Fortran Compiler for Linux 7.1
When I try to compile one of my codes, with only -O3 or -xW option, it can be done within 20 seconds. But when I use both options simultaneously, it could take more than 15 minutes. Any idea with this?
With the combination of options, you must be engaging an opportunity for some complicated analysis. I tend to avoid -O3 unless I have a reason for using it. If -xW is able to vectorize all the loops which are important for performance, you may not need -O3. If not, you may look for reasons why in-lining or interprocedural analysis might be needed, and whether you could make the optimization possible at -O2.
If you compile in 20 seconds with the other combinations, it seems the program is not too large to attempt this combination. It might be a useful submission as a premier.intel.com support issue.