<Fallenou> lekernel: hi
<Fallenou> I am trying to re organize the ethernet driver
<Fallenou> first to put everything in the softc instead of global variable
<Fallenou> but I have an array of rxbuffers
<Fallenou> how can I put this into the struct ?
<Fallenou> because it muse be initialized with the addresses of the rxbuffers
<Fallenou> must*
<Fallenou> and where do I put the rxbuffers declaration ?
<Fallenou> in the struct ? it would be ugly
<Fallenou> to have 4 rxbuffers in the struct
<Fallenou> and then to put their address in an array within init()
<Fallenou> how can I put the rxbuffN in the struct since they *have* to be declared as global
<Fallenou> well I think I will initialize everything in the init() and put dma memory in static in the init function
<CIA-31> rtems-milkymist: Yann Sionneau master * r03a674b / c/src/lib/libbsp/lm32/shared/milkymist_framebuffer/framebuffer.c : indent fix - http://bit.ly/hUYyu2
<CIA-31> rtems-milkymist: Yann Sionneau master * r11a52f6 / c/src/lib/libbsp/lm32/shared/milkymist_networking/network.c : driver clean-up, we now use a softc and no global variables - http://bit.ly/eDWxa9
<Fallenou> hop hop some cleanup
<Fallenou> I just checked that it compiles
<Fallenou> time to go to bed, tell me if it's cleaner for you lekernel :)
<Fallenou> next step, add a pool of rx buffers
<methril> hi guys!!
<methril> i just paid the MMOne order (few weeks and is in brazil)
<wolfspraul> I hope we are faster than 'a few weeks'
<wolfspraul> it will ship on Monday, then let's see how lucky we are with Brazil customs
<methril> just going to get the most recent gnu-crap (as lekernel name it)
<methril> wolfspraul, we'll see
<lekernel> hmm? what gnu-crap?
<lekernel> the gnu-crap is fortunately a very small amount of what's used in milkymist :)
<methril> binutils/gcc
<methril> ;)
<kristianpaul> dislikes the gnu-crap word
<lekernel> methril: you can try gcc svn (4.6) - it needs testers :)
<methril> ok, i hope it works in qemu-lm32
<kristianpaul> 4.6? does it compilesfinally ?
<lekernel> the c++ compiler doesn't build (but it's not such a big issue as it seems didn't even work before) but the c compiler should be ok
<kristianpaul> ah, good :-)
<methril> lekernel, is this a gcc issue or only for lm32 port?
<methril> for the ISE 13, we need all full installer_
<methril> ?
<kristianpaul> webpack
<lekernel> methril: it's only for the lm32 port
<lekernel> but lm32 is included in gcc for almost a year now
<methril> i know
<methril> i compile some 4.5.x in my own, and deal with some Coldfire v1 problems (+newlib)
<methril> i'm following the milkymist project since the beginning
<lekernel> if you can fix lm32 problems in 4.6, I can commit your changes
<methril> i'll try to get a toolchain working, and later i try to deal with them
<kristianpaul> lekernel: You think is a valid but ask for Cycle Counter support in lm32-gcc?
<kristianpaul> yeah i already now, thats it not defined for lm32 in mm1
<kristianpaul> s/now/know
<lekernel> kristianpaul: what does that cycle counter have to do with the compiler?
<kristianpaul> is not the compiller supposed to support it?
<lekernel> to do what?
<lekernel> read from application? you can use inline assembly
<kristianpaul> i need compile fftw for rtems :-)
<lekernel> why does fftw need a cycle counter?
<kristianpaul> In order to perform runtime self-optimization
<kristianpaul> well i can skip that. i just was asking just in case ;-)
<kristianpaul> FFTW needs a high-precision timer to measure the cost of different code choices
<lekernel> and can't you read the cycle counter with inline assembly? why do you need special compiler support?
<kristianpaul> hmm, yes i could use assembly
<kristianpaul> yeah, nv
<wolfspraul> lekernel: one of the cheap cmos cameras I bought for 15 usd says in the manual "programmable ic bus function: color saturation, brightness and contrast ratio, self balance, exposure time and 2 times gain"
<wolfspraul> do you know what this means?
<lekernel> well there's no standard for cycle counters afaik, so it shouldn't be in the compiler
<wolfspraul> there are 2 analog cables coming out, I would think one for composite video, one for audio
<lekernel> inline assembly is the way to do it
<wolfspraul> so how can the IC on the other side be 'programmed'?
<lekernel> let's not add more cruft to gcc than what it already has :)
<wolfspraul> is there a standard to feed data back into the camera?
<kristianpaul> lekernel: yes, it make sense for me know ( no standard )
<wolfspraul> over the composite video line?
<lekernel> wolfspraul: not as far as i know
<wolfspraul> ok. need to find out how this 'programmable ic bus function' is supposed to work then...
<lekernel> btw, if your camera has audio out, it'd be nice that it could be plugged to the line in with the supplied cables/adapters
<wolfspraul> could be just a copy/paste manual of course, who knows
<wolfspraul> I thought you said that's a not-important use case, and we have an internal micro?
<lekernel> wolfspraul: well, maybe IC= chinese name for I2C and they just mean that some chip inside the camera has I2C :)
<lekernel> I guess you know how chinese technical docs are sometimes :)
<lekernel> yeah, it's not important
<wolfspraul> ok, in 3 lines you first say "it'd be nice" and then "it's not important". so what now?
<wolfspraul> the microphone will not be very far away from the internal one
<wolfspraul> m1 has no fans or other noise
<wolfspraul> and the one in those cheap cameras for sure won't be better than the internal one either, I would think
<wolfspraul> so that only leaves that it may be a bit more exposed, up in the air a little
<wolfspraul> will that make a difference?
<lekernel> well, it just means that if audio out is for free, and if you have a choice of plugs, pick one that is compatible (e.g. jack)
<wolfspraul> or rather: how much of a difference will that make?
<wolfspraul> ah ok
<kristianpaul> wonders what is the usabillity of the internal mic
<wolfspraul> of course I keep that in mind
<lekernel> kristianpaul: well I tested it on stage with the acrylic case already, and it does a decent job
<kristianpaul> lekernel: yeah, but i mean, how is related with flickernoise, ie, if you clamp you a get a custom effecton screen?, or if clap twice i swich a patch..
<lekernel> the intended use is beat sync, but you could code something like that too
<kristianpaul> Any on had tested the Amsterdam Compiler Kit ?
<kristianpaul> wow pcc 1.0 was released just days ago
<lekernel> we need to find gsoc mentors... who's in?
<rejon> and
<rejon> ok bbiab
<wpwrak> (cycle counter) they do seem to be in in POSIX: man 3 clock
<wpwrak> (cycle counter) you're allows to implement it as  return (clock_t) -1;  though :)
<wpwrak> (attacks on irc) may also be some amount of bot-net testing. easily available target that gives feedback and doesn't call the cops
<lekernel> rejon: that's totally outdated
<lekernel> almost everything on this list is done except the linux tasks that were replaced with rtems ones
<lekernel> Gurty: hi
<lekernel> is that you who has the gcc compile farm?
<Gurty> hi lekernel, no :o
<lekernel> uhm, why does the spartan6 have a SPI_ACCESS block that is normally used to read the on-chip flash of FPGAs that have it (spartan6 does not)?
<lekernel> there's also a completely undocumented "OCT_CALIBRATE" site
<Fallenou> lekernel: hi
<Fallenou> lekernel: is this better about the softc/global variables problem in the driver ?