#milkymist IRC log for Friday, 2012-01-13

frdmincSo got DMX working, just had to set dmx1 to the start address of my light, dmx2 to that +1, etc.00:03
frdmincWondering what "Chain" does at the moment...00:03
wpwrak(dmx working) whee ! congratulations !00:06
wpwrak(chain) forward input to output ?00:06
frdmincwpwrak: hmm that would make sense. Thanks. So I'd need a different DMX controller to test that I guess :)00:07
frdmincor another M1...00:07
wpwrakwolfspra1l: btw, how's the epic usb-midi battle ? still waiting for the update ?00:12
frdmincwow the fades between effects are awesome00:14
GitHub166[flickernoise] wpwrak pushed 6 new commits to symtab: http://git.io/njOHiA00:21
GitHub166[flickernoise/symtab] ptest: put a blank line between system and user symbols, like symtab_dump did - Werner Almesberger00:21
GitHub166[flickernoise/symtab] compiler: change forall_<plural> to more common foreach_<singular> - Werner Almesberger00:21
GitHub166[flickernoise/symtab] src/Makefile: suppress ctype.h warnings with -Wno-char-subscripts - Werner Almesberger00:21
GitHub86[milkymist] wpwrak pushed 1 new commit to symtab: https://github.com/milkymist/milkymist/commit/9f8e94702f6ec5c26b6405a9a238e7cfcddff47200:28
GitHub86[milkymist/symtab] libfpvm/x86-linux/Makefile: support silent builds as well - Werner Almesberger00:28
GitHub36[flickernoise] wpwrak pushed 1 new commit to symtab: http://git.io/elOy7g00:28
GitHub36[flickernoise/symtab] ptest/Makefile: added "silent" copy operation - Werner Almesberger00:28
frdminc_xiangfu: hi, wpwrak tells me you are the DMX expert05:35
xiangfufrdminc_, not expert. I have try some DMX fixtures with Milkymist One :)05:35
frdminc_xiangfu: Have you gotten multiple DMX lights to do different things at the same time with any of the current patches?05:36
frdminc_(I have 3 lights each of which has 4 addresses - R G B and on/off/brightness/strobe05:37
xiangfufrdminc_, Hi there is one patch that support DMX out: Rozzor & Aderrasi - Canon (DMX out).fnp05:38
xiangfubut there are only dmx1/2/3/4 mean four channels.05:38
xiangfufrdminc_, now Milkymist one support 8 output channels.05:39
frdminc_xiangfu: ah okay so I'm not missing anything. So it would be possible to write a patch that controlled 2 lights seperately, but not 3.05:39
frdminc_xiangfu: although I'm a bit confused as with the "DMX Desk" it looks like you can control up to 256 channels manually.05:39
xiangfu(controlled 2 lights seperately) totally ok.05:40
xiangfuDMX Desk. yes. we can control  all 512(not only 256 :) channels manually,05:41
xiangfuM1 can act like a DMX device or DMX controller.05:42
xiangfuunder the DMX settings: there two columns , idmx1/2/3/4/5/6/7/8   and dmx1/2/3/4/5/6/7/805:43
xiangfuidmx* mean dmx input.05:44
xiangfudmx* mean dmx output.05:44
frdminc_xiangfu: so 8 channels programmatically or 512 channels manually, is that right?05:44
xiangfuyes. that is correct.05:44
frdminc_xiangfu: is the 8 channels programmatically a hardware or software limitation?05:45
xiangfusoftware limitation.05:45
xiangfuat begin we support 4 channels. now we have 8. will have more in future.05:45
frdminc_xiangfu: thanks a lot. So it looks like I want to work on something pretty simple getting an effect to use all of the available 8 DMX outputs might be a simpler thing. But for party tomorrow will just set all 3 lights to use the same DMX addresses :)05:46
xiangfuyes. 3 lights use the same DMX addresses.05:48
xiangfuplease share your patches with us. we need more DMX patches. for now there are only one.05:48
xiangfuI want finish one dmx patch. I am slow. still not write it. :(05:49
xiangfufrdminc_, when you switch patches. from DMX patch to a non-dmx patch. all dmx variables will set as 0.05:50
xiangfumostly mean the light will off.05:50
frdminc_xiangfu: actually I think I found 1-2 other patches that did some DMX stuff.05:51
xiangfufrdminc_, if you already prepared some patches. a word around for now is add a default dmx* value in all your non-dmx patches.05:51
xiangfufrdminc_, great where is them?05:52
frdminc_xiangfu: I'm far away from actually modifying or writing anything05:52
frdminc_xiangfu: They were just included with the set of patches it shipped with. They had DMX in the name somewhere. I'm not in the same place as the M1 at the moment so can't take a look right now.05:53
xiangfuoh. correct. there are two patches that have dmx output. sorry.05:54
xiangfufrdminc_, there are two dmx video here: http://www.youtube.com/user/xiangfuliu?feature=mhee05:56
xiangfufrdminc_, in this patch 'Rozzor & Aderrasi - Canon (DMX out).fnp' I think you can just map dmx2/3/4 to R G B. then it will work out of box.06:00
frdminc_xiangfu: Need to sleep now, 1am my time... but thanks a lot... I'll need to look at "MIlkymist One & DMX fixtures" closer later today, looks like you got Drawing Board to do something somehow :)06:01
xiangfufrdminc_, and map the dmx1 to ' on/off/brightness/strobe' . set it to ON. for in that patch is set it '1' mean '255'06:01
xiangfufrdminc_, sure. goodnight.06:01
frdminc_xiangfu: yeah that one works fine, want to change dmx1 to have more variation oh as my lights start to strobe at a certain dmx1 level.06:01
xiangfufrdminc_, have a nice party tomorrow. :)06:01
GitHub79[scripts] xiangfu pushed 2 new commits to master: http://git.io/GfBQlA07:30
GitHub79[scripts/master] rtems is changed to git. cvs not working anymore - Xiangfu Liu07:30
GitHub79[scripts/master] compile-milkymist-firmware.sh: no needs handle cvs rtems - Xiangfu Liu07:30
