Topic for #qi-hardware is now Copyleft hardware - http://qi-hardware.com | hardware hackers join here to discuss Ben NanoNote, atben / atusb 802.15.4 wireless, and other community driven hw projects | public logging at http://en.qi-hardware.com/irclogs
wolfspraul joined #qi-hardware
<wolfspraul>
back :-)
<wolfspraul>
wpwrak: you screenshot motivated me to look for an xmas gift for myself in the form of a large-screen high-res new notebook :-)
aw_ joined #qi-hardware
aw joined #qi-hardware
xiangfu joined #qi-hardware
<wolfspraul>
xiangfu: good morning
<xiangfu>
good morning
<wolfspraul>
cool Adam is also there - good morning aw_ :-)
<wolfspraul>
I don't follow legacy tech that much anymore, but if Microsoft also systematically bans copyleft that may be something we can use to make people interested in Qi
<wolfspraul>
maybe only for mobile now, don't know
<wolfspraul>
well if they don't like quality software, we will receive it with open arms
<kristianpaul>
floss on windows is popular, people installing vlc, openffice 7z..
<kristianpaul>
if they cut off, yeah good for us
<pabs3>
this is nothing new though, there was something about the Apple store being incompatible with the GPL before that
<pabs3>
btw, are Qi folks in touch with the opencores/openrisc folks?
<wolfspraul>
pabs3: unfortunately too little [opencores]
<wolfspraul>
if you can help bridge the gap, please do so
<wolfspraul>
naturally I believe Milkymist One is an exciting SoC development platform and there should be more stuff from opencores that we at least play with, learn more about, etc.
<pabs3>
I'm not involved in either project
<wolfspraul>
but it also depends on people doing so, right now Milkymist SoC is really only 1 active HDL developer, that's Sebastien
<wolfspraul>
ah ok. but at least you are lurking, good :-)
<wolfspraul>
I think we need to try out more, move it from the realm of talk to the realm of usage.
<wolfspraul>
I mean that in terms of opencores & milkymist (One)
<wolfspraul>
Sebastien I'm sure has been doing this all along, but not many others yet...
<pabs3>
its interesting that OpenRISC support was merged into Linux 3.1
<wolfspraul>
the ASIC direction is good technically, sooner or later we will definitely 'tape out' freely licensed chips in the same way that we 'gerber out' freely licensed pcbs
<wolfspraul>
but to get there we need to open up and understand the process, and zoom in on a particular product where the asic/structured asic/whatever process makes sense economically
<wolfspraul>
use case first, then tech
<wolfspraul>
otherwise the only guidance you have is to 'catch up' with whatever asic that is currently on the market, and that is a completely hopeless proposition
<wolfspraul>
we've seen the same kind of path in the ogd1 (opengraphicsproject)
<wolfspraul>
every time I use my m1 and admire the visuals and play with music and effects, I know we are on the right path :-)
<wolfspraul>
let's see where the OGD1 is now... one sec
<wolfspraul>
should check this once a year :-)
<wolfspraul>
"Full specifications will be published and open source device drivers will be released. All RTL will be released. Source code to the device drivers and BIOS will be released under the MIT and BSD licenses. The RTL (in Verilog) used for the FPGA and the RTL used for the ASIC are planned to be released under the GNU General Public License (GPL)."
<wolfspraul>
is that true?
<wolfspraul>
that sounds like from an English class about 'future tense'
<wolfspraul>
now the opencores donation, where is that?
<wolfspraul>
20k ?
<wolfspraul>
man my guess is good :-) 20,800 USD
<wolfspraul>
pabs3: milkymist one (and milkymist soc) is years ahead this stuff
<wolfspraul>
we should definitely try to collaborate and exchange as much as possible with opencores, but Sebastien and others are aware of opencores since forever so I'm sure this is already happening to the extent that we find useful stuff and knowledge
<wolfspraul>
I am subscribed to the opencores newsletter :-)
<wolfspraul>
and then I recommend buying a Milkymist One, of course :-) and making the jump into the cold water and start IC design hacking...
