Releases: pnnl/DM-Sim
DM-Sim-V2.0
We made the follow major updates in V2.0:
Provide python APIs, see the adder_n10_omp.py example in src.
Provide C++/CUDA APIs, see the adder_n10_omp.cu example in src.
Support single-GPU fast running (up to 14 qubits on a V100 GPU).
Support single-node-multi-GPU execution managed by OpenMP.
Support multi-node cluster execution managed by MPI and mpi4py.
In this new version, a python-API based circuit can be directly launched and executed without the need to recompile (which is the practice in V1.0). This will significantly benefit emerging quantum-classical hybrid algorithms where the quantum circuits are adjusted and executed online guided by classical optimization processes.
DM-Sim-V1.0
This is the release of DM-Sim associated with our SuperComputing (SC-20) paper titled "Density Matrix Quantum Circuit Simulation via the BSP Machine on Modern GPU Clusters". In the this version, all the gates of the target quantum circuit needs to be embedded & inlined into a GPU global function. We discussed the details in the SC-20 paper. In the following release, in order to flexibly use the simulator in Python, we will change the design of the implementation so that when a new circuit needs to be executed, we can directly launch & run without the need to compile. This will facilitate the patterns of quantum-classical hybrid algorithms where the circuit needs to be generated & adjusted at runtime.