Browse Teach Curate

GPU-Accelerated Combinatorial Summation of Feynman Diagrammatics

Porting the combinatorial summation (CoS) algorithm for Feynman diagrammatic series onto GPUs. Because CoS evaluates many near-identical diagram contributions independently, it maps naturally onto the GPU SIMT model; organizing threads to avoid warp divergence under CUDA yields at least three orders of magnitude speedup over CPUs for Diagrammatic Monte Carlo, even on consumer hardware, making high-order probes of strongly correlated systems far more accessible.

KITT Text

Exploiting Parallelism for Fast Feynman Diagrammatics

This work dramatically accelerates the evaluation of Feynman diagrammatic series in quantum many-body theory by porting the recently introduced combinatorial summation (CoS) algorithm to GPUs. Diagra…