Re: slow running NCARG or NCL program?

Pretty significant speedups, I'd say ...

Dave

On Jan 18, 2008, at 9:42 AM, Mary Haley wrote:

>
> Just FYI: compiling Dave's program with ifort versus g95 gave me some
> pretty good speed ups for some parts of the code (you print out
> timings for various pieces). For example:
>
> g95 ifort
> --------------------------------
> TIME GENERATING TRIANGULAR MESH
> 48.19966 14.51000
>
> TIME DRAWING RECTANGULAR GRID
> 50.769966 28.94000
>
> TIME DRAWING TRIANGULAR MESH
> 49.855415 28.51000
>
> TIME DRAWING SIMPLE CONTOURS
> 22.296738 14.23000
>
> TIME DRAWING FILLED CONTOURS
> 18.633148 11.54000
>
>
> --Mary
>
>
> On Thu, 17 Jan 2008, David Kennison wrote:
>
>>
>> Mary,
>>
>> A program that comes to mind is one I put together in 2006 as part
>> of my "scalability" testing of CONPACKT. If you crank up the
>> value of the parameter IDIM (which occurs in three places) to 2001
>> (search for occurrences of "IDIM=2001"), it runs for something
>> close to ten minutes on my machine. (And, if that value doesn't
>> work, you can crank it up or down, as needed)
>>
>> I've attached the files "Testit.2.f", which is the FORTRAN
>> program, "ctpopg.dat", which is a required data file, and a plot
>> showing timing as a function of IDIM.
>>
>> Dave
>>
>>
Received on Fri Jan 18 2008 - 10:11:24 MST