| 01:55.48 | *** join/#picogui Zensunni (rnd@n109.umd.com.au) |
| 01:55.56 | Zensunni | morning |
| 01:59.10 | Zensunni | anyone awake? |
| 01:59.30 | Zensunni | hi |
| 01:59.37 | gonkulator | what can I do for you? |
| 01:59.55 | Zensunni | oh, well, a while ago I was hoping to port PicoGUI to our little embedded board... |
| 02:00.21 | Zensunni | unfortunately, the kernel we use is crappy with libc/posix compliance, so it was too major an effort |
| 02:00.48 | Zensunni | instead (seems stupid, now that I explain it) I'm developing a very minimal system to allow us GUI-like funcitonality |
| 02:01.17 | Zensunni | and I was hoping to talk about CPU effective ways to handle a few basic things, like dirty-region repainting, etc... |
| 02:01.33 | gonkulator | heh |
| 02:01.45 | Zensunni | basically, I have written a lib to work on "canvases"... |
| 02:02.00 | Zensunni | ah, ok |
| 02:02.13 | gonkulator | so all my knowledge is picogui specific :/ |
| 02:02.17 | Zensunni | see, our board is very tight on the memory constraints... |
| 02:02.24 | Zensunni | 256K RAM, 128K ROM |
| 02:02.38 | gonkulator | yeah, thats pretty small |
| 02:05.04 | Zensunni | I think I chatted with micah about some of this some time ago |
| 02:05.12 | Zensunni | about replacing the network layer with message pipes... |
| 02:06.22 | gonkulator | yeah, he would be the person to talk to |
| 03:15.21 | Zensunni | gah |
| 03:15.33 | Zensunni | I had a big recursive function written to handle it |
| 03:15.43 | Zensunni | but I think I figured it would blow the stack :( |
| 06:20.02 | *** join/#picogui jojo (jojo@io.infidyne.com) [NETSPLIT VICTIM] |
| 06:20.02 | *** join/#picogui gonkulator (~brandon@c-24-9-213-236.client.comcast.net) [NETSPLIT VICTIM] |
| 06:20.02 | *** join/#picogui Talez (~talez@talez.arach.net.au) [NETSPLIT VICTIM] |
| 06:20.02 | *** join/#picogui TD[zZz] (~mike@bdsl.66.13.154.244.gte.net) [NETSPLIT VICTIM] |
| 06:20.02 | *** join/#picogui njs (njs@c-24-4-201-128.client.comcast.net) [NETSPLIT VICTIM] |
| 06:20.03 | *** mode/#picogui [+b *!*@*.dip.t-dialin.net] by ballard.freenode.net |
| 06:31.25 | *** join/#picogui FunkyBob (~tasty@c211-28-138-224.eburwd3.vic.optusnet.com.au) |
| 06:33.16 | Zensunni | back again... |
| 06:39.24 | *** join/#picogui CIA-4 (~CIA@navi.cs.colorado.edu) |
| 06:39.55 | Zensunni | morning |
| 07:10.46 | Zensunni | anyone awake, yet? |
| 09:53.34 | *** join/#picogui FunkyBob (~tasty@c211-28-138-224.eburwd3.vic.optusnet.com.au) |
| 10:00.03 | Zensunni | anyone awake, yet? |
| 10:06.23 | *** part/#picogui Zensunni (~tasty@c211-28-138-224.eburwd3.vic.optusnet.com.au) |
| 10:28.27 | *** join/#picogui Zensunni (~tasty@c211-28-138-224.eburwd3.vic.optusnet.com.au) |
| 10:38.18 | *** part/#picogui Zensunni (~tasty@c211-28-138-224.eburwd3.vic.optusnet.com.au) |
| 10:59.42 | *** join/#picogui Zensunni (~tasty@c211-28-138-224.eburwd3.vic.optusnet.com.au) |
| 10:59.45 | Zensunni | ... |
| 11:28.13 | *** join/#picogui azer (~azer@h75n3fls22o1062.bredband.comhem.se) |
| 11:28.51 | Zensunni | morning, azer |
| 11:29.16 | azer | morning =) |
| 11:29.27 | azer | the clock is 12.30 pm here, but almost morning =D |
