SPONGE (Simulation Package Toward Next GEneration of molecular modelling) is a GPU-accelerated molecular dynamics simulation engine supporting CUDA, HIP (AMD GPU / Hygon DCU), and various CPU SIMD backends.
SPONGE uses pixi to manage dependencies and build workflows.
curl -fsSL https://pixi.sh/install.sh | bashChina mirror:
curl -fsSL https://conda.spongemm.cn/pixi/install.sh | bashpowershell -ExecutionPolicy ByPass -c "irm https://pixi.sh/install.ps1 | iex"China mirror:
powershell -ExecutionPolicy ByPass -c "irm https://conda.spongemm.cn/pixi/install.ps1 | iex"If you want to use a prebuilt SPONGE binary, install one of the published packages instead of building from source.
| Package | Hardware | Platforms |
|---|---|---|
sponge-cuda13 |
NVIDIA GPU (driver >= 570) | Linux x86_64, Windows x64 |
sponge-cuda12 |
NVIDIA GPU (driver >= 525) | Linux x86_64, Windows x64 |
sponge-cpu |
CPU only | Linux x86_64 / aarch64, Windows x64, macOS ARM64 |
sponge-cpu-mpi |
CPU + MPI | Linux x86_64 / aarch64 |
The commands below use sponge-cpu as an example. Replace it with another
package from the table when appropriate.
Create a Pixi project, add the SPONGE channel, and install the package into the project environment:
pixi init sponge-project
cd sponge-project
pixi project channel add https://conda.spongemm.cn
pixi add sponge-cpu
pixi run SPONGE -vIf you are already in a Pixi project that uses conda-forge, only the last
three commands are needed.
Install SPONGE into a Pixi-managed global environment when you want to run it outside a project:
pixi global install \
--channel https://conda.spongemm.cn \
--channel conda-forge \
sponge-cpu
SPONGE -vMost users should install a binary distribution above. To compile SPONGE from source or set up a development environment, follow the Build Guide.
Prepare a TOML input file mdin.spg.toml:
md_name = "NVT water"
mode = "nvt"
step_limit = 50000
dt = 0.002
cutoff = 8.0
default_in_file_prefix = "WAT"
constrain_mode = "SHAKE"
target_temperature = 300.0
write_information_interval = 1000
[thermostat]
mode = "middle_langevin"
tau = 0.1
seed = 2026Use the command that matches how SPONGE was installed.
For a global installation:
SPONGE -mdin mdin.spg.tomlFor an installation in a Pixi project:
pixi run SPONGE -mdin mdin.spg.tomlFor an optional source build in a development environment:
pixi run -e dev-cuda13 SPONGE -mdin mdin.spg.tomlYou can also enter that development environment first:
pixi shell -e dev-cuda13
SPONGE -mdin mdin.spg.tomlThe benchmark tasks are available from a source checkout with a development environment installed:
pixi run -e dev-cuda13 perf-amber # AMBER force field performance
pixi run -e dev-cuda13 perf-nonortho # non-orthogonal box
pixi run -e dev-cuda13 vali-thermostat # thermostat validation
pixi run -e dev-cuda13 vali-barostat # barostat validation- Build Guide — detailed multi-platform build instructions
- Input Reference — full TOML input parameter reference
- Contributing — code style and contribution guidelines