Skip to content

feature: NVIDIA Jetson Thor jetpack 7 support#7

Merged
amadeuszsz merged 2 commits into
autowarefoundation:mainfrom
pixmoving-auto:feature/thor-jp7
Feb 13, 2026
Merged

feature: NVIDIA Jetson Thor jetpack 7 support#7
amadeuszsz merged 2 commits into
autowarefoundation:mainfrom
pixmoving-auto:feature/thor-jp7

Conversation

@ahuazuipiaoliang
Copy link
Copy Markdown
Member

Description

support for NVIDIA Jetson Thor jetpack 7.

environment

OS: Ubuntu 24.04
JetPack: 7.0
g++: (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0
nvcc: Cuda compilation tools, release 13.0, V13.0.48/Build cuda_13.0.r13.0/compiler.36260728_0

@ahuazuipiaoliang ahuazuipiaoliang self-assigned this Feb 12, 2026
@ahuazuipiaoliang ahuazuipiaoliang added the enhancement New feature or request label Feb 12, 2026
@ahuazuipiaoliang
Copy link
Copy Markdown
Member Author

@amadeuszsz this branch may not be compatible with other platforms, please create a new branch for Jetson Thor/Jetpack 7.

Copy link
Copy Markdown
Contributor

@amadeuszsz amadeuszsz left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Hi @ahuazuipiaoliang
Thank you for your contribution!

I think CUDA_STANDARD and CXX_STANDARD are ok, but for sake of easier maintenance, could you try to apply similar "trick" as I did in e.g. autowarefoundation/autoware_universe#12053? For the instance:

set(GENCODE_FLAGS 
    -gencode=arch=compute_75,code=sm_75
    -gencode=arch=compute_80,code=sm_80
    -gencode=arch=compute_86,code=sm_86
    -gencode=arch=compute_87,code=sm_87
    -gencode=arch=compute_89,code=sm_89
    -gencode=arch=compute_90,code=sm_90
    -gencode=arch=compute_100,code=sm_100
    -gencode=arch=compute_120,code=sm_120
)

if(CMAKE_CUDA_COMPILER_VERSION VERSION_LESS "13.0")
    list(APPEND GENCODE_FLAGS -gencode=arch=compute_101,code=sm_101)
else()  # CUDA 13.0 renamed SM101 to SM110
    list(APPEND GENCODE_FLAGS -gencode=arch=110,code=110)
endif()

target_compile_options(spconvlib_cumm_conv_main_Ampere_f16f16f16f16f16tnt_m128n64k32m64n32k32A1T1688_400_C301LLL_SK_ConvKernel PRIVATE $<$<COMPILE_LANGUAGE:CUDA>:
     -DTV_CUDA
    ${GENCODE_FLAGS}
    -Xcudafe "--diag_suppress=implicit_return_from_non_void_function"
    --expt-relaxed-constexpr
    -Xcompiler="-fPIC"
    -Xcompiler='-O3'>)

I didn't test it, so please make sure used flag (e.g.: CMAKE_CUDA_COMPILER_VERSION or CUDA_VERSION) actually stores any value. I usually check via message() call in CMakeLists.

After merging this PR, I will update docs and release +cu130.

@ahuazuipiaoliang
Copy link
Copy Markdown
Member Author

ahuazuipiaoliang commented Feb 12, 2026

@amadeuszsz I've modified CMakeLists for the compatibility, and successfully compiled on Thor/JetPack 7/CUDA 13.0.

Copy link
Copy Markdown
Contributor

@amadeuszsz amadeuszsz left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM!

Thank you for your work! In separate PR I will make a tiny adjustment and then release new binaries.

@amadeuszsz amadeuszsz merged commit 111ffdf into autowarefoundation:main Feb 13, 2026
1 check failed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

enhancement New feature or request

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants