1 Reply Latest reply on Sep 7, 2010 9:39 AM by tedk

    Linux, Multi threads per core and memory protection


      I have just joined the community and have been reading the documentation for a day or two, but do not yet have permission to log onto a machine due to the University not having completed the paperwork. But reading the documentation a number of questions occur to me about the relationship between Linux and the RCCE and the LUTs on the cores.


      1. If one choses the option of having linux running on each core, do linux programmes on the cores have full access to files and file read write calls?
      2. If one choses to have Linux running on the cores can a single core fork off several processes and if so can more than one process have access to RCCE communication. If so, how does the send recieve blocking in RCCE interact with linux process scheduling?
      3. The documentation describes how the LUT maps 32 bit addresses to 36 bit addresses. Does this occur after normal Pentium segmentation and paging have occured? If so what can we assume about the state of the paging and segment registers when an RCCE_AP program starts running?
        • 1. Re: Linux, Multi threads per core and memory protection

          Linux on the cores .. full access to files ... I think so ... I've created files, written to them, read from them. If you come across limitations, let us know.


          RCCE users operate with one RCCE process per core, and that is its intended operation. Having said that, I don't see anything inside RCCE that would prevent you from running more than one RCCE process per core. It's a change to the usage model, but may be worth trying out.


          The addresses in the LUTs are physical addresses.