I'm happy to report the the latest IVF (9.1.025) download fixes the problem I was having. As expected, IVF now produces the fastest version of my application. This combined with its OpenMP support means I am "forced" :-) to buy a (Standard) copy!
If only the Mac version could produce universal binaries (x86 & PPC), I'd be willing to send some more money Intel's way.
Al Greynolds Chief Scientist www.ruda.com
Sun, 23 Nov 2008 05:04:43 GMT
awgreyno.. #2 / 5
WAS: IVF continues to disappoint (relatively)
Oops, I forgot to qualify my "fastest" statement. IVF produces the fastest version of my application on genuine Intel chips. Lahey actually produces faster code on my AMD Turion64 laptop.
It seems to me that Intel has intentionally hamstringed IVF when producing code for non-Intel chips. Even though the Turion64 supports SSE3, there appears to be noway to get IVF to use it.
Quote:
> Hi all,
> I'm happy to report the the latest IVF (9.1.025) download fixes the > problem I was having. As expected, IVF now produces the fastest > version of my application. This combined with its OpenMP support means > I am "forced" :-) to buy a (Standard) copy!
> If only the Mac version could produce universal binaries (x86 & PPC), > I'd be willing to send some more money Intel's way.
> Al Greynolds > Chief Scientist > www.ruda.com
Sun, 23 Nov 2008 21:53:28 GMT
#3 / 5
WAS: IVF continues to disappoint (relatively)
Wed, 18 Jun 1902 08:00:00 GMT
Steve Lione #4 / 5
WAS: IVF continues to disappoint (relatively)
Quote:
> If only the Mac version could produce universal binaries (x86 & PPC), > I'd be willing to send some more money Intel's way.
You can use Intel fortran on Mac to create universal binaries. You have to also build with a PowerPC compiler and then use the "lipo" tool to combine them.
As for your application that is faster with another compiler on the Athlon, please send an example to Intel Premier Support. We'd like to see it.
Steve
Sun, 23 Nov 2008 23:24:27 GMT
awgreyno.. #5 / 5
WAS: IVF continues to disappoint (relatively)
I am aware of the "lipo" route. I should have added "by itself" to my desire to have IVF produce universal binaries on the Mac. Its simpler (and cheaper) to deal with as few compilers (and their quirks and multitude of disimilar options) as possibe when creating production binaries.
On the other hand during development and testing, I have found it advantageous to routinely pass my application through as many different compilers/platforms as I can afford. Tends to guarantee more robust code.
Quote:
> > If only the Mac version could produce universal binaries (x86 & PPC), > > I'd be willing to send some more money Intel's way.
> You can use Intel Fortran on Mac to create universal binaries. You have > to also build with a PowerPC compiler and then use the "lipo" tool to > combine them.
> As for your application that is faster with another compiler on the > Athlon, please send an example to Intel Premier Support. We'd like to > see it.
WAS: IVF continues to disappoint (relatively)