<lekernel> hmm i'll recommend to leave the source untouched in the first place
<lekernel> also your host CPU needs to load a program in the avr
<jackgassett> I should be able to use data2mem to load the BRAM with the program.
<jackgassett> I don't mean to be bugging you with silly questions. :)
<jackgassett> but I looked at the top level signals and it looked pretty straighforward. But I get errors trying to synthesize.
<lekernel> if you really want to use that proprietary thing... and in every case, the AVR is in reset mode by default and needs to be started with a CSR write to the right register
<jackgassett> both when I do nothing to modify the project and when I try to comment out the usb modules.
<jackgassett> when you say it would not work by just dropping into ISE project what issues are you referring too?
<jackgassett> that it needs external way to program the code and USB hardware?
<lekernel> well, you need a host system, and support software
<jackgassett> ok, I think I could work through those issues. but I see this which I've never seen before when I comment out usb modules.
<jackgassett>     Using one-hot encoding for signal <dmem_sel>.
<jackgassett> INTERNAL_ERROR:Xst:hdltool.c:4862:1.209 -  Signal dmem_sel, 0 ones:XXXXXXXXXXXXX
<jackgassett> INTERNAL_ERROR:Xst:hdltool.c:4863:1.209 -  To resolve this error, please consult the Answers Database and other online resources at http://support.xilinx.com
<lekernel> yeah, that's a xst bug for spartan3 that we don't have for spartan6
<jackgassett> ok. I see.
<lekernel> I told you about that and the solution when you came on IRC the first day ... search the logs
<jackgassett> oh. ok.
<jackgassett> well thank you for your help.
<jackgassett> I appreciate it.
<CIA-94> flickernoise: Sebastien Bourdeauducq master * r862c11f / src/Makefile : Build LZMA image - http://bit.ly/hoTsbO
<jackgassett> hmmm, just reviewed chat logs and maybe you are confusing me with someone else. We never got into any really technical discussions.
<jackgassett> I don't see anything about a spartan3 bug. can you point me in the direction to start searching on google for a solution?
<lekernel> you need to disable signal re-encoding in the xst option
<jackgassett> ok, thank you.
<lekernel> but to do so, you need to switch to advanced mode in the process properties and then pass that as "extra xst command line"
<lekernel> the exact option name (command line) is in the xst user guide
<jackgassett> ok, that should put me in the right direction. thank you again. I know how it is to be busy and people start bugging you. :)
<lekernel> maybe there's a possibility to only disable reencoding on this signal e.g. with a special comment in the verilog file
<jackgassett> last time we talked you said navre uses 30% of 250K chip and I just wanted to take a look.
<lekernel> this wouldn't prevent the synthesizer from performing the re-encoding optimization where it doesn't crash
<jackgassett> btw, did you see this on hackaday today? http://hackaday.com/2011/03/02/gameduino/
<jackgassett> looks like a cool project.
<lekernel> yup. and maybe I should rename this project "milkuino" so they'll post it at last.
<lekernel> (or not)
<larsc> fpgarduino
<jackgassett> :) have they never posted your project?
<jackgassett> they wont post my papilio either!
<kristianpaul> lzma, nice :-)
<kristianpaul> upgrades to ISE 13.1
<kristianpaul> xiangfu: Hi
<kristianpaul> xiangfu: How is your mm1 gointg?
<kristianpaul> going**
<xiangfu> kristianpaul: Hi. finish assemble the case. play flicknoise a little.
<kristianpaul> good
<xiangfu> kristianpaul: next step. try to reflash ;)
<kristianpaul> ah thats not so long, but what really take some time is get up all the toolchains and flicernoise compiled
<kristianpaul> wonders if the lzma support is intentionally related to avoid the "Invalid flash boot image length" when booting last flicernoise versions ;-)
<kristianpaul> 8,2M > 2,8M , and if remenber well bios limit is 4M
<xiangfu> kristianpaul: 1. reflash. 2. compile toolchians and flicernoise. :)
<kristianpaul> ha, a have small typo in build_bitstream.sh, LOGFILE or LOGFILEHOST ?, please define !
<kristianpaul> choosed LOGFILEHOST
<kristianpaul> xiangfu: can i ask about your current interest with milkymist?
<kristianpaul> rtems, linux, hdl part, flickernoise?
<xiangfu> kristianpaul: I think I like rtems. I would like stick lekernel's plan for start  :)
<kristianpaul> :)
<kristianpaul> ok lzma add some seconds to boot time..