<wolfspraul>
when you are frustrated, listen to some great music and visuals thanks to M1 itself :-)
<wolfspraul>
btw, related in a way. semiconductor industry has about 300 billion USD worldwide revenues this year
<wolfspraul>
I need to learn more about the revenue and margin trends inside that 300 billion USD bucket, which types of chips are responsible for what share of the total
<wolfspraul>
that will be important for us to decide how to design products and where to do something ourselves and where to buy from outside
skynet-2000 joined #qi-hardware
skynet-2000 joined #qi-hardware
panda|x201 joined #qi-hardware
aw joined #qi-hardware
aw_ joined #qi-hardware
skynet-2000 joined #qi-hardware
aw_ joined #qi-hardware
aw_ joined #qi-hardware
wolfspraul joined #qi-hardware
<wpwrak>
wolfspraul: (big screen) hehe ;-) (an) external screen(s) can also help. a smaller screen on the laptop itself makes it travel more easily and break less
<wolfspraul>
:-)
<wolfspraul>
I shall find out
<wolfspraul>
1920x1080, ratpoison
<wolfspraul>
actually I would like to build one myself using shenzhen parts, but too much trouble right now
<wolfspraul>
that would be a good excuse to work on some nice case :-)
<wolfspraul>
good point about breaking, my Asus has a metal behind the lcm and even though I took it around a lot, never felt uneasy about breaking the screen
<wolfspraul>
and it got squeeze all over in buses, subways, chairs, and so on
<wolfspraul>
wpwrak: about our earlier SoC discussion
<wolfspraul>
Ingenic does not compete with TI, Freescale, Qualcomm
<wolfspraul>
not today, not tomorrow
<wolfspraul>
they compete with other low-cost Chinese SoC vendors, whose names you will never want to learn :-)
<wolfspraul>
that keeps them trapped in terms of slowly acquiring more knowledge, no matter how much cash they have
<wolfspraul>
I think the idea of Chinese SoC makers will not go very far, I can already see the limits to that
<wpwrak>
well, if the volume is right, they'd still have enough left for a top-notch sw team. sw scales nicely :)
<wpwrak>
of course, all this assuming they care
<wolfspraul>
I really talk to a lot of Chinese manufacturers, designers, here there.
<wolfspraul>
yes
<wolfspraul>
so everybody knows - don't buy Chinese connectors, for example
<wolfspraul>
it's just shit, really
<wpwrak>
oh :)
<wolfspraul>
so the Chinese companies that want to make quality products buy Taiwanese connectors, or Japanese!
<wolfspraul>
PCB, same
<wolfspraul>
passive parts, same
<wolfspraul>
power ics, same (lots of good stuff there from the US)
<wolfspraul>
the political agenda put "CPU" on the table at some point, but just as prestige projects
<wolfspraul>
state of IC design in China is horrible, similar to software
<wolfspraul>
so look at Ingenic's sources :-)
<wolfspraul>
(they stop publishing recently, nobody cares in all the junk anyway)
<wolfspraul>
in very friendly terms you could say "will never go upstream"
<wpwrak>
well, at least the 4740 family seems to work :)
<wolfspraul>
yes, it's shared among many Chinese low-cost SoC makers
<wolfspraul>
this stuff is all blindly copy/pasted
<wolfspraul>
zero talent at those companies
<wolfspraul>
the government sets up pools, shared 'centers' etc.
<wolfspraul>
then everybody runs
<wpwrak>
you admire your masters and copy the best work from them :-)
<wolfspraul>
nah
<wolfspraul>
that's a legend
<wolfspraul>
scratch the 'admire masters'
<wolfspraul>
that would require that you know who a master is, and 'admiring' means you understand something
<wolfspraul>
for example memory controllers
<wolfspraul>
one of the most critical parts in the Milkymist SoC, arguably
<wpwrak>
don't you think they think they know who the masters are ? :)
<wolfspraul>
no
<wolfspraul>
so the state of memory controllers in CHina is catastrophic
<wolfspraul>
even the famous Loongson has to license that part from abroad
<wolfspraul>
that's the problem with political projects that they have to go for the 'big' thing first (CPU), but their talent is not enough to make it work down into the details
<wolfspraul>
so there are a lot of problems with that licensed memory controller, and the end result is that a comparable Intel chip can be put on a 6 or max. 8 layer PCB, but you need a 16-layer PCB for the Loongson :-)
<kristianpaul>
details like proper video graphics !! (yes lemote :()
<kristianpaul>
s/proper/better supported etc..
