09:50 <_whitenotifier> [sv-tests] piotr-binkowski opened issue #381: Remove blacklist from Basejump STL generator config - https://git.io/Jeui9

20:58 <hackerfoo> litghost: I'll try to get a PR by the end of the day, but here's the branch I've been working on: https://github.com/HackerFoo/vtr-verilog-to-routing/tree/wip-testing

20:58 <tpb> Title: GitHub - HackerFoo/vtr-verilog-to-routing at wip-testing (at github.com)

21:46 <litghost> hackerfoo: Is that an integrated master+wip branch?

22:06 <hackerfoo> litghost: Yes. I still need to rebase onto the upstream master.

22:07 <litghost> hackerfoo: I expect you will be able to rebase nicely onto https://github.com/SymbiFlow/vtr-verilog-to-routing/tree/connection_box_serdes_merge_squashed

22:07 <tpb> Title: GitHub - SymbiFlow/vtr-verilog-to-routing at connection_box_serdes_merge_squashed (at github.com)

22:07 <hackerfoo> litghost: Okay, I'll do that instead.

22:09 <litghost> hackerfoo: What you need to do is:

22:09 <litghost> git rebase --onto a56cb11fbff1eadac737f4a6edce341af333c1a6 a81d5eab19f179767210096f4441f22e1a5cff32 <your branch name>

22:09 <litghost> a56cb11fbff1eadac737f4a6edce341af333c1a6 is https://github.com/SymbiFlow/vtr-verilog-to-routing/tree/connection_box_serdes_merge_squashed

22:09 <tpb> Title: GitHub - SymbiFlow/vtr-verilog-to-routing at connection_box_serdes_merge_squashed (at github.com)

22:10 <litghost> a81d5eab19f179767210096f4441f22e1a5cff32 should be the old https://github.com/SymbiFlow/vtr-verilog-to-routing/tree/connection_box_serdes_merge_squashed

22:10 <litghost> This is assuming your are on a feature branch on top of the old https://github.com/SymbiFlow/vtr-verilog-to-routing/tree/connection_box_serdes_merge_squashed

22:10 <tpb> Title: GitHub - SymbiFlow/vtr-verilog-to-routing at connection_box_serdes_merge_squashed (at github.com)

22:10 <litghost> If you are not, then some of what I wrote is inaccurate

22:15 <hackerfoo> I'm looking for the commit.

22:17 <hackerfoo> No, that wasn't in my parallel-connection-box-lookahead branch yet.

22:18 <litghost> Okay, well replace a81d5eab19f179767210096f4441f22e1a5cff32 with whatever your "root" commit on the feature branch was

22:48 <hackerfoo> I've rebased my parallel-connection-box-lookahead branch.

22:48 <hackerfoo> Now I'm going to try it out, then check it over again and submit a PR.

23:30 <litghost> hackerfoo: What graph's have you been testing on?

23:30 <litghost> hackerfoo: I just tested on the full 50T, and got:

23:30 <litghost> Failed to route connection from '$false' to '$auto$alumacc.cc:485:replace_alu$2737.slice[2].carry4_full' for net 'murax.system_uartCtrl.uartCtrl_1.tx.io_configFrame_stop' (#0)

23:30 <litghost> Cannot route from SYN-GND.GND[0] (RR node: 878 type: SOURCE location: (3,1) class: 0) to BLK-TL-SLICEL.BX[0] (RR node: 1922176 type: SINK location: (95,121) class: 13) -- no possible path

23:31 <hackerfoo> The ROI and full graph, and I've been testing murax.

23:31 <litghost> Ya, same

23:32 <litghost> What is your symbiflow-arch-def githash?

23:32 <litghost> Mine is 880a06e840dc7d96d72461f2acb4fcd5e7ef580c merged with c427fe0b4645560df51160881d3cde84725e7dd5 (e.g. master)

23:32 <litghost> I was testing with VTR @ 0477b874ad1be65626c99ae13f707586a7dae116

23:35 <hackerfoo> arch-defs @ 5d68d4e8e82a3d36d6b9fb26d95038105afc5679

23:35 <hackerfoo> With this patch: https://gist.github.com/HackerFoo/76989afd10f4f75d10226c91e90814d9

23:35 <tpb> Title: remove-graph-limit.diff · GitHub (at gist.github.com)

23:35 <litghost> Right, I was about to ask about that

23:36 <litghost> 5d68d4e8e82a3d36d6b9fb26d95038105afc5679 doesn't have the graph limit removed

23:36 <hackerfoo> building murax_basys_x1y0_fasm

23:36 <litghost> Same target on my side, that produced the above error

23:37 <litghost> How many sample points did your code have?

23:37 <hackerfoo> I'm working on the octopus merge right now, so I haven't run the new branch yet.

23:37 <hackerfoo> 3x3

23:37 <litghost> grep "Expanded sample" xc7/archs/artix7/devices/vpr_stdout.log | wc

23:37 <litghost> 476 3808 27216

23:38 <litghost> ok

23:39 <hackerfoo> It sorts the points by the index of the first point to hopefully be more cache friendly.

23:39 <hackerfoo> At least, there's no reason to group the expansion per segment type anymore.

23:41 <hackerfoo> Each `SamplePoint` is a group of the same segment type, though.

23:42 <hackerfoo> So it should be 3 * 3 * segment types * connection box types, roughly.