i think i have seen this error before. Can you retrain the cores and try your experiment again?
I got the same error code: 126. I just retried the procedure after some time, and it worked. I do not know whether this will work all the time. I also do not know whether resetting the board, or retraining the cores will definitely do the trick.
The error codes from pssh are the bash exit codes. We started collecting some in the file Error Codes on the first page of this site. You can also check out http://www.museum.state.il.us/ismdepts/library/linuxguides/abs-guide/exitcodes.html
"share" example seems to work. I guess error code 126 has to do with file having the executable bit not set. Now things seem to go well. Another question: what purpose does a redirect inwards from /dev/null serve?
Thanks. I noticed also that error code 139 occurs when my program actually gets a seg fault.
About your shared example ... Not all the shared memory assigned by default to the SCC can be used by an application programmer. Some of this shared memory is used by SCC Linux itself, and you must be careful not to step on this memory.
This means that in the default configuration, you actually have less than 64MB.
The post http://communities.intel.com/message/101172#101172 by Michael Riepen specifies exactly what shared memory is used by the system. You can avoid this memory by adding an offset to the start of shared memory specified in RCCE_shmalloc_init() as
If you do step on the memory, you will eventually see bad behavior. We're looking at a way of increasing the amount of shared memory by modifying the default LUT values. You can do this modification in RCCE_init(). We have a RCCE prototype that shows how to do this. It is available at
http://marcbug.scc-dc.com/svn/repository/trunk/rcce/ Note that this is the RCCE trunk. We tag official releases in http://marcbug.scc-dc.com/svn/repository/tags/
All the modifications are under #ifdef SHMADD. The current symbols.in specifies SHMADD but if you modify symbols.in to remove the -DSHMADD and reconfigure, you should be able to build the released RCCE with what comes from the trunk.
This is a temporary solution with a better one in the works.
Oh, that null redirect .... I think it's a artifact of pssh ... which may not be necessary any longer.