Monday, 2019-07-22

*** tpb has joined #tomu00:00
*** andi- has quit IRC02:54
*** andi- has joined #tomu02:57
xobsSo that was fun. I discovered the USB core had some metastability in it, and about 50% of the boards that we produced could only run on my machine (due to USB host core timing).05:05
xobsI tracked it down and fixed it.05:06
xobsTo prove it was fixed I spent the weekend synthesizing bitstreams with random seeds to try and get them to fail.  With the fix, not a single design (that met timing) failed.05:07
ssbawesome, heh05:10
xobsI'm going to have the factory re-test all the boards and put v1.8.7 on, which has the good firmware on it.05:11
*** rohitksingh_work has joined #tomu05:24
MadHackerxobs: How many failed timing, out of curiosity? Is the process struggling to meet it?08:21
MadHacker(i.e. what %age of seeds)08:21
MadHackerAlso, d'you think it'd work on an HX-series part too? I'd be keen to get USB on the iCE40 novena boards too. Some extra USB ports for sniffing and suchlike would be nice. :)08:22
xobsMadHacker: When I was using 5278/5280 LCs, about 30% were good.  Now that the memory crossbar is fixed in litex, it's down to about 5140 LCs, and about 80% of seeds work.08:23
xobsAnd I definitely think it'd work on the HX-series.  You just need a stable clock and a 1.5k pullup, and that's about it.  That'll get you the wishbone bridge and everything.08:23
MadHackerAye. I can run a bridge across from the Xilinx on the novena to that, then.08:25
MadHackerThe same thing'd work directly on that but for me personally it's less interesting (because I can't build for the S6 on the novena directly).08:25
MadHackerCan export a clock from the novena and could use the S6's RAM as a buffer for USB sniffing, for example.08:26
MadHacker(obviously under-utilising the core's abilities a bit, but still)08:26
MadHackerAnyway. Fun. :)08:26
xobsI wonder if the Raspberry Pi packages I built will work on Novena...08:27
MadHackerLikely to, modulo libc differences.08:27
MadHackerRaspbian + novena kernel would be OK.08:27
xobsI was contemplating getting an autobuilder working where a Novena would clone the repo, build it, and run tests on a Fomu.08:27
MadHackerSince the build it a bit nondeterministic, I'm not sure how I feel about autobuilders for this job.08:28
MadHackerYou'd want some sort of threshold of maximum re-seeds before declaring it a fail.08:28
xobsTrue, but I ran it through 200 seeds last night and had no failures, so I'm considering the problem fixed.08:29
MadHackerAh, I meant more even getting timing closure.08:30
MadHackerThe 80% from above.08:30
xobsAh, right.  Yes, 10 seeds should be more than enough.08:30
*** rohitksingh_work has quit IRC13:50
mithroxobs: Looks like someone is hacking on toboot?15:07
xobsmithro: looks like it!15:09
mithroxobs: I'll leave you to respond?15:09
*** xkapastel has joined #tomu15:18
*** emeb has joined #tomu15:18
mithroxobs: So for camp -- We can start putting together a shared Google Present slide deck15:37
mithroxobs: I think it would be good to attempt to run a Fomu workshop every day at camp if your up for that?15:38
mithroAt one of the villages?15:38
xobsSounds good.  Does Google Present work offline?15:45
pollomithro, xobs: if you merge the toboot PRs and make a new release, please sign them (issue #38) !15:48
pollothat way it'll be easier for me to update the Debian package15:48
xobspollo: alright, I'll sign the tags.15:50
xobsmithro: I'll have 2000 Fomu cases and ~1000 Tomu cases.15:51
mithroxobs: Should work offline...15:53
xobsmithro: I'll be going to bed soon, and Github is having issues so I can't comment on / merge any patches now :(16:02
*** rohitksingh has joined #tomu17:25
*** rohitksingh has quit IRC19:20
*** xkapastel has quit IRC20:38
*** xkapastel has joined #tomu21:05
futarisIRCcloudmithro: camp?23:48

Generated by 2.13.1 by Marius Gedminas - find it at!