<wolfspraul>
that's what I'm downloading atm, but not the original source
<wolfspraul>
I'm looking for a download of Werner's copyleft hardware talk too, if one exists
<wolfspraul>
sh4rm4: but thanks, yeah! I found that one already as well :-)
<sh4rm4>
:)
<sh4rm4>
oh btw, since i heard that you work on a linux port (again), you may be interested in the bloat-free musl libc @ http://www.etalabs.net/musl/
<sh4rm4>
porting it to another cpu arch is pretty straightforward
<wolfspraul>
thanks for the link
<wolfspraul>
you mean running the Linux kernel on m1?
<sh4rm4>
yeah...
<wolfspraul>
I don't plan any effort in that direction, I'm very excited about the possibilities of a pure video synthesizer, and RTEMS is just fine for that
<wolfspraul>
and all feedback I get regarding Linux is that Linux without MMU is not a good idea
<sh4rm4>
the guy in the speech said something about a port thats currently being done
<kristianpaul>
but have drivers.. well
<wolfspraul>
maybe but why do you think that's me? I am mostly working on manufacturing issues
<wolfspraul>
dealing with vendors, quality issues, getting the whole package together, production testing, etc.
<sh4rm4>
well, "you" in english can be singular and plural
<kristianpaul>
really?
<sh4rm4>
i meant it in the plural sense
<kristianpaul>
wow
<wolfspraul>
oh I see :-)
<wolfspraul>
you speak to the whole group here
<sh4rm4>
yep
<wolfspraul>
I think Takeshi Matsuya's Linux effort has halted
<wolfspraul>
and I also haven't seen Lars doing much on m1 lately
<kristianpaul>
but mwalle and larsc are doing some things i think
<wolfspraul>
ah ok
<wolfspraul>
that would be fantastic :-)
<kristianpaul>
just not ready i guess
<wolfspraul>
there was some serious MMU planning, which is archived on the list
<wolfspraul>
(for reference, if anyone is interested...)
<wolfspraul>
sh4rm4: so that's all I know about Linux on m1 :-)
<kristianpaul>
imo will be cool to run linux inside an arm cpu inside fpga
<wolfspraul>
why?
<sh4rm4>
oh thanks for the links
<sh4rm4>
having linux could attract more developers
<wolfspraul>
you are welcome!
<wolfspraul>
sh4rm4: if you can help with the MMU, you would make a lot of people happy
<kristianpaul>
because a generic os is a basis for mosts platforms
<kristianpaul>
my opinion of course
<sh4rm4>
sorry i have no HW knowledge at all
<wolfspraul>
I am focusing on the manufacturing side and really making this a great video synthesizer, so I only have time and energy to motivate people about such tasks, can't do anything myself...
<wolfspraul>
kristianpaul: you mean a soft-arm or an arm+fpga combo chip?
<kristianpaul>
wolfspraul: arm+fpga combo
<wolfspraul>
I can't see either one going very far
<kristianpaul>
sofr arm no
<kristianpaul>
why?
<wolfspraul>
kristianpaul: when thinking about this stuff, keep one thing in mind. The decisions about where things are moving are made on economics, not technology. Especially not free software.
<kristianpaul>
good point
<wolfspraul>
so in order to understand where these combo chips are moving, one would need to see the exact numbers of what the different companies are investing, in what timeframe they want to recoup their investments, which kinds of volumes that requires and in which markets they believe they can generate these volumes.
<wolfspraul>
nothing else matters
<wolfspraul>
I have no insight into any of those numbers, I think nobody but those few companies themselves has.
<kristianpaul>
so you think it wil die soon?
<wolfspraul>
but I feel uneasy throwing myself into such a pool and becoming totally dependant on it.
<wolfspraul>
I don't know. that's my point.
<wolfspraul>
but when I get up in the morning, I want to be excited about the day
<wolfspraul>
I don't want to be in a mood like "let's just wait through the day, in the evening we will know a little more"
<kristianpaul>
dependant, yeah..
<wolfspraul>
maybe I first try to get hired by those corps, go to my cozy cubicle every day, then I don't mind :-)
<wolfspraul>
I would never want to participate in this from the outside as a small independant business. too risky.
<wolfspraul>
kristianpaul: hey come on, you just want to be lazy and not get the MMU & Linux port going on m1 :-)
<wolfspraul>
(kidding)
<kristianpaul>
hahah
<kristianpaul>
dont like MMUs
<kristianpaul>
too complicate, but awesome confort for software development
<kristianpaul>
(lazy) may be, but dont image how usefull this linux will be when done
<wolfspraul>
that's why we have RTEMS today, no?