GitHub9[scripts] xiangfu pushed 1 new commit to master: http://git.io/nreMwA07:34
GitHub9[scripts/master] compile-milkymist-firmware.sh: track rtems git revision - Xiangfu Liu07:34
cladamw( current rc4 sch, audio & dvi-i not finished yet): http://downloads.qi-hardware.com/hardware/milkymist_one/sch/tmp/08:01
cladamwthe whole details of modification/histories, I'll record them here: http://en.qi-hardware.com/wiki/Milkymist_One_Schematic_Change_History08:18
xiangfuwpwrak, just sent out the audio indicate patches. :)09:10
wpwrakxiangfu: do the bars grow from the top ? if yes, can you make them grow from the bottom ?09:38
xiangfuwpwrak, yes. it grow from the top. I don't know how to make it from bottom.09:38
wpwrakfor the levels, i was thinking of showing line in and mic levels separately. so that one can also see how they relate to each other. that may need extra processing.09:39
xiangfuthe MTK only have two option.: 'horizontal' or 'vertical'09:39
wpwrakbut let's see how a combined meter does. maybe it's already good enough09:40
xiangfuwpwrak, ' line in and mic levels separately' have no idea how to do this. but when I mute the Line-in. it all about MIC. when I mute Mic it's all about Line-in. but don't know how to separate them :(09:41
xiangfuwpwrak, you mean we have 9 bars. 3 for Line-in, 3 for Mic. 3 for combined audio.09:41
xiangfu?09:42
wpwrakah, i was thinking of just two, one for mic and one for line. that would be a more traditional arrangement09:42
wpwrakbut perhaps what you've done works well in practice, too. it would have the advantage of using things we already have, so we don't need to make the poor little lm32 work too hard :)09:43
wpwrak/opt/milkymist/mtk.git/doc/mtk-users-guide.txt line 518 ;-)09:45
xiangfuread three more lines is always good. :) I only read the 517 line. :(09:48
xiangfuwill try that.09:48
xiangfuthanks wpwrak09:48
wpwrakand i think it would look better if the indicators were on the right side not on the left side09:49
wpwrakthat's for the usual left-to-right reading direction: first we set the input levels (left), then we show the resulting values (right)09:50
xiangfuI like the left. because the left part is READ-ONLY. :)09:53
wpwrakalso, do the ranges make sense ? i see that you have 0-300 and 0-600, but you scale *10009:53
xiangfuin the commit log there are some values of those three.09:53
wpwrakoh, i see. not just 0.0-1.009:54
xiangfuI just manually set them to 0-300, 0-600. I don't what is the MAX value of those three.09:54
xiangfuhave to offline. back in about ~2 hours.09:54
wpwraklet's see what the flickernoise manual says ...09:55
wpwrakhmm, 1.3 is already "load". so perhaps not more than 0.0-1.5 or 0.0-2.009:56
wpwrakmaybe we need to rescale things elsewhere too09:56
Action: xiangfu is back :)12:40
cladamwlekernel, wpwrak, wolfspraul , ( m1 h/w version naming ): M1v4.x ? or any idea ?13:09
wolfspraulah13:10
wolfspraulgood point13:10
wolfspraulplease make sure that the product name is always "Milkymist One"13:10
wolfspraulthat's important13:10
wolfspraulthen we have a datecode13:10
wolfspraulbut the problem is, if we only have product name + datecode, it's hard for internal communication13:10
cladamwAlthougt wiki on qi or milkymist wiki, they are something named rc2 already. but now it's time to change it.13:11
wolfspraullike we are working towards what now?13:11
wolfspraulI think we keep it simple, and just can say "rev 4" or "R4" instead of RC4 ?13:11
wolfspraulrc = release 'candidate'13:12
wolfspraulsome people may get hung up on that13:12
wolfspraulhow about just remove the 'C'13:12
wolfspraulR413:12
cladamwwith datacode is surely markable to trace even like while not final but still discussing or reviewing.13:12
wolfspraulwell datecode for sure, datecode is great13:12
wolfspraulyyyymmdd13:13
wolfspraulI suggest "R4" instead of RC413:13
lekernelthe FN "about" dialog already displays a "revision" which is the bare encoded value on the PCB, i.e. n-1 for RCn13:13
cladamwsome kinds of like xiangfu's release version with datacode13:13
wolfspraulwe need a name for releases we are working towards, to facilitate internal communication13:13
lekernelso it starts at 0 for our first 6 prototype series13:13
wolfspraulso we can talk about rc2, rc3, and we know what we mean13:13
lekernelwe can call RC4 R313:14
wolfspraulif the 'candidate' thing makes some people unsafe, let's remove the C13:14
wolfspraulno '3' please13:14
wolfspraulR413:14
lekernelby the way - is there support for ethernet in the linux port?13:16
cladamwcurrent "about" is marked => Board: M1 (PCB rev. 2) for original rc313:16
cladamwthat's being the final released version, for intenal reviewing control we can add like wolfspraul said that usually we want to name schematic communications.13:18
xiangfualso I think we should remove PCB.13:19
cladamwso M1 (PCB rev. 3.20120113) ? for internal discussion for better?13:19
xiangfuBoard: M1 R4  <=== is better.13:19
xiangfuwhat is 'PCB' here meaning. (sorry. )13:20
lekerneldoes this work? https://github.com/milkymist/linux-milkymist/blob/master/drivers/net/milkymist_minimac.c13:20
lekernelcladamw: it's the raw encoded value from the PCB, straight from the resistors13:21
lekernelthere's no way to encode a date in thee13:21
cladamw"M1 R4" I would like it being the final name, but before we release it, how about during the reviewing sche period?13:21
wolfspraulsame, R413:21
cladamwlekernel, sure, can do like that.13:22
wolfspraulthe software should say "R4" as well in the GUI13:22
lekernelhmm there are still cache flushes which are unnecessary now13:23
cladamwfor example, i just put my current sch work under: http://downloads.qi-hardware.com/hardware/milkymist_one/sch/tmp/13:23
wolfspraulcladamw: no "PCB rev. 3.20120113"13:23
wolfspraulR413:23
cladamwI'd like to give it a new naming code. :-)13:23
wolfspraulthe purpose of these things is to facilitate internal communication13:23
wolfspraulfor focus13:23
lekernelwolfspraul: ok, let's do like that13:24
wolfspraulyes, from now on we remove the 'C'13:24
wolfspraulthat's all13:24
wolfsprauland the software needs to write it like that in the GUI too13:24
wolfspraulmaybe a +1 is needed somewhere, that should be possible :-)13:24
cladamwso if final then we call "M1 R4", right?13:24
wolfspraulwe are working on R4 now, yes13:24
wolfsprauland when you make the pcb, it will also get a datecode, but that's only on the silkscreen13:24
wolfspraulmeanwhile the GUI will say "Milkymist One (R4)" or any other pretty printing like "revision 4" or "hardware version 4" or whatever13:25
wolfspraulbut that's just the GUI13:25
cladamwso the "about" will be "PCB rev.3" or just "M1 R4"?13:25
wolfspraulfor internal communication like here it's R413:25
cladamwmmm13:25
wolfspraulthe about should be easily understood, so it should definitely have a '4'13:26
wolfspraulother than that it's about making people understand, not us, but users looking at this about dialog13:26
wolfspraulso I suggest "Milkymist One (R4)" or "Milkymist One (rev 4)" or "Milkymist One (revision 4)"13:26
wolfspraulprobably I would pick the first one since it's shortest13:27
cladamwalright, me => also pick first = "Milkymist One (R4)"13:28
cladamwany else vote?13:28
cladamwseems done, tks !13:29
xiangfuMilkymist One (R4) +1 :)13:30
wolfspraulthe "+1" is the vote, not a proposal for a string extension :-)13:30
wolfsprauljust saying13:30
wolfspraulR4+1=R5 ?13:30
wolfspraul:-)13:30
cladamw;-O13:31
wolfspraulunless there are complaints, I will start referring to earlier releases with just "r" as well13:34
wolfspraulso I will not say "rc2", I will say "R2"13:35
wolfspraulI think this should still be clear, and consistent with our naming going forward13:35
wolfspraulso the one we work on now is R413:35
wolfspraulwhatever we don't want to get into R4 is pushed out to R5. etc.13:35
wolfsprauljust R from now on...13:35
wolfspraulhey 'R' can also stand for 'run'13:35
wolfspraulrun, revision, release13:35
cladamwokay ;-)13:36
wolfspraulonto R4!!!13:36
wolfspraul:-)13:36
cladamwwpwrak, about latest emails on: "consider scrapping the AUX* and VIDEO* input paths. That's R213:37
cladamw  through R9, and C4 through C8. Maybe keep the DC block footprints13:37
cladamw  (C4 through C8), so that one could solder something external13:37
cladamw  (just in case)."13:37
wolfspraulah there's a collision. R4 could be mistaken for a resistor.13:38
wolfspraulbut I think from the context it will always be clear whether the resistor or the board revision is meant13:38
xiangfuwpwrak, another 3 patches send out.13:39
wolfspraulso I still go with R4 (for the board revision)13:39
cladamwfor schematic compiler problems, I'll still put another real "part" on that one terminal of those DC block footprints. so the part i currently used a "TPxx" instead.13:39
wolfspraulif anybody feels bad about that naming collision, the board revision could be BR13:40
wolfspraulbut I think that's ugly13:40
wolfspraulR413:40
xiangfulekernel, from Lars email: So VGA, Keyboard and network will work13:40
wolfspraulthe context will always clarify between resistor and board13:40
xiangfulekernel, but I am not sure the new minimac will. I think the kernel is old.13:41
cladamwwolfspraul, BR? "Board Revision"?13:41
lekernelR. period13:41
wolfspraulR.4 ?13:41
lekernelR413:41
wolfspraullekernel: you mean that?13:41
wolfspraulah :-)13:41
wolfspraulyes sure, R413:41
wolfspraulcontext makes it clear13:41
wolfspraullekernel: I had another question/idea that just popped to my mind13:41
wolfspraulmany notebooks still have VGA output13:42
wolfsprauli.e. analog video13:42
wolfspraulis there a way to feed a vga output into m1 ?13:42
wolfspraulmaybe with the help of some cable we could build?13:42
lekernelwith the new adv7181c maybe13:42
wolfspraulor is there no way to make the ADC convert that back?13:42
lekernelup to 1024x76813:42
lekernelthis will need some rework of the FPGA and software though13:42
wolfspraulit might add value to some people13:42
wolfspraulyeah no rush or anything, I just realized13:43
wolfspraulwith the adv7181b definitely not?13:43
lekerneldefinitely not13:43
wolfspraulok13:43
wolfspraulanother reason why to upgrade to C - great13:43
lekernellet me check for the -C ...13:43
wolfspraulthanks!13:43
lekernelyes, that would work (with the proper FPGA and software support, which isn't close to existing right now)13:44
lekerneleven 1080i13:44
wolfspraulwith what cable though?13:44
lekerneland 720p13:44
lekernelRGB VGA pins to RCA13:45
lekerneland sync-on-green13:45
wolfspraula standard cable that goes from 15-pin vga to 3*rca? or 1*rca ?13:45
lekernel3*RCA13:45
wolfsprauldo such cables exist? or we would need to build one?13:45
wolfspraulwell a lot of work in between, but good stuff13:46
wolfspraulfirst update to adv7181c - moving in R413:46
wolfspraulthen a lot of work in fpga and software13:46
wolfspraulthen a cable, easily sourcable or custom-made13:46
wolfspraulbut good to know that it's possible13:46
xiangfulekernel, I should require both RESOURCE_AUDIO, and RESOURCE_SAMPLER?13:47
lekernelxiangfu: exact same as the variable monitor, since you are spawning the same task13:50
