03:31 <_whitenotifier-f> [sv-tests] mithro opened issue #451: Test failing on kokoro - https://git.io/Je6fh

06:30 <_whitenotifier-f> [sv-tests] mithro opened issue #453: Tag not present in the database: basejump and 16.9 - https://git.io/Je6Uz

09:22 <_whitenotifier-f> [symbiflow-arch-defs] kgugala opened issue #1159: Add support for multiple parts in families - https://git.io/Je6IE

09:40 <_whitenotifier-f> [symbiflow-arch-defs] kgugala opened issue #1160: Create make install target - https://git.io/Je6Ij

10:15 <_whitenotifier-f> [symbiflow-arch-defs] mkurc-ant opened issue #1161: Inference of 3-state IOBUFs in Yosys - https://git.io/Je6LD

11:49 <_whitenotifier-f> [python-symbiflow-v2x] glatosinski opened issue #21: Errors during Travis CI tests - https://git.io/Je6q1

13:46 <bunnie> so...i got through step 7 (i think) on the quickstart, I have a lot of data in 072-ordered_wires. But when I run step 9, i get a "no tilegrid.json" found error. What am I doing wrong?

13:48 <bunnie> I've got a "run.ok" in a few directories too now, so something seems to be working, but for example I don't see an entry for spartan7 in the database directory, so I'm worried I've got something configured wrong? The export XRAY_DATABASE is spartan7 in my environment.

13:58 <bunnie> yes, it definitely seems the lack of a database/spartan7/partnumber.yaml file is holding things up. What's the right way for me to go about building the spartan7 database .yaml/.json files?

14:43 <miek> bunnie: it looks like those should be built by 001-part-yaml?

15:25 <_whitenotifier-f> [python-symbiflow-v2x] glatosinski opened issue #22: Changing outputs of the `vlog_to_pbtype` function in Python 3.5 - https://git.io/Je6ZU

15:31 <litghost> bunnie: Ya, 001 generates the partnumber.yaml

15:31 <litghost> bunnie: And any fuzzer that touches the bitstream needs it

15:31 <litghost> bunnie: It is a very fast fuzzer, you should be able to run it

15:33 <litghost> bunnie: FYI, the fuzzer order info is located here; https://github.com/SymbiFlow/prjxray/blob/master/fuzzers/Makefile#L57

15:33 <tpb> Title: prjxray/Makefile at master · SymbiFlow/prjxray · GitHub (at github.com)

15:42 <vaughnbetz> Some info on artix7 optimization quality: router time is down to 29 s and 71 s for murax and picosoc respectively with the Nov. 17 Symbiflow build. The router is using its defaults of astar_fac = 1.2 and bb_factor = 3; critical path delay actually improved a little (5%) vs. the older build we were using. Top 3 routines from gprof are

15:42 <vaughnbetz> https://www.irccloud.com/pastebin/SJGpv0JL/

15:42 <tpb> Title: Snippet | IRCCloud (at www.irccloud.com)

15:43 <vaughnbetz> Keith, should we try again with a larger region of interest? How do we get the larger region of interest?

15:53 <bunnie> ah ok. I see the problem. I first have to mkdir spartan7 in database. Everything breaks without that directory being made first. I thought the errors saying file can't be found was complaining about them not existing in the first place, but in fact the root cause is there is no spartan7 directory.

15:54 <bunnie> for some reason I thought that would be automatic as part of the script.

15:56 <bunnie> yah. things running much better now. :)

16:16 <_whitenotifier-f> [sv-tests] mithro opened issue #456: Tar up the outputs on Kokoro - https://git.io/Je6nF

18:11 <litghost> vaughnbetz: That is true. The full 50T graph is postfixed with "_full"

18:11 <litghost> vaughnbetz: So the 1/5 50T graph picosoc is picosoc_basys_bin

18:12 <litghost> vaughnbetz: The full 50T graph picosoc is picosoc_basys3_full_bin

18:12 <litghost> Sorry type above the 1/5 50T is picosoc_basys3_bin

18:13 <litghost> vaughnbetz: The ROI is no longer present for two target types no, instead we have the idea of a "graph limit", which dynamically limits the imported graph size

18:13 <litghost> vaughnbetz: In the absence of the graph limit it uses the whole part

18:14 <litghost> vaughnbetz: We current have a 2/5 50T graph which is called "xc7a50t-bottom", defined here: https://github.com/SymbiFlow/symbiflow-arch-defs/blob/master/xc7/archs/artix7/devices/xc7a50t-bottom-virt/CMakeLists.txt#L28

18:14 <tpb> Title: symbiflow-arch-defs/CMakeLists.txt at master · SymbiFlow/symbiflow-arch-defs · GitHub (at github.com)

18:15 <litghost> vaughnbetz: The primary reason we are using the bottom two CMT's is because VPR doesn't have clock network specific placement implemented (per https://github.com/verilog-to-routing/vtr-verilog-to-routing/issues/1045#issuecomment-553949121)

18:15 <tpb> Title: Placer awareness of dedicate routing resources · Issue #1045 · verilog-to-routing/vtr-verilog-to-routing · GitHub (at github.com)

18:15 <litghost> vaughnbetz: The PLL <-> BUFG relationship requires special placement requirements

18:16 <litghost> vaughnbetz: We avoid having VPR know about them by only importing the portion of the graph that is valid

18:53 <bunnie> hmm...i'm getting stuck in 005-tilegrid now, the specific issue being that IOBX0Y101 doesn't exist in this part. I checked in the layout editor, and it's not there, but for some reason it's being extracted into the tiles.txt file. I tried adjusting the XRAY_ROY IOB range to IOB_X0Y0:IOB_X0Y49 in the settings file, but that didn't seem to do the trick. Any hints appreciated :)

18:58 <litghost> 005-tilegrid has some specialized logic, did you redefine the part from artix7 to spartan7?

18:59 <litghost> ah, "make_project_roi" in https://github.com/SymbiFlow/prjxray/blob/master/fuzzers/005-tilegrid/generate_tiles.tcl#L37 needs to be replaced

18:59 <tpb> Title: prjxray/generate_tiles.tcl at master · SymbiFlow/prjxray · GitHub (at github.com)

19:00 <litghost> All the stuff about pblocks in https://github.com/SymbiFlow/prjxray/blob/015b04ce90e7c8899933f6b6f54da730a212b202/fuzzers/005-tilegrid/util.tcl#L144 can be removed

19:00 <tpb> Title: prjxray/util.tcl at 015b04ce90e7c8899933f6b6f54da730a212b202 · SymbiFlow/prjxray · GitHub (at github.com)

19:11 <bunnie> thanks. looks like i can just edit utils.tcl's make_iob_sites function to excise the ones that don't exist..i'll give that a try.

23:08 <vaughnbetz> Thanks Keith.