Hi Adam, Yes we are using the casper_xps tool flow but with a bit of a twist. The main design is encapsulated in a Simulink subsystem and consists of hdl synthesizable blocks. Hdl coder is used to generate vhdl from the design. Another Simulink design contains all of the yellow blocks, the system generator token and one Xilinx black box containing the hdl synthesized design. This design is then compiled using casper_xps. This allows us to use vanilla FFT’s and such that hdl coder generates. Getting back to the timing error, In the design, there is a heavy lifting block containing a number if FFT’s and IFFT’s. If I generate this block with the FFT/IFFT architecture mask parameter set to radix2 the design compiles and meets timing. If I set this mask parameter to radix2^2 (which is supposed to be faster/better according to mathworks) I get the timing error in the one_GbE zone. As I mentioned before, the design is using only occupying 50% of the slices (plenty of dsp48e’s available as well) and running at a fabric clock rate 125 MHz. My initial thought was this should be no problem. I believe floor planning is a way to deal with issues like this and adding constraints is also a good suggestion. The problem is I haven’t done this before. Is there any information you know of that explains how to floor plan? I have pulled the ngc files into plan ahead and pushed the green play button to implement but don’t know what to do from there. Also any tips on adding timing contraints would be appreciated. Regards, Michael Movius
From: Adam Isaacson [mailto:[email protected]] Sent: 27 September 2016 12:13 PM To: Mike Movius Cc: [email protected] Subject: Re: [casper] one_Gbe timing errors Hi Mike, I have had a look at this. I can definitely tell you this is not a pipe-lining issue (as logic levels/interconnect = 3 stages, so no need to insert flip-flops to reduce the timing between the paths. It is difficult for me to analyse the path without seeing the floor plan, but I believe you will need to reduce the data path delay if you want to meet timing. The best is to open your design in the floor planner and highlight this failing path - hopefully you will see a long delay for the data path routing. It may be as easy as assigning an area constraint in your ucf file for the MAC itself, so that the data path is shorter by a few ns i.e. a pblock constraint. I am not sure if you are using the CASPER_XPS tool flow at all or if you have your own flavour, but maybe there are existing constraints for the 1GbE that you could use? I am not sure if any of this helps, but it is a start anyway :). Kind regards, Adam On Fri, Sep 23, 2016 at 10:49 AM, Mike Movius <[email protected]<mailto:[email protected]>> wrote: Hi All, I have a Roach2 design using the one_Gbe yellow block. I intermittently have timing errors in this block seemingly unrelated to the design. Below is a TS_mac_rx_clk timing constraint failure. The FPGA is running at 125 MHz and approximately 50% of the resources are currently being utilized. The design only sends data out of the interface. Any ideas on what is causing this or how to fix it? ================================================================================ Timing constraint: TS_mac_rx_clk = PERIOD TIMEGRP "mac_rx_clk" 125 MHz HIGH 50%; For more information, see Period Analysis in the Timing Closure User Guide (UG612). 