#milkymist IRC log for Thursday, 2011-11-17

wolfspraulwpwrak: aha, looks like full-speed USB is (almost) solved?00:46
wpwrakyeah, i think we're on a good track00:46
wpwrakhow do you like the PR text ? (i just sent a patch)00:49
wolfspraulwpwrak: hmm. :-) I don't think it will achieve anything.01:09
wolfspraula better text will not get it published01:10
wolfspraulbut lemme read it thoroughly, now that it's written :-)01:10
wpwrakit's not so much about a better text but about having a bit of a narrative, and covering the essentials that make this project intersting01:16
kristianpaulhey, indeed a HHKB can save lot of space.. actually where is always hard have the M1 + logic analizer + +... next to you03:00
kristianpaullars_: you manage to run that make timing ?04:21
kristianpaulinteresting http://www.myhdl.org/doku.php/users:cfelton:projects:zpu?s[]=zpu04:26
kristianpaulokay, running 'trce -v 12 -fastpaths -o design_timing_report system.ncd system.pcf' inside boards/milkymist-one/synthesis/build seems generate the .twr file04:54
kristianpaularggh, more non-free software we use :/04:55
kristianpauloh, http://pastebin.com/4Hup1djM slowest paths points mostly  ddram/hpdmc04:57
kristianpaulnotice this build was with ALL no esencial cores disabled04:58
kristianpaulhe and same time fastest past is sdram. :_S04:58
kristianpaulbetter ignore this :)04:58
wolfspraul [05:27] [wolfspraul(+i)] [5:freenode/#milkymist [Act: 3,4,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,33,34,35,36,37,38,39,40,41,42,43]05:27
wolfsprauloops :-) sorry my notebook is breaking apart... speaker crackling, touchpad going crazy and mouse wildly selecting and clicking stuff, etc.05:28
kristianpaulwildly :-)05:36
stekernyou have a lot of windows open in irssi at least ;)05:38
kristianpaulokay all settle for 65535000 Hz clock :-), including that libbase/board.c i missed before05:40
Action: kristianpaul feels bad when writing CLOCK_DEDICATED_ROUTE = FALSE; :-|05:43
kristianpaulbut it should work anyway05:44
kristianpaulgood now mm soc seems to work with gps receiver cloc :)06:32
kristianpaulindeed, using clk from memcard06:32
kristianpaulokay tomorrow try namuru soc and custom bios for it and see if the memory test pass ;-)06:33
kristianpauls/soc/core ergg not yet06:34
kristianpaulhum mr 0xa0000000 system got freeze.. oh well, more todo :)06:42
aw_lekernel, do you know what's the current choice of video decoder for rc4? still 7181B or another?08:05
wpwrakhmm. i've started to try to read this in "leet" :) "libic" ... ?08:26
wolfspraulaw_: definitely switch to 7181C, if you haven't sourced them yet please add to the list. should be a drop-in replacement plus removing the external antialias filter (capacitors and ferrite beads near the video connectors)08:27
aw_wolfspraul, okay08:28
wolfspraulI am a bit confused as to the final reset circuit proposal now. I think it's the 4.0v reset ic plus new USB power switch.08:29
wolfspraulthat definitely needs a thorough design verification08:30
wpwrakagreed :)08:30
wpwrakstill needs lekernel's consent08:31
wolfspraulbut it seems the 3-input reset ic idea is off the table right now?08:37
xiangfuwpwrak, can you post your latest flickernoise.fbi somewhere. I just want test my 2 itf keyboard and full-speed mouse.08:38
wpwrakunless someone wants to bring it back. the single-input seems sufficient, once we protect the 5 V rail.08:39
wolfspraulwpwrak: can we already safely assume we will not find USB-related electrical problems and improvements?08:39
wpwrakwolfspraul: no, not yet08:39
wpwrakxiangfu: this is for 2 itf (fs still broken): http://downloads.qi-hardware.com/people/werner/tmp/flickernoise.fbi.bz208:40
wpwrakxiangfu: i'll clean up the one with full-speed in a bit08:40
aw_hmm..ADV7181CBSTZ is only available One piece in Digikey. Else vendor this time I need to try. like Verical, Avnet Express or Arrow.08:47
wolfspraultry octopart, you can get a quick overview who else might have it08:50
aw_yes, tried octopart already and listed above i can source. ;-)08:51
wolfspraullooks available here and there08:51
wolfspraularrow maybe?08:51
xiangfuwpwrak, (fs still broken) can't boot to GUI. stop at "I: Booting..."08:52
aw_I can combine them to my cart, but not now. since my mouser cart had have some parts to be ordered. now I need to decide transfer those parts to that vendors who also have 7181C.08:53
aw_no worries, but I just don't know the new USB power switch that if we need to prepare/source few pcs now? or still needs lekernel's consent?08:54
aw_wolfspraul wpwrak so i think that I can stay for a while to place orders until USB-related problem becomes more clear then.09:05
