> On Aug 6, 2015, at 11:13 AM, Dalton Fury <daltonfur...@gmail.com> wrote: > > I see, but why boot from SD card, which is slower when there is eMMC?
easy to flash and wanted to keep ubuntu around. > > Anyways, I flashed it successfully. But now I have no clue as to how to > enable CUDA. > > I want to do some image processing with OpenCV/CUDA. What I tried is to > compile opencv with USE_CUDA=ON flage set in cmake. Then I tested to see if I > an use the gpu with opencv. But can't, and get 0 when I call > gpu::getCudaEnabledDeviceCount(). This clearly indiacates that cuda is not > installed. So how do I install CUDA? We use precompiled drivers package from nvidia for this. Do you have /etc/ld.so.conf on target ? if not create one with following content /usr/lib/arm-linux-gnueabihf/tegra/ and ensure that libcuda.so is there on target > > On Thu, Aug 6, 2015 at 9:29 AM, Khem Raj <raj.k...@gmail.com > <mailto:raj.k...@gmail.com>> wrote: > >> On Aug 3, 2015, at 11:15 AM, Dalton Fury <daltonfur...@gmail.com >> <mailto:daltonfur...@gmail.com>> wrote: >> >> Hi, >> >> I am a hobbyist, and liked yocto very much since I used in on an old >> Pandaboard ES. Now I got my hands on a Nvidia Jetson TK1, and want to get a >> yocto running on it. I intend to use it for heavy image processing and I >> prefer not to do it with the Ubuntu which comes with it loaded with GUI and >> Unity and all. So I build the image, but could not find any information on >> how to flash it to the eMMC on the board. >> >> Nvidia provides scripts for doing this. The process is as follows: >> >> 1) Download the driver package and extract it. >> >> 2) Download the sample rootfs and extract it to `rootfs` folder in where the >> driver package is extracted. >> >> 3) Run the apply_binaries.sh script. Upon inspecting the script, I found >> that it copies NVIDIA user space components, Nvidia firmware files, BSP test >> tools, gst test applications, create a lot of symbolic links within the >> rootfs extracted to the 'rootfs' folder. It also places the firmwares and >> kernel modules, and also copies some configuration files for the boot >> loader(U-Boot). >> >> 4) Run flash.sh. >> >> What I did is to copy the rootfs that was made with yocto, and entirely skip >> Step 3 assuming that all the necessary software is already made with yocto >> as a result of adding the BSP. This attempt of course, failed. Thought the >> kernel booted, it froze halfway. No error message. Just a flack screen with >> a while "_" on the top left corner. >> >> This is the bsp I used: https://github.com/kraj/meta-jetson-tk1 >> <https://github.com/kraj/meta-jetson-tk1> >> >> This is the guide I used as reference: >> http://elinux.org/Tegra/Downstream_SW/Gentoo_From_eMMC >> <http://elinux.org/Tegra/Downstream_SW/Gentoo_From_eMMC> >> >> If anyone has experience performing this(i bet you do), please help me out. >> Thanks. > > I have booted Angstrom XFCE-nm-image on it from SD card. Firstly you need to > flash new kernel and also modify the boot loader arguments so it boots from > SD card first. > all instructions to update the boot loader cmdline and kernel are documented > > see > https://cyclicredundancy.wordpress.com/2014/05/10/flashing-the-rootfs-on-a-nvidia-jetson-tk1/ > > <https://cyclicredundancy.wordpress.com/2014/05/10/flashing-the-rootfs-on-a-nvidia-jetson-tk1/> > > you might need > > sudo ./flash.sh -S 14580MiB jetson-tk1 mmcblk0p1 > > >> -- >> _______________________________________________ >> yocto mailing list >> yocto@yoctoproject.org <mailto:yocto@yoctoproject.org> >> https://lists.yoctoproject.org/listinfo/yocto >> <https://lists.yoctoproject.org/listinfo/yocto> > >
signature.asc
Description: Message signed with OpenPGP using GPGMail
-- _______________________________________________ yocto mailing list yocto@yoctoproject.org https://lists.yoctoproject.org/listinfo/yocto