12625 paths analyzed, 2332 endpoints analyzed, 4 failing endpoints 4 timing errors detected. (4 setup errors, 0 hold errors, 0 component switching limit errors) Minimum period is 8.157ns. -------------------------------------------------------------------------------- Slack: -0.157ns (requirement - (data path - clock path skew + uncertainty)) Source: fort_adm_rsp_bb_one_GbE/fort_adm_rsp_bb_one_GbE/enable_cpu_tx.cpu_buffer_tx/BU2/U0/blk_mem_generator/valid.cstr/ramloop[0].ram.r/v6_noinit.ram/TRUE_DP.SIMPLE_PRIM36.ram (RAM) Destination: temac_inst/temac_inst/emac_wrapper_inst/v6_emac (CPU) Requirement: 8.000ns Data Path Delay: 7.893ns (Levels of Logic = 3) Clock Path Skew: -0.229ns (1.919 - 2.148) Source Clock: mac_mac_rx_clk rising at 0.000ns Destination Clock: mac_mac_rx_clk rising at 8.000ns Clock Uncertainty: 0.035ns Clock Uncertainty: 0.035ns ((TSJ^2 + TIJ^2)^1/2 + DJ) / 2 + PE Total System Jitter (TSJ): 0.070ns Total Input Jitter (TIJ): 0.000ns Discrete Jitter (DJ): 0.000ns Phase Error (PE): 0.000ns Maximum Data Path at Slow Process Corner: fort_adm_rsp_bb_one_GbE/fort_adm_rsp_bb_one_GbE/enable_cpu_tx.cpu_buffer_tx/BU2/U0/blk_mem_generator/valid.cstr/ramloop[0].ram.r/v6_noinit.ram/TRUE_DP.SIMPLE_PRIM36.ram to temac_inst/temac_inst/emac_wrapper_inst/v6_emac Location Delay type Delay(ns) Physical Resource Logical Resource(s) --------------------------------------------------------------- ------------------- RAMB36_X6Y36.DOBDO1 Trcko_DOB 2.073 fort_adm_rsp_bb_one_GbE/fort_adm_rsp_bb_one_GbE/enable_cpu_tx.cpu_buffer_tx/BU2/U0/blk_mem_generator/valid.cstr/ramloop[0].ram.r/v6_noinit.ram/TRUE_DP.SIMPLE_PRIM36.ram fort_adm_rsp_bb_one_GbE/fort_adm_rsp_bb_one_GbE/enable_cpu_tx.cpu_buffer_tx/BU2/U0/blk_mem_generator/valid.cstr/ramloop[0].ram.r/v6_noinit.ram/TRUE_DP.SIMPLE_PRIM36.ram SLICE_X127Y181.C6 net (fanout=2) 1.727 fort_adm_rsp_bb_one_GbE/fort_adm_rsp_bb_one_GbE/cputx_tx_rd_data<1> SLICE_X127Y181.C Tilo 0.068 fort_adm_rsp_bb_one_GbE/fort_adm_rsp_bb_one_GbE/gbe_tx_inst/mac_data_reg<1>2 fort_adm_rsp_bb_one_GbE/fort_adm_rsp_bb_one_GbE/gbe_tx_inst/mac_data_reg<1>2 SLICE_X127Y181.D3 net (fanout=1) 0.333 fort_adm_rsp_bb_one_GbE/fort_adm_rsp_bb_one_GbE/gbe_tx_inst/mac_data_reg<1>1 SLICE_X127Y181.D Tilo 0.068 fort_adm_rsp_bb_one_GbE/fort_adm_rsp_bb_one_GbE/gbe_tx_inst/mac_data_reg<1>2 fort_adm_rsp_bb_one_GbE/fort_adm_rsp_bb_one_GbE/gbe_tx_inst/mac_data_reg<1>3 SLICE_X133Y180.D5 net (fanout=1) 0.719 fort_adm_rsp_bb_one_GbE/fort_adm_rsp_bb_one_GbE/gbe_tx_inst/mac_data_reg<1>2 SLICE_X133Y180.DMUX Tilo 0.191 fort_adm_rsp_bb_one_GbE_app_dbg_data<0> fort_adm_rsp_bb_one_GbE/fort_adm_rsp_bb_one_GbE/gbe_tx_inst/mac_data_reg<1>4 TEMAC_X0Y1.CLIENTEMACTXD1 net (fanout=1) 2.527 fort_adm_rsp_bb_one_GbE_app_dbg_data<1> TEMAC_X0Y1.CLIENTEMACTXCLIENTCLKIN Tmacdck_TXD 0.187 temac_inst/temac_inst/emac_wrapper_inst/v6_emac temac_inst/temac_inst/emac_wrapper_inst/v6_emac --------------------------------------------------------------- --------------------------- Total 7.893ns (2.587ns logic, 5.306ns route) [cid:[email protected]] Please consider the environment before printing this e-mail View the Reutech Radar System online disclaimer at http://www.rrs.co.za/links/E-maildisclaimer.asp<http://www.rrs.co.za/links/E-maildisclaimer.asp> -- Adam Isaacson DBE: FPGA Engineer SKA-SA 3rd Floor The Park Park Road Pinelands 7405 Tel: +27215067300 (W) Fax: +27215067375 (W) Cell: +27825639602 Please consider the environment before printing this e-mail. DISCLAIMER: This message is intended for the use of the addressee and may contain information that is privileged and/or confidential. If you are not the intended recipient, you are hereby notified that any dissemination, distribution or copying of the information contained in this message is strictly unauthorized and prohibited. If you have received this message in error, please notify the sender by reply e-mail and delete the message from your system. Opinions, conclusions, or other statements in this message that do not relate to the business of REUTECH Radar Systems, a Division of Reutech Ltd., its subsidiaries or affiliates, are neither given nor endorsed by REUTECH Radar Systems, a Division of Reutech (Pty)Ltd.