wpwraklemme see about that switch ...09:09
wpwrakah wait. it's not thw switch but the codec. good :)09:10
wolfspraulaw_: you can get 5 of that USB power switch now09:10
wolfspraulit only costs a few USD, shipping costs will be far higher if that's the one thing missing one day09:10
wolfspraula few USD for all 509:10
wolfspraulso even if it's only under consideration now, let's say likelihood to use 50%, it's already best to just include at the next good opportunity, just 3-5 or so09:11
wolfspraulthe same may be true for the proposed 3-state power monitor ic, if it's easy to buy just include a few09:11
wolfsprauleven if the likelihood of using that one in rc4 only looks like 10% to me now09:12
wolfspraulcan you confirm the exact part numbers for those 2?09:12
wpwrak(3-way power monitor) we'd have to choose which one. depends on the tolerances of the various voltage rails, etc.09:12
GitHub119[scripts] xiangfu pushed 1 new commit to master: http://git.io/ElwhoA09:13
GitHub119[scripts/master] compile flickernoise: make clean before compile test image - Xiangfu Liu09:13
aw_hmm...so seems that we can include few parts both 3-state power monitor and usb-related09:13
wpwrakaw_: you already have the 4.0 V version of the single input reset chip ?09:15
wolfspraulok, sounds like leave the 3-state ic out now, too many uncertainties09:16
wolfspraulbut the usb power switch is easier, and more likely to be used09:16
aw_wpwrak, in my Mouser cart now. ;-) haven't clicked 'purchase' btn. ;-)09:17
aw_wpwrak, do you think that what type of EN <enable pin> acted as HIGH or LOW? for http://www.diodes.com/datasheets/AP2142A_52A.pdf09:21
aw_wpwrak, or still too far away to know it now. ;-)09:22
wpwrakwhat state are the FPGA's pins in after reset ? Z ? pull-up ?09:23
aw_hmm..good question. don't know after reset.09:24
wpwrakif it's pull-up, pick active-low. if it's Z, toss a coin and add a pull up/down :)09:27
wpwrakgrmbl. and now xiangfu is hiding again :-(09:33
wpwrakanyway, here's a binary for double itf and fs: http://downloads.qi-hardware.com/people/werner/tmp/flickernoise-fs.fbi.bz209:33
wpwrakfull-speed is very shaky. loses characters, can get garbage, can get endless auto-repeat. but if you type very slowly, it almost works ;-)09:34
wpwrakof course, all these problems are what a stack that now has zero error detection richly deserves :)09:35
lekernelyes, let's try the USB power switch09:46
lekernelbut can it just power the ports all the time?09:46
wpwrakyou mean you don't want to be able to switch them off ?09:47
lekernelwhy bother?09:47
wpwraki'd keep them off when starting. make sure we don't have unnecessary drain09:47
wpwrakthen flip them on (and keep them on)09:48
lekernelwe can't detect the pull up resistor if we don't have VBUS09:48
wpwraksure. first thing you'd do once you get out of standby would be to turn VUSB on09:48
lekernelwhy turn it after standby?09:49
wpwrakalso, the ability to power-cycle troublesome devices may be handy at times09:49
lekerneljust unplug/replug09:50
wpwrak(stby) well, or when entering the stby loop. though i see littler point in powering USB at this time. but okay, there are usb-powered gadgets you may want to use. LED fans and such ;-)09:50
wpwrakby software :)09:51
lekernelno, that's additional complications09:51
wpwrakc'mon :)09:51
lekernelmany laptops don't even have this capability09:51
wpwrakbut the mighty M1 will have it ;-)09:52
lekerneland some also immediately turn off if you short VBUS, btw :)09:52
wpwrakyeah. there are plenty of examples out there one better doesn't copy from ;-))09:52
lekernelI don't want involvements with glitches etc. messing up devices when the FPGA reconfigures or such09:53
lekernelwe power the ports all the time09:53
lekernelthis is what the current hardware does, and there hasn't been any problem with this09:53
wpwrakhow about 0402 footprint to unused pins and pull-up/down at the switch. then the on/off can be added later. i.e., when we hit that situation where it is a problem ;-)09:54
lekernelthat switch should simply act as an inrush current limiter09:54
wpwrakdoes that sound good ? that way, you get your "always on" switch, but if we ever think that we need a real switch, it's a 30 seconds upgrade09:57
wpwrakand all at a cost of a few picocents :)09:58
lekernelso, by default the switch won't be connected to the fpga?09:59
lekernelalso, putting more traces under the BGA takes time09:59
lekerneland increases risk if there are other traces to be moved ...09:59
