#milkymist IRC log for Tuesday, 2012-05-15

cladamw_(J27 Keep-out layer/ring ) wpwrak, http://downloads.qi-hardware.com/people/adam/m1/tmp/m1r4/m1r4_j27_keepout_layer.png01:53
cladamw_except adding this 6.6 mm keep-out ring, I'll also keep J27 away from J21 about 1.5mm and still move to south 1.5 mm. The 0.32 mm you measured was traces in inner layers, but it's very nice a precaution discovery on using keep-out layer to clean. great.01:57
wpwrakcladamw: great. maybe make the ring 7 mm in diameter. that way we can be sure it's large enough, even if something in the process changes slightly02:06
wpwrakand perhaps you want to make all the 7 mm copper02:07
cladamwhow about we let all mounting holes with a total 6.5 mm copper and plus 0.5 extra keep-out layer ? then total is 7 mm02:11
cladamwwpwrak, see my last sentences. :-) Is it okay to you? or you want 7 mm copper then we add extra 0.5 mm keep-out layer, so total 7.5 mm for mounting hole dia.02:18
wpwrakroh suggested 7.0 mm minimum and likes 8 mm even better02:24
wpwrakso 7.5 mm would fit that nicely :) do we have enough room ?02:24
cladamwwpwrak, four corners I think there's no problems for 8 mm, i just want to let J27 with suitable dia and make it reserved to other routes, but overall i think J27 should be also okay for 8 mm in total.02:41
cladamwwpwrak, so let's 7 mm for copper ring and 1 mm for extra keep-out ring02:42
wpwrakyes, there should be enough room for 8 mm around J2702:42
wpwrakhehe, cool. 7+1 = 8 then :)02:42
GitHub196[board-m1] adamwang pushed 1 new commit to master: https://github.com/milkymist/board-m1/commit/5ca97efb803b81e1ecc5acdcab2a57f81cbabeb403:25
GitHub196[board-m1/master] Misc.sch: removed text of "Placement (TP32, TP38-TP41) at the bottom of PCB." - Adam Wang03:25
GitHub139[rtems-yaffs2] sebhub pushed 1 new commit to master: https://github.com/milkymist/rtems-yaffs2/commit/3294f1c700e52f1029f09b221143451ca371726309:16
GitHub139[rtems-yaffs2/master] rtems_yaffs: Update due to RTEMS API changes - Sebastian Huber09:16
lekernelnow, writes ...10:02
wpwrak"ddrrow" -> "Precharged". that what i call low-level control ;-)10:15
guyzmolekernel - hey, how are you doing ? :)10:27
guyzmolekernel - I had some adventures working the last version of milkymist (didn't know about the UI changes on right-click/esc keys :-S)10:28
guyzmolekernel - anyway, things have gone swell, though I only had less than ten vjs that went to see the workshop, and only ~4 did actually try to write FN patches10:29
guyzmoall were impressed by the realtime-ness of the stuff, but most of them said they don't want it, as their "usual" vj software can do everything that do the MM and more stuff, and they don't care that much about latency... while they did care about writing complicated lines of code to do stuff10:32
guyzmo(one said "I prefer to stick with Puredata programming")10:32
lekernelso, does someone want to begin the development of the "rf reactable" surface?10:38
guyzmoand last thing, they all said it's cool, but it'd be better if it could output HD !10:40
lekernelthis I'm on just now10:41
guyzmosweet :)10:41
guyzmoso in the end, I'm quite sad to say that I don't think they will buy one10:43
guyzmoto those guys10:43
guyzmobut people working on VJ/video software were interested in the product to integrate it with they software so it can run and control stuff "blindly"10:44
lekernelyou mean resolume/modul8/... developers?10:46
lekerneltell them they should instead consider running their software on a milkymist device :)10:47
guyzmonope, they weren't there, but the milumin dev guy was interested10:48
lekernelhttp://www.millumin.com/ ?10:49
guyzmoit's a software that helps manage a whole stage scenography10:49
guyzmoand there was also some guys working on a 3D mapping software10:49
guyzmothat thought about it for a R&D project they're working on10:49