<yizhang> hi, i tried to compile ftmem, and got an error /bin/sh: xst: not found
<yizhang> which package should i install?
<yizhang> sorry, fjmem
<yizhang> or do i need to install anything from xilinx?
<yizhang> do I need Xilinx ISE?
<yizhang> I hope any of you can give me a hint, i'm stuck here ;)
<yizhang> resolved, thanks xiangfu. use pre-synthesized bitstream.
<aw> is there anyone know what's currently supported resolutions on M1's VGA out? I meant s/w ported already.
<terpstra> lekernel, i split the multiplier into parts so that the multiply and carry-chains are in different stages. altera has 18x18->36 multiplier blocks so performing a 32x32->32 multiplication requires inferring the multiplier and adders in the same cycle. my manual splitting makes those land on different cycles. obviously, one could do further to pipeline it, but not with inferred logic. and for the cyclone3, the mul and add on different cycles was eno
<terpstra> ugh to remove it from the critical path.
<terpstra> for the arria2 it's still one of the slowest paths, and there is a way to tell quartus to make the multiplier span two clock cycles, but i just haven't been fussed to get it faster than 175MHz yet
<terpstra> my current project is to get a reliable gdb connection going!
<xiangfu> Hi where I can find more .fnp files ?
<kristianpaul> wolfspraul is yizhang the same Yi the same from sharism ?
<wolfspraul> kristianpaul: yes, same
<Gurty> scrts Happy Birthday ^_^
<lekernel> wolfspraul: how many M1's do you have left?
<Fallenou> yep happy birthday scrts !!
<CIA-94> milkymist: Sebastien Bourdeauducq master * r22569e9 / software/bios/boot.c : Run CRC before decompression - http://bit.ly/eiZkcj
<CIA-94> milkymist: Sebastien Bourdeauducq master * rf634db3 / software/libhal/brd.c : Claim to support PCB revision 2 (RC3 run) - http://bit.ly/hYOu3l
<CIA-94> milkymist: Sebastien Bourdeauducq master * r67f4f7b / boards/milkymist-one/rtl/setup.v : Enable back all cores in rescue mode (works with ISE 13) - http://bit.ly/ftThYs
<CIA-94> milkymist: Sebastien Bourdeauducq master * rf890883 / boards/milkymist-one/standby/Makefile : Generate swapped image for standby bitstream - http://bit.ly/dSes2e
<wolfspraul> lekernel: which projector do you think works best with Milkymist One?
<wolfspraul> there seem to be different technologies like LED, LCD, DLP
<wolfspraul> and different resolutions, it seems the price goes up dramatically with higher resolution
<lekernel> I don't really know
<lekernel> luminosity is important
<lekernel> resolution, not that much
<wolfspraul> in fact the cheapest projectors are all the ones that max out at 800x600, which is perfectly fine for m1, afaik
<lekernel> if you get a 800x600 projector with very high luminosity, it would look better than a HD one with low luminosity
<lekernel> fyi, the big panels at the Stade de France have PAL (768x576) resolution :)
<lekernel> the LED ones have very low luminosity
<lekernel> seems DLP is the best
<wolfspraul> LCD also
<wolfspraul> looks like around 300 EUR, you can get LCD or DLP ones with ca. 2000 ansi-lumen
<wolfspraul> for the same 300 EUR, you can get a LED one with 150 ansi lumen, Samsung Pocket Image SP-P400B
<wolfspraul> (if you sort say by resolution, you get a fairly nice overview)
<wolfspraul> I would guess (from the pictures) the LED one is probably much lighter/smaller and needs less power?
<wolfspraul> ok, I think I got an idea. good :-)
<lekernel> also try to get a good dynamic range
<lekernel> it doesn't make sense to have such a high luminosity if the thing also projects a gray background all over the picture (ie black is gray)
<wolfspraul> how about camera? what works well / any guidance?
<wolfspraul> lekernel: what kind of output on the camera is preferable? component video with 3 cables, or composite with just one? are both supported?
<lekernel> component video has better quality and should be supported but has never been tested afaik
<wolfspraul> do you have any recommendations what kind of camera makes sense?
<wolfspraul> which camera features are important/unimportant when used with m1?
<lekernel> so i'd stick with composite unless we're ready to potentially spend time on sw fixes
<wolfspraul> that's clear :-) composite
<lekernel> it depends what you are doing with it, but good performance in low luminosity is nice when used on stage
<wolfspraul> is there a cheap 'camera' that just sort of streams composite down the wire and that's all?
