scottk/meta-tegra
1
0
Fork 0
mirror of https://github.com/lingble/meta-tegra.git synced 2025-10-29 19:42:41 +00:00
Code Issues Packages Projects Releases Wiki Activity

README: update for 32.1.0 work-in-prorgress

Browse Source 
Signed-off-by: Matt Madison <matt@madison.systems>
This commit is contained in:
Matt Madison
2019-03-24 15:03:45 -07:00
parent 4808280284
commit 5af2e7bdce
1 changed files with 47 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

55
README
View File

@@ -1,30 +1,69 @@
OpenEmbedded/Yocto BSP layer for NVIDIA Tegra X1/X2/K1
======================================================
OpenEmbedded/Yocto BSP layer for NVIDIA Tegra X1/X2/AGX/K1
==========================================================
**NOTE** this is a 'work-in-progress' branch. Please do not
rely on the presence of this branch, it will be deleted once
the work is completed and changes have been merged.
Boards supported:
* Jetson-TK1 development kit (Linux4Tegra R21.7)
* Jetson-TX1 development kit (Linux4Tegra R28.2, JetPack 3.3)
* Jetson-TX2 development kit (Linux4Tegra R28.2.1, JetPack 3.3)
* Jetson-TX1 development kit (Linux4Tegra <tbd>)
* Jetson-TX2 development kit (Linux4Tegra R32.1, JetPack 4.2)
* Jetson AGX Xavier development kit (Linux4Tegra R32.1, JetPack 4.2)
* Jetson Nano development kit (Linux4Tegra R32.1, JetPack 4.2)
This layer depends on:
URI: git://git.openembedded.org/openembedded-core
branch: master
Current WIP Status
------------------
* Recipes have been updated for L4T R32.1.0.
* Jetson-TX2: passes basic tests with core-image-sato
(camera capture, video playback, Vulkan, GLX, CUDA).
Weston image boots but does not start the compositor.
* Jetson AGX Xavier: still to be tested
* Jetson Nano: still to be tested
* Jetson-TX1: not actually supported by NVIDIA with 32.1,
but it may work. If not, will need to figure out how to
continue to support it with R28.2 in parallel with Nano
(also tegra210) using R32.1.
Limitations
-----------
* CUDA 9 supports up through gcc 6 only, so if you intend
to build CUDA applications, stick with gcc 6.
* Starting with JetPack 4.2, packages outside the L4T BSP cannot
be downloaded from NVIDIA's web site without being logged in
to a developer network account. So to use CUDA 10, cuDNN, and
any other packages that require a devnet login, you must first
create an devnet account and install the NVIDIA SDK Manager
on your host so you can download the packages you need for
your builds. Once downloaded, you can set the variable
NVIDIA_DEVNET_MIRROR to "file://path/to/the/downloads" in
your local.conf file to make them available to your bitbake
builds.
* CUDA 10 supports up through gcc 7 only, and some NVIDIA-provided
binary libraries appear to be compiled with g++ 7 and cause linker
failures when building applications with g++ 6, so it looks
Selecting the toolchain version
-------------------------------
Toolchain version selection is usually a distro configuration setting,
but you can also set this in your build/conf/local.conf file. To use
gcc 6 instead of gcc 7, set:
gcc 7 instead of gcc 8, set:
GCCVERSION = "6.%"
GCCVERSION = "7.%"
but you will also need the gcc 7 toolchain recipes in one of your layers,
since it was retired from OE-Core in favor of gcc 8.
Contributing