Hi Xavier, I don't think there is. There should be! I'll add it to the infinite todo list.
Here's a preview, which maybe someone more knowledgable with augment / correct -- You can always send packets from the FPGA without tap_start (which starts the tgtap CPU process). Either you can configure the core parameters from the block options in simulink, or set them manually via software registers which are embedded in the ethernet blocks. If you select the "enable interface on startup" option, the interface should come up ready to send as soon as you program the FPGA. Note -- there is no way to set destination MAC addresses from simulink, so without some level of user/cpu intervention, all traffic you send will be destined to the default mac --- 0xffffffffffff If you want to configure your core after programming, corr's FpgaClient object has a method "config_10gbe_core" had a method which sets the Ethernet interface's MAC/IP/port and arp-table in a single, straightforward command. This is what I usually use to set up cores. What tgtap does is start a process on the CPU which will read and respond to packets from the interface which aren't addressed to the port the FPGA-side of the interface is bound to. This is handy, because it means the CPU will handle arp traffic, and will automatically discover the MAC addresses of devices on the network so you don't have to handle the arp table yourself. I actually don't know whether tgtap works with a 1GbE interface (I've never tried), but if it doesn't manually configuring the core parameters is fairly straightforward. Hope that helps a little, Jack On Wed, 28 Jun 2017 at 09:18 Xavier Bosch <[email protected]> wrote: > Hi, > Thank you very much Jack. > Is there a documentation regarding the oneGbE and tenGbE packages where > those interactions between the Ethernet blocks and the ROACH2 CPU are > explained? > Just curious, is there any way to send GbE packages without the tap_sart, > using the fabric configuration only? I am trying to learn how these > interfaces work and I feel I am missing something. > > thanks, > XB > > > > On Thu, Jun 22, 2017 at 12:46 PM Jack Hickish <[email protected]> > wrote: > >> Hi Xavier, >> >> Yeah, that's a bug. I'll have a fix shortly, but for now I would just >> leave the RX/TX CPU interfaces enabled. This is probably not a bad idea >> anyway -- the CPU interfaces are distinct from the interfaces which you see >> in simulink -- they allow the core to offload packets that aren't destined >> for your simulink design to the ROACH2's CPU, which can handle things like >> ARP so that you don't need to manually specify the MAC address of whatever >> device you want to send data to. >> In my opinion the only reason not to enable the CPU interfaces is if they >> negatively impact your timing / resource utilization. Sometimes it's nice >> not to worry about things like arp traffic on the network, but even if the >> interfaces are compiled in to your design, they won't send any packets >> unless you do a tap_start call via katcp. >> >> Cheers >> Jack >> >> >> On Thu, 22 Jun 2017 at 11:25 Xavier Bosch <[email protected]> >> wrote: >> >>> Hi, >>> I managed to compile and synthesize the project using the one_GbE >>> "yellow block". It worked when I unchecked the "Disable CPU Rx" option, see >>> attached schreenshot. I am using the board only for transmitting so, I >>> thought it was a good idea to disable the receiving part of the block. The >>> same project but with this option enabled it crashes. The first error is: >>> >>> >>> >>> >>> >>> >>> *Mapping design into LUTs...ERROR:MapLib:979 - LUT5 symbol >>> "epb_opb_bridge_inst/epb_opb_bridge_inst/opb_reply1" (output >>> signal=epb_opb_bridge_inst/epb_opb_bridge_inst/opb_reply) has input >>> signal "opb0_OPB_retry" which will be trimmed. See Section 5 of the Map >>> Report File for details about why the input signal will become undriven.