Intel® Fortran Compiler
Build applications that can scale for the future with optimized code designed for Intel® Xeon® and compatible processors.

Compilation speed for x64

Simo
Beginner
941 Views
Since version 10 of Intel Fortran, I have compiled 32 and 64 bit versions of my code on both Windows and Linux. I have noticed that compilation for x64 platform takes muuuch longer. I typically use the default optimizations for release version. I would like to know if there are simple tips for reducing the compilation time on x64 platforms.

I do not use OpenMPI.

Simo

0 Kudos
8 Replies
Alexander_C_Intel
941 Views
The set of default optimizations does not remain the same for different releases. What you see may be a result of more optimizations enabled.

If you use inter-procedure optimization (IPO) you may try to reduce the compile time with -(Q)ipo_jobs option.
It's description is a part of built-in help message:
-ipo-jobs
specify the number of jobs to be executed simultaneously during the
IPO link phase

n is usually set in [1(default), 2*(number of cores)]
0 Kudos
Alexander_C_Intel
941 Views
I have just foun similar topic below - "Slow release build with 11.0.066", http://software.intel.com/en-us/forums/showthread.php?t=63541

Check which version you are using and try to update if possible.

Alexander
0 Kudos
jimdempseyatthecove
Honored Contributor III
941 Views

Simo,

Search this forum for a reply from Steve relating to this issue.

The jist of the reply was one of the versions of IVF, after addition of License, was still calling and checking for the Demo License expiration. The work a round (assuming you have a valid license) was to delete the license check .DLL files.

Sorry I cannot recall the thread or names of the dll files. However, with this information you should be able to locate this yourself.

Hint,

Use Google for searchand qualify your query with "site:software.intel.com"

If Steve reads this, he might provide you with the link.

Jim Dempsey
0 Kudos
Simo
Beginner
941 Views
Thanks, I found the the previous post. Unfortunatelly, deleting the _libFNP.dll files did not help.

Simo
0 Kudos
jimdempseyatthecove
Honored Contributor III
941 Views

Are you by chance using /gen-interfaces /warn:interfaces?

If so, consider making a build configuration without the /gen-interfaces /warn:interfaces. Use that for most of your compiles then when adding code with new interface, build once witht he /gen-interfaces /warn:interfaces configuration. You only need to perform this extra work when you have added/changed code that has the potential for goofing up the interfaces.

Jim Dempsey
0 Kudos
peter_cool14
Beginner
941 Views
Simo,

I've got the same problem and I've resolved it. All you need to do is change your platform to x64. Fo that go to

Build>cofiguration manager>Active solution platform>New>x64.

I am assuming that you are running on win32. Ignore this if you are already running x64.

Also you might wanna see IVF vs CVF speed thread.

0 Kudos
Simo
Beginner
941 Views
If you use inter-procedure optimization (IPO) you may try to reduce the compile time with -(Q)ipo_jobs option.

No, the problem is the long compilation time of the fortran files.


0 Kudos
TimP
Honored Contributor III
941 Views
If you don't want to use Makefile to control optimization level by subroutine, there is still the option to set -O level by comment per subroutine. Our usual experience is that when we set -O level by subroutine for best performance, compilation time is less than half the time required at default optimization level.
0 Kudos
Reply