03:37 <sb0> whitequark: do you have any recommendations on how to stream logic analyzer data from one of those Saleae devices (https://sigrok.org/wiki/Saleae_Logic) into python?

03:37 <whitequark> sb0: well... sigrok? it has python bindings iirc

03:38 <whitequark> yeah sure https://sigrok.org/api/libsigrok/unstable/bindings/python/annotated.html

03:38 <sb0> I want a continuous stream of data (analyzer continuously sampling without glitches), ideally passed into numpy arrays some thousands samples deep

03:38 <sb0> okay and that sigrok lib works correctly? have you tried it? i saw you reimplemented many things yourself in glasgow...

03:39 <whitequark> glasgow is orthogonal to sigrok though, no?

03:40 <whitequark> it pays off to reimplement things in glasgow because i can use python to share code between gateware and software, so i don't have to e.g. add some complicated serialization to explain sigrok what channels, sample rates, etc I need. also sigrok doesn't work well with glasgow's internal logic analyzer because it doesn't support multibit signals

03:40 <whitequark> and glasgow's LA shows you FIFO reads and writes

03:41 <sb0> ok

03:41 <whitequark> anyway, i didn't try that part of sigrok's python bindings myself

03:41 <whitequark> i only really wrote decoders for it

03:43 <whitequark> I suppose you could also use sigrok-cli to stream data into an unix FIFO in straight binary, or something like that

04:05 <mtrbot-ml> [mattermost] <sb10q> Hmm I'll see. For that FIFO approach, Python I/O seems a bit slow, and I need at least 5Msps on 3 channels

04:06 <mtrbot-ml> [mattermost] <sb10q> Something that writes numpy arrays directly and efficiently sounds safer

04:10 <whitequark> 15 MB/s?

04:10 <whitequark> I easily hit 30 MB/s with Python, libusb, and glasgow

04:10 <whitequark> and that's without any numpy arrays, just regular Python memoryview, and some pretty crappy code too

04:11 <whitequark> s/30/40/, actually

04:11 <whitequark> wait, no, 3 channels fit into 1 byte, so it'd be just 5 MB/s

04:11 <whitequark> that's completely trivial

04:16 <mtrbot-ml> [mattermost] <sb10q> Well OK, but you need to separate them afterwards

04:16 <mtrbot-ml> [mattermost] <sb10q> Anyway I can benchmark that easily

04:16 <whitequark> sure, use numpy for that

07:19 <sb0> okay, this sigrok stuff just works out of the box. nice!

07:20 <sb0> no firmware to compile, fxload, etc.

07:20 <whitequark> yes, sigrok is quite good. unfortunately it is badly understaffed.

07:21 <whitequark> the core needs some significant refactoring to really get to its full potential and there's not enough people working on it

07:21 <whitequark> e.g. pulseview ought to be better than gtkwave. and decoders should not choke on 100Ms of mostly idle data points.

10:34 <_whitenotifier-3> [nmigen] whitequark opened issue #160: pysim is very slow - https://git.io/fjDdl

10:35 <_whitenotifier-3> [nmigen] whitequark opened issue #161: pysim should have a way to reset user/sync signals - https://git.io/fjDd8

17:05 <isengaara-tomoko> When using NMigen I get UnusedElaboratable warnings, meaning that

17:06 <isengaara-tomoko> class Fragment: @staticmethod def get(obj, platform): while True: if isinstance(obj, Fragment): return obj elif isinstance(obj, Elaboratable): obj._Elaboratable__used = True

17:06 <isengaara-tomoko> is never called

17:07 <whitequark> indeed. any elaboratables that are created but not synthesized display this message

17:12 <isengaara-tomoko> so I just need to connect each Elaboratables output

17:13 <isengaara-tomoko> I currently working on a unittest for a TLB, and all inner modules have already been tested

17:13 <whitequark> are you adding the modules that show UnusedElaboratable warning as submodules somewhere?

17:13 <whitequark> that's the bug this warning is supposed to guard against

17:19 <isengaara-tomoko> I think yes, the modules are added as submodules

17:20 <isengaara-tomoko> however the code was generated using a systemverilog to nmigen converter and had to be modified manually to run inside nmigen

21:46 <cr1901_modern> sb0: I sent a privmsg and email when you get the chance.