Saturday, 2019-01-19

*** tpb has joined #timvideos00:00
mithrohttps://github.com/timvideos/litex-buildenv/wiki/Xilinx-ISE00:00
tpbTitle: Xilinx ISE · timvideos/litex-buildenv Wiki · GitHub (at github.com)00:01
mithropavelow: The setup script should probably check for that...00:06
mithropavelow: The enter-env.sh script already can tell the difference between Vivado and ISE here -> https://github.com/timvideos/litex-buildenv/blob/master/scripts/enter-env.sh#L200-L23300:07
tpbTitle: litex-buildenv/enter-env.sh at master · timvideos/litex-buildenv · GitHub (at github.com)00:07
* tumbleweed told pavelow that he doesn't need any xilinx stuff, if he just uses download-prebuilt00:08
*** TheAssassin has quit IRC00:08
tumbleweedmithro: do you have some spare cycles?00:10
tumbleweedhdmi2usbd and unbind methods00:11
mithrotumbleweed: only some types of cycles....00:11
mithroDo you mean modeswitch?00:12
tumbleweedI mean hdmi2usbd, and (separately) unbind methods in modeswitch00:13
*** TheAssassin has joined #timvideos00:13
tumbleweedhdmi2usbd doesn't seem to work with current opsis. This is the kin dof thing I was talking about when I asked about maintained00:21
* tumbleweed moves on to tally lights00:23
CarlFKhttps://github.com/timvideos/HDMI2USB-litex-firmware00:28
tpbTitle: GitHub - timvideos/HDMI2USB-litex-firmware: A version of the HDMI2USB firmware based around LiteX tools produced by @Enjoy-Digital (based on misoc+migen created by @M-Labs) (at github.com)00:29
mithrotumbleweed: unbind stuff -> https://github.com/timvideos/HDMI2USB-mode-switch/blob/master/udev/99-hdmi2usb-permissions.rules#L15-L2900:38
tpbTitle: HDMI2USB-mode-switch/99-hdmi2usb-permissions.rules at master · timvideos/HDMI2USB-mode-switch · GitHub (at github.com)00:38
mithroewen: Thinking of going fo a walk to get lunch...00:39
ewenmithro: Happy to go for a walk :-)00:40
mithroewen: https://github.com/m-labs/migen/blob/e6ff283a467dc0657019dedadc5b06fa7130b744/migen/fhdl/tracer.py00:41
tpbTitle: migen/tracer.py at e6ff283a467dc0657019dedadc5b06fa7130b744 · m-labs/migen · GitHub (at github.com)00:41
tumbleweedmithro: sooo we have that00:43
* tumbleweed investigates why they don't do what we need00:44
CarlFKtumbleweed: gst-launch-1.0 v4l2src device=/dev/video0 ! jpegdec ! fakesink01:14
mithroewen: Ran 19 tests in 559.027s01:15
ewenmithro: That's around 10 minutes?!01:51
ewenmithro: https://github.com/m-labs/migen/blob/e6ff283a467dc0657019dedadc5b06fa7130b744/migen/fhdl/tracer.py#L65-L70 I think is where it's walking back to find "not good" names.01:52
tpbTitle: migen/tracer.py at e6ff283a467dc0657019dedadc5b06fa7130b744 · m-labs/migen · GitHub (at github.com)01:52
ewenSeems to assume "derived from us" is an indication of "skip this one".01:52
mithroewen: But they all passed!01:53
xobsmithro: he mentioned to me that he got it working!01:58
mithroxobs: He did the same 12mhz clock domain thing01:58
xobsas you can see, 19 MHz doesn't meet timing at all, and is kinda like how NeTV1 could do 1080p rock-solid on about 40% of the boards we shipped.01:58
ewenmithro: All passed is indeed great progress.  Well done!01:59
ewenmithro: I believe I know the root cause of the FSM signal name issue now: https://github.com/mithro/valentyusb/issues/18#issuecomment-45573873102:00
tpbTitle: Fix the resetinserter / clockdomainrenamer names in vcd output · Issue #18 · mithro/valentyusb · GitHub (at github.com)02:00
mithroewen: Yeah02:00
ewenJust need to figure out how to actually fix it.  (As I'd guessed the decorators are *not* being skipped, due to the class hierachy of them / the skipping code).02:00
xobsmithro: though there is a lot of interesting stuff to learn from in there.02:00
