sb0 changed the topic of #m-labs to: ARTIQ, Migen, MiSoC, Mixxeo & other M-Labs projects :: fka #milkymist :: Logs http://irclog.whitequark.org/m-labs
sb0 has quit [Remote host closed the connection]
sb0 has joined #m-labs
<sb0> rjo, set a watchdog and the master will terminate the worker
<rjo> then the keepalive stuff is only used to evict hung/crashed workers.
<rjo> (from the core device side)
<sb0> disconnected network cables, OS crashes, etc.
<sb0> the synthnv is remarkably buggy
<sb0> connecting it via a USB hub makes its protocol timeout due to the increased USB latency
<sb0> connecting it directly exceeds the USB power budget of some ports
<sb0> the result of the protocol timeout is, of course, incorrect values being programmed for obscure reasons
<rjo> what is the usb-to-spi chip in there?
<sb0> the protocol I'm talking about is what goes on top of serial
<sb0> and delimitates each command
<rjo> sure. but do you know what the chip is?
<sb0> I don't know... and is it really usb-to-spi? it's detected as a ACM device on Linux, and the commands are ASCII based
<rjo> re idle experiment: that falls apart if the master and a worker hang.
<rjo> the synthezier is spi iirc.
<rjo> i guess it is some µP.
<rjo> that does the protocol
<rjo> they could have just piped spi over a random ftdi chip.
<sb0> and some custom firmware
<sb0> ATmega32U4
<sb0> "programmed with Windfreak Technologies proprietary firmware"
<rjo> blarg.
<sb0> yes, if there is nothing to enforce the watchdogs, it'll break
<rjo> it will fix itself when we have the possibility to manage a little "kernel queue" on the core device. then the core device would just run idle experiments when the queue is empty.
<sb0> you can still crash it e.g. if you RPC a tight loop
<sb0> also, this kernel queue is incompatible with smooth handover
<rjo> if the master+worker hang, the rpcs would trigger the on-device watchdog, no?
<sb0> (as the point of the latter is to execute host code between kernels, and that host code will determine what kernel is next)
<rjo> what is smooth handover? seamless handover?
<sb0> yes
<rjo> seamless as in "no idle exp" or as in "zero dead time"?
<sb0> obviously you cannot run the idle exp between those, as it would insert additional rtio events
<rjo> ah. we would need something like the guard time/buffer in the fifos. only "commit" an experiment when now() is within guard time.
<sb0> rjo, how trustworthy is the ad9914 hardware wrt keeping the magic smoke inside the kc705?
<sb0> I still haven't powered it
<sb0> 12V and all that ...
anthony has joined #m-labs
anthony is now known as Guest80466
Guest80466 is now known as bentley`
rjo_ has joined #m-labs
<rjo_> sb0: ha. we run that hardware on a few zedboards and i don't think there has been loss with those setups.
mumptai has joined #m-labs
fengling has quit [Ping timeout: 245 seconds]
fengling has joined #m-labs
cr1901_modern has quit [Read error: Connection reset by peer]
mumptai has quit [Remote host closed the connection]
rjo_ has quit [Ping timeout: 244 seconds]
bentley` has quit [Ping timeout: 256 seconds]
<GitHub121> [artiq] fallen pushed 1 new commit to master: http://git.io/vsndt
<GitHub121> artiq/master e6cc041 Yann Sionneau: conda: give 10 more minutes to ISE to achieve routing
fengling has quit [Ping timeout: 245 seconds]
travis-ci has joined #m-labs
<travis-ci> m-labs/artiq#421 (master - e6cc041 : Yann Sionneau): The build has errored.
travis-ci has left #m-labs [#m-labs]
<ysionneau> ah crap
<GitHub147> [artiq] fallen pushed 1 new commit to master: http://git.io/vsceY
<GitHub147> artiq/master c6bba4d Yann Sionneau: Revert "conda: give 10 more minutes to ISE to achieve routing"...
<ysionneau> oh, seems like someone is doing automated conda builds on windows for artiq?
<ysionneau> I'm seeing win32 builds being uploaded
travis-ci has joined #m-labs
<travis-ci> m-labs/artiq#422 (master - c6bba4d : Yann Sionneau): The build passed.
travis-ci has left #m-labs [#m-labs]
FabM has quit [Quit: ChatZilla 0.9.92 [Firefox 40.0.2/20150812163655]]
<ysionneau> or maybe it's just a date issue on anaconda.org website
FabM has joined #m-labs
<mithro> ysionneau: some time when I'm not about to give a lightning talk, I'd like to chat about conda
<ysionneau> sure :)
<GitHub65> [artiq] fallen pushed 1 new commit to master: http://git.io/vsC8z
<GitHub65> artiq/master df7489f Yann Sionneau: travis/conda: add automated builds for kc705 nist_qc2 subtarget
chiggs has quit [Quit: WeeChat 0.4.2]
bentley` has joined #m-labs
rjo_ has joined #m-labs
rjo_ has quit [Client Quit]
<GitHub177> [artiq] fallen pushed 1 new commit to master: http://git.io/vsWWB
<GitHub177> artiq/master 646721d Yann Sionneau: artiq_flash: add missing semicolons
travis-ci has joined #m-labs
<travis-ci> m-labs/artiq#424 (master - 646721d : Yann Sionneau): The build has errored.
travis-ci has left #m-labs [#m-labs]
cr1901_modern has joined #m-labs
mumptai has joined #m-labs
<GitHub187> [misoc] enjoy-digital pushed 1 new commit to master: http://git.io/vslpX
<GitHub187> misoc/master 4acab79 Florent Kermarrec: sdram/module: cleanup indent
<cr1901_modern> Has anyone ever dealt with vvp infinite loops? Having significant trouble debugging why my simulation outright refuses to run to completion.
_whitelogger has joined #m-labs
mumptai has quit [Remote host closed the connection]
travis-ci has joined #m-labs
<travis-ci> m-labs/artiq#424 (master - 646721d : Yann Sionneau): The build passed.
travis-ci has left #m-labs [#m-labs]