Abstract:We present novel algorithmic solutions together with implementation details utilizing non-Abelian symmetries in order to boost the current limits of tensor network state algorithms on high performance computing infrastructure. In our in-house developed hybrid CPU-multiGPU solution scheduling is decentralized, threads are autonomous and inter-thread communications are solely limited to interactions with globally visible lock-free constructs. Our custom tailored virtual memory management ensures data is produced with high spatial locality, which together with the use of specific sequences of strided batched matrix operations translates to significantly higher overall throughput. In order to lower IO overhead, an adaptive buffering technique is used to dynamically match the level of data abstraction, at which cache repositories are built and reused, to system resources. The non-Abelian symmetry related tensor algebra based on Wigner-Eckhart theorem is fully detached from the conventional tensor network layer, thus massively parallel matrix and tensor operations can be performed without additional overheads. Altogether, we have achieved an order of magnitude increase in performance with respect to results reported in <a class="link-https" data-arxiv-id="2305.05581" href="https://arxiv.org/abs/2305.05581">arXiv:2305.05581</a> in terms of computational complexity and at the same time a factor of three to six in the actual performance measured in TFLOPS. Benchmark results are presented on Hilbert space dimensions up to $2.88\times10^{36}$ obtained via large-scale SU(2) spin adapted density matrix renormalization group simulations on selected strongly correlated molecular systems. These demonstrate the utilization of NVIDIA's highly specialized tensor cores, leading to performance around 110 TFLOPS on a single node supplied with eight NVIDIA A100 devices. In comparison to U(1) implementations with matching accuracy, our solution has an estimated effective performance of 250-500 TFLOPS.

YASTN: Yet another symmetric tensor networks; A Python library for abelian symmetric tensor network calculations

TensorNetwork: A Library for Physics and Machine Learning

The Cytnx Library for Tensor Networks

Supercomputing tensor networks for U(1) symmetric quantum many-body systems

Tensor Network Python (TeNPy) version 1

Tnsp: A Framework Supporting Symmetry and Fermion Tensors for Tensor Network State Methods

QSpace - An open-source tensor library for Abelian and non-Abelian symmetries

Efficient 2D Tensor Network Simulation of Quantum Systems

Boosting the effective performance of massively parallel tensor network state algorithms on hybrid CPU-GPU based architectures via non-Abelian symmetries

Approximately-symmetric neural networks for quantum spin liquids

ANTN: Bridging Autoregressive Neural Networks and Tensor Networks for Quantum Many-Body Simulation

SySTeC: A Symmetric Sparse Tensor Compiler

Tensor Network enhanced Dynamic Multiproduct Formulas

Tensor Network State Algorithms on AI Accelerators

MPSDynamics.jl: Tensor network simulations for finite-temperature (non-Markovian) open quantum system dynamics

TNSPackage: A Fortran2003 Library Designed for Tensor Network State Methods

Tensor Network Efficiently Representing Schmidt Decomposition of Quantum Many-Body States

Quantum correlation functions through tensor network path integral

One-dimensional many-body entangled open quantum systems with tensor network methods

Automatic transformation of irreducible representations for efficient contraction of tensors with cyclic group symmetry