| 05:40 | sudonhim | i think the framebuffer is using triple buffering... does anybody know of a way to make it use double or single buffering to save memory? |
| 06:28 | jnettlet | sudonhim, by default only enough memory for a single framebuffer is allocated. If you enable double buffering then the gpu will expand the size of the framebuffer by doubling the height. |
| 06:28 | sudonhim | oh... so it increases in size when I start up the opengl context? |
| 06:29 | jnettlet | if you have the environment variable set to use multiple buffers |
| 06:29 | jnettlet | although it doesn't shrink it after that. |
| 06:29 | sudonhim | I can tell its at least double buffering, because I need to swap the buffers before I can access the contents |
| 06:30 | jnettlet | well some apps might |
| 06:30 | jnettlet | that may be the default now. |
| 06:31 | sudonhim | is that why you have a virtual and a physical height in fbdev? |
| 06:31 | jnettlet | yes |
| 06:31 | sudonhim | ahhhhh |
| 06:31 | jnettlet | that is how fbdev does it. |
| 06:32 | jnettlet | kms/drm actually allocates multiple buffers |
| 06:38 | jnettlet | sudonhim, are you using the linux-fslc kernel? |
| 06:38 | sudonhim | yeah |
| 06:38 | jnettlet | everything pretty stable? |
| 06:38 | sudonhim | going well :) |
| 06:38 | jnettlet | good. |
| 06:38 | jnettlet | spdif is broken. I need to fix it. |
| 06:40 | sudonhim | I think every issue we've had has been either a) not reproducible on the hummingboard, or b) also present with the linaro |
| 06:40 | sudonhim | ... a.k.a, not your fault! |
| 06:41 | sudonhim | (and probably ours) |
| 06:41 | jnettlet | well let me know any issues you are having. If it is a kernel or driver problem I would like to have it fixed. |
| 10:54 | topi` | jnettlet: do you provide GPT support with U-BOOT these days? :) |
| 10:58 | topi` | jnettlet: it seems your linux-linaro-stable-mx6 is still on 3.14.14 ... do you have another repo somewhere with newer linux kernel? |
| 11:04 | jnettlet | topi`, no I still don't include GPT by default. Since it is so hard to setup because SPL sits on the disk where GPT wants to be. |
| 11:04 | jnettlet | It is just too difficult to support for those not very experienced with it. |
| 11:05 | jnettlet | topi`, the latest kernel we are working on is the linux-fslc kernel. Since freescale has released a 3.14 kernel we are switching to that so hopefully more platforms can share patches and fixes. |
| 11:06 | jnettlet | Although we later found out that fsl did a miserable job implementing a base kernel, and I have spent more time fixing their bugs rather than integrating new features. |
| 11:06 | jnettlet | once I have spdif fixed I think we are at full hardware support now. |
| 11:10 | vpeter | jnettlet: Can I assume openelec will switch to this linux-fslc kernel you are working on? |
| 11:11 | jnettlet | vpeter, I can neither confirm nor deny that. I would hope they would do that, but that also requires them switching to the new vivante v5 drivers. |
| 11:11 | jnettlet | although sraue did ping me yesterday. I haven't had a chance to get back to him yet. |
| 11:11 | vpeter | Yes, he answered me also. |
| 11:12 | jnettlet | vpeter, this should help your situation as I am hoping to merge my changes back to yocto once I am certain all the bugs have been sorted out. |
| 11:12 | vpeter | I have to check if my patches applied there. |
| 11:13 | jnettlet | although right now yocto is not so happy with me because I didn't patch the existing galcore driver I move mine over to a new location. |
| 11:13 | jnettlet | trying to maintain my patches on top of the mess fsl was providing was just too time consuming. |
| 11:13 | jnettlet | and my driver works better :) |
| 11:14 | vpeter | I'm not surprised :) |
| 11:14 | suihkulokki | :P |
| 11:14 | jnettlet | vpeter, if you want to rebase your patches onto our tree i will gladly merge them. |
| 11:15 | jnettlet | like I said, right now only spdif is not working for some reason. |
| 11:15 | jnettlet | and I need to finish the TI hdmi decoder integration. |
| 11:22 | vpeter | Let me see how will they behave and when they merges I send PR. |
| 11:23 | jnettlet | great. If you do that then I will bring that into my test OE images |
| 11:23 | jnettlet | I still have to go through all the extra OE patches and figure out which ones we need now. |
| 11:25 | vpeter | Some are mines. |
| 11:26 | vpeter | And in few minutes I will see how they go. |
| 11:26 | jnettlet | yes I saw that |
| 11:27 | vpeter | It is nothing major. |
| 11:46 | vpeter | jnettlet: The only patch in question is linux-330-udoo-chipidea-clock-for-an-USB-phy-hub.patch. Not sure if this will be needed. |
| 11:47 | jnettlet | okay, I will take a look. |
| 11:47 | vpeter | And linux-334-udoo-lvds-display-resolution.patch I have to check where it belongs. Seems some things are different on this kernel. |
| 11:47 | vpeter | I will make one test image with only this kernel and see if all hw works for me. |
| 11:47 | jnettlet | yeah freescale changed a bunch of things. I just recently fixed a bunch of things that they broke. |
| 11:48 | jnettlet | spdif won't if udoo has that. |
| 11:48 | vpeter | doesn't have it. |
| 13:18 | topi` | jnettlet: I have already solved the SPL/GPT mismatch problem ;) I will compile my own GPT enabled U-BOOT and can then send patches to you |
| 13:18 | jnettlet | topi`, that would be great thanks. Do you have documentation as well? |
| 13:24 | topi` | jnettlet: it's just a nasty hack that slightly "bends" the GPT specification which states that GPT tables *ought* to start at sector #34 but I move them to #1024 and it seems to work with linux 3.14 |
| 13:24 | topi` | "seems to work" - when I add "gpt=force" to the kernel command line :) |
| 13:25 | topi` | rabeeh: do you know when you expect to ship the first Edge boards? |
| 13:26 | topi` | we should probably place an early order to get our foot in :) |
| 15:35 | topi` | jnettlet: btw is it possible to run the MX6Q cpus at 1.2GHz as opposed to 1.0 as they seem to run now? |
| 15:35 | topi` | being still thermally throttled, of course |
| 15:35 | jnettlet | topi`, our chips are only 1ghz socs. They are pinned down at the factory. |
| 15:35 | topi` | OK |
| 15:36 | topi` | they should have a Turbo mode like Intels ;) |
| 15:36 | topi` | however, for our purposes we just need to do a lot of work, so maybe just try to parallelize the code for 4 cpu cores |
| 15:40 | topi` | the A9 doesn't do well in Geekbench, but in some of my own benchmarks (like a Python bench) it's not much slower than a 1.7 GHz Krait 300 core |
| 15:40 | topi` | which is a bit surprising |
| 15:41 | topi` | the difference is very small |