guyzmobut anyway, I don't think that will end in real stuff10:51
guyzmoand in the end, vision-r people were really nice10:51
GitHub91[milkymist-ng] sbourdeauducq pushed 1 new commit to master: http://git.io/0VaHeA11:21
GitHub91[milkymist-ng/master] asmicon/multiplexer: fix read tag delay - Sebastien Bourdeauducq11:21
wpwrakguyzmo: code complexity is a bit of an issue indeed. and debugging is also unnecessarily hard at the moment. e.g., you can't trace most variable values and tracing those you can is inconvenient.11:36
wpwrakguyzmo: did you show the M1 with MIDI ? or did you use it just as a non-interactive effect generator ?11:36
guyzmowpwrak - I did not have a controller, neither did the vision-r guys could give me one11:38
guyzmobut that was not an issue11:38
guyzmopeople understood very well how a controller could interact11:38
guyzmoeither to switch the effects, or to modify an effect behaviour11:38
wpwrakhmm. seeing is believing, though :)11:39
wpwrakwhen showing the M1 to friends, the reaction i get is a polite "nice geek gadget. pretty colors." (with a stifled yawn) when i show it just generating things, to "THIS IS SO COOL !" when i just run one patch but with MIDI control11:41
wpwraki can't imagine that it wouldn;t have a similar effect on more experienced users, too. also, with MIDI, you start to care a bit more about latency :)11:42
guyzmowpwrak - well, I had the camera on11:46
guyzmoso the "WHOA ! it's fast" effect was there11:46
guyzmoso every time I say something is fast on the MM1, they just then believed me :)11:46
lekernelwrites work :)11:49
lekernelwe got DRAM11:49
lekernelwpwrak: with the next UI, there shouldn't be any way to run it without controllers ;)11:52
lekernelhmm... should we maintain any current SoC support in Linux?11:53
lekernelor just switch everything to -ng11:53
Fallenouyes !!!11:58
Fallenouawesome !11:59
lekernelI wouldn't be surprised if there were a couple of subtle bugs lurking in the darkness though12:00
Fallenoulekernel: I pushed a few commits about exception handling, and I added if (rst_i) in always blocks12:08
FallenouI will do more intensive testing soon12:09
Fallenouto check if exception handling is robust12:09
lekernelor not-so-subtle, eg accesses above 4MB do not work12:10
lekernelI suspect this is when crossing a bank boundary ...12:11
Hodappit warms my heart that people still exist who care about memory amounts that are less than 4 MB.12:17
HodappFirst programming class in undergrad, I heard plenty of "Why should I care about that? Computers have gigs and gigs of memory."12:18
lekernelthis is only true for computers that use DRAM...12:19
HodappFirst engineering job, I was writing for a Cypress PSoC that had RAM measured in kilobytes12:19
lekernelyes :)12:19
lekernelSRAM is extremely expensive12:19
wpwrakHodapp: my first "home computer" had 1 kB in total. including screen memory :)12:19
Hodappwpwrak: I'm just going to halt this discussion immediately with: http://www.youtube.com/watch?v=Xe1a1wHxTyo12:20
wpwrakwhen your program got too big, you had to sacrifice screen space ..12:20
HodappOh, you had a screen? We used to DREAM of having a screen!12:20
Hodapplekernel: This is all a bit foreign to me, but I do vaguely recall that DRAM has more complex refresh and access patterns... right?12:21
lekernelamong other complexities, yes12:21
HodappI've not written enough stuff at this level... I need to.12:22
lekernelessentially, writing a good performance controller for a modern DRAM (>= DDR) will likely make you a FPGA/computer architecture expert12:23
wpwrakHodapp: good one ;-)12:24
Hodappwpwrak: what, the Monty Python?12:24
Hodappit's one of my favorites12:24
wpwraki like the series that goes something like this: when i was a kid, we had to walk 10 km to school each day. through 2 m of snow. barefoot. ... and so on ... uphill. both ways.12:29
