FusionRCG: Orchestrating Recursive Computation Graphs across GPU Memory Hierarchies

TL;DR

FusionRCG is a framework that optimizes recursive computation graphs on GPUs, reducing memory usage and significantly improving performance in quantum chemistry calculations.

Contribution

It introduces a novel approach combining graph orchestration, algebraic reduction, and multi-tier kernel architecture to enhance GPU efficiency for high-dimensional integrals.

Findings

Up to 3.09x speedup over existing GPU implementations.

Shrinks intermediate footprints by up to 7.7x.

Maintains 75% parallel efficiency at 64 GPUs.

Abstract

Evaluating high-dimensional integrals via deep hierarchical recurrences is a dominant cost in quantum chemistry. While CPUs manage these efficiently, GPUs suffer a critical mismatch: limited per-thread memory is quickly overwhelmed by an explosion of simultaneously live intermediate variables. As recurrence scales, this forces massive data spilling to global memory, collapsing performance into a severe memory-bound regime. We present FusionRCG, a framework that jointly optimizes computation graph structure and GPU memory mapping. Exploiting the inherent topological flexibility of recurrence graphs, using electron repulsion integrals as an example, we contribute: (1) liveness-aware graph orchestration to minimize peak live intermediates; (2) algebraic dimensionality reduction via stepwise Cartesian-to-spherical fusion, shrinking intermediate footprints by up to $7.7\times$; and (3) an adaptive multi-tier kernel architecture routing graphs across the memory hierarchy. Evaluated on NVIDIA A100 GPUs, FusionRCG achieves up to $3.09\times$ end-to-end SCF speedup over GPU4PySCF and maintains $75\%$ parallel efficiency at 64~GPUs, successfully rescuing these workloads from memory-bound limits.