Code and experiment data of the SoCS 2026 paper "Parallelizing Classical Planning: Critical Path Heuristics on the GPU"

Zenodo (CERN European Organization for Nuclear Research)

Abstract

Code The file code.zip contains an extended version of the Fast Downward planning system (http://fast-downward.org). The code for the $h^m$ heuristic used for the CPU is stored in `src/search/heuristics/h2_heuristic.cc/h` and the GPU implementation in `src/search/heuristics/hm2_torch.cc/h`. The scripts used to run the experiments can be found in "experiments/socs26/". Please see http://www.fast-downward.org for detailed instructions on how to compile the planner. Note that the CUDA toolkit is strictly required for compilation. Please, follow the instructions: libtorch 2.1.2 (CUDA 12.1, cxx11 ABI) wget https://download.pytorch.org/libtorch/cu121/libtorch-cxx11-abi-shared-with-deps-2.1.2%2Bcu121.zip unzip libtorch-cxx11-abi-shared-with-deps-2.1.2+cu121.zip export LIBTORCH=$PWD/libtorch CUDA 12.1 must be installed (nvcc + libcudart_static, libcudadevrt). export CUDA_HOME=/path/to/cuda-12.1 export LIBRARY_PATH=$CUDA_HOME/targets/x86_64-linux/lib:$LIBRARY_PATH Build Fast Downward (requires cmake + gcc/g++ 12.3, C++20)cd downward_h2CMAKE_PREFIX_PATH=$LIBTORCH \LD_LIBRARY_PATH=$LIBTORCH/lib:$CUDA_HOME/lib64:$LD_LIBRARY_PATH \./build.py release An example to run $h^2$ on the GPU without cost partitioning is: LD_LIBRARY_PATH=$LIBTORCH/lib:$CUDA_HOME/lib64:$LD_LIBRARY_PATH \./fast-downward.py domain.pddl problem.pddl \ --search "astar(hm2torch(m=2,use_gpu=true,cp=0,n_cp=1))"

Authors 4

1. Markus Fritzsche

   Linköping University

   Affiliation as printed

   Linköping University

2. David Speck

   University of Basel

   Affiliation as printed

   University of Basel

3. Daniel Gnad

   Heidelberg University

   Affiliation as printed

   Heidelberg University

4. Simon Ståhlberg Aachen

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen University