| *** tpb <[email protected]> has joined #yosys | 00:00 | |
| *** adjtm <[email protected]> has quit IRC (Quit: Leaving) | 01:23 | |
| *** Lord_Nightmare <Lord_Nightmare!Lord_Night@user/lord-nightmare/x-3657113> has quit IRC (Quit: ZNC - http://znc.in) | 01:50 | |
| *** Lord_Nightmare <Lord_Nightmare!Lord_Night@user/lord-nightmare/x-3657113> has joined #yosys | 01:52 | |
| *** bq <[email protected]> has quit IRC (Quit: Floating point exception (core dumped)) | 06:12 | |
| *** phillid <[email protected]> has joined #yosys | 06:13 | |
| *** phillid is now known as bq | 06:13 | |
| *** emeb_mac <[email protected]> has quit IRC (Quit: Leaving.) | 07:05 | |
| corecode | oh duh, of course | 09:59 |
|---|---|---|
| *** adjtm <[email protected]> has joined #yosys | 10:29 | |
| *** kraiskil <[email protected]> has joined #yosys | 10:32 | |
| corecode | LUT4 are just not so good for MUXes, it seems | 11:07 |
| corecode | xdot complains about error: unknown xdot opcode '0]' for this line that yosys generates: n4 [ shape=diamond, label="$0\\counter[23:0]" ]; | 11:08 |
| corecode | seems it is related to the \\ | 11:09 |
| *** kraiskil <[email protected]> has quit IRC (Ping timeout: 260 seconds) | 11:35 | |
| corecode | oh it's a bug in graphviz. | 11:45 |
| *** emeb_mac <[email protected]> has joined #yosys | 12:51 | |
| corecode | it's a bug that was fixed in graphviz, but xdot didn't release yet since. | 12:51 |
| corecode | it would be nice if show could collapse busses with the same structure | 13:02 |
| corecode | oh cool i found a segfault in yosys | 13:18 |
| corecode | https://github.com/YosysHQ/yosys/issues/3317 | 13:33 |
| *** skipwich <skipwich!~skipwich@user/skipwich> has quit IRC (Quit: DISCONNECT) | 13:44 | |
| *** skipwich <skipwich!~skipwich@user/skipwich> has joined #yosys | 13:57 | |
| whitequark | flowmap isn't... very functional | 14:52 |
| whitequark | (i wrote flowmap) | 14:52 |
| whitequark | there's a few issues with flowmap -relax it that basically require a complete rewrite | 14:52 |
| whitequark | without -relax it should work but i'm not surprised there are other bugs in it | 14:53 |
| *** kraiskil <[email protected]> has joined #yosys | 15:05 | |
| corecode | hi | 15:25 |
| corecode | yea i saw experimental, but i figured it shouldn't crash with what's likely an over/underflow | 15:25 |
| Sarayan | flowmap is what, a synthesis methodology? | 15:33 |
| *** kraiskil <[email protected]> has quit IRC (Ping timeout: 240 seconds) | 15:58 | |
| lofty | Sarayan: it's a LUT mapper | 16:08 |
| lofty | I've actually written a few, and I am kind of surprised how bloody difficult it is to get right | 16:09 |
| Sarayan | must be fun | 16:12 |
| Sarayan | if a pain | 16:12 |
| *** kraiskil <[email protected]> has joined #yosys | 16:48 | |
| corecode | lofty: difficult not to make mistakes? | 17:26 |
| lofty | corecode: there are a lot of invariants that are difficult to articulate in such a way that one can meaningfully test them | 17:26 |
| *** peeps[zen] <peeps[zen]!~peepsalot@openscad/peepsalot> has joined #yosys | 17:32 | |
| *** peepsalot <peepsalot!~peepsalot@openscad/peepsalot> has quit IRC (Ping timeout: 260 seconds) | 17:33 | |
| corecode | interesting | 17:38 |
| corecode | i would have thought that transforming a set of logic should be easy to test for equivalence | 17:42 |
| lofty | Oh, it is | 17:58 |
| lofty | But equivalent does not mean bug free | 17:58 |
| lofty | This is not strictly relevant to FlowMap because it has a different working methodology, but ABC needs "reference counters", and these represent how many best LUTs at a node use this node as an input | 18:01 |
| lofty | ... Sorry, I should rephrase that: these represent how many nodes use this node as an input to their best LUT | 18:01 |
| lofty | When a LUT is selected as the best at a node, it increments the reference counters of the inputs to that LUT | 18:03 |
| lofty | When the mapper changes its mind, it has to decrement the reference counters of the inputs to its previous-best | 18:03 |
| lofty | Now, suppose you have a bug somewhere where you decrement a reference counter when you shouldn't | 18:05 |
| lofty | You'll only discover that there was a mistake when you decrement the reference counter when you should have | 18:05 |
| lofty | Which means failure cases are highly nonlocal | 18:06 |
| *** ec <ec!~ec@gateway/tor-sasl/ec> has joined #yosys | 18:46 | |
| *** kraiskil <[email protected]> has quit IRC (Ping timeout: 240 seconds) | 18:58 | |
| corecode | i see | 20:50 |
| corecode | thanks | 20:50 |
| *** cr1901 <cr1901!~cr1901@2601:8d:8600:911:d2:685c:f1e:7edb> has quit IRC (Ping timeout: 252 seconds) | 21:18 | |
| *** peeps[zen] <peeps[zen]!~peepsalot@openscad/peepsalot> has quit IRC (Quit: Connection reset by peep) | 21:40 | |
| *** peepsalot <peepsalot!~peepsalot@openscad/peepsalot> has joined #yosys | 21:56 | |
| *** nonchip <[email protected]> has quit IRC (Quit: https://quassel-irc.org - Chat comfortably. Anywhere.) | 22:12 | |
| *** nonchip <[email protected]> has joined #yosys | 22:12 | |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!