<kristianpaul>
yeap
<kristianpaul>
anyway... back to re-use and find new ways to make things work in milkymist
<wolfspraul>
yes!
<kristianpaul>
like having no place to save 1Gb of raw sampled data :)
<wolfspraul>
still having that problem?
<wolfspraul>
maybe just try to process it live
<kristianpaul>
yes
<wolfspraul>
if it's too hard to move it elsewhere, so what
<kristianpaul>
i'm that
<kristianpaul>
the problem is
<kristianpaul>
about sige 4162
<wolfspraul>
instead of spending weeks/months on that, since it is not the final goal anyway, add the on-board processing first, then the moving data problem will go away
<kristianpaul>
thats true
<kristianpaul>
in part
<kristianpaul>
why?
<kristianpaul>
because osgps dont have support for complex data from sige, i changed the chip mode to real
<kristianpaul>
that means more data per second, but osgps and some matlba code david and i ported to octave know is helpfull
<kristianpaul>
but you are right, i alread had spended something that is not part of the main goal
<kristianpaul>
s/something/lot of time in
<kristianpaul>
okay i'll write that mail to namuru developer and finish the port
<kristianpaul>
and find later how to ran osgps inside rtems and give support for namuru or end doint something my self to drive the namutu correlator, and process navigation data
<wolfspraul>
is it possible to run milkymist one rendering in qemu?
<kristianpaul>
if you managed too ;)
<kristianpaul>
i tried but my CPU go 100% and well.. i got my real m1 and i stoped trying again
<kristianpaul>
make sure you can do 3d acceleario i think
<kristianpaul>
just sent mail to namuru developer, including a intro about milkymist SoC of course
<kristianpaul>
s/a/an
<wolfspraul>
ok got it [render in qemu] it sounds like too much hassle
<wolfspraul>
I was wondering whether I can use that way to record performances, but with real-time and audio in etc. I'm sure it will all be quite a headache
<wolfspraul>
maybe one could slow down everything, say to 1fps, and then put it back together into a 30fps recording? :-)
<wolfspraul>
but the audio-in would need to be in sync still, not sure whether that's possible
<kristianpaul>
better buy a video card and plug vga out from mm1 :-)
<GitHub187>
[extras-m1] yizhangsh pushed 2 new commits to master: http://bit.ly/k0uCBM
<GitHub187>
[extras-m1/master] 3 new label files for printing factory - Yi Zhang
<GitHub187>
[extras-m1/master] removed 3 new added label files - Yi Zhang
<GitHub140>
[extras-m1] yizhangsh pushed 1 new commit to master: http://bit.ly/kPg5GC
<GitHub140>
[extras-m1/master] remove the file mistakenly added. - Yi Zhang
<GitHub158>
[extras-m1] shiyele pushed 2 new commits to master: http://bit.ly/lTC4oX
<GitHub158>
[extras-m1/master] box_artwork update - Wolfgang Spraul
<GitHub158>
[extras-m1/master] Merge branch 'master' of github.com:milkymist/extras-m1 - Wolfgang Spraul
<lekernel>
wolfspraul, I dumped the stream... usually the fastest way to get videos out of a conference
<lekernel>
so there's no other way to download it atm (as far as i know)
<wolfspraul>
i see
<lekernel>
but yeah the video quality isn't very good and the beginning is missing
<wolfspraul>
maybe the download link shows up soon. I also want werner's talk anyway.
<GitHub101>
[extras-m1] shiyele pushed 1 new commit to master: http://bit.ly/mOm7tG
<GitHub101>
[extras-m1/master] cosmetic update - Wolfgang Spraul
<GitHub106>
[flickernoise] sbourdeauducq pushed 1 new commit to master: http://bit.ly/mFhzPT
<GitHub106>
[flickernoise/master] Web update: count local and remote patches - Sebastien Bourdeauducq
<lekernel>
kristianpaul, what to expect from a project like namuru?
<lekernel>
any new (and probably illegal or at least US-irritating) features, or just an "open" clone of the small modules found in phones today?
<wolfspraul>
illegal features :-)
<wolfspraul>
lekernel: what do you mean with 'expect'? you don't understand what namuru does?
