FusionRCG: Orchestrating Recursive Computation Graphs across GPU Memory Hierarchies
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866911679097339904 |
|---|---|
| author | Zhang, Yihong Wei, Xinran Chen, Junshi Ju, Fusong Hu, Wei Yang, Jinlong Xia, Huanhuan |
| author_facet | Zhang, Yihong Wei, Xinran Chen, Junshi Ju, Fusong Hu, Wei Yang, Jinlong Xia, Huanhuan |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_10312 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | FusionRCG: Orchestrating Recursive Computation Graphs across GPU Memory Hierarchies Zhang, Yihong Wei, Xinran Chen, Junshi Ju, Fusong Hu, Wei Yang, Jinlong Xia, Huanhuan Computational Physics Distributed, Parallel, and Cluster Computing Chemical Physics 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. |
| title | FusionRCG: Orchestrating Recursive Computation Graphs across GPU Memory Hierarchies |
| topic | Computational Physics Distributed, Parallel, and Cluster Computing Chemical Physics |
| url | https://arxiv.org/abs/2605.10312 |