mithroxobs: He hasn't updated the pull request info02:01
mithroxobs: But see the latest commits - @osresearch - Tomu: use 12 MHz clock for USB endpoint02:01
mithroxobs: Clock crossing success - endpoint runs at 12 Mhz02:02
xobsoh, that's really cool!02:05
xobsGoing to flash Hacker Boards, then?02:06
mithroxobs: Haven't thought about it yet02:06
ewenmithro: Looks like maybe that bit of the tracer is only being used for naming the *clocks*, not the Signals.  At least based on trying to trace it.  (And the clocks all have overrides.)02:34
ewen(or maybe migen has some cache of things I don't know about?)02:36
cr1901_modernmithro: Is Trammel also working on Tomu USB?02:38
mithrocr1901_modern: Apparently02:38
mithroewen: Hrm?02:38
ewenmithro: Best guess is that it's implicitly done by this bit of code tracer.trace_back: https://github.com/m-labs/migen/blob/e6ff283a467dc0657019dedadc5b06fa7130b744/migen/fhdl/tracer.py#L90-L13602:40
tpbTitle: migen/tracer.py at e6ff283a467dc0657019dedadc5b06fa7130b744 · m-labs/migen · GitHub (at github.com)02:40
ewen(Rather than get_obj_var_name() which I was assuming before.)02:40
mithroewen: Maybe ask in #m-labs ?02:40
CarlFKmithro: hdmi2usb jpeg error: last good version: v0.0.4-399-g33b53ba   next version is v0.0.4-404-g57fb82102:42
mithroCarlFK: Great! that is useful info02:42
ewenmithro: Looks like maybe this is the relevant line: frame = inspect.currentframe().f_back.f_back.02:42
ewenWhich obviously assumes there's exactly two levels to go back.  And a decorator will add another level...02:43
mithroCarlFK: can you log a bug?02:43
mithroewen: Can you just skip things which inherit from the decorator class or something?02:43
ewenmithro: Yes, that's my plan.  I think I'll need to create a flagname SuperClass to be able to detect that.02:48
tumbleweedpavelow: https://github.com/timvideos/litex-buildenv/wiki/Networking03:31
tpbTitle: Networking · timvideos/litex-buildenv Wiki · GitHub (at github.com)03:31
mithroAt a hackfest previously, we get *very* close to making DHCP work on the Opsis03:31
*** Kripton has quit IRC03:45
*** Kripton has joined #timvideos03:46
ewenmithro: decorator skipping kludge looks plausible on my machine, although I'm not 100% sure the resulting name is the best yet (get_fragment -- does that make sense?).03:57
ewenmithro: https://github.com/mithro/valentyusb/issues/18#issuecomment-455746736 has a bunch of details on what I'm getting / how I've kludged it.03:57
tpbTitle: Fix the resetinserter / clockdomainrenamer names in vcd output · Issue #18 · mithro/valentyusb · GitHub (at github.com)03:57
ewen(I'm not really happy with how it's done -- it's definitely a kludge -- but if the results are useful maybe we find a way to do this?)03:57
mithroewen: functools.wraps ?03:59
mithroewen: https://docs.python.org/3/library/functools.html#functools.wraps04:00
tpbTitle: functools — Higher-order functions and operations on callable objects Python 3.7.2 documentation (at docs.python.org)04:00
ewenmithro: Isn't that for *defining* a decorator?  Here we want to *detect* that a class in the call stack *is* a decorator (and thus that we do not care about its name, we care about something else)04:10
ewen(I'm also not sure whether we want to ignore *all* decorators.  Or just these certain magic ones that insert extra Signals...)04:11
mithroewen: Well, I was wondering if it did something to the stack to was detectable in some way04:11
ewenLooks like if they *are* using functools.wraps *and* we assume Python 3.2+ there's an attribute we can look for: "New in version 3.2: Automatic addition of the __wrapped__ attribute."04:17