<qi-bot>
kristianpaul meant: "details like better supported etc.. video graphics !! (yes lemote :()"
<wpwrak>
(16 layer) awww ...
<kristianpaul>
*g*
<wolfspraul>
well it's the typical story
<wolfspraul>
hide the dirt under the rug
<wolfspraul>
under the pcb layers in this case :-)
<kristianpaul>
but this politica thing is up to 2020 if i remenber, longsoon 16 cores cpus :)
<wolfspraul>
fine
<wolfspraul>
all meaningless really
<wolfspraul>
I'm here to study, and I learn every day
<wpwrak>
kristianpaul: each core gets its own layer :)
<wolfspraul>
kristianpaul: did you buy a lemote?
<wolfspraul>
the problem comes from lousy memory controller I was told
<wolfspraul>
Intel is excellent at memory controllers (just relaying)
<wolfspraul>
so with Ingenic, we have to see
<wolfspraul>
I am very open and excited about the relationship
<wolfspraul>
and as long as Liu Qiang wants to meet with me, I will do so once or twice a year
<kristianpaul>
wolfspraul: more o less yes, 2 years ago
<wolfspraul>
I learnt so much already, it's great for me
<wolfspraul>
but right now I just don't think this will take off, we both don't understand and don't need each other
<wolfspraul>
he has to run, throw whatever random SoC together, tape out, sell, squeeze out every penny, and then again
<wolfspraul>
if I see things correctly, the idea of indendently designed Chinese SoCs will go roughly where the idea of Linux/free software allowing Chinese to make 'Chinese software' went in the last 10 years
<wpwrak>
if the USD 2 CPU needs an USD 100 PCB, then we have of course a problem with the "low cost" approach
<wolfspraul>
that is: nowhere
<wpwrak>
if, however, they manage to commoditize the hardware as such, i don't see too much of an issue with them not understanding software. as long as someone is around who does.
<wolfspraul>
yes but in embedded SoCs that's very hard
<wpwrak>
whether ingenic have to be part of that someone's revenue stream or not, seems secondary
<wolfspraul>
that's my point
<wolfspraul>
Ingenic cannot survive on revenues coming from us
<wolfspraul>
they would go out of business right away
<wolfspraul>
so whatever strategy they have, it must fit the market overall
<wolfspraul>
and I think it will be hard for independent Chinese SoC makers
<wpwrak>
i was more thinking of ingenic benefiting from a well-maintained kernel / system
<wolfspraul>
because a good SoC is so product and software dependant
<wolfspraul>
yes but they only care about sales volume
<wolfspraul>
so we don't need to talk with them, we just buy chips when we need them
cladamw joined #qi-hardware
<wolfspraul>
they blew their chance to enter phones
<wolfspraul>
that's too late now
<wpwrak>
well, maybe you could approach them from the "why amazon/quanta didn't pick up your chip but went with expensive TI instead" angle
<wolfspraul>
Qualcomm is totally unreachable
<wolfspraul>
MTK also
<wpwrak>
for all we know, quanta may have even suggested ingenic to amazon, or may - unsuccessfully - do so in the future
<wolfspraul>
the Chinese MTK clone Spreadtrum will embed their own ARM cores into their baseband chips
<wolfspraul>
oh no, no way
<wolfspraul>
they probably get flooded with spam from unknown Chinese vendors trying to sell them whatever
cladamw joined #qi-hardware
<wolfspraul>
it will all be discarded
<wpwrak>
kristianpaul: nice HPC ranking ;-)
<wolfspraul>
there's nothing behind
<wolfspraul>
so Ingenic's media player market is going away
<wolfspraul>
they are trying to get into tablets, but except for extremely optimistic people like Ron nobody thinks that that crap can sell in any quantities
<wolfspraul>
you can only sell so much of a broken product :-)
<wpwrak>
(amazon access) well, wouldn't a "western" company doing the base software make it look more friendly ?
