On Jan 7, 2009, at 10:39 AM, Brian Helenbrook wrote:
Eric & Ian & sci-tech,
Thanks to your help, I have the floating point counter working on my
G4 laptop. I have gone through several tests to make sure that the
results are roughly independent of compiler options etc... Its not
entirely, but it is close. For example, I've been examining an
integration routine I wrote. With optimization, the count is
typically a little less than a factor of two smaller than without
it. (The cpu time's are different by a factor of 20 roughly). I
assume this is because the compiler is using fused multiply & add in
the second case, but I haven't gone through the assembly to check.
In any case, this amount of variation is tolerable and I think this
metric will still be useful.
Now, I want to see what happens when I change processors. I am
trying the same thing on a Intel Core 2. Reading through all the
performance monitor descriptions, I think the one I want is (193,
0xC1) X87_OPS_RETIRED which has the description:
X87_OPS_RETIRED.ANY
MacOS X uses SSE, not x87, for single and double precision scalar
arithmetic.
Ian
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Scitech mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden
