#milkymist IRC log for Tuesday, 2011-06-14

GitHub159[scripts] xiangfu pushed 3 new commits to master: http://bit.ly/iPtlSA08:07
GitHub159[scripts/master] fix the libpng download URL - Xiangfu Liu08:07
GitHub159[scripts/master] remote rtems-patches - Xiangfu Liu08:07
GitHub159[scripts/master] build lm32-rtems4.11-gdb - Xiangfu Liu08:07
lekernelxiangfu, why have you removed the rtems patches from the build scripts?08:44
lekernelyou download them from rtems.org?08:44
xiangfuyes. download them from rtems.org08:44
xiangfuI am debug the libcurl timeout.08:45
xiangfuI found when I exit the rendering. it's needs 2~4 mins for wait the 'rtems_semaphore_obtain',08:46
xiangfuwhich is in libcurl.08:46
xiangfulibcurl not exit. :(08:46
xiangfualso the timeout in libcurl should be shorter then the 'rsswall_refresh' right?08:47
xiangfuor the feed_get_last will be 're-enter'08:48
xiangfuright?08:48
lekernelno08:58
lekernelthe loop goes like this: while(terminate_event || refresh_period_elapsed) { download(); display(); }08:59
xiangfuthanks.09:00
lekernelthe refresh period is actually configured period + download time...09:00
xiangfuhmm... I only found:09:05
xiangfuwhile(rtems_event_receive(RTEMS_EVENT_1,09:05
xiangfu  RTEMS_EVENT_ANY | RTEMS_WAIT,09:05
xiangfu  rsswall_refresh, &events) == RTEMS_TIMEOUT) {09:05
xiangfu 09:05
xiangfuthere is no `terminate_event`09:05
xiangfuanother question. when I hold the left and middle button. then press the right. it's will goto crash screen, the serial output is :09:10
xiangfuu: inc. DQ delay  // d: dec. DQ delay t: test (small)   // T: test (large)....09:10
lekernelRTEMS_EVENT_1 is the terminate event09:11
lekernelthis is not crash screen, it's the dram test data - dram test mode is entered when powered up with the left button held09:12
lekernelit's normal behaviour, even if it looks like a crash09:13
xiangfugot it.09:14
lekernelmaybe we will replace that with a jumper on J21, if people complain :p09:14
GitHub148[flickernoise] sbourdeauducq pushed 1 new commit to master: http://bit.ly/mNsxzn11:00
GitHub148[flickernoise/master] French translation: add flasher states - Sebastien Bourdeauducq11:00
lekernelwolfspraul, "Windows logos can be removed (+1 RMB), but printing 'Milkymist' brand on it would have MOQ of 300" -> well, the windows key does nothing in the software anyway...11:13
lekerneland I can't think of a useful action to bind to it11:13
lekerneland the gui isn't functional without mouse. in fact, the mouse is more important than the keyboard.11:15
rohmaybe a kbd with integrated touchpad?11:19
lekernelthe USB firmware doesn't support hubs11:19
wolfspraullekernel: xiangfu will try to make the gui fully functional with keyboard only, should not be too hard11:19
wolfspraulwe looked at keyboards with integrated mice, they do exist, but the problem is that the pointer accuracy is so bad that it would either be a big software effort, or unusable11:20
lekerneland given the incredible amount of overengineering that went into USB, making them work is extremely painful11:20
lekernelwolfspraul, include a mouse then11:20
lekerneland no keyboard11:20
lekernelmh, or simulate the mouse pointer with windows key + arrows?11:21
wolfspraulI haven't found a mouse yet we wouldn't support, but we don't support quite a few varieties of keyboards.11:22
lekernelthat's the only easy option I see11:22
wolfspraulso keyboard first, then mouse if necessary11:22
wolfspraulin the long run we don't know very well how users will use the product, what feedback we get, and how the software evolves. so all sorts of things can change then.11:22
wolfspraulright now I can say that: keyboard with integrated mouse: couldn't find one that worked well11:23
wolfspraulmouse only: will only expose a subset of functionality, most likely ever11:23
wolfspraulkeyboard only: maybe it's a small effort to make the gui 100% controllable without mouse. if not, we can always throw in a mouse as well, or see whether people have trouble finding a mouse themselves, given that I think all will work...11:24
wolfspraulkeyboard+mouse: still an option, if we think so a little later11:24
