CuPBoP-AMD: Extending CUDA to AMD Platforms

Copyright © 2023

Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at

http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.

Introduction

CuPBoP-AMD (Extending CUDA to AMD Platforms) is a extension of the framework of CuPBoP following similar architecture. However, CuPBoP-AMD has its over version of kernerl and host translators and runtime implementation. CuPBoP-AMD currently supports many of the Rodinia benchmarks and more support than AMD HIPIFY.

• SC23-W - CuPBoP-AMD: Extending CUDA to AMD Platforms
• https://doi.org/10.1145/3624062.3624185

Install

Prerequisites

• Linux system
• LLVM 16.0.1
• CUDA Toolkit

Installation

1. Clone from github

   git clone --recursive git@github.com:gthparch/CuPBoP-AMD.git
   cd CuPBoP-AMD

2. Build LLVM using the llvm-16-init branch and the ROCm suites

   Edit the build-rocm.sh script under the scripts/ folder, replace BUILD_LLVM=0 with BUILD_LLVM=1.

   ./scripts/build-rocm.sh

   Ensure the compiled LLVM binaries are in the PATH variable.

3. Build CuPBoP-AMD

   mkdir build && cd build
   cmake .. \
      -DLLVM_CONFIG_PATH=`which llvm-config` \
      -DCUDA_PATH=$CUDA_PATH
   cmake --build . -j$(nproc)

4. Set the environment variables

   export ROCM_PATH="<path_to_rocm>"
   export HIP_PATH="$ROCM_PATH/hip"
   export DEVICE_LIB_PATH="$ROCM_PATH/amdgcn/bitcode"
   export HIP_DEVICE_LIB_PATH="$ROCM_PATH/amdgcn/bitcode"
   export HIP_CLANG_PATH="<path_to_compiled_llvm>"
   export HIPCC_COMPILE_FLAGS_APPEND="--rocm-path=$ROCM_PATH"
   export CMAKE_PREFIX_PATH="$ROCM_PATH/hip:$ROCM_PATH:$CMAKE_PREFIX_PATH"
   export ROCMINFO_PATH="$ROCM_PATH"
   export LD_LIBRARY_PATH="$ROCM_PATH/lib64:$ROCM_PATH/lib:$LD_LIBRARY_PATH"
   export PATH="<path_to_compiled_binaries>:$PATH"

Run Rodinia Benchmark example

In this section, we provide an example of how to use CuPBoP-AMD to execute a CUDA program.

# Run rodinia benchmarks. Ensure all environment variables are set.
$ cd benchmarks/rodinia

# Download Rodinia benchmark dataset
# This script automatically downloads the dataset required to run the rodinia
# suite of benchmarks to the rodinia-data folder
$ ./download-data.sh

# Take the `bfs` benchmark for an example
$ cd bfs && ./run.sh

How to contribute?

Any kinds of contributions are welcome. Please refer to Contribution.md for more detail.

Related publications

If you want to refer CuPBoP in your projects, please cite the related papers:

• CuPBoP-AMD: Extending CUDA to AMD Platforms
• CuPBoP: CUDA for Parallelized and Broad-range Processors

Contributors

• Xule Zhou
• Jun Chen
• John Lu
• Hyesoon Kim

Acknowledgements

• This research was partially supported by AMD. We also want to thank John Lu, Ruobing Han, and Jaewon Lee for their support and expertise.
• CuPBoP: CUDA for Parallelized and Broad-range Processors
• Rodinia Benchmark