Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Support for H100 and sm_90 #265

Open
maueki opened this issue Jul 18, 2024 · 2 comments
Open

Support for H100 and sm_90 #265

maueki opened this issue Jul 18, 2024 · 2 comments

Comments

@maueki
Copy link

maueki commented Jul 18, 2024

Hi,

Does this project work on the H100?
It seems that libspconv.so does not support sm_90. Is it possible to get a binary that supports it?

$ cuobjdump --list-elf libspconv.so
ELF file    1: libspconv.1.sm_80.cubin
ELF file    2: libspconv.2.sm_86.cubin
ELF file    3: libspconv.3.sm_80.cubin
ELF file    4: libspconv.4.sm_86.cubin
ELF file    5: libspconv.5.sm_80.cubin
ELF file    6: libspconv.6.sm_86.cubin
ELF file    7: libspconv.7.sm_80.cubin
ELF file    8: libspconv.8.sm_86.cubin
ELF file    9: libspconv.9.sm_80.cubin
ELF file   10: libspconv.10.sm_86.cubin

Thank you.

@Icecream-blue-sky
Copy link

Same problem. Current libspconv.so doesn't support H100 sm_90:
-- Configuring done (0.1s)
-- Generating done (0.1s)
-- Build files have been written to: /perception/icecream/code/Lidar_AI_Solution-master/CUDA-BEVFusion/build
[ 3%] Linking CXX shared library libcustom_layernorm.so
[ 7%] Linking CXX shared library libbevfusion_core.so
[ 11%] Built target custom_layernorm
[ 70%] Built target bevfusion_core
[ 74%] Linking CXX executable bevfusion
[100%] Built target bevfusion
Create by resnet50int8, int8
CUDA Runtime error init_quantization_scaleFeaturesType::INT8<<<grid, block, 0, stream>>>( input_scale, output_scale, weight, prod_DHWI, int8weight_.ptr(), weight_scales_.ptr(), alpha_.ptr(), (half *)bias, bias_scaled_.ptr(), numel) # no kernel image is available for execution on the device, code = cudaErrorNoKernelImageForDevice [ 209 ] in file src/spconv/implicit-gemm.cu:241
tool/run.sh: line 44: 2057630 Aborted (core dumped) ./build/bevfusion $DEBUG_DATA $DEBUG_MODEL $DEBUG_PRECISION

@hopef
Copy link
Collaborator

hopef commented Sep 11, 2024

Hi,

libspconv with SM90 has been supported.

$> cuobjdump --list-elf Lidar_AI_Solution/libraries/3DSparseConvolution/libspconv/lib/x86_64_cuda12.6/libspconv.so
ELF file    1: libspconv.1.sm_80.cubin
ELF file    2: libspconv.2.sm_86.cubin
ELF file    3: libspconv.3.sm_90.cubin
ELF file    4: libspconv.4.sm_80.cubin
ELF file    5: libspconv.5.sm_86.cubin
ELF file    6: libspconv.6.sm_90.cubin
ELF file    7: libspconv.7.sm_80.cubin
ELF file    8: libspconv.8.sm_86.cubin
ELF file    9: libspconv.9.sm_90.cubin
ELF file   10: libspconv.10.sm_80.cubin
ELF file   11: libspconv.11.sm_86.cubin
ELF file   12: libspconv.12.sm_90.cubin
ELF file   13: libspconv.13.sm_80.cubin
ELF file   14: libspconv.14.sm_86.cubin
ELF file   15: libspconv.15.sm_90.cubin

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants