#qi-hardware IRC log for Monday, 2013-03-04

K0D3Rhello 02:25
K0D3Rwhere can I buy a Milkymist 02:25
K0D3Rsharism.cc is down02:29
K0D3Rthe links to hackabe-devices  returns 40402:32
whitequarkmaybe you should ask wolfspraul02:47
K0D3Rwhitequark: who is wolfspraul03:28
whitequarkwolfspraul: K0D3R wants to buy an M104:15
wolfspraulwow nice!04:18
whitequarkbut sharism.cc is down04:19
wolfspraulthere were some hickups but it should be back up04:20
K0D3Rwolfspraul: hey!04:20
whitequarkwolfspraul: oh. some folks here decided that you've had "enough of it"04:21
whitequarkI'm glad that's not true04:21
K0D3Rwolfspraul: whitequark said that I should talk to you04:22
wolfspraulsure that's nice04:22
K0D3Rhow can I buy a Milkymist?04:22
wolfsprauljust at sharism.cc04:22
wolfspraulwe ship from Taipei and it should be with you in 3-4 days or so, depending on where you are04:23
wolfspraulwe have shipped a lot to US, Australia, Europe04:23
K0D3Rbut it is down04:23
K0D3RI meant: sharism.cc is down04:23
whitequarkK0D3R: up for me already04:23
K0D3Rwhitequark: oh! it's up now! thanks!04:24
K0D3Rsomeone turned the switch on! :)04:25
K0D3Rwolfspraul:  do you ship it yourself?04:26
wolfspraulwhitequark: thanks, I think I answered all his questions05:09
wolfspraulthanks for making the connection...05:09
whitequarkwolfspraul: you're welcome :)05:46
Fallenouhi wolfspraul , good that sharism.cc is back on :)06:21
wpwrakkristianpaul: not sure what you mean with moving things between application and kernel. i think an fpga should probably be assigned to a single task (process or group of cooperating processes), at least for now.12:51
wpwrakif someone can figure out a way to actually multitask with them, even better. so that's another item for the research topic list :)12:52
wpwrak(multitask) in this case, have independent tasks using them, e.g., a video game and a bitcoin miner12:54
paul_boddiewpwrak: How did you do the video for http://en.qi-hardware.com/wiki/UBB-VGA (hint: not very descriptive page)? Is it all generated in the CPU's "spare moments" or is there some kind of backing store?14:02
wpwrakpaul_boddie: you may find this page more useful: http://downloads.qi-hardware.com/people/werner/ubb/vga/web/14:22
wpwrakand it's unfortunately not in "spare moments". the cpu is working on it full-time. well, most of it is spent waiting for DMA, but it wouldn't be easy to let it do useful work instead of waiting.14:23
paul_boddieI guess the aim is to do live screen updates, rather than just push out still images, or am I wrong?14:27
paul_boddieThe reason why I ask is because you were discussing FPGAs and the whole thing reminds me of some ULA-like circuitry from the microcomputer era that I've become interested in over the past year or so.14:28
Action: paul_boddie looks at http://downloads.qi-hardware.com/people/werner/ubb/vga/web/14:29
wpwrakthe initial aim was just a proof of concept. see how far i can take it in terms of image quality/resolution. that works pretty well now. but it only does still images.14:31
wpwrakfor live updates, you'd have to move the DMA control and hsync processing into an interrupt handler and then time interrupts such that you never miss the time for the next line. (the cpu needs to intervene for each line, because the operation stalls the mmc controller)14:34
paul_boddieIt's a shame that the 8:10 slot isn't like the expansion bus on various computers (or Firewire, maybe).14:35
wpwrak(and the mmc controller also spits out some "noise" that you don't want in the image. maybe you could work around this, though)14:35
wpwrakwell, it's quite flexible14:35
paul_boddieIn the sense that with various microcomputers, you could make peripherals that can access memory and stuff like that.14:36
paul_boddieI guess all that is formalised in various standard interfaces, though, but not for the 8:10 port.14:37
wpwrakway too complex :) and you don't want to expose a memory bus. not since 1990 or so, when they just started to get too fast for such things14:39
paul_boddieI was just thinking of the DMA aspect, but as I noted, I imagine that various standard interfaces provide that, and Firewire is notorious for doing so.14:41
paul_boddieA kind of slow-motion access to memory would be cool, but I agree that it would complicate things.14:42
wpwrakyeah, but that also means that you need a firewire controller on your boards. i prefer the simplicity of ubb :)14:42
wpwrakwhat would be nice is to have a cpld or small fpga between cpu and 8:10 card(s). that way, you could implement your own protocols.14:43
paul_boddieI guess you continuously drive the display instead of transferring a bitmap for output, then.14:43
wpwrakwell, you have to continuously generate the video signal. even though the screen would probably have an internal frame buffer, it's not used in that way. and of course, even less so with vga.14:45
paul_boddieYes, I have a small LCD panel for use with Arduino that probably has a more powerful CPU and more internal memory than the ATmega.14:46
paul_boddieFor the video signal, I'd have thought we would be in FPGA territory.14:47
wpwrakwhat ubb-vga does is that it precomputes the bit pattern for the image. then it sends this pattern in an MMC block transfer. at the beginning of a line, the cpu needs to generate a start bit for MMC and mask the output for the mmc command (which would look like noise in the image).15:03
wpwrak(look like noise. or, worse, throw off the image position detection in the screen)15:04
kristianpaulwpwrak: yes basically what you said i meant15:46
--- Tue Mar 5 201300:00

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