12:34 <bunnie> hiya! can anyone point me to the docs about how the ROIs are defined for a given chip? e.g. in the settings/artix7.sh, certain slice ranges are noted to be of interest for things like CLBs or BRAMs. thanks!

14:15 <hackerfoo> bunnie: We are in the process of removing the need for an ROI. We can implement designs on a full 50T now, but it's slow.

14:15 <hackerfoo> So we are working on improving routing time.

15:59 <bunnie> ah i see. the reason I'm asking is I wanted to do a small experiment to look at a device you're not currently looking at (Spartan7) without bothering you folks too much, so I'm trying to extract ROI parameters for that device

16:00 <bunnie> so I'd like to figure out how configure a custom settings for the Spartan7, basically, without causing too much of a distraction for y'all...

16:41 <litghost> bunnie, gotcha

16:43 <litghost> bunnie: So one thing to watch out for is what the Spartan7 fabric looks like compared with the Artix7/Kintex7/Zynq7. The A7/K7/Z7 parts share almost all of their structure and bitstream information, I don't know if Spartan7 is the same

16:44 <litghost> bunnie: To answer your specifiic question, the ROI harnesses are generated and initially defined here: https://github.com/SymbiFlow/prjxray/tree/master/minitests/roi_harness

16:44 <tpb> Title: prjxray/minitests/roi_harness at master · SymbiFlow/prjxray · GitHub (at github.com)

18:52 <mithro> hzeller[m]: https://github.com/SymbiFlow/conda-packages

18:52 <tpb> Title: GitHub - SymbiFlow/conda-packages: Conda build recipes for the toolchains needed by LiteX / MiSoC firmware (at github.com)

18:54 <mithro> bunnie: I would suggest just starting with seeing if the fuzzers just work for spartan7 at https://github.com/SymbiFlow/prjxray

18:54 <tpb> Title: GitHub - SymbiFlow/prjxray: Documenting the Xilinx 7-series bit-stream format. (at github.com)

18:59 <bunnie> mithro: right, i started on that path but then I was wondering how to configure the ROI. are you suggesting basically just to change the part number info at the top, and then maybe just running the fuzzer and seeing if it does something?

19:00 <daveshah> I wouldn't be surprised if Spartan 7 and Artix 7 (at least where one density is in both families) were the same dice; albeit with broken transceivers in the spartan7 case

19:00 <mithro> bunnie: That would be how I go about it - I'm pretty sure you'll hit an error pretty quickly but working from an error is probably easier

19:13 <bunnie> ok. i'll give it a try. thanks!

19:28 <litghost> FYI, the ROI isn't used very much during fuzzing, so I wouldn't stress about it too much

19:36 <bunnie> cool. hasn't crashed yet :)

19:37 <litghost> If you create a PR with the structure for a spartan7 database, that would be awesome

20:03 <mithro> bunnie: If you send an initial PR, I'm more than happy to set up the CI and stuff too

20:03 <mithro> Spartan 7 support has been something we wanted to do but just didn't have the bandwidth