lekernelmeh I doubt the minimac driver really works. if it does, it's with bugs.13:50
xiangfuok got it.13:50
lekernelseems it's still for the old core13:51
lekernelanyway - it's better than nothing, it's a decent starting point for a driver for the new core13:52
xiangfuok. now it's better.13:57
xiangfulekernel, I think it time to do a push. :)14:08
xiangfulekernel, the last patch I didn't add the resmgr_release under 'OK' event. already fixed in local.14:08
lekernelxiangfu: don't use the window title in resmgr_acquire, just one word14:14
lekernelxiangfu: otherwise be consistent with all other places where resmgr_acquire is called14:14
lekernelxiangfu: also, I think LoadDisplay can take floating point values directly (but check)14:15
lekernelother than these two points it's fine14:15
xiangfuabout LoadDisplay values: I would like keep them for now. I think when we dig into more about the values. I will create another patch. :)14:17
xiangfuI would like check all resmgr_acquire. make them all using windows titles. that make people easy understand.14:17
lekernelxiangfu: it only has to do with code simplicity. I think the LoadDisplay already has a "scaler" built-in, no need to duplicate it14:18
lekernelxiangfu: so you'd give it the range you expect for bass/mid/treb, and then feed the values to it directly14:19
xiangfujust checked. after the changed Variable monitor. all resmgr_acq is using windows titles. except 'rendering' mode.14:20
xiangfulet me check the LoadDisplay now.14:21
xiangfuI think the 'LoadDisplay' is better then 'Scale' :)14:27
xiangfusorry. just ignore my last message.14:28
xiangfutest the float value now...14:28
xiangfuyes. float works fine.14:32
wpwrak(R4) can we change R3 in R4 to 10 kOhm ? ;-)14:32
wpwrakbut i guess we could just call it "r4". reduces the difference to rc4 by one bit :)14:33
wpwrak(resistor) ah, i see that you already noticed :)14:34
xiangfuwpwrak, I will do a git push. there are two small patches not send to mailing list. 1. using float value directly. 2. add the resmgr_release in both OK and Cancel event.14:35
xiangfuwpwrak, then goto sleep. :)14:35
xiangfuwpwrak, the screenshot is here: http://downloads.qi-hardware.com/people/xiangfu/tmp/Audio-indicate-Screenshot-Right-01.png14:36
lekernelxiangfu: ok14:39
wpwraklooks nice !14:39
GitHub121[flickernoise] xiangfu pushed 8 new commits to master: http://git.io/qkmxYg14:40
GitHub121[flickernoise/master] gui/audio add Bass Mid Treb indication bar - Xiangfu Liu14:40
GitHub121[flickernoise/master] gui/audio: connect the indicate bar to bass,mid,treb - Xiangfu Liu14:40
GitHub121[flickernoise/master] gui/audio: make the bar from bottom to top - Xiangfu Liu14:40
wpwraklekernel: do the ranges of the values look weird to you, too ? or is this a documentation problem ?14:40
lekernelboth I think :-)14:43
lekernelwhat it does is FIR-filter the incoming audio into three bands, then take the sum of squares of each output14:45
lekerneltaking the square root of that would probably help14:46
wpwrakwhat does milkdrop do ? and do they document their real range ?14:46
lekernelsomething similar except it does a more expensive Fourier transform14:47
wpwraksqrt would help against outliers, yup14:47
wpwrakhmm, fft is a bit suckish indeed14:47
wpwrakunless there's some fun way to do this in the fpga :)14:47
lekernelbtw the main reason why we do not have sqrt atm is that we didn't have libm or anything when I first wrote that code (in late 2008/early 2009)14:48
wpwrakheh :)14:48
lekernelwpwrak: yeah, but 90% of geeks confuse "signal processing" and "FFT" :-)14:49
Scopeukconfuse or get the sub set wrong?14:50
lekernel"my winmodem does a FFT" etc.14:50
wpwrakthat's probably the same people who'd say the Windows(tm) on your M1 looks funny :)14:53
wolfspraulwpwrak: in those cases you would want to say "can we change R3 in rev4 to xxx"14:57
wolfspraulbut in 99% of cases R4 should be clear from context14:58
wolfspraulI looked at some usb board connectors I found here, they are all 5-pin, 100 mil headers14:58
wpwrakthe most common is 5x2, for two ports. two pins have no signal and one of them is removed, for keying15:01
wolfspraulone is D-, one is D+, one GND, one power15:02
wpwrakyup. just 1:1 full-size usb15:03
wolfspraulin my cables I think the gnd is going to 2 pins15:04
wolfspraulthe colors on the board side are red-white-green-black-black15:05
wolfspraulso we want a 4*1 or 5*1 header?15:05
wpwrak5x2, keyed ? :)15:05
wolfspraulseems 5*1 (or rather 5*2)15:05
wpwraka 4x1 plug or a 5x1 plug should fit on 5x2 keyed as well.15:06
wolfspraulyes15:06
wolfspraulI just want to make sure we don't screw this up, it's easier than anyone would think15:07
wolfspraulseems 5*2 header, exact pin assignment should be standardized then (to be confirmed)15:07
wpwraklike this: http://search.digikey.com/us/en/products/TMM-105-01-SM-D-SM-009-P-TR/TMM-105-01-SM-D-SM-009-P-TR-ND/268539315:08
wolfspraul2mm?15:10
wpwrakor maybe better this: http://search.digikey.com/us/en/products/TSW-105-26-H-D-009/TSW-105-26-H-D-009-ND/268593015:10
wolfspraulthe first one is not 100mil I believe15:10
wolfspraulI'm not so worried about the mechanical header, we'll find a cheap one (the second one is still 2.50 USD, sounds crazy expensive to me)15:11
wpwrakoh, right. sorry about that. it's also a bit short, 3 mm15:11
wolfspraulmore worried about pin assignment15:11
wpwrakthe 2nd one looks right15:11
wolfspraulexcept for the price15:11
wolfspraulchina price (and china quality) this would be what? 5 cents? :-)15:11
