 2.494
INTEGER INDEX : 2.127
FLOATING-POINT INDEX: 2.421
Overall benchmarks raised by 3% after recompiling world with new CFLAGS.
-
Nathan
To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-current" in the body of the message