<wolfspraul>
out for dinner, bbiab
<lekernel>
is it just a normal GPS, or would it e.g. bypass civilian altitude and speed restrictions, or provide functions to break the Y-code ?
<wolfspraul>
what is an 'illegal feature' btw? I found the news about Fedora excluding some SQL attack tool the other day quite interesting, made me think
<lekernel>
wolfspraul, the drawing is missing from your pdf export
<lekernel>
the back of the box appears entirely blank
<lekernel>
also, remove "128MB DDR SDRAM", as people will think "computer" and "we have DDR3 now"
<lekernel>
you definitely should print the drawing - if you think the transparency will create problems at printing, just make an export of it with a white background ...
<lekernel>
worst case export to bitmap
<lekernel>
a 300dpi drawing is better than no drawing
<lekernel>
[{#«~#{[! libcurl won't automatically encode spaces in urls ...
<lekernel>
only insist that the URL is enough conformant so that libcurl can use it,
<lekernel>
    accepting all possible violations you can do while still making it mostly
<lekernel>
    look like a URL to libcurl.
<lekernel>
gnah gnah gnah
<lekernel>
let's strictly respect the standard and waste everyone's time
<lekernel>
and of course the somehow equally stupid C library doesn't provide a search and replace function for strings
<GitHub116>
[flickernoise] sbourdeauducq pushed 1 new commit to master: http://bit.ly/l3LkVB
<GitHub116>
[flickernoise/master] Web update: patch pool download - Sebastien Bourdeauducq
<lekernel>
ah, they finally added an option to libcurl
<lekernel>
[{#«~#{[!!!! this crap function encodes the / in http://
<wolfspraul>
I removed the drawing, for now, because throwing it on the back is not right
<wolfspraul>
I wanted to make a first draft, and there it is
<GitHub44>
[flickernoise] sbourdeauducq pushed 1 new commit to master: http://bit.ly/mLbiEZ
<GitHub44>
[flickernoise/master] Web update: encode special characters in filenames - Sebastien Bourdeauducq
<wolfspraul>
it's not a great design, but at lease somewhat consistent
<wolfspraul>
I will now make a second one, and the drawing comes to the front side
<lekernel>
the front side will become a bit loaded - there's already the photo on it
<wolfspraul>
well, I see
<lekernel>
imo it's better on the back
<wolfspraul>
also it may be quite big
<wolfspraul>
nobody will see it there
<lekernel>
just throwing it on the back is better than nothing
<wolfspraul>
maybe I replace the device part in the drawing with a small photo?
<wolfspraul>
he :-)
<wolfspraul>
alright alright, probably you are right
<lekernel>
no, there should be a large picture on the front
<lekernel>
definitely
<wolfspraul>
if we cannot get our act together and make a really nice box, then of course the drawing on the back is better than an empty back
<wolfspraul>
you beat me :-)
<wolfspraul>
I put it back there...
<wolfspraul>
but I still hope this is only a draft, and somehow we get a better design. your drawing is a good start.
<wolfspraul>
Jon is flying around a bit, I'm not sure when he sees this and whether he has time on a flight for some design...
<lekernel>
and this diagram must be there - it does make a big difference wrt whether people understand what this thing is about or not
<wolfspraul>
absolutely
<wolfspraul>
I didn't remove it because I didn't like it, but because I thought we should design the entire box around it
<wolfspraul>
I tried to email the stuff before Jon's long flight out of Brazil, but I'm not sure whether I made it, and whether Jon is in work mood on the flight :-)
<lekernel>
interesting: dividing by zero immediately freezes the board (without error message) and GDB gives a wrong program counter in 0x100...
