Hey Jeff, Ioana Probably worth mentioning that that python script mainly existed because the ROACH 1 didn't have access to a 100mb Ethernet connection except via the powerPC. ROACH 2 has a dedicated fpga-side 1000(?)mb/s interface. Depending on your application, you might consider using this for outputting data, and foregoing powerpc-side python scripts altogether.
Cheers, Jack On Jan 9, 2013 5:58 AM, "David MacMahon" <[email protected]> wrote: > Hi, Jeff and Ioana, > > To get the /proc/*/hw/io_reg/* devices you will need to run a "BORPH > enabled" linux kernel. This will lock you into an older kernel (unless you > want to migrate the BORPH linux changes to a newer kernel, but that would > be a lot of work). I am not sure what level (or longevity) of support the > older BORPH-enabled kernel will have on ROACH2, but if that's adequate for > you then you should be all set. > > The new approach on ROACH2 is to interface to the FPGA via a kernel module > that maps the FPGA (and all the gateware "yellow block" devices) directly > into the PPC address space. This is much faster than the BORPH approach, > but does not create the /proc files that I suspect your python scripts use > on ROACH1. Instead, the new approach uses a new version of > "tcpborphserver" (even though the only thing left from BORPH is the "bof" > file format) to provide access to the gateware devices via KATCP. > > If you really have stuff that you want the PPC to do (besides provide > access to the devices via tcpbporphserver), you could create your own > extensions to tcpborphserver or create some sort of python extension > (written in C) that mimics much of what tcpborphserver does. > > I would recommend going the tcpborphserver extension route if 1) you > decide to go with the new kernel module approach and 2) you need to run > custom code on the PPC (e.g. for fringe rotation updates or something like > that). > > Hope this helps, > Dave > > On Jan 8, 2013, at 6:00 PM, Haoxuan Zheng wrote: > > > Dear Casper Group, > > > > We are working in MIT's Omniscope group, which until now has been using > roach-1's. Four days ago, we got our first roach-2 board! Following the > mail archive (thank you so much for your help!), we set a static ip for the > roach-2 through the serial port, and we managed to program the fpga > through the telnet connection using katcp commands, with the PowerPC > booting soloboot from its flash drive. > > > > However, in our roach-1 system, which had NFS setup, we made use of a > python script running on the PowerPC, whose role was to take the data from > the FPGA, and send it through the UDP protocol to the server PC for storage. > > > > We are now trying to use the same code for the roach-2; for this, we > need to have python running on the PowerPC. > > > > Is there any way to get python going if we have the PowerPC booting in > soloboot from the flash drive? > > > > If not, what do you recommend us to do? > > > > If the answer is that we need to set up netboot to boot via > dhcp+tftp+nfs, there is another constraint: we want our final system to > make use of a combination of roach-1 and roach-2. > > We are going to have the server PC connected to both. Is there going to > be a conflict between the versions when we try to set up the NFS for > roach-2 as well? > > > > We weren't part of the group when the first roaches in our group were > configured, and this is all new to us, so we really appreciate your help. > > > > Thank you so much! > > > > Ioana, Jeff > > >