wolfspraulthen for taiwan/japan quality let's say 50 cents - enough15:11
wpwrakand it's non-stock at digi-key15:12
wpwrakmaybe they have the same in a different category15:12
wolfsprauldon't worry about that15:12
wolfspraulthat part is just some leg work15:12
wolfspraulI'm worried about pin assignment15:12
wpwrakthe usb stuff often shows up in 4-5 categories15:12
wolfspraulwe may not find out until it's way too late...15:13
wolfspraulthen it's the big "oooh" moment15:13
wpwrakwe can check against the cables we have.15:19
wpwrakmine has VBUS on the unkeyed side, then D-, D+, GND (connector order), and the NC/unloaded position15:23
wpwrakthat's for some PC mainboard15:24
wpwrakthe unloaded position is pin 9, not pin 1 as one may expect15:26
GitHub150[milkymist-ng] sbourdeauducq pushed 1 new commit to master: http://git.io/HNCxSA16:13
GitHub150[milkymist-ng/master] convtools -> tools - Sebastien Bourdeauducq16:13
lekernelwolfspraul: btw, since you asked about migen tasks, there is rewriting existing cores in FHDL or encapsulating them16:32
GitHub150[milkymist-ng] sbourdeauducq pushed 1 new commit to master: http://git.io/wZ-wHg16:33
GitHub150[milkymist-ng/master] Wishbone: omit fixed LSBs - Sebastien Bourdeauducq16:33
GitHub155[migen] sbourdeauducq pushed 1 new commit to master: https://github.com/milkymist/migen/commit/20425703fa67239a8ed180baa6a5eb80ba42e4df16:33
GitHub155[migen/master] Wishbone: omit fixed LSBs - Sebastien Bourdeauducq16:33
wolfspraullekernel: yes that's a good one, thanks! [migen conversion task]16:48
wolfspraulI don't expect to find anybody, and I don't expect to be up to the task myself, but it gives us a good story to go around with, something valuable and worthwhile we can invite people to wholeheartedly16:48
lekernelhttp://www.ddr-phy.org/17:38
lekernel"NOTWITHSTANDING ANYTHING ELSE WILL CADENCE'S TOTAL AGGREGATE LIABILITY FOR ANY CLAIM, SUIT, PROCEEDING OR OTHERWISE, RELATING IN ANYWAY TO THE DFI SPECIFICATION EXCEED $1.00USD"17:39
lekernellol17:39
wpwrakthere'll probably be a multi-billion lawsuit over that clause alone :)17:43
lekernelhttp://www.ddr-phy.org/page/2351641:Page:30117:44
lekernelfor once, there doesn't seem to be anything evil in that, is there?17:44
wpwrakclause 6 seems a little odd17:50
wpwrakclause 3 also leaves it open whether there are any patent traps. e.g., the whole thing would be compatible with RAND/FRAND patents17:51
wpwrak(held by cadence & associates)17:51
wpwrakclause 6 could perhaps be construed as including the licensee's own implementations17:52
lekernel"and provided by Licensee for the purpose of further developing, modifying or changing the DFI Specification."17:54
wpwrakno, that one has "as a result of the use of the Communications in developing ..."17:55
wpwrakwhatever that means17:55
wpwraki.e., if you implement the DFI spec, do you "develop" it ?17:56
wpwrakbut it's probably not an issue. you could only be hassled over that by your contract partner, which seems to be either just candence or maybe the whole gang of participants. but not just any random rogue troll17:58
wpwrakand since there are no disreputable companies among the participants, e.g., apple, kodak, and such, it's probably okay17:59
wpwrakthat clause 3 leaves the door open for (F)RAND is more problematic. at least an assurance that the participants hold no patents affecting DFI (unlikely) or o do not intend to enforce those they hold against licensees would be welcome18:01
wpwrakthis is a big problem with standards and organizations that accept flat out patented stuff without any waivers18:02
wpwrake.g., don't touch anything from ETSI that's not at least ~18 years old. like good wine, technology gets more palatable with age :)18:03
lekernelThe DFI specification supports a 1:2 or 1:4 MC to PHY frequency ratio,18:07
lekernelmeh18:07
lekernelso if you have DDR3 at 800MHz (1600Mbps/pin) you must run the memory controller at 200MHz at least18:08
lekernelthat's not easy in those slow and crappy FPGAs18:08
wpwrakif it was easy, where would be the fun ? :)18:08
lekernelwell... in fact it seems straightforward to "extend" DFI to support abitrary clock multipliers anyway18:13
lekernel"for frequency ratio systems, the control signal interface, the write data interface and the read data enable signal will all be suffixed with a _pN where N is the18:15
lekernelphase number."18:15
lekernelphew... lack of structured signals in *HDL kicks again18:15
wpwrak;-)18:16
wpwrakbe happy that you can name them :) "connect signal 124 to 743" :)18:16
lekerneli'll rather write a Migen DFI wrapper that takes care of this stuff automatically :)18:17
lekernelthen I have a Python object for each phase number18:17
wolfspraulthe specification from ddr-phy.org is free and open for Milkymist to use?18:18
lekernelseems ok... http://www.ddr-phy.org/page/2351641:Page:30118:19
lekernelthere's even a right to "distribute and copy" the document18:20
wolfspraulnice18:24
lekernelmeh, seems this Spartan6 DDR PHY allows only one DRAM command per DFI clock (even though it multiplies the clock by 2)19:04
lekerneland it definitely makes sense to send e.g. a read and a precharge command to different DRAM banks in the same DFI cycle19:05
lekernelWTF? this thing appears to send all commands to the DRAM *twice*19:17
lekernelha! no, they inhibit the chip select signal on every second cycle19:20
lekernelthat's lousy19:20
kristianpaulworkaround? :)19:34
lekernel_ask kristianpaul to write a better PHY20:09
kristianpaullol, i dint complain about it ;)20:09
