Quantum Simulation with Sparse Tensors
summary
A compiler framework called Lightweight Functional Quantum Simulator (LFQS) converts gate-level linear algebra into an intermediate language (SDQL/SQL) and generates parallel C and CUDA code that exploits tensorized, structured sparsity instead of forming full OTMs. Micro- and macro-benchmarks on 26-qubit circuits show LFQS is ~3.5x faster on CPU and ~1.66x faster on GPU than QuEST, and beats Qiskit, Catalyst and NumPy; adding a new gate drops from ~40–50 lines to 1 line, and new-backend support falls from ~500 to ~90 lines.
AI-written summary · about
chapters (13)
- 0:00Quantum simulation intro
- 2:31Why quantum simulation matters
- 5:03Quantum bits and gates
- 8:17Linear algebra in circuits
- 10:32Sparsity in quantum matrices
- 12:51Maintainability challenges
- 14:41Existing approaches trade-offs
- 15:24LFQS two-stage approach
- 16:51Compiler optimization
- 17:33Low-level code generation
- 18:37Performance results
- 19:22Maintainability vs Quest
- 20:03Conclusion and future
related talks
59mUtility, Computational Cost, and Effective Quantum Volume of NISQ Experiments
02h 41mLightning Talks
19mQuantum Computing for Simulation in Manufacturing
26mQrisp: A Framework for Compilable High-Level Programming of Gate-Based Quantu...
45mQuantum application development for multi-processor systems using CUDA-Q
46mQuantum Machine Learning 101: An Introduction for Beginners with...
27mQuantum Machine Learning With TensorFlow Quantum
24mDeriving a Kronecker-Free Functional Quantum Simulator