1 of 1 people found this helpful
It should be possible to make your own modules.
Pre-22.214.171.124, users built a completely new kernel. Although you can ssh to a core and see insmod, I wasn't successful using it. Pre-126.96.36.199 SCC Linux was a monolithic kernel. With 188.8.131.52 I think you can load modules after booting, but I haven't tried it.
With 184.108.40.206 we have a much newer kernel. That new Linux source is on our SVN but under a different project, namely
The Linux build directions for 220.127.116.11 are different from what they used to be. http://communities.intel.com/docs/DOC-6869
I looked into the running system and 18.104.22.168 has the eMac driver as kernel module (not compiled in the kernel).- **edit $HOME/.bashrc to comment out the $PATH-extension for i386-unknown-linux-gnu stuff** (otherwise you get an error about "ncurses header files not found")- screen -R -D- **unset LD_LIBRARY_PATH**- cd sccLinux_V22.214.171.124- edit configure.sh and replace >& by &> in lines 22 and 23- make sccDemo_defconfig- make ctng-menuconfig and deactivate all build-tools (automake, autoconf, m4 and so on) in the second last submenu (otherwise you will get an error about bad m4)- make (and make coffee)
finished compilation of the system successfully.
Good news. Just curious ... do you use RCK MPI? I'm seeing some difficulty with RCK MPI with the provided 126.96.36.199 SCC Linux that I don't think is there with the latest Linux built from the repository.
Basically, one needs to be able to ssh from one core to another .. like ssh into rck00 and then from rck00 ssh to rck01.
No, can't help with that. We never used RCK_MPI. We start our TACO just with pssh from the management PC and then it communicates with our own communication protocols over shared memory.
Now I just compiled the sccLinux. Probably it is sufficient for my purpose, to use the included configured kernel sources to compile own modules. Then it is not necessary to actually boot the self-compiled system.