frdmincIt turns out that the (non-bluetooth, propriatary USB dongle) keyboard I got to control the M1 remotely doesn't work with the M1. Has anyone gotten a different kind of remote keyboard/mouse working - eg something like x2x or x2vnc?21:34
wolfspraulfrdminc: hmm, sorry to hear that. what model did you get?21:35
wolfspraulWerner is using a Rii Remote (same tech as yours) successfully, and we are even thinking about including such a type of keyboard/pad in the box in the future.21:36
wolfspraulwpwrak: do you think you could have fixed keyboard related bugs after our last 2011-11-29 release?21:37
frdminclogitech k270 kb and anywhere mouse. They use the same USB dongle. On Linux it shows up as a USB HID; I'll get the exact messages if you'd like. I'm trying an older wireless kb/dongle atrm.21:37
wolfspraulyes please do so21:37
wolfspraulin the past finding a keyboard that would work was a hit and miss game, but slowly we are fixing bug by bug21:38
wpwrakrii should be okay around that time. there may be more fixes after nov 29.21:38
wpwrakbut mose likely it's a HID report format problem21:38
wolfspraulcan he upload something that would help us fix it?21:39
wpwrakjust a format we don't recognize with the overly primitive parser we have now21:39
wolfspraulfrdminc: here's the current 'support map' http://en.qi-hardware.com/wiki/Milkymist_One_accessories#keyboard21:39
wolfspraulif you have a minute you can edit the wiki and add yours as not working21:39
wpwrakfor npw, i'd just wait :) the fix for the whole class of problems is known. just needs implementing.21:39
wolfspraulyeah but frdminc has a party tonight21:39
wpwrakthen i'd look for a keybaord/mouse that works with what he has21:40
wpwrakthere's no telling what may break in a last second hotfix21:40
wolfspraulyes, no update now21:40
wolfspraulhow can one dump the report format?21:42
wpwrakand the proper fix is about half a week of work. sometimes, you can hack something quickly, but it's risky since it may break other formats21:42
wpwrakbah, nobody reads my postings :)21:43
wpwraksearching ...21:43
wpwrakhttp://lists.en.qi-hardware.com/pipermail/discussion/2011-November/009206.html21:44
frdmincwolfspraul: BTW Star Simpson is here, we were talking about M1 and of course it turns out she's met all you guys in CN :-)21:46
wolfspraulah great, say hello! (although I didn't meet her in person)21:46
frdmincI'm just getting one of those USB Ethernet externder things for the future.21:46
frdmincAh yeah I think she mentioned rejon, thought she mentioned you but guess not.21:47
frdmincSweet the Adesso Inc WKB-4000US wireless keyboard does work... can just return all the logitech stuff :)21:48
wolfspraulthey should all work, maybe in a few months :-)21:48
wolfspraulif you have a little time, please update the wiki21:48
wolfspraulif not, also ok21:48
wolfspraulwe are on it21:48
frdminchmm may have spoke too soon... stopped working.21:49
frdmincII'm planning on spending quality time with the wiki after the party21:49
frdmincIf you un/replug it starts working again... looks like maybe the sleep state is "confusing" the M1.21:51
wolfspraulpossible :-)21:51
wolfspraulwherever you look - bugs and missing features21:51
wolfspraula developer's dream21:51
wolfspraul:-)21:51
frdminc:-)21:51
wolfspraul"if you don't find a bug or problem on your first day of using the product, you have a special right to return for a full refund and free beer..."21:52
wpwraksome keyboards go to suspend if you don't use them for a short while21:54
wpwraktypically there's a button that wakes them up. e.g., on the rii it's shift21:55
wpwrakwe don't implement usb suspend but ... that shouldn't bother the device21:55
wpwrakcould be that something else trips it, though21:55
wolfspraulif frdminc really sees a sleep-related bug, that's the first time I hear about it wrt keyboards and quite important I think21:55
wolfspraulhoepfully after your party we have some time to get dumps or logs from those devices21:56
wolfspraulcan you find some good developers who might be interested in joining Milkymist? even explaining and talking about the project may attract some...21:57
wpwraki hope to kill a lot of those remaining usb problems in the next 1-2 weeks. we still have a few unsolved issues. e.g., 1) only a few hardcoded report formats; 2) no CRC; 3) for some reason, lots of RX timeouts (which may be false alarms, but then we should at least suppress the alarm)21:58
wolfsprauldozens or more people could easily hack on m1 without any fear of overlap or redundancy :-)21:58
wpwrakso it's fragile. but works surprisingly well ;-))21:58
frdminchmmm keyboard has been working fine for a while now, and itself has issues, so perhaps if treated delicatly / not moved it'll work for the party.21:59
wolfspraulgood luck!21:59
wpwrakmaybe try to find a 2nd one and bring it too, in case the other quicks22:01
wpwraks/quicks/quits/22:02
frdmincwpwrak: oh I'm doing a permanent installation in a space; so plenty of wired KBs avail if needed :)22:05
Action: frdminc will take pics when done22:05
lekernel_isn't USB suspend only activated at the host's request?22:09
lekernel_did they ALSO felt the need, that suspend mode can be activated by the device?22:10
lekernel_is Rube Goldberg the chair of the USB design committee?22:10
wpwraki've only heard of host-activated22:11
lekernel_we should have used PS/2 instead. much much easier to get to work with all devices :p22:11
wpwrakand no, i don't think it was mr. goldberg. i think usb was kafka's project for a phd in engineering :)22:12
wpwrakps/2. bah ! use the full-sized DIN connector of the original IBM PC:)22:13
