-
Notifications
You must be signed in to change notification settings - Fork 84
Build on NVIDIA platform
Install CUDA
Debian repository - apt Add the ROCm apt repository
For Debian based systems, like Ubuntu, configure the Debian ROCm repository as follows:
wget -qO - http://repo.radeon.com/rocm/apt/debian/rocm.gpg.key | sudo apt-key add -
sudo sh -c 'echo deb [arch=amd64] http://repo.radeon.com/rocm/apt/debian/ xenial main > /etc/apt/sources.list.d/rocm.list'
sudo apt-get update
sudo apt-get install hip_nvcc
Notice: DO NOT sudo apt-get install rocm
See here for details.
cd /opt/rocm/hip/bin
./hipconfig
It will shows
== hipconfig
HIP_PATH : /opt/rocm/hip/bin
HIP_PLATFORM : nvcc
CPP_CONFIG : -D__HIP_PLATFORM_NVCC__= -I/opt/rocm/hip/bin/include -I/usr/local/cuda/include
add nvcc and /opt/rocm/hip/bin
in your environment variable
Unfortunately, the install.sh has bug and cannot run on NVIDIA platform. You need to manually build rocFFT.
mkdir -p [ROCFFT_BUILD_DIR]/release
cd [ROCFFT_BUILD_DIR]/release
# Default install location is in /opt/rocm, define -DCMAKE_INSTALL_PREFIX=<path> to specify other
# Default build config is 'Release', define -DCMAKE_BUILD_TYPE=<config> to specify other
CXX=/opt/rocm/bin/hipcc ccmake [ROCFFT_SOURCE]
make -j$(nproc)
sudo make install # sudo required if installing into system directory such as /opt/rocm