- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hello there,
I have a code that I used to run using Fortran powerstation. And, because of array size limitation of powestation, I had to go for Intel Fortran. What I noticed is that Intel fortran is very slow compared to Powerstation. My question is "Am I misusing Intel Fortran or it is possible that the Powerstation comiler is slower?" Any suggestions about how to speed up Intel Fortran.
Thank you,
Ashraf
Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
You are almost certainly not using Intel Fortran correctly. Which options are you using to compile? If you are using Visual Studio, are you building a Debug or Release configuration?
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hello,
Thank you for your fast response. I am currently using debug configuration. And, I am comaring it to powerstation debug configuration.
The options I use are:
"/nologo /debug:full /Od /gen-interfaces /warn:unused /warn:interfaces /module:"Debug\\" /object:"Debug\\" /traceback /check:bounds /libs:static /threads /dbglibs /c"
Thanks you,
Ashraf
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Please compare release configurations. We generate a lot of extra code for debugging that PowerStation did not.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
i would also ask what does your program that you are benchmarking do? meaning what types of subroutine, functions, calculations does it do? and how large are the arrays? and if Powerstation can't handle them how do you bench between the two (use the smaller value for the array or simply compare large v small arrays?) On the web there are benchmarks comparing current Fortran compilers:
http://www.polyhedron.com/pb05-win32-f90bench_SBhtml
in the benchmarks INTEL rocks!. So I'm not sure whay you'd seea performance degradation between the (old) Powerstation and Intel compiler. Comparing Release builds will help but me thinks something else is wrong in your comparison.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hello,
You guys have made some good points. I will have to use the release libraries for a better comparison. I'm sure that the intel library will be faster. What I'm really asking is what are the factors that might cause my library to run slower than powerstation so that I can fix the issues. What are some of the broad issues that I should look into?
Thank you,
Ashraf
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Come back and ask us when you try the release configuration.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
What I'm really asking is what are the factors that might cause my library to run slower than powerstation so that I can fix the issues.This should be rather obvious. 1. First fix the bugs in your program so that you do not have to use the debug configuration. If, for example, you have array overruns or usage of uninitialized variables, it is pointless to compare run timings. 2. When you are sure that there are no bugs left, and you are able to obtain essentially the same results with both compilers, compile with the best optimization flags for your system/compiler combination, and then compare run timings.

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page