* >>> >>> Not sure if that "Disable CPU Rx" option requires further selection in >>> the configuration panel to work properly. i.e.: I am configuring the block >>> wrong, or the "yellow block" has a problem. >>> >>> If it is of any help, I am using mlib_devel master branch commit >>> 225aa9ed21aa914078546d2256c56f77f912c90a from Wed Apr 19 09:49:41 2017 >>> >>> Thank you, >>> XB >>> >>> >>> >>> On Tue, Jun 20, 2017 at 2:59 PM Jack Hickish <[email protected]> >>> wrote: >>> >>>> Oh good! >>>> >>>> Which mlib devel repository are you using? If you scroll up through the >>>> errors, can you find the first error message, and does it say anything >>>> useful? >>>> >>>> Cheers >>>> >>>> >>>> Jack >>>> >>>> On Tue, Jun 20, 2017, 2:56 PM Xavier Bosch <[email protected]> >>>> wrote: >>>> >>>>> Hi all, >>>>> Thank you Jack for your reply I followed your advice and tested the >>>>> one_gbe "yellow block". It does not give me an error in the compilation >>>>> but >>>>> it does in the synthesis using XPS. >>>>> At the beginning of the compilation it says that: >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> *############################### Block objects creation >>>>> ###############################misc_ports = app_clk: {[1] 'in' >>>>> 'sys_clk'} mac_tx_rst: {[1] 'in' >>>>> 'one_geb_xb_one_GbE_app_tx_rst'}misc_ports = app_clk: {[1] 'in' >>>>> 'sys_clk'} mac_tx_rst: {[1] 'in' 'one_geb_xb_one_GbE_app_tx_rst'} >>>>> mac_rx_rst: {[1] 'in' 'one_geb_xb_one_GbE_app_rx_rst'}* >>>>> >>>>> and after a while it gives the following error: >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> *Error found in mapping process, exiting...Errors found during the >>>>> mapping phase. Please see map report file for moredetails. Output files >>>>> will not be written.Design Summary--------------Number of errors : >>>>> 86Number of warnings : 4ERROR:Xflow - Program map returned error code 2. >>>>> Aborting flow execution...gmake: *** [__xps/system_routed] Error >>>>> 1ERROR:EDK >>>>> - Error while running "gmake -f system.make bits".: XPS failed.* >>>>> >>>>> Any idea what I could be doing wrong? >>>>> Has anyone worked with that "yellow block" recently? >>>>> Is there any example available? >>>>> Thank you, >>>>> XB >>>>> >>>>> >>>>> >>>>> >>>>> On Mon, Jun 19, 2017 at 4:43 PM Jack Hickish <[email protected]> >>>>> wrote: >>>>> >>>>>> Hi Xavier, >>>>>> >>>>>> The ROACH2 (unlike the ROACH1) has a dedicated 1GbE connection >>>>>> straight to the FPGA fabric -- it's the "other" RJ45 connector in >>>>>> addition >>>>>> to the one you're using to talk katcp. You can access it using the >>>>>> one_gbe >>>>>> yellow block -- I think this is probably the way to go. >>>>>> >>>>>> Cheers >>>>>> Jack >>>>>> >>>>>> On Mon, 19 Jun 2017 at 16:39 Xavier Bosch < >>>>>> [email protected]> wrote: >>>>>> >>>>>>> Hi, >>>>>>> >>>>>>> I have a system that generates ~100 Mbit/s data throughput in a >>>>>>> ROACH2 board. >>>>>>> >>>>>>> Until now I have been using the KATCP and software registers to read >>>>>>> those values at a much smaller data rate. The only option that I found >>>>>>> in >>>>>>> the documentation to use the design at full resolution are the 10 GbE >>>>>>> boards, that we already have in hand. The 10 GbE boards look to me that >>>>>>> is >>>>>>> a little overkill for what I would like to do. Is there any other >>>>>>> method to >>>>>>> get data out from a ROACH2 board faster than the software register but >>>>>>> slower than the 10 GbE boards? >>>>>>> >>>>>>> I am thinking in something like an USB 2.0 port (max 480 Mbit/s) or >>>>>>> a 1GbE connection. >>>>>>> >>>>>>> Thank you, >>>>>>> XB >>>>>>> >>>>>>> >>>>>>> -- >>>>>>> You received this message because you are subscribed to the Google >>>>>>> Groups "[email protected]" group. >>>>>>> To unsubscribe from this group and stop receiving emails from it, >>>>>>> send an email to [email protected]. >>>>>>> To post to this group, send email to [email protected]. >>>>>>> >>>>>> -- You received this message because you are subscribed to the Google Groups "[email protected]" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected].

