08:52 <joseng> FIFO on write and read ports.

08:52 <joseng> Hi guys, in all my designs so far with VHDL for xilinx devices I needed to use the inverted clock for the FIFO ports, otherwise lets say the FSM is on the "sys" clock and changes the WE signal of the fifo, the signals would change with the rising edge of the fifo. So when the fifo should latch. This is why I every time used the inverted clock for a

08:52 <joseng> But I never saw this anywhere in migen/litex where a FIFO is used.

08:55 <daveshah> There's no need for that. The place and route tooling will ensure that setup/hold times are met

08:55 <daveshah> Using the opposite edge of the clock only serves to reduce the available timing period from one clock cycle to half a clock cycle, and can actually halve Fmax in the worst case scenario

08:59 <joseng> Ah interesting, didn't know that. Thanks

09:01 <_florent_> daveshah: Hi, I saw on discord you were eventually interested in trying to get LiteSATA running on ECP5, just want to share some info that can be useful I think:

09:02 <_florent_> I initiated some work recently on that, but only had a few hours to work on this, so there is still work to do

09:02 <_florent_> I did some test with LiteICLink and the Versa ECP5 at the linerates used for SATA1 and SATA2: 1.5Gbps and 3.0Gbps

09:02 <_florent_> here: https://github.com/enjoy-digital/liteiclink/tree/master/bench/serdes

09:03 <daveshah> Thanks, thats useful

09:03 <_florent_> ./versa_ecp5.py --linerate=1.5e9 (or 3.0e9) --build --load

09:03 <_florent_> sorry

09:04 <_florent_> ./versa_ecp5.py --linerate=1.5e9 (or 3.0e9) --connector=pcie (or sma) --build --load

09:04 <_florent_> you can then use a pcie loopback module or sma cable to do the loopback

09:04 <_florent_> the RX leds should be blinking with the counter generated on TX

09:05 <_florent_> the PRBS test can also be run with litex_server and ./test_prbs.py

09:07 <_florent_> I've been able to get 0 errors with the PRBS test, so it shouldn't be too bad

09:07 <_florent_> but that's possible some adjustments still need to be done on the SerDes parameters

09:08 <_florent_> I also created a branch in LiteSATA: https://github.com/enjoy-digital/litesata/tree/ecp5

09:08 <_florent_> where I started the integration and started working on OOB

09:09 <_florent_> sata-specification.3.3.gaps-filled-by-sata-qvip_vh-v14-i1.pdf could be useful if you want to understand OOB on SATA

09:10 <_florent_> I was able to see the COMINIT to see and detect the COMINIT from the device and send the COMWAKE but was not seeing the COMAWAKE from the device

09:10 <daveshah> Great, I'll find some time to give it a try when my ECPIX arrives

09:15 <_florent_> daveshah: great

09:16 <daveshah> Being able to capture to a SATA SSD definitely opens up a lot of interesting ideas

09:18 <_florent_> Indeed, Artix7/Ultrascale(+) support has been added recently, but EPC5 would be really nice

09:19 <_florent_> For these tests I was using a minimal bench with just the PHY and LiteScope in place: https://github.com/enjoy-digital/litesata/blob/ecp5/bench/versa_ecp5.py

09:23 <_florent_> I suspect the OOB from the FPGA are not received by the SSD (COMINIT are generated spontaneously by the SSD, even without sending COMRESET)

10:19 <zyp> ah, ECPIX is available to order now, nice

10:41 <acathla> Wow, nice board!

11:02 <keesj> seen nice one also? https://shop.lambdaconcept.com/home/46-1-ecpix-5.html#/1-ecpix_5_fpga-ecpix_5_45f

11:02 <tpb> Title: ECPIX-5 (at shop.lambdaconcept.com)

11:17 <zyp> I just ordered one :)

15:03 <joseng> Am I right in the assumption that the async FIFOs in migen are First Word Fall Through (by default)? Seems to from my tests with it and and watch the output with litescope

15:05 <daveshah> Yes fwft defaults to true

15:05 <daveshah> https://github.com/m-labs/migen/blob/master/migen/genlib/fifo.py#L99

15:06 <joseng> For the SyncFIFO I saw that there is a parameter, but I mean the AsyncFIFO. Need to cross from the SerDes clock domain to sys

19:05 <roboknight> Okay, I've gotten the vexrisc running a very short piece of blinky code (not the LedChaser, but some actual software manipulating Leds). I can see the LEDs running. I can even get serial output that I can see. But if I output more than about 16 or so characters, it freezes (no more sweeping LEDs) and no output.

19:05 <roboknight> I was wondering if there might be an issue with multi-level interrupts or something along those lines.

19:07 <roboknight> The ice40-hx8k is VERY small (maybe too small), but I can pack the code in with about 2K ram and about 10K of rom (that nearly exhausts my 32 ram blocks)... I haven't moved to XIP flash yet, but that's where I'm trying to head.

19:23 <roboknight> I guess I'll stick to UART polling for a while... seems like that works. For now anyway.

22:32 <joseng> Is there any way to reset/empty FIFOs?

22:38 <zyp> I don't think so, but I guess it might work with a ResetInserter

23:14 <joseng> Hmm ok, then I have a look at it, thanks