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
mumptai has quit [Quit: Verlassend]
sb0 has quit [Read error: Connection reset by peer]
bentley` has quit [Remote host closed the connection]
bentley` has joined #m-labs
<mindrunner>
i am trying to copy my development environment from a gentoo box to a ubuntu virtual machine. Everything looks pretty much okay, but when flashing the bios/bitstream with the ubuntu box, the pipistrello board does not boot correctly. I am getting the follofing error: Memtest failed: 66/532480 words incorrect, Memory initialization failed. i have absolutely no idea what the reason could be. Anyone can point me to the right direction?
aeris has quit [Quit: en a pas]
acathla has quit [Changing host]
acathla has joined #m-labs
<mindrunner>
it seems like i tracked down the issue a bit further. the ubuntu box is on migen/misoc HEAD, whereas my gentoo dev system still on older commits is (migen 73a1687, misoc: 130fd19). using the older source on the ubuntu system makes it working! maybe a regression somewhere?
imrehg has joined #m-labs
imrehg has quit [Quit: Leaving]
aeris has joined #m-labs
_florent_ has joined #m-labs
_MiW is now known as MiW
sb0 has joined #m-labs
<sb0>
mindrunner, or some intermittent ise bug. can you bisect it?
ysionneau has quit [Ping timeout: 256 seconds]
ysionneau has joined #m-labs
<mindrunner>
yep, can do. how is migen and misoc related to each other in terms of the commits? Should i bisect misoc or migen? or both at the same time? whats best practice>
<mindrunner>
sb0,
<sb0>
mindrunner, there is no best practice and bisecting that is a mess
<sb0>
which is another reason we should move most stuff into misoc and stabilize the migen api...
<mindrunner>
:)
<mindrunner>
will see what i can do
<mindrunner>
the most recent commits are related to sdram, so just guessing that one of these is the problem
bhamilton has joined #m-labs
<_florent_>
mindrunner: do you always have 66/532480 errors? (memtest command in BIOS)
<_florent_>
if not, it's probably more related to timing issues than functionnal changes
<mindrunner>
it is always thar
sb0 has quit [Quit: Leaving]
<mindrunner>
_florent_, its always that error, but different numbers (60,66,...)
<mindrunner>
havent tried the memtest command yet.
<mindrunner>
will report back later
<_florent_>
OK, so I think it's related to timings
<_florent_>
you can check the timing report (if generated by ISE), or run timingan
<_florent_>
but it can also be a timing that is not correctly constrained...
<rjo_>
ysionneau: if you want to try other ci tools, you can have a look at circle-ci. no windows either though.
rjo_ is now known as rjo
<ysionneau>
ah thanks
<ysionneau>
I had a very quick look at AppVeyor also
<ysionneau>
which supports Windows
<ysionneau>
didn't have enough time to make it work yet though
<rjo>
i noticed. what about the binstar build service?
<rjo>
and iirc for non-binary python packages you could usually also build them under linux for windows targets
<ysionneau>
yep
<ysionneau>
I should have a look yes at binstar build service
<ysionneau>
don't know exactly how it works yet
<ysionneau>
Binstar build provides everyone with a free linux-64 build VM < ok it's linux-64 only
<rjo>
ah. i thought it was windows as well.
<ysionneau>
ah, you're right, but one need to pay for that
<ysionneau>
With a paid subscription, you can also create windows, osx or any other machine type workers that you want. <-
<rjo>
ack
mumptai has joined #m-labs
<rjo>
ysionneau: good that you added the -or1k suffix to the llvmlite package.
<rjo>
ysionneau: but: it would be really great if we could also install that under that name (llvmlite_or1k) in the package tree
<rjo>
because then you can actually install your regular native llvmlite in parallel with the one linked against the or1k-llvm
<rjo>
afaict there would just be a bunch of s/llvmlite/llvmlite_or1k/ in the artiq source.
<rjo>
the native llvmlite is used by numba, which i tend to use.
<ysionneau>
ah ok
<rjo>
ok if i file an issue for this to keep track?
<ysionneau>
yes sure
<ysionneau>
but I'm not sure I understand exactly how to do that
<ysionneau>
in $HOME/miniconda3/pkgs the name of the package is already -or1k
<rjo>
yes. but in the install tree.
<ysionneau>
ah in pkgs/llvmlite-or1k-0.2.1-py34_0/lib/python3.4/site-packages
<ysionneau>
ok
<ysionneau>
I guess it's a setup.py option?
<ysionneau>
to change the package name
<rjo>
less an option. more a field.
<rjo>
exactly.
<rjo>
would need to patch that.
<ysionneau>
ah, not cool
<ysionneau>
I could apply a patch during the conda build
<rjo>
you already have a patch for llvmlite
<ysionneau>
in the "patches: " part of the .yml
<ysionneau>
ah yes
<ysionneau>
indeed
<ysionneau>
seems ok then :)
<rjo>
thanks. low priority. just a nice coexistence feature.
<ysionneau>
ok!
<cr1901_modern>
ysionneau: Might be interested to know litescope (minus RLE) works on Mercury. I can use it to diagnose some of my failing vintage hardware.
<ysionneau>
o/
<ysionneau>
nice!
<cr1901_modern>
I've managed to confirm it works with one of my 8088 boards. 8088 uses a 1/3 duty cycle clock, and it's obvious from the dump that the analyzer is "seeing" 1/3 duty
cr1901_modern has quit [Quit: Leaving.]
cr1901_modern has joined #m-labs
sb0 has joined #m-labs
attie has quit [Ping timeout: 256 seconds]
attie has joined #m-labs
<rjo>
.is!
cr1901_modern has quit [Read error: Connection reset by peer]
<GitHub15>
[misoc] enjoy-digital pushed 3 new commits to master: http://git.io/vLFWn
<GitHub15>
misoc/master a3c0e5c Florent Kermarrec: liteeth/core/arp: fix missing MAC address in ARP reply
<GitHub15>
misoc/master 369cf4c Florent Kermarrec: liteeth/example_designs: add false path between clock domains (speed up implementation) and use automatic PHY detection
<GitHub19>
[misoc] enjoy-digital pushed 1 new commit to master: http://git.io/vLFgG