Saturday, 2019-01-12

*** tpb has joined #symbiflow00:00
litghostlong time00:05
litghostI agree that get_nodes is the likely culprit again00:05
litghostthere is more happening in that tcl though00:05
litghostmight be worth breaking up the sites from the tiles00:06
litghostor something along those lines00:06
litghostthere are some straight forward split points00:06
nats`yep what I'm looking :)00:06
nats`and in fact it seems the tcl interpreter from vivado is even worse than I thought00:06
nats`unsetting vars for example doesn't seem to free memory at all....00:07
nats`at least for those I tested get_nodes get_pips get_tiles etc....00:07
nats`litghost, I realized something could be broken in my 072 fuzzer !00:42
nats`easy to fix00:42
nats`if the number of pips is not always a multiple of 6400:43
litghostah00:49
nats`the artix works because the numbe rof pips is 64n00:51
nats`I wrote a patch I'll test it against artix to verify it works and I would like to test against other target is it possible ?00:51
nats`ok I can test with kintex01:15
nats`I do it right now01:16
nats`litghost, do you have the sha1 of kintex file ?01:20
litghostI don't have one handy01:24
nats`I fixed it for non integer multiple01:25
nats`it doesn't totally optimize you can end with a last block of few index to browse but it shouldn't matter a lot01:25
mithronats`: Do you have to call the gc in some way?01:56
nats`nop02:02
nats`I simply start new instance of vivado02:02
nats`mithro, you don't have the fuzzer 072 built for kintex with file ?02:03
nats`real    41m2,012s02:05
nats`user    135m58,549s02:05
nats`sys     22m59,056s02:05
nats`time to build 072 for kintex 7002:05
nats`(env) [email protected]:~/project/symbiflow/prjxray/fuzzers/072-ordered_wires$ sha1sum build/specimen_001/downhill_wires.txt02:06
nats`c32830d20ab25241eb560fa3c96302532f5e037a  build/specimen_001/downhill_wires.txt02:06
nats`(env) [email protected]:~/project/symbiflow/prjxray/fuzzers/072-ordered_wires$ sha1sum build/specimen_001/uphill_wires.txt02:06
nats`490f19862d57c506a4ed7a945c295d2bcc73fceb  build/specimen_001/uphill_wires.txt02:06
nats`I'll open a new PR with the fix to be tested02:06
nats`is there a lighter way to push a fix than a PR ?02:11
mithronats`: What is so heavy weight about a PR?02:15
nats`nothing really heavy but I was wondering when you have to push a small fix of few lines if there are other ways02:16
*** citypw has joined #symbiflow02:19
nats`mithro litghost https://github.com/SymbiFlow/prjxray/pull/50702:24
tpbTitle: Fixing Fuzzer 072: supporting all pips count by natsfr · Pull Request #507 · SymbiFlow/prjxray · GitHub (at github.com)02:24
nats`if this one is confirmed I'll continue on 07402:25
nats`I realized my mistake while working on that one02:26
nats`because it has odd num of tiles :D02:26
nats`mithro, I restarted a kintex test with the change you told me03:05
nats`but I'll go to sleep it's 4am here :D03:05
nats`will push that tomorrow03:05
mithronats`: Yes, sounds like a good idea!03:07
*** _whitelogger has quit IRC05:24
*** _whitelogger has joined #symbiflow05:27
nats`good morning09:30
nats`finishing to fix the 072 mistake I made it should be good after10:03
nats`for a reason I don't understand near the end python only made thread one by one oO10:51
nats`real    51m48,687s10:51
nats`user    138m34,724s10:51
nats`sys     23m25,369s10:51
nats`litghost, mithro it should be good now sorry for the mistake :)11:58
*** Qbix_ has joined #symbiflow16:04
Qbix_hi all16:33
digshadowhi Qbix_19:01
*** Rahix has quit IRC19:42
*** Rahix has joined #symbiflow19:43
*** Rahix_ has joined #symbiflow19:53
*** Rahix has quit IRC19:54
*** _whitelogger has quit IRC21:03
*** _whitelogger has joined #symbiflow21:06
*** Qbix_ has quit IRC21:21
nats`litghost, are you here ? I just came back home and I have question about your change request21:43
*** Rahix_ has quit IRC22:35
nats`ping litghost you accepted change but I'm not sure it's good I still doesn't understand what you suggested23:07
nats`I just saw the comment oky :)23:17
nats`someone could test that fuzzer to have a definitive answer :)23:17
nats`I want to know if in all case it generates the same file as former fuzzer23:17
nats`I made an other run with the kintex and it took about 39 minutes23:25
nats`so maybe at one time because of cpu use the pool was used differently23:25

Generated by irclog2html.py 2.13.1 by Marius Gedminas - find it at mg.pov.lt!