wpwraka "sleep issue" may also come from the usb device doing its own suspending. e.g., the rii suspends after only a few seconds. it'll wake up if you press shift or power-cycle it (it has a little switch on the side to cut battery power)22:14
wpwrakif you don't realize what's happening, you may think it is malfunctioning22:15
lekernel_too bad you have to pay for http://www.rubegoldberg.com/?page=contest_form3, otherwise I'd have sent them a copy of the USB spec :-)22:15
wpwraki guess each generation must have its algol 68 :)22:20
wpwrakgeeks breed slowly, so the time between generations (algol 68, 1968 vs usb, 1995) is about right. expect the next complexity explosion around 2020-202522:23
wpwraklekernel_: i think you should be around the right age for chairing a design standardization committee around that time :)22:24
wolfspraulhmm, my RC2 froze again in rendering, I think I can reproduce it after about a few hours at most right now23:06
wolfspraulsince we never specifically tracked down the root cause of the freezes, I am wondering whether we should we worried going forward that they might reappear?23:07
wolfspraulI don't remember any report of RC3 freezing, but I do feel a little uneasy about it :-)23:07
wolfspraulto worry or not to worry?23:07
wpwrakif it's just a lone rc2, perhaps not :)23:12
wpwrakmy rc3 has been rendering pretty much endlessly without freeze23:12
lekernel_same here (both R2/R3 ...)23:12
wpwrakof course, there could be device variations or the different power supply could have an effect23:13
wolfspraulit's definitely not a lone rc2, I remember several reports of rc2 freezing23:13
wpwrakheh :)23:13
wolfspraulI'm not overly worried right now, just as we dig in and fix and polish everywhere, the freezes I am seeing are a constant reminder for me that that issue seemingly went away in RC3 without any specific action against it taken23:13
wpwrakwell, you can go though the changes from rc2 to rc3, see if anytbing looks suspicious ...23:14
wolfspraulyes exactly, I don't think so :-)23:14
lekernel_wolfspraul: are the ferrite beads removed?23:14
wolfspraulI mean there was a lot, could be anything23:14
wolfspraullekernel_: no idea23:14
wpwrakwhich software version do you have now ?23:14
lekernel_wolfspraul: well, look at your board23:14
wolfspraulI could send it back to adam for checks & service ;-)23:14
wpwrakhehe, a lot + no idea sounds like a good start for debugging ;-)))23:14
wolfspraulwait need to open the case...23:14
lekernel_he, you can tell if a ferrite bead is there or not, can't you?23:15
lekernel_no need for Adam's intervention I hope23:15
kristianpaulmt rc2 never froze rendering23:15
kristianpauls/mt/my23:15
wolfspraulI cannot - where do I need to look?23:15
frdmincsweet, everything is working except I underestimated how hard the UI would be to use in 640x480 in a massively non-right-angled  trapizoid...23:15
wolfspraulI take a pic :-)23:16
lekernel_frdminc: as far as the UI is concerned you can switch to 1024x76823:16
lekernel_wolfspraul: L19 close to the video input and L3 close to J323:16
lekernel_wolfspraul: you should have molten solder blobs in both places, not actual components23:17
frdminclekernel_: it's running to a piece-of-shit unfocusable projector via a vga to s-video converter that needs 640x480. VNC would be really useful right now. The option now is, what, send OSC commands?23:17
kristianpaulyeah i had those L* fixes23:17
wpwrakor maybe actual components drowned in solder23:17
frdminc(option for remote control via ethernet)23:17
wolfspraulhmm. I could have sworn I applied L3/L19 at some point, but L19 is still there :-)23:18
wolfspraulthen probably L3 as well, can't find it right now, one sec23:18
wpwrakfrdminc: how many hours do you have left ? :)23:18
wpwrakfrdminc: there's also the IR remote control. not sure what it can do, though. (haven't tried mine yet)23:19
lekernel_wolfspraul: problems related to those beads did crash my M123:19
wolfspraulah here, for dummies http://en.qi-hardware.com/wiki/File:M1_PCB_L19_L3.png23:19
wpwraklekernel_: did it also crash when just running ? or only when connecting things ?23:20
wolfspraulL3 also still there, yes23:20
frdmincwpwrak: party o'clock is 10pm (+4h). For now I'm going to grab a IPKVM switch I have hanging around and set up a performance / IR / etc. Later it looks like I can leave the IPKVM there and get a signal splitter for like $25.23:20
wolfspraulyes, I am very careful with the connectors, but not sure whether L3/L19 can also freeze the board without being touched (which is my case)23:20
wpwrakwolfspraul: (L3, L19) you and jon may want to visit adam :) (or has jon his fixed now ?)23:20
wolfspraulI don't know, I try to get old boards fixed whenever possible23:21
lekernel_wolfspraul: time to use your soldering iron? :)23:21
wolfspraulbut it's a lot of work, I guess I never cared for my own board enough!!! ;-)23:21
kristianpaulfor not fixing L* seems not ;)23:21
wolfspraulyes I should try, that's really easy23:22
wpwrakfrdminc: btw, why do you have to go to the GUI during performance ? the idea is generally that you shouldn't need that23:22
wolfspraullekernel_: do you think L3/L19 can also be responsible for freezes when m1 is untouched, just running?23:22
frdmincSweet looks like it's a $5 part .. http://www.amazon.com/Premium-GOLD-Splitter-Cable-PTC/dp/B001EQ363M/ref=sr_1_24?s=electronics&ie=UTF8&qid=1326496888&sr=1-2423:22
lekernel_L19 may give some trouble if you have a weak soldering iron, because the ground plane sucks the heat23:22