<lekernel> oh sure
<lekernel> you can take a cheap CCTV camera
<lekernel> I saw some at around 15 euros or so
<lekernel> big brother has some advantages too :)
<wolfspraul> I can imagine. Chinese police state helps the whole world with low cctv prices, thanks to driving the volumes of those critters into the tens of millions and more.
<wolfspraul> oh you should see the cctv building in shenzhen, it's sickening.
<wolfspraul> that's one field where china definitely is top of the world already
<lekernel> beats London?
<wolfspraul> but anyway, I got it
<wolfspraul> oh, many times i think
<wolfspraul> there must be millions and millions and god knows how many cameras here
<wolfspraul> you are being photographed all the time, the cameras are EVERYWHERE
<wolfspraul> I don't even want to list typical places, just believe me when I say EVERYWHERE.
<wolfspraul> and in 5 years they will not be easy to spot either anymore - nice.
<kristianpaul> cctv is just white and black right? should the camera support colors also?
<lekernel> oh, B&W is for old CCTV :) innovation never stops there...
<kristianpaul> heh, ok
<lekernel> btw, Chinese cameras might be NTSC. AFAIK, this hasn't been tested either
<lekernel> so if you want to be sure not to have to mess in the code, find one that has PAL, which I tested many times and works
<lekernel> otoh, getting ntsc (as well as component) to work would be valuable too
<lekernel> is reworking all the mm soc scripts to totally do away with impact
<wolfspraul> understood about ntsc and component - this is helpful, thanks!
<CIA-59> flickernoise: Sebastien Bourdeauducq master * rcb4808e / (6 files in 2 dirs): Flash scripts - http://bit.ly/hM9omj
<CIA-59> milkymist: Sebastien Bourdeauducq master * r13b3cd8 / (7 files in 3 dirs): UrJTAG flashing - http://bit.ly/gS3shs
<wolfspraul> lekernel: when I connect a music player to line-in, and a speaker to line-out, I should get the same music coming out on line-out?
<lekernel> yeah
<lekernel> but only when flickernoise is running
<wolfspraul> it seems there is an ongoing noise or something
<lekernel> (you can click "audio" in the gui to get the mixer table)
<lekernel> yeah, that lm4550 produces a noticeable amount of noise with some speakers
<lekernel> that's one of the hw problems we have
<wolfspraul> are there any requirements for the equipment so that this works? you think the hardware is bug free? (to the degree that we care)
<lekernel> it's normal line audio
<lekernel> yeah, we don't care about the output noise, especially since it seems we can't fix it easily
<lekernel> audio output is an extra feature
<wolfspraul> why can't we fix it?
<lekernel> because I don't know where it comes from, and investigation sinks time
<wolfspraul> is this noise with all speakers - you say 'some', can you elaborate a bit moer?
<wolfspraul> you have seen some with less/no noise?
<lekernel> yes
<wolfspraul> but you don't know what the requirements are that will make it noise free?
<lekernel> but it could simply be a false impression depending on the configuration of the speakers etc.
<lekernel> I did not measure anything
<lekernel> yeah... if you care, maybe replace the lm4550 with a more high-end ac97 codec (they have some degree of footprint compatibility)
<lekernel> but that's not something worthy of time imo
<wolfspraul> is the lm4550 configurable to some degree? or is it definitely a hardware problem?
<lekernel> fyi the ML401, which has the same codec chip, also had a nice amount of output noise
<wolfspraul> yes, replace codec doesn't sound right to me
<lekernel> and it was not only white noise like the M1, you could hear other parts working like the DRAM
<lekernel> i think it's a hardware problem, but definitely not an important one
<wolfspraul> well, first we need to know our product well, and manage expectations.
<wolfspraul> anyway, you answered all my questions - thanks!
<lekernel> if I were to fix it the first thing i'd try is replace the codec chip
<wolfspraul> the line-in is not affected by this?
<lekernel> it seems to be affected as well, but to a lesser degree
<lekernel> but quite frankly I haven't done measurements or investigations on this very minor issue
<lekernel> the only thing I tried was additional power supply filtering (that 4.3V regulator)
<lekernel> but it did not improve
<lekernel> btw I think you can simply leave off that regulator on rc3, and simply use the shorting 0 ohm placeholder that was made for that
<wolfspraul> have you ever recorded line-in and played it back on a noise-free system?