<wolfspraul>
I suggested we should look at new categories like lamps, laser projectors, etc. but that is completely beyond their imagination.
<wolfspraul>
they can just sell us some plastic, that's all
<wolfspraul>
they barely understand what's inside the package ;-)
<wolfspraul>
trust me
<wolfspraul>
that's the downside of the 'attractive' price
<wolfspraul>
it's no different from the rusting connector
<wolfspraul>
that every Chinese manufacturer themselves knows you should never touch
<wolfspraul>
yes it may sound strange and unbelievable to compare a 50 cent connector to a 50 cent SoC, but that's how I see it
<wolfspraul>
they are both designed and manufactured with roughly the same amount of brain input :-)
<wpwrak>
the common item is longevity
<wpwrak>
hmm. connectors look fiendishly complex to me. for most of them, i wouldn't even begin to understand how to manufacture them.
<wolfspraul>
fair enough
<wolfspraul>
so no China :-)
<wolfspraul>
I just met this 1-person Chinese hardware maker who made the Icarus bitcoin mining board
<wolfspraul>
(all freely licensed and in github)
<wpwrak>
wow :)
<wolfspraul>
and it's the same story as always, when going through his board he proudly says "all connectors sourced from abroad"
<wolfspraul>
"look, US power ICs"
<wpwrak>
;-)
<wolfspraul>
well they know, they manufacturer here all the time
<wpwrak>
so it's "designed in the US, manufactured in China" ?
<wolfspraul>
to me it looks right now as if the idea of a Chinese SoC is as flawed as the idea of Chinese data converters (adc/dac), Chinese connectors, Chinese passive parts, Chinese power ICs, ...
<wolfspraul>
or RedFlag Linux :-)
<wpwrak>
(soc) well, there's hope for some "learning by doing" ...
<wolfspraul>
not in me :-)
<wolfspraul>
that's happening in Milkymist SoC, yes
<wolfspraul>
big time
<wpwrak>
maybe you should approach the government. make great linux support for ingenic chips, and hire groups interns for, say, one year, to learn on the job :)
<wpwrak>
s/groups/groups of/
<qi-bot>
wpwrak meant: "maybe you should approach the government. make great linux support for ingenic chips, and hire groups of interns for, say, one year, to learn on the job :)"
<wolfspraul>
won't work
<wolfspraul>
I am mentally moving past the idea of us finding a great Chinese SoC maker to work with
<wolfspraul>
I will keep trying to find a way to collaborate, but once you learned something by really doing it, there is little need to do that again and again.
<wolfspraul>
so in other words, I think the lack of understanding how this could work mutually will be stronger than any initial motivation to just do it regardless
<wpwrak>
well, for mid-range devices, ingenic don't look so bad. one problem is their (non-)publication policy
<wolfspraul>
nah it's deeper
<wolfspraul>
they think their problem is that Quanta hasn't picked up their phone call yet
<wolfspraul>
even though they sent some samples! :-)
<wpwrak>
but for the chip, they seem to be well-positioned. e.g., from the system software side, they're not more trouble to work with than, say, samsung. perhaps even less.
<wolfspraul>
you mean for the Ben?
<wolfspraul>
yes sure, we did well and I hope we can continue with the Ben
<wolfspraul>
I am talking about future products
<wpwrak>
ben and similar
<wolfspraul>
and I will continue to work with Ingenic, I am just describing the status quo as clearly as possible
<wpwrak>
do you expect that 4740 successors will be inferior in terms of overall system integration cost ?
