| 16:06 | MMlosh | Hmm.. I haven't checked anything for a while.. My cubox seems to be running a kernel 2.6.32.9-dove-5.4.2 and has ports.ubuntu.com / lucid in the sources.list. I remember that I was advised not to upgrade, because of neon instructions mixed in regular ubuntu binaries. Is this still the case? What are the current deb-based distro possibilities? |
| 16:29 | MMlosh | oh my.. I was gone from the forums that long.. I forgot where the post datetime is displayed, so I was looking at "join date" of posters instead.. |
| 16:47 | MMlosh | Ah, that phoronix drm news is current, awesome.. |
| 19:14 | _rmk | 19:14 * _rmk_ grumbles, just trying to get h/w cursor working and it looks like its one pixel out :( |
| 19:14 | _rmk_ | and I can't find a reason for it |
| 19:17 | _rmk_ | thing is, it's not the hardware |
| 19:17 | _rmk_ | and not the cursor theme either |
| 19:22 | shesselba | _rmk_: I posted a follow up to your RFC, as in-reply-to yours .. looks like mailing-lists doesn't like it |
| 19:25 | shesselba | about the "one pixel out" above, I noticed the whole frame being one pixel out (set blank color != black) when poking with dove lcd.. haven't tested on your RFC though |
| 19:25 | shesselba | may be a bug in dove lcd HW |
| 19:29 | MMlosh | btw: is it easy to get a fresh kernel booting for a noob, or I'd better wait? |
| 19:52 | _rmk_ | shesselba: well, I'm currently programming the cursor with the first displayed column/row in cyan so I know where the top,left corner of it is. |
| 19:52 | _rmk_ | when the horizontal position is set to 0, the vertical cyan line aligns with the left edge of the background |
| 19:53 | _rmk_ | so that much is correct |
| 19:54 | _rmk_ | if I load up xpaint, move it so the window is partially off the LHS, and position the pointer/pencil at the very left edge, then it's one pixel more than it should be |
| 19:57 | shesselba | _rmk_: I just mentioned it because I had the whole active area with one pixel offset once |
| 19:57 | shesselba | you can see it when you set blanking color to non-black |
| 19:58 | shesselba | and it could be possible that the lcdc has a bug in "counting" pixels that could also be in overlay part |
| 20:05 | shesselba | _rmk_: With your driver working from DT, I can measure sync polarity on d2plug for you |
| 20:06 | shesselba | it is not that tiny as the cubox and sync is easier to access with an oscilloscope |
| 22:06 | felisuco | hi does anybody was able to install gnome-panel in ubuntu 12 10 cubox? |
| 22:47 | _rmk_ | shesselba: I think this is an X server bug :( |
| 22:48 | _rmk_ | shesselba: I guess Jean-Francois is peeved about my driver |
| 22:50 | shesselba | _rmk_: I guess he is pissed because of me |
| 22:51 | _rmk_ | oooh, just gone back to sw cursor |
| 22:51 | shesselba | I am about to have video on d2plug |
| 22:51 | shesselba | looking for a mode that my lcd likes with only axiclk (400MHz) available |
| 22:51 | shesselba | well 400MHz with integer div of lcdc |
| 22:55 | _rmk_ | btw, wrapping the eSATA cable in tin foil fixed my wireless kbd/mouse problems :) |
| 22:55 | MMlosh | oh.. was that an eSata-eSata or eSata-sata cable? |
| 22:56 | shesselba | hehe, now wrap yourself up also.. who knows what that bad bad rays can cause you ;) |
| 22:56 | _rmk_ | that was the eSATA cable that came with my icybox enclosure |
| 22:56 | MMlosh | (esata-esata cables should be and are shielded, but the esata-sata I picked up is not) |
| 22:56 | dv_ | true |
| 22:56 | _rmk_ | right. I will stop chasing this bug, because it's *not* my problem :) |
| 22:57 | dv_ | I have an icybox enclosure that actually does not use esata, but regular one |
| 22:57 | dv_ | its from 2005 tho |
| 22:57 | _rmk_ | with s/w cursor in xpaint, the hotspot is one pixel to the right and one pixel below the point of the pencil. |
| 22:57 | _rmk_ | in exactly the same place with my h/w cursor |
| 22:58 | dv_ | and I had to keep it shielded (because the PC also didnt have esata, just an extension for the regular one) |
| 22:58 | shesselba | _rmk_: so you are really going to hunt that bug down to X? |
| 22:59 | dv_ | _rmk_: when reading about all of your efforts, I wonder how many of the difficulties are related to the dove/vivante stuff, and how much is because of X |
| 22:59 | _rmk_ | shesselba: it could well be an error in the X cursor data file for it |
| 23:00 | _rmk_ | dv: well, yes, but I kind of expect hardware not to be perfect :) |
| 23:00 | shesselba | can you dump the data passed to either hw or sw cursor (i guess you already did) |
| 23:00 | dv_ | perfect hardware would be scary. so unfamiliar :) |
| 23:01 | _rmk_ | shesselba: yep, I've already dumped the cursor image back out of the LCD controller and compared it to what the X server passed :) |
| 23:03 | dv_ | btw. off the top of your head, are you aware of something in vfp or neon for scalar products? |
| 23:04 | dv_ | (and iwmmxt for the cubox) |
| 23:04 | dv_ | recently finished an audio filter that uses several biquads, and for these, such a thing would be really handy |
| 23:07 | dv_ | oh, nevermind, stupid me. googling for "dot product" when "fused multiply-add" gives me much better results. |
| 23:12 | shesselba | _rmk_: ok got a mode running (1400x1050p60) .. by coincidence one affected by the sync inversion issue. |
| 23:13 | shesselba | the active region is of by some 20-30 pixel which should be hsync duration - and therefore misinterpretation of hsync edges |
| 23:13 | shesselba | exactly what you are looking for, right? |
| 23:15 | shesselba | I can open up the d2plug and put an oscilloscope on hsync and vsync provided by lcdc at hdmi encoder.. you know the pins by heart? |
| 23:49 | _rmk | 23:49 * _rmk_ looks back at irc :) |
| 23:49 | _rmk_ | I just want to know whether the pulses on vsync and hsync are +ve going or -ve going |
| 23:51 | shesselba | ok, gimme some minutes more to write it all down |
| 23:52 | _rmk_ | also if you do this to confirm the register setting: grep 01b8 /sys/kernel/debug/dri/0/reg |
| 23:54 | _rmk_ | it'll be good to get some of these questions finally answered :) |
| 23:54 | shesselba | 0: 0x01b8: 0x7000000c |
| 23:54 | _rmk_ | that's the one... so both the vsync and hsync invert bits are set. |
| 23:55 | _rmk_ | hmm, but it's blanked |
| 23:56 | shesselba | the display? |
| 23:56 | _rmk_ | apparantly so. |
| 23:56 | shesselba | it just blanked before I grep'd |
| 23:57 | _rmk_ | I thought I'd mention that if you were just about to put the scope on the signals... |
| 23:57 | _rmk_ | if you prod it so it unblanks, it'll become 6000000d |
| 23:58 | shesselba | you have a shell cmd ready for unblank? I ve no kbd attached |
| 23:58 | _rmk_ | should I assume it's running X or console? |
| 23:58 | shesselba | console |