<kristianpaul> Fallenou: yeah, i'm havin lots of  power problems this weak :(
<kristianpaul> damit
<kristianpaul> Fallenou: is up now
<kristianpaul> arggh, powr failure. data storage failure... :/
<xiangfu> kristianpaul: I upload the makefile to a new repo : http://projects.qi-hardware.com/index.php/p/m1s/source/tree/master/
<xiangfu> kristianpaul: fyi :)
<kristianpaul> good
<kristianpaul> I'll have more time to try out on before friday
<kristianpaul> ok, now where is the dma arbiter...
<kristianpaul> "You may not (and may not allow anyone else to) disclose the results of any benchmarking of a Licensed Product (whether or not the results were obtained with assistance from [the EDA vendor]) to any third party" patetic
<kristianpaul> lekernel: can i say dma is mm soc is an implementation of a wishbone swich topology, and of course the cpu is the master 1?
<kristianpaul> morning btw :-)
<lekernel> it's a multi-master wishbone with (partial) crossbar switch
<Fallenou> lekernel: is fml crossbar switch or pipeline or shared bus ?
<Fallenou> topology
<lekernel> hmm
<lekernel> what would be the point of a crossbar?
<Fallenou> I don't even understand the crossbar
<Fallenou> it's just to know which one it is
<lekernel> well, then read up on crossbars first, then ask yourself how many slave peripherals there are on FML and this should answer your question
<Fallenou> I thought I could have a quicker answer here, obviously a mistake
<lekernel> crossbars allow simultaneous communication between different masters each talking to a different slave
<Fallenou> ok
<lekernel> here there's only one slave, the sdram controller
<lekernel> so it's obviously a shared bus
<Fallenou> ok thanks
<terpstra> is lm32 support in mainline gcc now?
<Fallenou> it's been in mainline gcc for some time now
<Fallenou> it's not recent
<lekernel> it was thrown into gcc 4.5 without much advertisement
<terpstra> great
<lekernel> maybe I should update the gcc website myself :)
<Fallenou> it's been two years ? something like that ?
<lekernel> i'm just a bit afraid of using CVS and breaking things...
<lekernel> one year
<Fallenou> ok
<lekernel> we also have serious issues with the C++ compiler - it seems to generate broken code in 4.5.x and no longer builds in 4.6
<Fallenou> even the gcc-core-4.5.2 from rtems.org with their patch ?
<Fallenou> it's generating kroken code ?
<Fallenou> broken*
<lekernel> yes
<lekernel> at least it seems so
<terpstra> lekernel, the g++ i have from lattice works
<lekernel> I ran into the Qt4 broken method table bug with that one, and Till also had broken code issues
<terpstra> though i've not built large programs with it
<lekernel> but it's 3.x, no?
<lekernel> or a modified 4.4
<terpstra> its 4.3.0
<lekernel> yeah ok :)
<lekernel> gcc seems to be such a mess to maintain, they regularly break stuff... :(
<lekernel> especially "minor" architectures
<Fallenou> I guess they don't even test those minor architectures
<Fallenou> do they even have testsuites for those ?
<Fallenou> lol
<lekernel> iirc someone spoke about a gcc build farm on the ML... somewhere around July 2009
<lekernel> I should contact him back to add some lm32 tests, even basic (just check that the compiler builds) to his suite
<Fallenou> Examples for this are plenty. Many of the GNU projects are ported to a wide variety of platforms, even to undeserving once like cygwin and mingw
<Fallenou> lol'ed
<lekernel> yeah, sometimes by rms himself :)
<lekernel> yeah, I read that too some time ago...
<lekernel> free software isn't exactly the ideal world some advocacy organizations depict :)
<Fallenou> yeah it seems so
<Fallenou> lol Xilinx was routing power to blockram even if the blockram block was un used
<Fallenou> they started to stop powering unused blockram in the 7 serie
<Fallenou> 30% of the power leakage was coming from unused blockram :)
<Fallenou> let's turn off the light please if you don't use it !
<lekernel> hmm???
<lekernel> from what I can tell, leakage power in current FPGAs isn't that big anyway, so there can be a justification for that :)
<lekernel> also, 28nm would leak more
<Fallenou> well they are saying 28 nm is leaking less
<Fallenou> Learn about the power benefits of the 28nm process Xilinx 7 Series #FPGA in this NEW white paper. http://bit.ly/WP28nm <=
<lekernel> probably B.S.
