OpenCL driver issue with Samsung Chromebook


I had an issue with the OpenCL Driver on my Chomebook.

I followed ARM's instruction and successfully boot into Linux. However, I got an error message said that missing. After install libxcb-dri2.0-dev, I can compile Mali OpenCL SDK sample codes. But they won't be executed correctly under X11 or via fbdev. I past error message below. Any suggestion? Thanks.

> ./hello_world_opencl


Creating an OpenCL context failed. common.cpp:248

Failed to create an OpenCL context. hello_world_opencl.cpp:39

> ldd hello_world_opencl => /opt/mali/fbdev/ (0x76efc000) => /usr/lib/arm-linux-gnueabihf/ (0x76e4a000) => /lib/arm-linux-gnueabihf/ (0x76e39000) => /lib/arm-linux-gnueabihf/ (0x76d55000)

        /lib/ (0x76f07000) => /opt/mali/fbdev/ (0x76304000) => /lib/arm-linux-gnueabihf/ (0x762e9000) => /lib/arm-linux-gnueabihf/ (0x7627e000) => /lib/arm-linux-gnueabihf/ (0x76270000) => /lib/arm-linux-gnueabihf/ (0x76265000)

Parents Reply Children
  • Hmmm interesting.  I'm using chrubuntu on the internal mmc-- boots fine w/graphics, sound, etc. though it's not stable enough to test much.  Some quick notes for anyone trying it on chrubuntu...

    • use mmcblk0p7 in the kernel command line instead of the mmcblk1 setting.  dd the resulting kernel .img to /dev/mmcblk0p6
    • for simplicity, I built xf86-video-armsoc right in chrubuntu-- no qemu needed.
    • the uboot-mkimage package is now u-boot-tools
    • manually had to add the built kernel modules to /lib/modules/3.8.11...
    • manually put the xorg libs in /lib/arm-linux-gnueabihf/  Xorg.0.log SEEMS to show them being picked up...  once I have network I'll paste it somewhere.
    • wifi doesn't work as described above, though I did have it working for a second somehow.  At least I saw the SSID pop up and say 'connnected' before everything crashed.  Problems I found so far wifi have included the regulatory domain not being set as well as the wifi modules not loading.  I need the lib80211_crypt_tkip one for example.

    I gotta figure this out too:

    • Really annoying issue:  kernel oopses in dw_mci_pre_dma_transfer.isra.13 that apparently takes out the mmc at seemingly random times.

    Any thoughts, especially on the kernel crasher? Unfortunately I haven't had enough time to track anything down- doing this in pieces when I get random time.