| 11:29.35 | Zensunni | it's always morning somewhere :) |
| 11:29.47 | Zensunni | hell, it's 10:28pm here |
| 11:30.36 | Zensunni | don't suppose you know much about repaint region clipping? |
| 11:30.51 | azer | nop, sorry |
| 11:30.57 | Zensunni | bugger |
| 11:31.54 | Zensunni | I'm trying to implement an ultra-light GUI system, and was hoping to find an efficient algorithm |
| 11:32.35 | njs | Zensunni: read about libart's uta's |
| 11:32.41 | njs | (= microtile arrays) |
| 11:32.47 | njs | or just steal the code |
| 11:33.17 | njs | you're _so_ memory constrained that even they might be too heavy, not sure; guess it depends on your resolution. |
| 11:33.18 | Zensunni | UTAs? |
| 11:33.22 | Zensunni | hmm |
| 11:33.29 | njs | but look into them. |
| 11:33.40 | Zensunni | current resolution is 320x240x1bit |
| 11:34.01 | Zensunni | but we're developing a controller for a 800x600x16bit LCD |
| 11:35.04 | Zensunni | for some reason, I understood that my first approach was pretty much how it was done... |
| 11:35.18 | Zensunni | so I figured someone, somewhere would have come up with an efficient way of doing it |
| 11:42.28 | Zensunni | no.. UTAs don't really look like a solution at all, I'm afraid... |
| 11:44.28 | njs | no? uh, okay. |
| 11:44.36 | njs | oh, _clipping_, not culling. |
| 11:44.43 | njs | ? |
| 11:46.15 | Zensunni | yer... |
| 11:46.49 | Zensunni | when a window invalidates a region, I need to know if it's occluded by a higher window |
| 11:46.56 | Zensunni | and if not, which window(s) cover that region |
| 11:47.18 | *** join/#picogui TD (~mike@cpc1-macc1-3-0-cust53.bagu.cable.ntl.com) |
| 11:47.27 | Zensunni | morning, TD |
| 11:47.57 | njs | the traditional thing is to keep a screen-level dirty region (using, say, a uta) |
| 11:48.14 | njs | and each time an area is dirtied, extend the screen-level region to cover it |
| 11:48.25 | Zensunni | sure... that bit's easy |
| 11:48.31 | Zensunni | but it's when it comes to actually redrawing... |
| 11:48.38 | njs | and then when you want to redraw, cycle through the display intersecting each thing to redraw with your dirty region |
| 11:48.46 | njs | and only redraw if it intersects |
| 11:49.32 | Zensunni | ok... now we get to the bit I'm actually asking about |
| 11:49.56 | Zensunni | as I go through each window in the list (front to back), it may only overlap parts of the dirty region... |
| 11:50.11 | Zensunni | in which case, I must cut the region, and check each of the resulting regions individualy... |
| 11:50.31 | Zensunni | I'm looking for an efficient way to do this |
| 11:51.34 | Zensunni | a computationaly cheap way to calcuate a set of rectangles representing the union and non-union parts of two rectangles |
| 11:52.22 | njs | uh, standard would be to go back to front and just use painter's algorithm |
| 11:53.31 | Zensunni | which would involve insane amounts of overdraw, which I can't afford |
| 11:54.43 | Zensunni | the idea was to eliminate overdraw |
| 11:54.58 | Zensunni | otherwise I might as well just repaint all the windows entirely... |
| 11:55.33 | njs | ah |
| 12:00.03 | Zensunni | surely PicoGUI doesn't use such a wasteful method? |
