- Отметить как новое
- Закладка
- Подписаться
- Отключить
- Подписка на RSS-канал
- Выделить
- Печать
- Сообщить о недопустимом содержимом
My code compiles. It is an MPI job compiled with Intel 10.1.
However, I am getting the following message when I run my job.
Symbol `__mth_sincos' causes overflow in R_X86_64_PC32 relocation
The job completes. I need to find out how to get rid of this message, before I trust the results of the simulation. Does anyone know how to deal with this?
Thank you.
Ссылка скопирована
2 Ответы
- Отметить как новое
- Закладка
- Подписаться
- Отключить
- Подписка на RSS-канал
- Выделить
- Печать
- Сообщить о недопустимом содержимом
As far as I can see, that function doesn't come from the libraries used by ifort. Is it a dynamic dependency (one which you see with ldd, or a dependency of a .so which appears in your ldd result?)
If you don't want to give details, you might at least show the result of "ldd yourexecutable," and divulge the nature of your MPI.
If you have libraries or objects built by a non-Intel compiler, other than gcc/g++, such as an MPI built for a compiler other than ifort, you take your chances. If you're asking for a wild guess, you may have pulled in a .so which has functions which you don't use, which were built to link against a conflicting math library.
If your build uses entirely Intel supported components, you might consider submitting a problem report on premier.intel.com.
If you don't want to give details, you might at least show the result of "ldd yourexecutable," and divulge the nature of your MPI.
If you have libraries or objects built by a non-Intel compiler, other than gcc/g++, such as an MPI built for a compiler other than ifort, you take your chances. If you're asking for a wild guess, you may have pulled in a .so which has functions which you don't use, which were built to link against a conflicting math library.
If your build uses entirely Intel supported components, you might consider submitting a problem report on premier.intel.com.
- Отметить как новое
- Закладка
- Подписаться
- Отключить
- Подписка на RSS-канал
- Выделить
- Печать
- Сообщить о недопустимом содержимом
This problem is the same enivironment the code was compiled under was not used to execute the program. Thus, the correct dynamic libraries were not established. The user had compiled his mpi code under Intel 10.1 and mpichgm 1.2.7..15, but was executing with the environment of PGI 6.0 and mpichgm 1.2.6..14b. When the enivornment was identical for both compilation and execution, everything worked AOK.

Ответить
Параметры темы
- Подписка на RSS-канал
- Отметить тему как новую
- Отметить тему как прочитанную
- Выполнить отслеживание данной Тема для текущего пользователя
- Закладка
- Подписаться
- Страница в формате печати