deepstream-tlt-apps: add recipe #4
Conversation
| inherit pkgconfig cuda | ||
| COMPATIBLE_MACHINE = "(tegra)" | ||
|
|
||
| TARGET_LDFLAGS += " -L${RECIPE_SYSROOT}/opt/nvidia/deepstream/deepstream-${DS_VERSION}/lib" |
There was a problem hiding this comment.
I'll post an alternative patch for the makefiles that works better and doesn't require you to add anything the LDFLAGS.
| TARGET_LDFLAGS += " -L${RECIPE_SYSROOT}/opt/nvidia/deepstream/deepstream-${DS_VERSION}/lib" | ||
| EXTRA_OEMAKE += " CC='${CXX}' CUDA_VER=${CUDA_VERSION} DS_VER=${DS_VERSION} TARGET_DEVICE=${TARGET_ARCH}" | ||
|
|
||
| TARGET_CC_ARCH += "${LDFLAGS}" |
There was a problem hiding this comment.
My alternative makefile patches also eliminate the need to modify TARGET_CC_ARCH.
| COMPATIBLE_MACHINE = "(tegra)" | ||
|
|
||
| TARGET_LDFLAGS += " -L${RECIPE_SYSROOT}/opt/nvidia/deepstream/deepstream-${DS_VERSION}/lib" | ||
| EXTRA_OEMAKE += " CC='${CXX}' CUDA_VER=${CUDA_VERSION} DS_VER=${DS_VERSION} TARGET_DEVICE=${TARGET_ARCH}" |
There was a problem hiding this comment.
The += operator automatically inserts a blank for you, so you don't need to put one at the beginning of the string.
|
I took a closer look at the build, and found a few more issues. Here's an alternative patch to use: With this patch, no more compiler, linker, or packaging warnings, as the right flags and paths get used throughout. And don't worry about the automated check failure, it was failing because I hadn't yet back-ported some fixes to the 32.4.3 branch from later branches. I've done that now, so rebase to the latest on the branch and the automated check should pass. |
This commit adds a recipe that builds several deepstream tlt sample apps as well as the libnvds_infercustomparser_tlt.so custom parser library from NVIDIA's https://github.com/NVIDIA-AI-IOT/deepstream_tao_apps repository. Signed-off-by: Bryan Cisneros <[email protected]>
bcis93
force-pushed
the
EDGE-600-ds-tlt-apps
branch
from
13a26b6
to
bdd4cc7
Compare
|
Thanks @madisongh for the new patch and the feedback. I made all the requested changes, so take a look again and let me know if there's anything else! |
This PR adds a recipe that builds several deepstream tlt sample apps as well as the libnvds_infercustomparser_tlt.so custom parser library from NVIDIA's deepstream_tao_apps repository.
I separated the samples and custom parser into separate packages so that the custom parser can be built by itself if desired.
I'm targeting the dunfell-l4t-r32.4.3 branch since that's what my image is based on, but I can also open another PR targeting the master branch once this one gets approved.