<wpwrak>
(i.e., the 16 layer pcb memory controller)
<wpwrak>
(or just not work or work unreliably)
<wpwrak>
in the ben, we had remarkably few problems. also, the documentation is adequate. not perfect, but when i compare this with some things samsung did, certainly in pretty good shape
<wolfspraul>
wait, you misunderstood me
<wolfspraul>
the reference to 16-layer was about a high-end Loongson CPU
<wpwrak>
yes yes
<wpwrak>
but you said that intel could have done the same with a lot less
<wolfspraul>
yes
<wolfspraul>
as an example of how strong Chinese companies actually are in IC design
<wolfspraul>
or how weak rather
<wpwrak>
so if ingenic do their own equivalent of the "16 layer memory controller", whichever subsystem or aspect of the chip that may be, e.g., could be power supply, then they would kill their own low-cost angle
<wolfspraul>
there's a lot of random guessing, and then if something actually has to work well they have to license/buy foreign stuff
<wolfspraul>
no I think the Ingenic chips are fine
<wolfspraul>
but that doesn't mean that Ingenic has the quality staff you would need for them to be in control of their own destiny
<wolfspraul>
on the IC design or software side
<wpwrak>
so you're implicitly predicting that a "16 layer memory controller" incident will happen
<wolfspraul>
think of it this way: how far is the Milkymist SoC from replacing any Ingenic chip entirely?
<wolfspraul>
Ingenic is licensing/copy-pasting a lot of IP blocks, yes
<wpwrak>
hmm, maybe 1-2 years
<wolfspraul>
and then they tape-out in the highest process they can get their hands on
<wolfspraul>
and then they dump the chip on the market fast and cheap because lots of other Chinese SoC makers have access to the same IP core pools
<wpwrak>
maybe that's the cruz
<wolfspraul>
you think they have time to think through the subtleties of something higher up in the Android stack working or not? :-)
<wpwrak>
crux even
<wolfspraul>
of course you can imagine me :-) I like to do the "what device are you using yourself" game
<wolfspraul>
always fun in China
<wolfspraul>
and I still love it :-)
<wolfspraul>
don't believe the 'save face' nonsense, jsut go right for the demasking
<wpwrak>
maybe the real problem is that any investment they make benefits their fiercest competitors just as much. so it depends on whether they believe in growing the pie instead of getting a bigger slice
<wolfspraul>
yes true, that would be one problem
<wolfspraul>
I think they are fine, that's my point
<wolfspraul>
not much we can do
<wolfspraul>
imagine you are a maker of cheap Chinese connectors, something simple like the memcard connector on the Ben you like so much
<wolfspraul>
tolerances are bad
<wolfspraul>
everything is bad
<wolfspraul>
you sell for 10 cents
<wpwrak>
which again would point to them being a sub-optimal tree to bark up. the source that finances that shared IP would be better.
<wolfspraul>
the good ones from Japan or Taiwan cost 50 cents
<wolfspraul>
what do you do now?
<wolfspraul>
you can just either continue or close your company
<wolfspraul>
you cannot genetically reinvent an entire company full of people
<wolfspraul>
company culture is self-selecting
<wpwrak>
well, you cuold perhaps try to see if there are small improvements you can make. before the name is ruined completely.
<wolfspraul>
you cannot start with farmers, and then 5 years into it when you have 100 farmers hacking into their keyboards suddenly reboot the entire company with a bunch of US-trained PhDs
<wpwrak>
of course if everybody flinches when they read "made in china" on a connector and doesnt' even look beyond that, then your chances are slim
<wpwrak>
oh, there are many ways to bring fresh blood into a company. of course, if you want to do a complete makeover, better shut down the old one and start something new
<wolfspraul>
the fact is that Ingenic blew their entrance into phones
<wolfspraul>
and Qi and Ingenic blew to setup something like Linaro
<wolfspraul>
now all they can hope for are tablets
<wolfspraul>
we'll see
<wpwrak>
and hope they survive tablets until the next big thing comes along :)
<wolfspraul>
I like Milkymist better every day :-)
<wolfspraul>
yes but they will be late/miss that next big thing again
<wolfspraul>
the cost advantage is all they have
<wolfspraul>
but if you think from a 'what works' perspective, that's so laughable
<wpwrak>
being in tablets must feel a bit like you have end-stage cancer, just dropped out of an airplane and are in free fall, and the only thing you see below you, where you may safely land, is the titanic, going full steam for the iceberg
<wolfspraul>
you really think someone wants to bring down their entire software *stack* so that the CPU is 10 USD cheaper?