| 12:02.41 | njs | picogui doesn't have overlapping windows :-) |
| 12:03.01 | Zensunni | ah |
| 12:03.05 | Zensunni | heh |
| 12:03.06 | Zensunni | sneaky. do |
| 12:03.07 | Zensunni | do8 |
| 12:03.22 | Zensunni | I thought it did |
| 12:04.31 | Zensunni | I can't seem to get to the PicoGUI home page |
| 12:05.38 | njs | actually, you can do what you want with utas |
| 12:05.50 | Zensunni | oh? |
| 12:06.12 | Zensunni | do tell |
| 12:06.31 | njs | cycle back through the window layers, and at each layer calculate the uta that conservatively estimates how much stuff is visible |
| 12:07.14 | Zensunni | and what to do with the bits that aren't covered? |
| 12:07.40 | njs | so basically you uta_layer_n-1 = intersection(uta_approx(complement(windows_at_layer_n-1)), uta_for_layer_n-1)) |
| 12:08.15 | njs | draw in them |
| 12:08.26 | njs | it's a little tricky because you have to do two passes, at least in the naive version |
| 12:08.33 | njs | there might be cleverness or caching or something that helps |
| 12:08.56 | Zensunni | hmm... I'm not really seeing it... |
| 12:09.21 | njs | not sure how to explain it better |
| 12:09.24 | njs | and I need to sleep |
| 12:09.25 | njs | sorry |
| 12:09.28 | Zensunni | :) |
| 12:09.31 | Zensunni | sleep well |
| 12:09.34 | Zensunni | thanks for trying |
| 12:09.42 | Zensunni | I'll be back other times, so we can try again :) |
| 12:11.03 | Zensunni | problem I see is when the UTA that's dirty and the UTA for the window don't totaly overlap.... |
| 12:11.11 | Zensunni | which brings us back to the original problem... |
| 12:11.23 | Zensunni | I'm just working on less limited UTAs |
| 12:26.25 | Zensunni | nobody else have any thoughts? |
| 12:27.37 | *** join/#picogui lalo (~lalo@200.102.128.117) |
| 12:27.48 | Zensunni | morning, lalo |
| 12:33.16 | lalo | hi |
| 14:09.34 | *** join/#picogui lalo (~lalo@200.102.120.32) |
| 14:16.43 | *** join/#picogui veli (veli@smooth.piipiip.net) |
| 14:17.02 | veli | hi... picogui website is down? |
| 14:20.18 | lalo | hmm |
| 14:20.22 | lalo | seems to be |
| 14:26.13 | lalo | memory error :-/ |
| 14:26.22 | lalo | I'll contact support, thanks for telling me |
| 15:08.29 | veli | ok, it works now.. |
| 15:08.32 | veli | great, thanks.. |
| 15:12.04 | lalo | thank you :-) |
| 18:39.07 | *** join/#picogui TD (~mike@cpc1-macc1-3-0-cust53.bagu.cable.ntl.com) |
| 21:28.17 | *** join/#picogui Zensunni (~tasty@c211-28-138-224.eburwd3.vic.optusnet.com.au) |
| 23:12.12 | Zensunni | gonkulator : you awake? |
| 23:12.40 | gonkulator | Zensunni: sure |
| 23:12.58 | Zensunni | someone in GGI helped me come up with a passable solution :) |
| 23:13.20 | gonkulator | cool |
| 23:13.33 | Zensunni | basicaly using a 1bit Z-buffer |
| 23:13.44 | Zensunni | and a masked blit function |
| 23:14.58 | Zensunni | each window keeps a single dirty region... and the repaint checks each one in turn... |
| 23:15.23 | Zensunni | after it's been repainted, it then fills in the Z-buffer (mask) with the window's position |
| 23:15.34 | Zensunni | each successive layer blits masked against that |
| 23:18.22 | Zensunni | does that make sense? |
| 23:19.08 | gonkulator | so, sure :) |
| 23:19.45 | Zensunni | heh |