<lekernel>
this should be made easier to debug
<GitHub67>
[flickernoise] sbourdeauducq pushed 1 new commit to master: http://bit.ly/k45cmQ
<GitHub67>
[flickernoise/master] Web update: do not divide by zero when there are no patches to download - Sebastien Bourdeauducq
<GitHub114>
[flickernoise] sbourdeauducq pushed 1 new commit to master: http://bit.ly/jTYNKO
<GitHub114>
[flickernoise/master] Web update: do not crash on zero length files - Sebastien Bourdeauducq
<lekernel>
ok web update should be done now
<GitHub161>
[extras-m1] yizhangsh pushed 3 new commits to master: http://bit.ly/lP2CIV
<GitHub161>
[extras-m1/master] add printing notes file for mac address label - Yi Zhang
<GitHub161>
[extras-m1/master] modified 5 label files - Yi Zhang
<GitHub76>
[extras-m1] yizhangsh pushed 1 new commit to master: http://bit.ly/l3tZlu
<GitHub76>
[extras-m1/master] saved file as plain svg instead of inkscape svg - Yi Zhang
<GitHub79>
[extras-m1] shiyele pushed 2 new commits to master: http://bit.ly/jtPB4i
<GitHub79>
[extras-m1/master] put drawing back on backside (as png for now to be safe) - Wolfgang Spraul
<GitHub79>
[extras-m1/master] Merge branch 'master' of github.com:milkymist/extras-m1 - Wolfgang Spraul
<GitHub180>
[extras-m1] shiyele pushed 1 new commit to master: http://bit.ly/jsClcp
<GitHub180>
[extras-m1/master] added missing graphic, removed memory ref as per Sebastien - Wolfgang Spraul
<lekernel>
mwalle, when IE.IE = 0, are debug exceptions enabled or not?
<lekernel>
right now, when there's a divide by 0 (or another exception)
<lekernel>
the RTEMS CRT simply branches to the corresponding exception vector in the debug ROM
<lekernel>
this is great if GDB is already attached, because it will send immediately a signal to it
<lekernel>
but if you attach GDB after the crash, GDB will send a break and go into the breakpoint vector _from the already running debug ROM_
<lekernel>
so you end up seeing the stack, code, etc. of the debug ROM instead of the program you want to debug
<kristianpaul>
lekernel: open clone of a small module used to be in gps receiver 10 years ago, and yes illegal features can be implemented as well :-)
<kristianpaul>
lekernel: Precision_code is not up to namuru, that first will need a gps frontend that works on the L2 band too, also crack the misterious 2 bit security on that code ;)
<kristianpaul>
lekernel: yes by pass altitude one of then
<lekernel>
i'm disabling break in the GDB stub ROM now. it works, but there is a (small) race condition window.
<GitHub174>
[milkymist] sbourdeauducq pushed 1 new commit to master: http://bit.ly/kXEgYp
<GitHub174>
[milkymist/master] gdbstub: disable break on entering debug ROM - Sebastien Bourdeauducq
<GitHub195>
[flickernoise] sbourdeauducq pushed 1 new commit to master: http://bit.ly/klmB5Q
<GitHub195>
[flickernoise/master] Web update: ensure that the patch pool directory exists - Sebastien Bourdeauducq
<GitHub143>
[flickernoise] sbourdeauducq pushed 1 new commit to master: http://bit.ly/m835kk
<GitHub143>
[flickernoise/master] Web update: ensure that the patch pool directory exists (2) - Sebastien Bourdeauducq
<lekernel>
mh, I also get freezes when using xiangfu's latest flash script
<lekernel>
nice. this sounds like a lot of production fun is upcoming
<lekernel>
omg :'((((
<methril>
hi all, what the autotest-m1 project does? it's for Factory Acceptant Tests and MMOne production?
<kristianpaul>
yes
<lekernel>
what the flying fuck is going on with urjtag now?!?
<lekernel>
I used this thing dozens of time, no problem
<lekernel>
now rejon and I cannot flash 2 times out of 3
<kristianpaul>
hum?
<kristianpaul>
but you not in fedora 15 right?
<lekernel>
i'm in fedora 15, and I remember a libusb update going by recently
<kristianpaul>
tell what you did i can do it here
<lekernel>
i'll try to find an old package and downgrade ...
<kristianpaul>
i reflash almost every day, no problem so far
<lekernel>
yeah it used to go flawlessly here as well
<lekernel>
phew. i'm so tired of problems like that :'(
<lekernel>
massive waste of time
<GitHub92>
[rtems-milkymist] sbourdeauducq pushed 1 new commit to master: http://bit.ly/mGT1Hb
<GitHub92>
[rtems-milkymist/master] ethernet: fallback MAC address - Sebastien Bourdeauducq
<mwalle>
lekernel: mh why does it jump to DEBA?
<mwalle>
instead of EBA
<mwalle>
if IE.IE=0 debug exceptions should still be enabled
<mwalle>
lekernel: (libusb) so -7 and -9 versions are FC package versions?
<mwalle>
div by zero should jump to EBA+offset, where sth useful should be done, eg print register dump, then loop forever, or just loop forever
<lekernel>
it initially jumps to EBA, but the RTEMS CRT just calls the DEBA entry