<lekernel> no
<CIA-59> flickernoise: Sebastien Bourdeauducq master * rf791278 / src/Makefile : Remove MCS - http://bit.ly/i99pAh
<wolfspraul> ok got it - thanks, all understood now
<lekernel_> #$!{!# when are they going to get that irc network to function properly...
<kristianpaul> lekernel_: what is J3 for?
<kristianpaul> audio expansion header?
<lekernel_> Fallenou: but who the hell is attacking the freenode servers?
<lekernel_> and continuously for years
<lekernel_> the hatred goes deep ...
<Fallenou> It's not always an attack, but sometimes it is
<Fallenou> a 15 years-old dumbass
<Fallenou> script kiddies
<kristianpaul> yeah
<Fallenou> I have attacks everyday on EpiKnet
<Fallenou> some are really strong botnet attacks
<Fallenou> with a lot of different IP address
<kristianpaul> what was that scrip to parse verilog modules to html?..
<kristianpaul> script*
<kristianpaul> ah. nv
<kristianpaul> v2html
<kristianpaul> need a better way of visualize core interconnect in mm1 than with vim
<Fallenou> you could load it in ISE
<Fallenou> it shows the block diagrams
<kristianpaul> I want avoid that
<lekernel_> opensparc is using something like that (iirc)
<Fallenou> kristianpaul: no target rtems when doing make rtems
<Fallenou> in your lua port
<Fallenou> oh it's in src
<Fallenou> you forgot to mention cd src
<Fallenou> :)
<kristianpaul> ah
<kristianpaul> jaja
<Fallenou> ok it's working nicely :)
<Fallenou> i printed hello :p
<kristianpaul> :-)
<Fallenou> (I don't know anything about lua)
<kristianpaul> me either, but i will learn ..
<Fallenou> so yes you should add the "cd src" in your blog article
<Fallenou> it may confuse someone
<Fallenou> even if it seems straight forward to you :)
<kristianpaul> before make rtems?
<Fallenou> yes
<Fallenou> or even better
<Fallenou> cd milkymist-lua ; git checkout milkymist ; cd src
<Fallenou> for replace ; by &&
<Fallenou> or*
<kristianpaul> done
<Fallenou> thanks
<lekernel_> hm, I thought all VHDL does is also add a delta delay to every assignment...?
<Fallenou> nice :)
<kristianpaul> well, at least is browsable
<Fallenou> yep
<lekernel_> afaik http://books.google.be/books?id=_VGghBpoK6cC&lpg=PR12&ots=F2kIzTXUA3&dq=sutherland%20gotcha%2029&hl=en&pg=PA64#v=onepage&q&f=false also applies to VHDL, though the blog post seems to suggest otherwise
<lekernel_> any expert with such details here?...
<Fallenou> verilog may have two assignment
<Fallenou> vhdl has signals avec variables
<Fallenou> I don't know if it's a good thing
<Fallenou> wow that's crazy
<lekernel_> yeah and rather dull. it's almost never useful in practice and always cause problems.
<Fallenou> so if I understand correctly
<Fallenou> for thinks like
<Fallenou> a = b xor c;
<Fallenou> without clock or anything
<Fallenou> we should use the =
<Fallenou> and for click sensitiv things we should use <=
<Fallenou> clock *
<lekernel_> yup. but blocking assignments (=) are also very useful in some places where there is a clock
<lekernel_> (like vhdl variables)
<Fallenou> hum ok
<Fallenou> I'm not sure to understand really the difference
<lekernel> urgh, the sigasi license is 1800¬
<lekernel> quite pricey for a IDE
<Fallenou> arg it's Eclipse based
<Fallenou> they did not integrate any synthesis tool :p
<Fallenou> they are not that crazy
<lekernel> asic tools are expensive :p
<Fallenou> well when you pay some millions for the asic mask, you can pay some more for the tools :p
<lekernel> yay, I managed to reduce the runtime of the antares xdlrc parser from 3 hours to less than 2 minutes
<lekernel> did you know? the smallest spartan6 has 333797 unique on-chip programmable wires :) (unless said parser has bugs)
<Fallenou> sounds like we have a "open source synthetizer" conference for the 28C3 lekernel ;)
<kristianpaul> Fallenou: i think will be more than a  synthetizer for that time ;-)
<mumptai> hi
<kristianpaul> hi
<kristianpaul> dammit
<kristianpaul> is me or gcc port for lm32 dint know about the mico32 cycle counter..