Home > Archive > Fortran > May 2004 > Poor performance with intel fortran compiler
You are viewing an archived Text-only version of the thread.
To view this thread in it's original format and/or if you want to reply to
this thread please [click here]
| Author |
Poor performance with intel fortran compiler
|
|
| Shriram Ramanathan 2004-05-04, 3:05 pm |
| Hi,
I have been trying to move to intel fortran compiler from portland group
for the last 1 month or so. And I keep having some problems or the
other.
A) I first used ifc v7.1 on an AMD athlon machine to compile my code
(with highest possible optimization. My code was running 10 times
slower when compared to the one compiled using pgf90.
B) I thought that probably intel compilers are bad at optimizing for AMD
chips. So, I used ifort v8.0 on an itanium machine (with -O3). The
compiler gave me a warning on one of the files :
fortcom: Warning: Optimization suppressed due to excessive resource
requirements; contact Intel Premier Support
I continued waiting inspite of the warning. The compiler finally
finished generating an executable after a 1/2 hour wait.
happy, I tried running the executable and the executable was running
almost 50 times slower than the one generated by pgf90.
Am I stupid or is ifort/ifc so bad at optimizing codes ?
Anyone have a reply ?
-shriram.
| |
| Steve Lionel 2004-05-04, 3:06 pm |
| On Tue, 4 May 2004 11:28:00 -0500, Shriram Ramanathan <ramanath@cems.umn.edu>
wrote:
>Am I stupid or is ifort/ifc so bad at optimizing codes ?
Most people find that Intel Fortran generates very fast code. It may be that
your particular program is not handled well by the compiler - the warning you
got suggests that is the case. Note that when you get that warning, all
optimization is disabled.
Please submit an example of the problem to Intel Support - we'll be glad to
look at it.
Steve Lionel
Software Products Division
Intel Corporation
Nashua, NH
User communities for Intel Software Development Products
http://softwareforums.intel.com/
Intel Fortran Support
http://developer.intel.com/software/products/support/
| |
| Dr Chaos 2004-05-04, 3:06 pm |
| hriram Ramanathan <ramanath@cems.umn.edu> wrote:
> Hi,
>
> I have been trying to move to intel fortran compiler from portland group
> for the last 1 month or so. And I keep having some problems or the
> other.
>
> A) I first used ifc v7.1 on an AMD athlon machine to compile my code
> (with highest possible optimization. My code was running 10 times
> slower when compared to the one compiled using pgf90.
>
> B) I thought that probably intel compilers are bad at optimizing for AMD
> chips. So, I used ifort v8.0 on an itanium machine (with -O3). The
> compiler gave me a warning on one of the files :
>
> fortcom: Warning: Optimization suppressed due to excessive resource
> requirements; contact Intel Premier Support
>
> I continued waiting inspite of the warning. The compiler finally
> finished generating an executable after a 1/2 hour wait.
>
> happy, I tried running the executable and the executable was running
> almost 50 times slower than the one generated by pgf90.
>
>
> Am I stupid or is ifort/ifc so bad at optimizing codes ?
> Anyone have a reply ?
Intel Fortran in general is not bad at optimizing, and in my
experience is very good almost all of the time.
One thing which is true is that if you are on a AMD chip, the
various -xW -axWPN whatever options will not be functional; the
executed code on an AMD chip will always be the basic "Pentium x87"
instructions. No SSE or SSE2, even if the particular AMD chip can do it.
That's a bummer.
The upside is that the AMD chips are lickety split fast even on
basic x87 Pentium instructions.
> -shriram.
I think you have found something peculiar specific to your code.
| |
| Tim Prince 2004-05-05, 3:36 am |
|
"Shriram Ramanathan" <ramanath@cems.umn.edu> wrote in message
news:Pine.LNX.4.33.0405041118080.8390-100000@lancelot.cems.umn.edu...
> Hi,
>
> I have been trying to move to intel fortran compiler from portland group
> for the last 1 month or so. And I keep having some problems or the
> other.
>
> A) I first used ifc v7.1 on an AMD athlon machine to compile my code
> (with highest possible optimization. My code was running 10 times
> slower when compared to the one compiled using pgf90.
>
> B) I thought that probably intel compilers are bad at optimizing for AMD
> chips. So, I used ifort v8.0 on an itanium machine (with -O3). The
> compiler gave me a warning on one of the files :
>
> fortcom: Warning: Optimization suppressed due to excessive resource
> requirements; contact Intel Premier Support
>
> I continued waiting inspite of the warning. The compiler finally
> finished generating an executable after a 1/2 hour wait.
>
> happy, I tried running the executable and the executable was running
> almost 50 times slower than the one generated by pgf90.
>
Optimization suppressed means the balance of the build is done at -O0. As
you noticed, that runs deadly slow on IPF. Options like -O1
and -override_limits look evident to try. How should we have a clue what
you have done in your source, which seems to interact so badly with the
compiler flags you have chosen? Do you have enough RAM installed for your
compile and run?
|
|
|
|
|