ewen(But I suspect migen is still trying to be compatible with Python 2.7...)04:18
ewenOh, and it appears get_fragment name is coming from https://github.com/m-labs/migen/blob/57c44674c2a6c38bd01804e69db61a9efd287524/migen/fhdl/decorators.py#L25 within the wrapper :-(04:20
tpbTitle: migen/decorators.py at 57c44674c2a6c38bd01804e69db61a9efd287524 · m-labs/migen · GitHub (at github.com)04:20
ewenmithro: So it appears I might need to move the "skip this stack frame" earlier.04:20
ewenmithro: Revised version, which gets more sensible results for the names: https://github.com/mithro/valentyusb/issues/18#issuecomment-45574888704:43
tpbTitle: Fix the resetinserter / clockdomainrenamer names in vcd output · Issue #18 · mithro/valentyusb · GitHub (at github.com)04:43
*** springermac has quit IRC04:57
*** springermac has joined #timvideos04:58
thaytanhiyas05:03
*** ewen has quit IRC05:10
CarlFKthaytan: hi!!05:23
CarlFKoff to dinner!05:23
thaytancya :)05:23
mithrothaytan: you in Christchurch yet?05:31
thaytanmithro, yes, we're in the pavilions hotel for the week05:34
thaytanwhere are you all holed up?05:34
mithroWill be in the uni accom from tomorrow morning05:44
mithroWant an FPGA Tomu?05:44
*** rohitksingh has joined #timvideos05:48
thaytanmithro, yes!05:50
thaytanmithro, any more hacking tonight?05:51
mithroProbably, at dinner now05:51
*** rails[I] has joined #timvideos05:58
*** rails[I] has quit IRC05:59
*** rails[I] has joined #timvideos05:59
*** rails[I] is now known as evilRails05:59
*** evilRails has joined #timvideos06:00
*** ewen has joined #timvideos06:49
*** rohitksingh has quit IRC07:40
mithrohttps://hackaday.com/2019/01/18/hackaday-podcast-ep2-curious-gadgets-and-the-fpga-brain-trust/07:45
ewenmithro: Updated attempt, using __wrapped__ attributes (in semi-random locations): https://github.com/mithro/valentyusb/issues/18#issuecomment-45575872708:05
tpbTitle: Fix the resetinserter / clockdomainrenamer names in vcd output · Issue #18 · mithro/valentyusb · GitHub (at github.com)08:05
mithrothaytan: Take a look at https://github.com/timvideos/litex-buildenv/wiki/HowTo-FuPy-on-iCE40-Boards08:07
tpbTitle: HowTo FuPy on iCE40 Boards · timvideos/litex-buildenv Wiki · GitHub (at github.com)08:07
*** evilRails has quit IRC08:08
*** rohitksingh has joined #timvideos08:12
*** rohitksingh has quit IRC08:40
mithrothaytan: https://en.wikipedia.org/wiki/Speaker_diarisation09:08
tpbTitle: Speaker diarisation - Wikipedia (at en.wikipedia.org)09:08
*** ewen has quit IRC09:16
nats`morning !10:03
*** rohitksingh has joined #timvideos10:08
*** ewen has joined #timvideos10:17
*** ewen has quit IRC10:22
mithrohttps://makecode.com/blog/one-chip-to-flash-them-all10:26
tpbTitle: One chip to flash them all - Microsoft MakeCode (at makecode.com)10:26
xobsBoarding now. See people tomorrow in Christchurch.11:38
mithroGetting closer! https://www.irccloud.com/pastebin/kst0YJgO/12:40
tpbTitle: Snippet | IRCCloud (at www.irccloud.com)12:40
*** Kripton has quit IRC18:00
*** Kripton has joined #timvideos18:16
*** rohitksingh has quit IRC18:19
*** CarlFK has quit IRC18:49
*** CarlFK has joined #timvideos19:21
*** ChanServ sets mode: +v CarlFK19:21
CarlFKmithro: coffee run? leaving in about .. now.19:30
mithroCarlFK: I won't be ready from ~20mins19:31
*** ewen has joined #timvideos19:53
*** ewen has quit IRC20:11
*** CarlFK has quit IRC20:24
mithroIf anyone is at Linux.conf.au 2019 I have FPGA Tomu boards and other give aways for people!20:38
railsoooh tomu boards22:07
*** CarlFK has joined #timvideos23:16
*** ChanServ sets mode: +v CarlFK23:16
*** ewen has joined #timvideos23:55

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