lekernel_maybe, let's find out23:23
wpwrakfrdminc: cute :)23:23
frdmincwpwrak: yeah I get that, but for this one I was going to be learning, and anyway I'll want to interact with the M1 in-place without having to move it all the time, and for that it needs to be usable in its current location of about 10'; off the ground and 15' from any wall.23:23
lekernel_frdminc: you want to use the GUI live?23:24
Action: frdminc contemplates going to a local store to get that cable...23:24
frdminclekernel_: yeah, that's the general idea, at least until I get more of a clue. And also be able to use the GUI without moving the M1 from its installed location all the time during non-performance.23:24
wpwrakfrdminc: (interact) hmm, tricky location indeed :)23:24
wpwrakwolfspraul: a case for RF keyboards :)23:25
lekernel_ah, that's also why you want wireless keyboards and such23:25
wpwrakor a rather long USB cable rather stylishly hanging from the ceiling into the middle of the dance floor :)23:25
wolfspraulI'm looking into them now, if I can find something cheap and well working, maybe we kick the silicone keyboard and remote control out in favor of one of them23:26
wolfspraulbut one by oen23:26
wpwraki wonder when we'll have the first case of wireless M1-jacking :)23:26
lekernel_I wonder if long USB cables might unearth some more timing related bugs in the softusb design23:26
wpwraklekernel_: just wait for hubs :)23:26
roh*eg*23:27
wpwrakin any case, there are still known issues with low-level timing. at least the stack complains bitterly and quite a lot. just happens to work. need to analyze that stuff ...23:27
Action: frdminc notes his wireless keyboard actually seems to be working fine atm23:28
wolfspraulfrdminc: which one is it?23:28
wpwraklekernel_: you may appreciate that USB close simplicity in the case of hubs and, instead of requiring complicate store-and-forward, conveniently extend the already extremely tight timing just across the whole hub chain23:30
wpwraks/close/chose/23:30
frdmincwolfspraul: the adesso one I mentioned earlier.23:31
wolfspraulhow can they do this 300 megabytes / sec superspeed stuff now?23:31
wolfspraulI wish I had more patches that are not highlighting the fact that there is a rectangular boundary around the whole thing23:32
frdmincre: case for RF keyboards, IMHO doing it in software would be preferable (eg a root-window vnc server or using something like pyjamas so the UI is available in exactly the same way locally with normal gtaphical toolkit and via the web with javascript)23:33
wolfspraulso basically no solid colors along the sides23:33
wpwrakprobably a somewhat different L2 protocol. already high-speed changes a few things, adds new packet types and so on23:33
wpwraknot sure if VNC wouldn't be quite a lot of work ..23:34
wpwrakbut yes, it would be handy. very.23:34
lekernel_wolfspraul: http://lists.milkymist.org/pipermail/devel-milkymist.org/2011-December/002492.html #3 ?23:34
lekernel_wpwrak: as long as you don't want it to work when rendering, it should be OK23:35
wpwraki think he just wants no explicitly drawn border around the patch23:35
wpwrak(vnc) is the protocol "friendly" ? never looked at it23:36
lekernel_alpha-mask the final result (not iteratively) then?23:37
wpwrakeasier: don't draw a border :)23:39
wpwrake.g., "tornado" would be "borderless", while the drunken boat wouldn't be23:40
lekernelwpwrak: just use some lib23:41
wpwrakof course, an output alpha mask could be cool, too23:41
lekernelhttp://libvncserver.sourceforge.net/23:41
lekernelthere's also the VNC server in QEMU that we can copy code from23:41
wpwrakparticularly if the mask could change dynamically :)23:41
lekernelwpwrak: what would be the needed feature set for the next TMU?23:42
lekernelsupport for RGB 10:10:10 and ARGB 8:8:823:43
lekernel:823:43
lekernelwhat primitive polygon? the current rectangles are convenient because they're easy to do without bugs :-)23:43
lekernelbut a bit limiting maybe23:44
wpwrakerr, what polygon ? for the overall shape ? or for the vertices23:44
lekernelatm the only supported configuration is a tesselation of straight rectangles23:45
lekernelbut if we switch to e.g. arbitrary triangles then we can draw any textured 2D polygon23:46
lekernelnow - we have to draw each of these triangles correctly in all configurations including messy corner cases, and without visible "seams" at common edges of triangles when alpha blending is enabled23:46
lekernelthis isn't easy23:46
wpwrakhmm, dunno how that would change things. i don't understand the logic of effect generation that well yet23:47
lekernelit doesn't change anything for the current process (that's why I took the rectangle shortcut)23:47
lekernelbut if you want fancy things like dynamic alpha mask23:47
lekernelyou could do it with a translucent 2D polygon overlaid on the picture23:48
wpwrakapropos easy. when you run the 8 bit starfield, do you see an array of red pixels ? if yes, are they artefacts or do they seem intentional ?23:48
lekernelhuh?23:48
lekernelwhere?23:48
wpwrakover the entire frame23:48
lekernelthose are white, no?23:49
wpwrakah sorry, it's the inner workings of my new computer23:50
wpwrakthere's a static array of red pixels23:50
wpwrak16x1623:50
wpwrakbut i'm not sure if this it the patch's doing or something going wrong in the lower layers23:51
lekernelcould be a bug... the black pattern on the bars of "slow shift matrix" shouldn't be there, and might be related23:58
--- Sat Jan 14 201200:00

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