lekernelthe only way to make it a small effort is to trigger mouse pointer movements with keystrokes11:24
lekerneleg win+arrows11:24
wolfspraulneither keyboard nor mouse: possible if the main method of interacting is something entirely different, like a web interface11:25
lekerneli'll add the win+arrows shortcut to the software roadmap11:27
lekernelwolfspraul, any news from the box btw?11:34
wolfspraulno news, be patient11:35
wolfspraulthat one is stuck on my end right now, I need to make sense of the dxf files we got then it moves again11:36
wolfspraulalso we need to make a few adjustments11:36
wolfspraulno worries, it's moving11:36
lekernelcan you forward me those dxf files?11:37
wolfspraulI've seen the price of the PCBs today, so I think that's moving too :-) (ca. 2000 USD)11:37
lekernelah, it's dxf from the printer?11:37
wolfspraulI cannot yet make sense of them, please wait a little longer of course they will be uploaded etc.11:37
wolfspraulI know too little, you need to wait.11:37
lekernel$2k for 95pcs? not too bad...11:38
wolfspraulhe11:38
wolfspraulas a ballpark, I can say that pretty much any PCB can be brought down to under 1 USD11:39
wolfspraulTaiwan is about twice as expensive as China, but the quality of employees is also about twice as good, and overall the resulting quality of the PCB is better11:40
wolfspraulif we make 10k of this thing in China, it'll cost 2 USD :-)11:40
wolfsprauland it will cost me a lot of nerves11:40
wolfspraulactually it's fairly big, so who knows the price may not come down that fast with higher volume, but it definitely can. yes, right now 20 USD / piece is good, I agree.11:42
wolfspraulin rc2 it was more like 40 USD / piece, mostly because we made less11:42
wolfsprauland in rc1 it was ca. 150 USD a piece, if I remember correctly11:43
lekernelare you still using immersion gold on rc3?11:43
wolfspraulrc1, 900 USD for 6 pieces11:43
wolfspraulrc2, 1600 USD for 4x pieces11:43
wolfspraulrc3 2000 USD for 95 pieces11:43
wolfspraulI don't know [immersion gold]11:43
wolfspraulAdam will probably err on the side of safety, it's still all low volume.11:44
wolfspraulso it may not be worth to take a risk to save 200 USD11:44
wpwrakENIG usually doesn't add much cost anyway11:44
wolfspraulif I can save 10k USD, it's worth some time and risk. to save 100 USD, nah, I rather pay them and sleep better.11:44
wpwrak(prices) one-time cost ~USD 800. per-board cost ~USD 15.11:48
wpwrak(ENIG saving) e.g., at 4pcb, ENIG or tin are the same once your board reaches a certain level of difficulty. seems that the gold isn't all that expensive :)11:49
wolfspraulwpwrak: there are definitely more numbers in the calculation, but I don't know the PCB process well at all11:51
wolfspraulI think you pay quite a bit for labor, so once you understand well where the labor exactly is, and you are willing to invest to automate that or move it to your side, the price you pay will drop and drop.11:53
wolfspraulit's not even worth to find out right now though11:53
wolfspraulwe have been testing several different pcb makers too, I think Adam had some fun :-)11:54
wolfspraulnow we use NewHeart http://newheart.com.tw/11:54
wolfspraulthey are a bit more expensive but have higher quality staff so communication is easier11:55
wpwrakthat alone probably saves a lot :)12:08
Fallenoulekernel: at the PES on Friday there will be some VJing performance, are you aware of that ? will they have some milkymist device ?12:44
rohlekernel: hey there13:26
rohlekernel: have you done anything with dsps yet/ know any open cores for that?13:27
larscroh: "Delayed sleep-phase syndrome"?13:42
rohlarsc: ?13:43
larscdsps13:44
rohhrr13:44
rohi am sure they can result in such syndromes13:45
larscah, now i get what you mean13:45
larscDSPs13:45
rohlarsc: yes. i am still thinking about what needs to be done to add a dsp core to mm.. so it could be used like the soundart chameleon13:46
lekernelFallenou, hackable devices is supposed to participate, yes15:47
Fallenouso the performer will use a M1 ? :)15:47
lekernelroh, just forget about this unless you want to program the fpga yourself15:48
kristianpaulor add code octave to export code in verilog ;)15:51
