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

ifort on MacOS High Sierra Bug

huang__wen
Beginner
455 Views
I installed the latest Intel Fortran complier 2018 update1 on my MacBook Pro which has updated to the latest High Sierra OS. I complier my Fortran program successfully. However when I run some data by my program, there will be some wrong processing and infinity cycle. The point is that my codes suppose to be right. Because it works very well on the Linux server of our institute. And also on my supervisor's MacBook. I tested my code on different system and compilers like below. 1. Linux --------------------------------------------------------ifort 12-----------------compile success------works well 2. My Macbook Pro 10.13---------------------------------ifort 18.0.1------------compile success--------bug 3. My supervisor's MacBook Pro 10.12 or lower-----ifort 14-----------------compile success-------works well 4. copy excitable file from my supervisor's Mac to mine---------------------------------------------------works well From the test above, I think the problem is not from my codes. Now, earlier version ifort can not be installed on MacBook 10.13. I can not downgrade my OS to 10.12 or earlier. What I can do to solve this problem? Thanks for any advice and help.
0 Kudos
3 Replies
Steve_Lionel
Honored Contributor III
455 Views

You can report the problem using the Intel Online Service Center, attaching a copy of the program source, instructions for building, and including any data files it needs. (A tarball works.) What you have said here is too vague for any of us to be helpful. It sounds to me as if ifort 18 is the issue, not the MacOS version. It might be an issue with the gcc libraries Apple makes available.

0 Kudos
huang__wen
Beginner
455 Views

Thank you for the reply and advice. But I need to ask for the permission of upload the program. Because it is from my supervisor and I just made some modifications. It is the copy right issue. And I am not sure about the results yet.

You mentioned the gcc libraries issue. So can you guide me to do some try by myself? In fact, I am also a beginner of command stuff and fortran. 

Steve Lionel (Ret.) wrote:

You can report the problem using the Intel Online Service Center, attaching a copy of the program source, instructions for building, and including any data files it needs. (A tarball works.) What you have said here is too vague for any of us to be helpful. It sounds to me as if ifort 18 is the issue, not the MacOS version. It might be an issue with the gcc libraries Apple makes available.

 

0 Kudos
Steve_Lionel
Honored Contributor III
455 Views

Sorry, I've exhausted my Mac knowledge. All I know is that the Intel compiler requires that gcc be installed so that its libraries can be used, and that Apple no longer installs these by default. But the gap between the 14 and 18 compilers is large. It may come down to a different sequence of code that reveals an instability in your program's algorithm. It could be improvements in the math library giving different results.

If you can't yet provide Intel with the program, you might at first try disabling optimizatioins and see if that changes the behavior. If it does, that doesn't necessarily mean the compiler is wrong.

What I would do is "instrument" the program to show you intermediate results and progress. Build and run it with both compilers and see where behavior starts to diverge. That will tell you a lot.

0 Kudos
Reply