<Fallenou> lol
<Fallenou> anyway, funny that they just happen to think about not powering what's not used by the design
<lekernel> what it could mean is the leakage power per transistor is lower, but *all* the power per transistor is lower in 28nm, including dynamic power
<lekernel> but the leakage/dynamic power ratio grows as process geometries shrink
<lekernel> if they reduced it, it's not by switching to 28nm
<lekernel> this actually makes things worse
<Fallenou> humm ok
<lekernel> btw, it'd be fun to get kintex7 support in llhdl before they generally sell the devices =]
<lekernel> the reverse engineering process should be largely automated, and k7 isn't very different from s6
<lekernel> so once s6 works great, supporting k7 should be a walk in the park
<Fallenou> 18:44 < lekernel> btw, it'd be fun to get kintex7 support in llhdl  before they generally sell the devices =]
<Fallenou> LOL
<kristianpaul> hehe
<Fallenou> they could be very happy about that
<Fallenou> I hope they will
<Fallenou> or they could wake up a bunch of lawyers :)
<kristianpaul> at least us will be happy :-)
<kristianpaul> or they can offer a big fat job to lekernel ;-)
<kristianpaul> ergg, ignore fat
<kristianpaul> lekernel: are curently automating reverse eng process for s6?
<Fallenou> his code parse Xilinx big fat information files about s6 arch
<kristianpaul> I was thinking in ulogic actually
<scrts> heh :)
<scrts> I have a friend that is able to crack any altera megacore
<scrts> which is encrypted
<scrts> maybe someone did that for xilinx cores?
<Fallenou> scrts: ask lekernel he told me about some xilinx core decryption one day
<Fallenou> the key was somewhere distributed with the ise
<scrts> well, for altera cores decryption the core ID is needed only
<scrts> like 4 symbols code
<Fallenou> lol
<Fallenou> and then you get the all hdl source code ?
<kristianpaul> netlist he get i think
<scrts> yes
<scrts> the source is available with quartus
<scrts> it is only encoded
<scrts> run the decoder and get the verilog files
<scrts> move them to the altera ip core dir
<scrts> regenerate the core
<scrts> and no time limits :)
<Fallenou> hehe :)
<Fallenou> I never used encrypted core before
<Fallenou> they have time limit ?
<scrts> yep, works ~1hour in silicon
<Fallenou> o_o
<Fallenou> must be a big register to hold that much time information
<scrts> hmm, I am not sure
<kristianpaul> you can make a clock div and count cycles
<scrts> afaik it is counted @ pc, because when You disconnect the jtag the core stops
<Fallenou> well how do you know the clock freq ?
<Fallenou> oh
<Fallenou> ok
<Fallenou> it sucks :(
<Fallenou> is it the same with Xilinx closed-source IP ?
<scrts> dunno
<scrts> I use xilinx at home and altera at work
<scrts> never used closed source xilinx cores
<Fallenou> Has your friend written a blog post about that ?
<kristianpaul> used to work with nios
<scrts> the cores needed @ work were SDI based
<Fallenou> SDI ?
<scrts> serial digital interface
<scrts> video standard :)
<Fallenou> oh ok
<scrts> mostly used in professional studies
<Fallenou> didn't know this interface
<lekernel_> scrts: I can also crack Xilinx encrypted IP, but there is nothing really interesting in there
<lekernel_> and it's not even hard to crack
<lekernel_> it's also illegal to use them... which makes it even less interesting
<Fallenou> lekernel_: do you have the source of your slides of 26C3 ?
<lekernel_> mom, I'm not sure... maybe lost them in a hdd failure last year
<Fallenou> I am rewritting some pages since I don't have the source code
<Fallenou> would help me :)
<Fallenou> ooh too bad
<Fallenou> otherwise I can continue rewritting
<lekernel_> I have the sources of the university conference last month though
<lekernel_> (but in french)
<Fallenou> oh
<Fallenou> I was not aware of this one
<lekernel_> with the source
<Fallenou> sure
<Fallenou> I based my slides on this
<Fallenou> and backported some other informations from the RMLL2010 talk
<lekernel_> but I can't find any trace of the 26c3 sources...
<Fallenou> and used the design from RMLL2010 (prefer the yellow)
<Fallenou> I just want to add some technical informations from 26C3
<Fallenou> ok no problem
<Fallenou> will just rewrite what I need
<lekernel_> kristianpaul: the s6 reverse engineering automation is being done by the ulogic authors
<lekernel_> who are on mm and llhdl lists btw
<lekernel_> and yeah for the chip geometry we can parse the xdlrc database
<lekernel_> also automatically
<lekernel_> the last difficult thing will be the timing model
<Fallenou> nice presentation, the masteri2l
<Fallenou> something weird
<Fallenou> you say you have 3.2 Gb/s of max bandwidth , from your ddr sdram chip
<Fallenou> and you say you need 3.3 Gb/s memory bandwidth
<Fallenou> how is this possibly working ?
