Hello everyone, I'm currently in the middle of some cuda fortran coding and I really need a bit of help.
First of all, a brief description: compiling cuda fortran code (usually *.cuf) requires the use of the proprietary compiler from the portland group (pgfortran). I've got access to such a compiler on a computer running ubuntu 10.04 and recently tried to install it with some success on Debian after some hacking. Specifically, I made a couple of symbolic links to help the gnu linker find crt1.o, crti.o and maybe one more crt*, as the location of said files wasn't properly set. (something to do with debian-specific gcc library file paths?) Now, when I try to compile code using pgfortran, the compiler fails as it requires "size_t" to be defined.. and I'm not really well versed on C or the location of it's libraries. Any ideas that could help? On a different note, I'd usually used the intel fortran compiler and I find it *much* faster for regular serial (non-cuda) code. Is it possible to link two object files produced by different compilers? Runtime could be much faster by cross-linking. In any case, I can ssh to work and run everything there ok. So, getting the compiler running on debian isn't my biggest wish. Many thanks. -- Daniel Jiménez -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected] Archive: http://lists.debian.org/cakuhbggm6gl0izfydirvgprdk6qbhqm0nxrmtyv+pmyxf+6...@mail.gmail.com
