nvidia / cudecomp Goto Github PK
View Code? Open in Web Editor NEWAn Adaptive Pencil Decomposition Library for NVIDIA GPUs
Home Page: https://nvidia.github.io/cuDecomp/
License: Other
An Adaptive Pencil Decomposition Library for NVIDIA GPUs
Home Page: https://nvidia.github.io/cuDecomp/
License: Other
Hello again,
I have been migrating to scikit-build-core for building and packaging.
After speaking with @henryiii , he suggested to not use CMAKE_INSTALL_PREFIX
since it causes issue for the scikit-build process.
He opened an issue scikit-build/scikit-build-core#786 but still recommended against using CMAKE_PREFIX_INSTALL
in the CMakeLists.txt
Hi, thanks for sharing this library. I'm trying to build it on the Dutch national HPC Snellius, but I have run into trouble with the compilation. The lib
stage of the Makefile
appears to complete without problems, but once it moves on to the tests
stage, it prints out many errors about not being able to find the libraries.
From the example config files, I believe I am pointing all the necessary variables to the right places, but the tests seem unable to find any of the NVIDIA libraries such as nccl
.
My config file is as below, and here is my log file from the make
command. Please let me know if you can see an obvious fix or have any suggestions.
# Having run
# module load 2022
# module load foss/2022a
# module load NVHPC/22.7
# NVHPC_HOME=/opt/nvidia/hpc_sdk/Linux_x86_64/2022
NVHPC_HOME=${EBROOTNVHPC}/Linux_x86_64/22.7
# Required variables to define
# MPICXX=mpicxx
# MPIF90=mpifort
CUDA_HOME=${NVHPC_HOME}/cuda
MPI_HOME=${NVHPC_HOME}/comm_libs/hpcx/latest/ompi
MPICXX=${MPI_HOME}/bin/mpicxx
MPIF90=${MPI_HOME}/bin/mpifort
NCCL_HOME=${NVHPC_HOME}/comm_libs/nccl
CUFFT_HOME=${NVHPC_HOME}/math_libs
CUTENSOR_HOME=${NVHPC_HOME}/math_libs
CUDACXX_HOME=${CUDA_HOME}
# Optional variables
CUDA_CC_LIST=61
BUILD_FORTRAN=1
ENABLE_NVTX=1
ENABLE_NVSHMEM=1
NVSHMEM_HOME=${NVHPC_HOME}/comm_libs/nvshmem
Hello cuDecomp team.
I am looking to package jaxDecomp into a pypi package (I am currently using cuDecomp as a CMake subdirectory)
Are you ok with me packaging the code or/and the binary of cuDecomp?
Thank you.
Hello,
I am trying to port and bind this library in Python
I was able to build it then link it to my library and use it from the library (Project is here : JaxDecomp )
But when I call it from Python via pybind11, I get this weird link error
cuDecomp/build/lib/libcudecomp.so: undefined symbol: __sync_val_compare_and_swap_4
I have tried to link libstdc++
, libatomic
, and libgcc_s
but nothing happend.
Do you have any idea why I am getting this error?
Following the removal of MPI, cutensor, and NCCL imports at the top of cudecomp.h in this commit
there is now an undefined reference to MPI_Comm
at this line of the include file:
Line 186 in e0cf871
This is a bit awkward because it forces the user to import manually MPI even in files that do not need to call the init function. Could it be possible to add back that MPI import?
First of all, thank you very much for adding a CMake build system in #15 .
I would like to point out that in your CMakeLists.txt
you set the compilers name (nvc++
) as the CXX compiler, while this works well when cuDecomp is the top project, it fails for us when we add cuDecomp as subdirectory with this error
"nvc++ " is not a full path and was not found in the PATH.
All you have to do is replace this code
# Use NVHPC compilers by default
set(CMAKE_CXX_COMPILER "nvc++")
set(CMAKE_Fortran_COMPILER "nvfortran")
# Locate and use NVHPC CMake configuration
find_program(NVHPC_CXX_BIN "nvc++")
With this one
# find NVHPC compilers
find_program(NVHPC_CXX_BIN "nvc++")
find_program(NVHPC_Fortran_BIN "nvfortran")
# Use NVHPC compilers by default
set(CMAKE_CXX_COMPILER ${NVHPC_CXX_BIN})
set(CMAKE_Fortran_COMPILER ${NVHPC_Fortran_BIN})
I would be happy to make PR if you want.
Thank you.
error message:
CMake Error at CMakeLists.txt:5 (find_package):
By not providing "FindNVHPC.cmake" in CMAKE_MODULE_PATH this project has
asked CMake to find a package configuration file provided by "NVHPC", but
CMake did not find one.
Could not find a package configuration file provided by "NVHPC" with any of
the following names:
NVHPCConfig.cmake
nvhpc-config.cmake
Add the installation prefix of "NVHPC" to CMAKE_PREFIX_PATH or set
"NVHPC_DIR" to a directory containing one of the above files. If "NVHPC"
provides a separate development package or SDK, be sure it has been
installed.
-- Configuring incomplete, errors occurred!
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.