<wpwrak>
so many ways to die, so little hope
<wolfspraul>
so if the SoC maker needs 5 or 10 more USD / pc so that it integrates well with a software stack, boy, that's worth it!
<wpwrak>
for smart people, the sw stack has little cost
<wolfspraul>
read the Ingenic sources a little :-)
<wolfspraul>
you may reconsider that
<wolfspraul>
on the Ben, 100% of the cleanup work was done outside of Ingenic and with zero support
<wolfspraul>
no documents were opened
<wpwrak>
e.g., why should amazon not use ingenic ? the sw stack is probably marginal. a much bigger worry would be whether ingenic really deliver those millions of chips on time, in good quality, etc.
<wolfspraul>
no single dollar changed hands
<wolfspraul>
nah, then Amazon better design their own SoC
<wpwrak>
yes. we did the smart thing - do our stuff without depending on ingenic
<wolfspraul>
a lot of companies try to avoid useless middlemen that can charge whatever margin they wish
<wolfspraul>
and you are suggesting to add one that wouldn't add anything but a tollbooth?
<wolfspraul>
I could not recommend that at all if I were on the Amazon side
<wpwrak>
if they made a 5740 tomorrow that's exactly like the 4740 but with an ARM core instead, it would take is mere days to switch
<wpwrak>
s/is/us/
<qi-bot>
wpwrak meant: "if they made a 5740 tomorrow that's exactly like the 4740 but with an ARM core instead, it would take us mere days to switch"
<wolfspraul>
yep
<wolfspraul>
the "Ingenic SoC" is really a "China SoC"
<wolfspraul>
you can get *a lot* of such chips from at least 10 or 20 companies
<wpwrak>
(middleman) well, see linaro. could they also do their job without being at excellent terms with ARM ? probably
<wolfspraul>
and that "China SoC" has some serious problems from my perspective, which is how to make great open hardware, fast and working and carrying forward the highest quality free software
<wpwrak>
let's hope they have at least the decency of naming them all <something>4740 :)
<wolfspraul>
so whether we go from the Ben to the Ya on another, newer, China SoC, or switch right to the Milkymist SoC, remains to be seen
<wpwrak>
you mean as in "we did it for the 4740, but don't count on it happening for the 4780" ?
<wolfspraul>
yes sure, we did excellent on the 4740
<wpwrak>
(as a message to ingenic)
<wolfspraul>
Lars largely
<wolfspraul>
oh they don't care, they don't even understand
<wolfspraul>
it's like a foreigner talking to you in some totally unintelligable language
<wolfspraul>
trying to tell you something VERY IMPORTANT
<wolfspraul>
ha ha
<wolfspraul>
what can you do...
<wolfspraul>
first another asado :-)
<wpwrak>
so where do their customers get the system software from ?
<wolfspraul>
from Ingenic
<wolfspraul>
totally dysfunctional Android builds
<wpwrak>
but they get it to work ?
<wolfspraul>
I spare you or anyone to copy/paste some reviewer comments from people who accidentally bought such tablets in the US :-)
<wpwrak>
;-)
<wolfspraul>
yay, it boots.
<wolfspraul>
and you can swipe!
<wolfspraul>
LOOOK!
<wolfspraul>
:-)
<wolfspraul>
swipe, swipe
<wolfspraul>
how many do you want to buy?
<wolfspraul>
swipe, swipe
<wpwrak>
do ingenic see these reviews ?
<wolfspraul>
I doubt it
<wolfspraul>
how many do you want to buy?
<wolfspraul>
swipe, swipe
<wolfspraul>
I give you a discount if you buy 100!
<wpwrak>
maybe you should bring them a few the next time :)
<wpwrak>
;-))
<wpwrak>
(a few reviews)
<wolfspraul>
like I said, I foudn the secret behind their low-cost :-)
<wolfspraul>
there's always a price to pay, now I know which one...
<wolfspraul>
bbiab
<wpwrak>
yes, but then there's the question of what they perceive as problems. even if their customers all fail, that may be okay, as long as they burn enough initial investment on buying chips
<wolfspraul>
yes
<wolfspraul>
it's a big world
<wolfspraul>
and that is how Ingenic sees it
<wolfspraul>
the US saying is "a sucker is born every minute", no?
<wpwrak>
yeah :)
<wolfspraul>
I grew up in Germany, until I started living in Asia I could not imagine that stores sell products that just plain DO NOT WORK.
<wolfspraul>
never did
<wpwrak>
of course, another problem would be that we don't have have what they need. their customers would want to have full android support.
<wolfspraul>
like an iron (to iron shirts) where the power cable inside is broken and was never connected
<wolfspraul>
not just on one, but on the entire lot
<wpwrak>
not some nice upstream patches and openwrt
<wolfspraul>
nobody cares, it's cheap, it sells
<wolfspraul>
maybe the poor buyers have to solder the missing cable themselves
<wolfspraul>
or whatever
<wolfspraul>
it doesn't matter
<wpwrak>
yeah, the infamous "china crap" :)
<wolfspraul>
well, the buyer is in control. he can choose to not buy.
<wpwrak>
not always
<wolfspraul>
you better make good use of that choice :-)
<wpwrak>
if all you find in the stores is china crap, you'r screwed
<wolfspraul>
well, you better keep that control
<wolfspraul>
yes their customers want full and powerful android, definitely
<wolfspraul>
and no, we cannot deliver that either
<wolfspraul>
our path is longer
<wolfspraul>
and requires understanding of how things connect and depend on each other in that big software stack
<penguinhead>
wpwrak: but i do not suppose it is possible to connect the serial console of the Ben through this microcontroller?
<penguinhead>
(i was looking for a hassle-free way of connecting to Ben's serial console)
skynet-2000 joined #qi-hardware
<wpwrak>
ah .. no, that's for giving the ben an interface to talk serial to something else, not for giving access to the console. well, unless perhaps you have two bens ;)
<penguinhead>
:p
<penguinhead>
but could the SDIO interface present itself as ttyS, in theory?
<wpwrak>
yes, it probably could. you'd have to bit-bang it, though. and timing is hairy
* kristianpaul
remebers bit-banging uart in pic16f84
<penguinhead>
better stick to the testpoints then
<penguinhead>
i thought i read somewhere that the exposed testpoints in the battery compartment share GPIO lines with the keyboard though, is that correct?
<kristianpaul>
bit-bang support inside kernel may be usefull then
<penguinhead>
ttyl! :: goes home for the weekend ::
penguinhead left #qi-hardware
kristoffer joined #qi-hardware
erikkugel joined #qi-hardware
Ayla joined #qi-hardware
<DocScrutinizer>
seen incredibly nifty scope probes today: shaped like a Y of 10cm height, with rubber feet at the 2 "upper" ends, and a pogo-pin like a stork bill 90° "into the table" from the lower end of the Y. the bill was made of transparent plastic and had colored LED to identify the probe and to illuminate the PCB area with the testpoint where the whole Y-probe would contact to by own weight
<wpwrak>
that sounds clever indeed
<wpwrak>
how are those probes grounded ?
<DocScrutinizer>
connected to a 1GHz scope with a special coax+power+$X cable
<DocScrutinizer>
the cable enters at one of the feet, well near it - IIRC
<DocScrutinizer>
the structure of the Y is made from ~15mm diameter black tubes. I didn't dare to touch, so can't say how much they weigh or what material the Y is made of
<DocScrutinizer>
the scope itself was some brand I never heard before, and it's been "tiny", size of a large shoebox
<DocScrutinizer>
but with a fan ;-D
<wpwrak>
what was the brand name ?