Hodappthere's a Slashdot thread that goes on for like 20 replies like that12:38
GitHub49[migen] sbourdeauducq pushed 1 new commit to master: https://github.com/milkymist/migen/commit/0bea1e25890908630f3cd403b0e41d4ecec042a112:49
GitHub49[migen/master] asmi: dat_wm high to disable data write - Sebastien Bourdeauducq12:49
lekernelah. that was a cache layout problem. now the CPU can access the whole 128M ...13:25
GitHub102[migen] sbourdeauducq pushed 1 new commit to master: https://github.com/milkymist/migen/commit/68cd445662c8d015ff4ec9a65e8e68e3427d366d13:25
GitHub102[migen/master] bus/wishbone2asmi: fix cache tag size - Sebastien Bourdeauducq13:25
lekernelelldekaa: when are you switching to migen? :)14:20
Last message repeated 1 time(s).14:45
elldekaaI cannot, it's to low level for me14:45
elldekaaBut I have to try it14:46
lekernelelldekaa: sounds like a sensible HDL back-end for ORCC14:48
elldekaayep, probably something to try14:49
elldekaait should not be to hard14:50
lekerneland you'll get features for free like actors accessing DRAM buffers at 10GBps off a $40 spartan-614:50
elldekaawith dma ?14:51
elldekaasounds interresting, i need something like that for my TTA backend14:51
elldekaafor the buffer of the motion compensation of the video decoder14:52
lekerneland we'll get a video decoder, which is also something we need. so... :)14:52
elldekaaseveral video decoders ;-)14:53
lekernelwell, after eliminating the patented and broken ones, I wonder how many will be left14:54
Fallenourm -rf h264 mpeg*14:56
elldekaaalive : AVC / MPEG4-SP / AVS ... unpatented: i don't know (probably nothing)14:56
lekernelgoogle webm?14:56
elldekaai don't know about the patent status of AVS14:57
elldekaayour dma is cross-platform or not ?14:59
lekernelnot yet, we don't have a good memory access abstraction in the dataflow system atm15:00
lekernelbut it should15:00
lekernelthere are tons of things that are still missing in DF now :)15:01
elldekaalekernel: I need a chapter four with more details :D15:22
Action: kristianpaul want to swich to llhdl ;D15:28
lekernelkristianpaul: llhdl and other bitstream hacks are too far away from the big M1 problem for me to be motivated to touch them atm15:32
kristianpauloh sure M1 as product has its own history15:33
lekerneluse your magic wand and make milkymist beat rpi on google trends, then I'll have time for llhdl15:34
kristianpaulthat will not happen, not us or anytime soon15:35
lekernelthen same for llhdl.15:35
kristianpaullol at leasy your port FN to rpi15:35
lekernelthis is one of the mindsets I want to avoid. and llhdl will do nothing towards this goal.15:46
lekernelsee Jon's email about the FN iPad application...15:50
kristianpaullekernel: i think we agree our goals may differ16:12
kristianpaullekernel: yup :)16:12
kristianpaul(Jon's email)^16:13
kristianpaulso why you started llhdl? what was your motivation at that time?16:30
kristianpaula weeked hack fpg internal project? ;-)16:30
lekernelthat was january-march 2011, we had just started selling the rc2 boards, and I was still young and innocent at that time16:37
lekernelthen drawerware, unpopularity, etc. started to gradually piss me off16:38
elldekaaI tried to advertise on the Milkymist One to one of my friend, researcher in 'new media art'16:42
elldekaahe seems interested16:42
elldekaaAnd I'm thinking about buying one for me :-)16:44
elldekaacould be a good motivation for the migen backend16:44
kristianpauli must said i bought a M1 because the first llhdl example worked on it :)17:03
lekernelmwalle: what's the latest on the fast UART? how fast did you get it to work?17:28
GitHub140[milkymist-ng] sbourdeauducq pushed 1 new commit to master: http://git.io/72ovIQ17:37
GitHub140[milkymist-ng/master] bios: automatically enable hardware memory controller and test memory - Sebastien Bourdeauducq17:37
--- Wed May 16 201200:00

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