aw_the two ENs for AP2142A/AP2152A yes needs low frequency(even stable as DC)  control bins which traces added under BGA, yeah...surely it will have a risk if add. I don't think adding these two routes will get serious trouble if really think it's a great idea.10:15
wolfspraulI think we should remove standby altogether10:52
Action: lekernel is packing bags for Norway10:58
wolfspraulhave you heard anything from the security conference and the m1 there?10:58
lekernelyes, piksel10:59
lekerneland no10:59
kristianpaulbtw bios have gdb vesion or something? i wonder now how to debug hangs/freezes when running commands (like simple mr mw) from bios11:53
wpwraksorry, needed one more nap. resume is difficult this morning :)12:17
wpwrak(switch) yes, unconnected by default. regarding the traces, there's major rerouting anyway with the DVI-I, so it shouldn't be a big deal. unless, of course, the overall complexity is already at the limit.12:18
wolfspraulyes I think for routing right now we just make a wishlist, then we see what the layout folks say12:23
wolfspraulthere is also the idea of leds at the ports...12:23
wpwrakyes. blinkenlights ! :)12:24
wpwrakalso on the full swing side of each path to a LED, add a TP12:24
wpwrakthat way, we can use LEDs as scope triggers12:25
wolfspraulfull swing side?12:25
wpwrakthe side before the LED-R combo. where you get 0-3.3 V (or whatever the voltage)12:26
wpwrakas opposed to putting the TP right next to the LED, where it may only see ~2 V peak-to-peak (depends on how the LED is connected to the R)12:26
lekernelhmm, LED capacitance could be a little problem here12:27
lekernelat least if you want precise triggers without unsoldering the LEDs12:27
wpwrakwas more thinking of "in the area" triggers12:28
wpwraka few ns more or less don't matter there. even a few us would be fine.12:28
wolfspraulok let's make it more real12:28
wpwrakthings like "when is a SETUP being sent", because the scope can't tell a SETUP from a SOF :)12:28
wolfspraulcan we enumerate the ports?12:29
wolfspraulwhich color?12:29
wpwrakwhich color looks good through the case ?12:29
wpwrakideally, it would be one LED per external connector12:30
wolfspraulbehind the connector? next to it? which side?12:31
wolfspraulon the bottom side of the pcb? :-) (just for completeness, I think we should not add more stuff to the bottom)12:31
wolfspraulcan you take a board png and just mark all places where you want to have one? :-)12:32
wpwrakhmm, tricky. i need to compare this with the real device to see where there is room. and my M1 can't be moved at the moment, with lots of cables hanging out of its brain12:34
wpwrakand yes, next to the connector, so that it can be seen from the outside12:35
wpwraklekernel: btw, doesn't the LV3 have USB in and out (for chaining) ? if yes, does this mean it has a hub inside ?12:42
lekernelno, it does not12:48
lekernelnot everyone is this perverted :p12:48
wpwrakheh ;-) but it does have an USB out port ? or does it use some proprietary bus ?12:50
wpwrak(or was that chaining interface a misunderstanding of a reviewer)12:50
lekernelit has nothing for chaining12:51
wpwrakah, i see. good :)12:51
lekernelwpwrak: how about computing the crc byte per byte in usb_rx? instead of just busy waiting12:55
lekernelwe have 32 AVR cycles per byte12:55
lekernelmost instructions are 1 cycle12:55
lekerneland yes, it's messy, but doing these things in verilog would be worse12:56
lekernelthe navre timing is 100% predictable, there's no DRAM, shared bus or interrupts12:59
wpwrakso navre runs at 48 Mhz ?13:00
lekerneland it might be possible to push the frequency a bit13:00
wpwrakah, good. that's better than i expected. hmm.13:00
wpwrakheh, the fastest avr in all mexico ;-)13:01
lekernelmaybe up to the 70-80MHz range13:01
lekernelbut if we do that, we should also keep a frequency that is a multiple of 12, otherwise the timing becomes less predictable13:01
lekerneland we pick up some latency in slightly messy clock domain transfers13:02
wpwraki suppose you could unroll usb_rx to solve problems one at a time. still a bit hairy, though. there's data FIFO, no ?13:02
wpwrak(12 Mhz) sure13:02
lekernelno, there's no FIFO13:03
lekerneljust one byte of buffering13:03
wpwrakah, interesting13:03
lekernelwell, the CPU timing is totally predictable, and the devices only transmit when asked. so there should be no need for a FIFO13:03
wpwrakyes, as long as the latency is under control, that's right13:04
wpwrakalso makes some of the timing easier to understand13:05
wpwrake.g., it would get messy if you got into usb_rx already with half the packet backlogged in a FIFO13:05
