We have started using Intel C 18.0 on Linux 64 bit.
We have seen issues where we have to reduce optimization level from O3 to O2 in order for the code to work correctly. The same code work fine with Intel C 16 using O3.
Has O3 been improved/changed in V18? I know this is very weak report but do have suggestions for what we can do?
- Development Tools
- Intel® C++ Compiler
- Intel® Parallel Studio XE
- Intel® System Studio
- Parallel Computing
Yes, it has. I would suggest to narrow down the issue by identify a file, which works correctly when compiled at -O2 and not at -O3; then try to construct a test case for us to investigate.
Did you get a different outputs or the program crashed?
Some computations are incorrect and assertions gets triggered.
As you most likely know your suggestion is easier said than done. For now I hope some one else makes a good bug report.