| 00:04 | otavio | dv_: this is the mirror |
| 00:04 | dv_ | well I'll git fetch manually now |
| 00:05 | dv_ | I've had some problems with kernel fetches in the past with oe/yocto in general |
| 00:05 | dv_ | I wonder if it were possible to clone only specific branches. that should be the default. |
| 00:16 | dv_ | otavio: either way , it is broken |
| 00:58 | jnettlet | dv_, you can use the --single-branch flag with git clone to do that. |
| 00:58 | dv_ | jnettlet: yeah, but OE doesnt use it |
| 14:06 | jnettlet | Anyone else tested the glmark-es2 benchs on the IMX hardware? |
| 14:06 | jnettlet | glmark2-es2 |
| 14:10 | jnettlet | I wish I had a reference. I am producing glmark2 scores close to what I see posted for the GC2000 |
| 14:10 | jnettlet | they shouldn't even be close. |
| 14:11 | jnettlet | it could be due to differing resolutions. |
| 14:19 | jnettlet | vivante GC880 (less than 1/4 of the GC2000 performances) can't wait to test the GC2000 with my changes. |
| 14:24 | dv | 14:24 * dv_ wonders what the results would be with etnaviv |
| 14:25 | jnettlet | wumpus, ping |
| 14:26 | wumpus | pong |
| 14:27 | jnettlet | hey, have you done any benchmarking of etnaviv using the latest vivante release from freescale? |
| 14:27 | wumpus | not really |
| 14:27 | wumpus | did they optimize the kernel driver? |
| 14:28 | jnettlet | nope I did over the last few games. |
| 14:28 | jnettlet | gtkperf was about a 25% improvement in 2d |
| 14:29 | jnettlet | just started testing 3d and I am getting glmark2-es numbers that are close to what I am seeing posted for GC2000 chips. |
| 14:29 | wumpus | nice :) |
| 14:29 | jnettlet | and it is really just testing the GPU because CPU usage running glmark2-es is about 10% user 10% system |
| 14:30 | jnettlet | my glmark2 fullscreen score is around 220 |
| 14:31 | jnettlet | although that could be the resolution. For some reason it is defaulting to 1024x768 instead of 720p need to figure out why |
| 14:33 | wumpus | in any case, if there is still so much room for optimization in the kernel driver, I wonder what we could do with a proper drm driver |
| 14:35 | jnettlet | I think there is still room for improvement. A lot of the problems were just lock contention trying to do powermanagement. |
| 14:35 | hste_ | jnettlet: I can test it when I get home. on my gk802. On what resolution do u want me to test |
| 14:36 | jnettlet | hste_, 640x480 and 720p if you could |
| 14:36 | jnettlet | I guess 1080p numbers would be useful as well |
| 14:37 | ssvb | jnettlet: are you testing gtkperf and glmark2-es2 on a marvell or imx hardware? |
| 14:38 | jnettlet | ssvb, this is on imx right now, but I am bouncing between the two |
| 14:38 | hste_ | jnettlet: Ihad about 330 on glmark2-es2 on 1080p on my gk802 running ubuntu with jas-hacks_ rootfs |
| 14:38 | jnettlet | hste_, what is in the gk802? |
| 14:38 | jnettlet | SOC I mean |
| 14:39 | hste_ | jnettlet: its a imx6q dongle |
| 14:39 | jnettlet | hste_, okay let me get 1080p running on this C1 then |
| 14:41 | jnettle | 14:41 * jnettlet thinks the 32-bit 400mhz memory is going to be the bottleneck. |
| 14:41 | jnettlet | will have to test with rabeeh's overclocking |
| 14:44 | hste_ | jnettlet: If I get time this evening I'll test both c1 and gk802 |
| 14:44 | jnettlet | thanks |
| 17:18 | hste | dv_: how is going with gstreamer-imx? |
| 17:21 | jnettlet | hste_, are you running xbmc on any of your IMX6 devices? |
| 17:22 | hste | jnettlet: yes just for testing |
| 17:23 | hste | I use stephans rootfs on my gk802 |
| 17:23 | jnettlet | hste_, did you compile it yourself? |
| 17:23 | jnettlet | ah okay. |
| 17:23 | rabeeh | hste: which build are you using? |
| 17:23 | hste | the latest from his blog |
| 17:25 | hste | rabeeh: http://stende.no-ip.info/files/xbmc-image-wandboard-quad.tar.bz2 |