<Fallenou> ok the cache
<Fallenou> it's the only thing I can see that could help
<Fallenou> if you access several times the same data
<Fallenou> it can save you sdram bandwidth
<Fallenou> lekernel_: how much sdram bandwidth are you using atm ?
<lekernel_> btw, HPDMC is scheduled to launch into space in January 2012
<Fallenou> :)
<Fallenou> the end of the world :p
<Fallenou> I knew it !
<lekernel_> some 3-4 Gbps I think
<lekernel_> haven't measured with the video input and everything
<lekernel_> but the soc has a core to do that... there's a memory statistics command in the demo firmware but it's not in the rtems shell yet
<Fallenou> ok so the sdram chip is capable of more than 3.2 Gb/s then ?
<lekernel_> each sdram chip has ~3.2Gbps peak bandwidth, and there are two
<Fallenou> oh :)
<Fallenou> which makes ~6.4 Gb/s peak ?
<lekernel_> yes
<Fallenou> ok
<lekernel_> at least when running at 100MHz, when it's 80MHz it's a bit less
<Fallenou> sure
<lekernel_> also, the DRAM control algorithm cannot reach the peak bandwidth in practice
<Fallenou> because of column changing ?
<lekernel_> if I have time and motivation, i'll redo the complete SoC design at some point, with out-of-order memory transactions, prefetching and QoS
<lekernel_> it'd potentially multiply the memory bandwidth available in reality by a factor of 4
<Fallenou> wo :)
<lekernel_> (and then we could probably process HD video)
<Fallenou> lekernel_: it is 16 bits sdram chips on the M1 ?
<lekernel_> yes
<Fallenou> so it's a 32 bits data bus, with 16 bits from each chip ?
<Fallenou> you sample at both rising and falling edge and you have 2 32 bits words per cycle ?
<lekernel_> yes
<Fallenou> ok
<Fallenou> when you say you do 4 reads bursts
<Fallenou> each read is 32 bits ?
<Fallenou> or 64 bits ?
<mwalle> hi
<mwalle> btw the avr core is a busmaster right?
<Fallenou> hi
<mwalle> we should route the 'break' signal to the avr core and write our own gdb stub :)
<mwalle> disable uart in main lm32
<mwalle> and use it from within the avr core
<mwalle> and we dont need any jtag2gdb bridge
<mwalle> would be a cool gsoc project :)
<lekernel_> ah yes, definitely
<lekernel_> do you want to mentor that? :)
<mwalle> maybe ;)
<lekernel_> Fallenou: each read is 64 bits
<lekernel_> mwalle: well, the deadline is tomorrow
<mwalle> narf :)
<mwalle> ill do it
<lekernel_> btw the avr core isn't a wishbone bus master, but it can send text messages to the lm32 which can forward them to the uart
<mwalle> mh
<lekernel_> there's actually a debug print() function in the avr firmware that does exactly that
<mwalle> ah i remember, theres a dual ported ram in between
<lekernel_> yes
<mwalle> sorry i'm to tired, going to bed now
<lekernel> feel free to edit http://www.milkymist.org/wiki/index.php?title=GSoC_application_2011 if you want to participate
<mwalle> when exactly is the deadline tomorrow?
<Fallenou> it's the deadline for mentoring organization
<lekernel> March 11: 23:00 UTC
<Fallenou> it's the 11 march 23
<lekernel> but we'll need to update the application in the google system before that
<Fallenou> for this dead line you have to list the mentors ?
<lekernel> yes
<Fallenou> oh crap :(
<Fallenou> the project list can be updated later on
<Fallenou> but you need the mentor list
<Fallenou> humm i'm not sure lekernel
<Fallenou> "Guide for Mentors"
<Fallenou> Once an organization has had its application accepted, anyone who has created a site wide profile can apply to that organization as a mentor
<lekernel> well they do ask for an initial mentor list
<Fallenou> oh ok
<lekernel> and probably attribute slots based on that
<lekernel> actually, they ask tons of stuff and writing an application is a pain
<lekernel> especially when they reject it with an automatically generated message
<Fallenou> slots are attributed after student proposals
<lekernel> and accept shitty projects like beagleboard instead
<Fallenou> so I guess they take into account the total number of mentors
<Fallenou> yes it's not perfect
<Fallenou> but it's still a good thing, if it can provide debugguer or mmu :)
<Fallenou> how many mentors do you have lekernel ?
<Fallenou> I guess there is mwalle + you , and then who else ?
<lekernel> maybe Eric Rannaud (one of the ulogic guys actually ;) and Jon