| 01:22 | _rmk_ | yay, agreement reins on my point :) |
| 09:56 | jnettlet | Interesting enough, galcore running under 3.0.35 kernel nice and fast no problems. galcore running under 3.10 kernel, almost 3 times slower. |
| 09:56 | jnettlet | I have to see if some sort of scaling is not happening. |
| 09:59 | hste_ | jnettlett: Do you use the latest vivante-libs? |
| 10:00 | jnettlet | hste_, yep both kernels are using the same userspace |
| 10:02 | hste_ | jnettlet: yocto master had some new libs that worked better on glmark2 than the dora version on my gk802 (imx6q) |
| 10:03 | hste_ | but I have only used 3.0.35... |
| 10:05 | jnettlet | this is definitely something to do with the kernel. The way freescale moves code between branches is a mess though so bisecting what caused it may be impossible. |
| 14:37 | jnettlet | _rmk_, ping |
| 14:56 | _rmk_ | hmm? |
| 15:00 | jnettlet | if you have a couple of seconds I was wondering your opinion on a theory. |
| 15:01 | jnettlet | I am tracking down this big performance slowdown between the fsl 3.0.34 and 3.10 kernels in regards to galcore. |
| 15:02 | jnettlet | The main difference between the two kernels is galcore uses CMA with 3.10 and the old DMA_ZONE_CONFIG with 3.0.35 |
| 15:04 | jnettlet | obviously the memory is still allocated through the dma_alloc_ interfaces, but is there anything you can think of as to why the memory allocation method would affect performance? |
| 15:08 | _rmk_ | is it the allocation which is slowing it down, or is it the use of the memory? |
| 15:08 | jnettlet | the use of the memory. |
| 15:09 | jnettlet | I am going through the driver diffs again more closely to see if I missed anything. But under CMA I am not even seeing some things flushed to the framebuffer properly. |
| 15:10 | _rmk_ | nothing springs to mind |
| 15:11 | jnettlet | okay I will keep digging. |
| 15:20 | hste_ | When running glmark2 in xorg with 1080P the Screens goes black and X have to be restarted: mxc_ipu mxc_ipu: IPU Error - IPU_INT_STAT_10 = 0x0008000, at 720P it runs without problem |
| 15:21 | _rmk_ | hste: black or loses sync? |
| 15:22 | hste_ | goes black. No hdmi signal |
| 15:22 | hste_ | I have seen someone had similar problems running xbmc |
| 15:23 | _rmk_ | so both then, and not black _with_ sync. |
| 15:25 | _rmk_ | the distinction is important, because it's the difference between the IPU still being alive but being fed black vs the IPU being dead |
| 15:30 | hste_ | I found this older commit http://git.freescale.com/git/cgit.cgi/imx/linux-2.6-imx.git/commit/?h=imx_2.6.35_11.09.01&id=bc8159027063aded47316b669b1293998337b2ee |
| 15:34 | rabeeh | hste_: those are ipu under runs |
| 15:34 | rabeeh | same like xbmc |
| 15:34 | rabeeh | i have a fix; checked it with geexbox team and needs some refinement |
| 15:35 | tazou | hi rabeeh do you received my postal card ? :) |
| 15:35 | rabeeh | tazou: hey |
| 15:36 | rabeeh | hmm... no one from the team told me; so it's probably on the way :) |
| 15:36 | tazou | h?h? ok |
| 15:36 | rabeeh | thanks a lot for it; i'm anticipating when it will arrive |
| 15:36 | tazou | with plesure ;) |
| 21:07 | jas-hacks_ | rabeeh: On C1 is the 'cd' pin wired for the micro sd card? |
| 21:43 | jnettlet | jas-hacks_, cd is wp isn't |
| 21:43 | jas-hacks_ | no wp (write protect) , cd is card detect |
| 21:45 | jnettlet | cd is wired, wp isn't |
| 21:45 | jas-hacks_ | thxs |
| 21:51 | cyball | hi @all ... I'm trying to boot my recently delivered board :-) .. are there some instructions how to get u-boot running ? thx :-) |
| 21:51 | jnettlet | cyball, http://imx.solid-run.com/wiki/index.php?title=Building_Carrier-One_U-boot_and_Kernel |
| 21:52 | cyball | thx |
| 22:21 | cyball | can I boot the board with the msata connector? |
| 22:22 | cyball | or do i need to put all the stuff on the microsd? |
| 22:47 | cyball | ok there is no mSATA on my board :D |
| 23:39 | mhoney | you mean eSata? |
| 23:41 | cyball | mhoney, no i mean mSata there (http://imx.solid-run.com/wiki/index.php?title=Carrier-One_Hardware#Detailed_hardware_specification) should be 3 different ifaces on the back of the board but there is only the microSD slot :-) |
| 23:46 | mhoney | Italic lines below are features not populated in the first batch for developers. |
| 23:47 | dv | oh my |
| 23:47 | dv | this is retarded |
| 23:47 | dv | apparently, if source and destination of an IPU blit task are identical (except for the physical addresses of course), in other word, if I just want to do a 1:1 pixel copy, the ipu device sees this as an error |
| 23:49 | mhoney | (insert a witty response here) |