<wpwrak>
and only 1 GHz ? now i'm a bit disappointed ;-)
<DocScrutinizer>
I haven't heard of a lot of brands
<DocScrutinizer>
I'm not exactly into market studies
<wpwrak>
the lecroys look cute. i wonder if their "density" display (number of beam passes changes color) is really as good as it looks on their pictures
<wpwrak>
ah, but lecroy are pretty well-known. sort of #3 after tek and agilent
<DocScrutinizer>
can't find the friggin Y-probes. Wonder if they were genuine LeCroy
<DocScrutinizer>
damn, all this nice nifty lab equipment around me and I'm tied to that friggin WinXP workstation all day, studying ClearCase with proprietary extensions, even worse OUTLOOK (BLAAAARRRGHHHH!!!)
<wpwrak>
;-)))
<wpwrak>
if you're nice during the day, maybe they'll let you play with the equipment at night ? :)
<DocScrutinizer>
it's obviously impossible in fscknf outlook to keep all your in and out msgs in one folder, nicely threaded
<DocScrutinizer>
nah, you get shot when they catch you after 21:00 in the building :-D
<DocScrutinizer>
or they trap you in the rotary tile or what that thing is called, or in the elevator
kuribas joined #qi-hardware
erikkugel left #qi-hardware
<kristianpaul>
now that you know libfpvm and the fpvm better than anyone here i was wondering you can tell more about how to use it
<kristianpaul>
for example this compiler.c is an interpreter or have more feaures?
<kristianpaul>
and eactually how you load the result of the copilation to the pfu
<wpwrak>
DocScrutinizer: sounds like those elevators with the alarm button making a telephone call ... but the defective technology dialing the wrong number. now, how many very long weekends are there per year and how long can the average human survive without water ?
<kristianpaul>
also how i can use this libs in other apps that may need to calculate heavelly sin/cos/atan
<wpwrak>
kristianpaul: compiler.c is a compiler. it generates pfpu "machine code".
<wpwrak>
but why here and not on #milkymist ?
<kristianpaul>
here is not lekernel and dont bother anybody again ecept you ;)
<kristianpaul>
except*
<wpwrak>
;-)
<kristianpaul>
of course if u dont mind :)
<kristianpaul>
dont want*
<wpwrak>
plus other ~43 people :)
<kristianpaul>
he
<kristianpaul>
okay... you fot the point
<wpwrak>
i haven't looked at the compiler-pfpu interface yet
<kristianpaul>
s/fpt/got
<wpwrak>
soi don't know how the code gets there, how the data gets in and how the data gets back out
<kristianpaul>
hum
<kristianpaul>
ok just cleanup for it,but are you considering go for it later?
<kristianpaul>
The idea of milkymist having a compiler for its own pfpu is quite interesting in think
<wpwrak>
so far, i just looked at code generation. code is generated in 3-4 steps: one is from strings containing an expressions to a parse tree. then, the parse tree to "virtual" machine code. and then, finally, the virtual machine code to physical machine code.
<kristianpaul>
at least if you have a lib that can boost and take off some heavy operation from main cpu
<wpwrak>
my focus at the moment is more at the top of that stack. i've already tackled the bottom end, the scheduler (that's virtual -> physical)
<wpwrak>
yeah, that would be the interface. i'll get to it, eventually.
<wpwrak>
i'll need that to make a proper symbol table. but that's just a performance optimization.
<wpwrak>
and i can't bring down compile time a lot anyway. i already know where some 70-80% of the time go and there's little potential for improving that. (i.e., the scheduler)
<kristianpaul>
rtems scheduler?
<wpwrak>
i expect to be able to squeeze something like 10% out of the top level of the compiler
<wpwrak>
no no ... pfpu instruction scheduler
<kristianpaul>
ah ok, sorry :)
<wpwrak>
libfpvm/gfpus.c or libfpvm/lnfpus.c
<wpwrak>
the former is easier to understand. the latter more efficient :)
<kristianpaul>
okay thanks, now i have a question for lekernel i hope :)
<kristianpaul>
and may be others here got caught from this talk and may get more interested of milkymist features