| 17:28 | hste | http://stephan-rafin.net/blog/2013/11/08/new-xbmc-yocto-image/ |
| 17:31 | jnettlet | hste_, so bumping to 1080p drops my glmark2 score down to 100 with cpu usage hovering around 30% |
| 17:32 | hste | jnettlet: on gk802 i got 208 on glmark2-es2 and 184 on glmark2 |
| 17:32 | jnettlet | at 1080p? |
| 17:32 | hste | yes |
| 17:33 | jnettlet | then we are seeing a good performance boost if the GC2000 is 4x the 880 |
| 17:34 | jnettlet | glmark2 on the C1 is no where close because a lot of GLES->GL calls requires more CPU |
| 17:34 | hste | :) but I guess in realworld it will be less |
| 17:35 | jnettlet | I think 1080p on the IMX6 solo is a stretch but lets see what we can do |
| 17:37 | hste | ok. got to go. I'll be back later. |
| 17:37 | jnettlet | ttyl |
| 17:39 | jnettlet | rabeeh, do you have an xbmc install that can run on an Ubuntu rootfs? |
| 17:59 | MarcusVinter | Hey Rabeeh. Just a question about the C1 image you sent me. By default does it output on hdmi or serial? |
| 18:00 | MarcusVinter | It was the Ubuntu 13.04 one. |
| 18:29 | rabeeh | MarcusVinter: serial port and after booting you should see the hdmi |
| 18:44 | MarcusVinter | Thanks rabeeh. And one more thing, we are expecting delays on the engineering sample due to the manufacturing delays. Are we looking at the middle of january still? |
| 18:55 | dv_ | hste: currently I am focusing on the eglvivsink |
| 18:56 | dv_ | its still that adapted eglglessink code, and could be considerably smaller and cleaner |
| 19:49 | rabeeh | MarcusVinter: ofcourse not |
| 19:49 | rabeeh | only who is ordering now might have some delay of the i2ultra model to mid Jan |
| 19:49 | rabeeh | MarcusVinter left |
| 20:46 | hste | jnettlet: found more code to fix? |
| 20:56 | jnettlet | hste, nope just trying to get xbmc built on ubuntu to see how it runs. right now the benchmarks appear to be doing what they are supposed to and just testing the GPU. |
| 20:56 | jnettlet | Even at 1080p I am not seeing more than 30% cpu usage when running glmark2-es |
| 20:58 | jnettlet | overall it is not so bad. With xrender compositing enabled I am still getting 70% performance numbers from GLES benchmarks |
| 20:59 | jnettlet | that can definitely be improved |
| 22:05 | jas-hacks_ | dv_: Hi, tested again on 1.2 the videos play fine :) |
| 22:06 | dv_ | nice! so 1.2 is a requirement |
| 22:09 | jas-hacks_ | dv_: possibly, my resolution is 720p and videos are 720p but they seem to launch in smaller window |
| 22:10 | dv_ | you are using the eglvivsink ? |
| 22:17 | jas-hacks_ | dv_: I set 'video-sink=eglvivsink' |
| 22:18 | dv_ | yeah, then the window frame will occupy screen space and therefore the video itself will be smaller |
| 22:18 | dv_ | I am working on it |
| 22:18 | dv_ | note though that typically applications embed the sink's window |
| 22:19 | jas-hacks_ | dv_: wonder if it works from totem |
| 22:19 | dv_ | I never tried it |
| 22:20 | jas-hacks_ | Let me try, I'm testing on ubuntu 13.10 |
| 22:22 | dv_ | its also been a while since I tried out seeking |
| 22:22 | dv_ | back then it worked though |
| 22:32 | jas-hacks_ | totem reverts to sw decoding |
| 22:46 | dv_ | jas-hacks_: hmm I'll build an image with totem and try it out myself |
| 22:48 | jas-hacks_ | dv_: I can upload a rootfs if required |
| 22:49 | dv_ | not needed - i can build it quickly here, I am anyway working at the eglvivsink atm, so building it will not block me |
| 22:52 | dv_ | oh, nevermind. OE does not have recipes for totem. |
| 22:52 | dv_ | so yes, upload please. |
| 23:02 | _rmk_ | so, do we have the new name for the carrier-1 ? |
| 23:28 | _rmk_ | okay, I think I'm just going to submit this stuff with it called "carrier-1" and be done with it |
| 23:29 | _rmk_ | I'm assuming that the renaming of it isn't happening anymore; the last update appears to be 14 December, almost a month old now. |