lekernelFallenou, are you talking about "SHOW : OUSK !" ?15:53
lekernelhttp://www.passageenseine.org/pes-201115:53
Fallenouyes15:53
lekernelMinky Mina doesn't have a M1, but she's definitely interested in it15:54
lekernelhopefully guyzmo will show her how nice it is :-)15:54
Fallenouhehe ok !15:55
lekerneli'll actually message her so she kicks guyzmo's ass and things get done... hehehe15:56
lekernelah she already knows about it :-)15:58
guyzmolekernel - we already talked her about it afaict :)16:00
Fallenouhehe k awesome16:00
Fallenougotta go see you16:00
lekernelguyzmo, btw with the latest software version (in git) you can control patch variables and switch patches from an arduino16:03
GitHub65[mtk] sbourdeauducq pushed 1 new commit to master: http://bit.ly/kPZhVE18:27
GitHub65[mtk/master] Use Meta+arrows/enter to move/click mouse pointer - Sebastien Bourdeauducq18:27
kristianpaulfpgaminer: with unrolling set to 0, it will fit on LX150 isnt?18:34
topi`lekernel: hi mate :) hope you enjoyed your trip to amsterdam. can you explain me why the price of the milkymist one is over 300 euros? I checked digikey.com and the cost of one Xilinx Spartan-6 LX45 is $66 ... are there other expensive components on the board?19:16
kristianpaultopi`: you should consider SMT and PCB too, but here you can check BOM  http://en.qi-hardware.com/wiki/Milkymist_One_RC2_BOM19:26
kristianpaulnot counting human labor i think ;)19:26
kristianpaulalso topi` , FPGA based boards are usually sponsored19:27
kristianpauli mean you dont pay for what the board realy costs, as it used as marketing19:27
kristianpaulie, the spartan avnet 49usd board19:28
lekerneltopi`, if you can make the M1 cheaper, we'd be happy to work with you.19:57
kristianpaul:D19:57
lekerneland yeah, there are tons of stuff which add cost everywhere19:57
topi`lekernel: I looked at the BOM and it indeed looks like it takes a lot of components, hence more complexity to the PCB20:32
topi`well. hardware design is something that is not my specialty.20:32
topi`will there be a new revision of the board at some point? do you think it's beneficial to go for a more expensive Spartan? i.e. with more LUTs20:34
topi`but I guess synthesizing the cpu core is not all about how many LUTs are available.20:34
lekernelwe're only using a bit less than half of the ressources of the current fpga20:46
topi`oh20:46
lekernelso - not anytime soon. the current focus is to finalize the software and sell lots of units.20:46
topi`if I find 20 buyers, would I get a discount? :)20:47
lekernelprobably20:47
lekernelthough by "lots of units", I rather mean something like 40000, like existing professional video consoles did.20:49
topi`what software do you use for the Xilinx chip? there's a free sw from Xilinx itself, but does it support bigger devices like LX45?21:00
kristianpaullekernel: topi`: an interesting option for people not wanting mm1 board with VJ related stuff and the most basic SoC functionallity is to just have the same board with unpopulated parts21:00
kristianpaultopi`: LX45 is supported, up to LX75 if i remenber well21:00
kristianpaulbigger will be somthing like LX150 btw ;-)21:01
topi`kristianpaul: I would think none of the RAMDACs etc are particularly expensive parts21:12
topi`but yes, it could be one way to offer two different price points21:12
topi`some bitcoin miners could be fooled to buy those ;)21:13
topi`what I'd like to do would be full emulation of a 68000 processor and the Amiga custom chips :)21:14
topi`there are software emulators, but they aren't very efficient21:14
GitHub97[llvm-lm32] jpbonn pushed 2 new commits to master: http://bit.ly/inHC6p21:22
GitHub97[llvm-lm32/master] Merge branch 'master' of http://llvm.org/git/llvm into mico32 - JP Bonn21:22
GitHub97[llvm-lm32/master] Updated for latest LLVM head. - JP Bonn21:22
GitHub36[llvm-lm32] jpbonn pushed 1 new commit to master: http://bit.ly/kDUKTE22:07
GitHub36[llvm-lm32/master] Merge branch 'master' of http://llvm.org/git/llvm into mico32 - JP Bonn22:07
lekernelusing a FPGA just to emulate those outdated archs is boring22:50
--- Wed Jun 15 201100:00

Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!