lekernelplane time... bbl13:18
sh4rm4is milkymist using an AVR cpu for usb ?16:41
wpwrakhmm, would be nice if "web update" would say from which date things will be. in fact, the version number seems kinda useless since the versions aren't being bumped16:43
wpwraksh4rm4: yes. it has "hardware" for the bit-level operations, but once there are bytes, the avr takes over16:44
kristianpaulsh4rm4: yes and not is called navre and is scratch implementationto be compatible with gnu-avr, so usb is implemented in a kind of "microcode"16:45
kristianpaulfrom scratch*16:45
kristianpaulhardware is the SIE no?16:45
sh4rm4so it is a custom softcpu ?16:45
kristianpauldunno if there are custom instructions16:46
sh4rm4custom as in IP-free16:47
kristianpauli hope ;)16:48
wpwrakkristianpaul: the mega-confusingly named softusb_sie ? yes, that's part of it16:49
kristianpaulhey we can actually said then milkymist soc is three core,lm32, fpu and navre16:53
wpwrak1, the only triple-core VJ workstation :)16:55
kristianpaulwpwrak: oh, yes you pixel procesor sounds really nice16:58
wpwrakup north ? :)17:12
sb0on the airport bus17:20
wpwrakdo they have snow yet ? ;-)17:20
sb0no, and it's actually a tad warmer than berlin17:22
wpwrakmust be the sea17:23
kristianpaul20 C are always perfect :)17:28
juliusbnot sure if this has been asked before, but are you guys able to run the Arduino/Wiring stuff on navre?20:29
juliusbI'm not too up on the AVR ISA and what the Navre supports, but that would be interesting20:29
kristianpauli think papillio boards are better example for such compatiblity/example20:30
juliusbwhat are those things? google just brings up links to birkenstocks :P20:33
juliusbi think i've heard of what you mean, but it's not papillio20:33
juliusbit's similar...20:33
kristianpaulhe, sorry typos20:35
kristianpauljuliusb: http://papilio.cc/20:35
kristianpauli think that what you want20:35
juliusbah right yeah20:36
juliusbok that looks cool, but I mean, could the Wiring stuff, as it is, run on say, the Navre in milkymist? Support for it it might make the M1 a good buy as an educational tool for beginners20:37
juliusbwould basically give you the simple IDE to push the LEDs/IOs/analog around20:38
juliusbanyway, just a thought, but I reckon it's pretty good, and the way it has drawn in a lot of users under the guide of the Arduino is impressive20:38
kristianpauli think already with wiring or not milkymist have a LOT of stuff for been an  educational tool20:43
juliusbyep, but with Wiring you can basically piggy-back onto the Arduino bandwagon in a way20:43
kristianpaulbut may be you mean been able to run processing on the M1, that will be killer, but kinda hard to port currently (java, c++, etc..)20:43
juliusbi think the accessibility, and simplicity of the interface is a winner, and has to be shown to be so20:44
juliusbmmm yeah actually id on't fully understand the Wiring/processing thing20:44
juliusbafaict the Wiring stuff provides the IDE and the I/O drivers20:45
juliusbwhat does processing do then?20:45
kristianpaulsure, simplify FN is important i think in the ML wpwrak mentioned something about aliases or self adjust to make things easier,20:46
juliusbah i see, processing is very java oriented20:49
juliusbmm, I guess I don't quite understand how it does what it does at the moment20:49
kristianpauleffects, shapes,, drawing and lots of libraries to make rich the process20:51
kristianpaulnot couting the examples and clean documentation in the web page20:51
juliusbmmm, fair enough20:53
mwallejust for the logs: +1 for software controlled USB enable signals23:13
mwalles/w controlled stuff makes test automation easier :)23:14
wpwrakmwalle: since sebastien isn't listening: my evil plan is to make it really easy to just drop in the connecting 0Rs. basically just one mouse click when preparing the fab files ;-)23:14
mwalleif its possible, just add the lines to an IO, and don't use it, so per default its always on23:16
mwallehopefully, this makes seb happy :)23:16
wpwraksebastien doesn't want it connected to the FPGA. he's concerned about glitches and such.23:16
mwalleuse some pullup or pulldown?23:18
mwalleexternal ones :)23:18
wpwrakwhat i proposed is to run the traces to the FPGA by have an unpopulated 0402 footprint in series with them. then add pull-up/down on the switch side to enable.23:19
wpwrakwell, 2 x 0402 footprint23:20
Action: mwalle votes for the same with popupated 0R resistors :)23:21
wpwrakheh :)23:22
rohusb-enable signals?23:53
rohyou mean for the 1.5k pullups?23:54
--- Fri Nov 18 201100:00

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