GEN-Graph: Heterogeneous PIM Accelerator for General Computational Patterns in Graph-based Dynamic Programming

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Chen, Yanru, Tian, Runyang, Li, Zheyu, Afarin, Mahbod, Xu, Weihong, Rosing, Tajana
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866915940781785088
author Chen, Yanru
Tian, Runyang
Li, Zheyu
Afarin, Mahbod
Xu, Weihong
Rosing, Tajana
author_facet Chen, Yanru
Tian, Runyang
Li, Zheyu
Afarin, Mahbod
Xu, Weihong
Rosing, Tajana
contents While graph-based dynamic programming (DP) is a cornerstone of genomics and network analytics, its efficiency is hampered by fundamentally conflicting computational patterns. Matrix-centric DP drives regular, compute-bound network analytics, while topology-centric DP handles irregular, memory-bound genomic traversals. These two categories of DP have substantially different computation patterns and dataflows, which makes it difficult for a single homogeneous processing-in-memory (PIM) architecture to efficiently support both. This work presents GEN-Graph, a novel heterogeneous PIM chiplet that integrates two types of specialized compute tiles within a 2.5D package: Matrix-tile, a processing-using-memory (PUM) tile optimized for matrix-centric workloads, such as all-pairs shortest path (APSP); and traversal-tile, a processing-near-memory (PNM) tile optimized for traversal-centric DP workloads, such as DNA sequence alignment. Our hardware-software co-design employs recursive partitioning and reconfigurable windowed bit-parallel logic to ensure exact computation. Results show the matrix tile achieves 42.8x speedup and 392x energy efficiency over the NVIDIA H100 GPU for APSP. For sequence-to-graph alignment, the traversal tile sustains 2.56 million reads/s (short-reads) and 39.3 thousand reads/s (long-reads), outperforming state-of-the-art accelerators by up to 2.56x in throughput. GEN-Graph provides the first scalable, exact solution for general DP dataflows by matching hardware specialization to algorithmic structure.
format Preprint
id arxiv_https___arxiv_org_abs_2604_15361
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle GEN-Graph: Heterogeneous PIM Accelerator for General Computational Patterns in Graph-based Dynamic Programming
Chen, Yanru
Tian, Runyang
Li, Zheyu
Afarin, Mahbod
Xu, Weihong
Rosing, Tajana
Hardware Architecture
Social and Information Networks
While graph-based dynamic programming (DP) is a cornerstone of genomics and network analytics, its efficiency is hampered by fundamentally conflicting computational patterns. Matrix-centric DP drives regular, compute-bound network analytics, while topology-centric DP handles irregular, memory-bound genomic traversals. These two categories of DP have substantially different computation patterns and dataflows, which makes it difficult for a single homogeneous processing-in-memory (PIM) architecture to efficiently support both. This work presents GEN-Graph, a novel heterogeneous PIM chiplet that integrates two types of specialized compute tiles within a 2.5D package: Matrix-tile, a processing-using-memory (PUM) tile optimized for matrix-centric workloads, such as all-pairs shortest path (APSP); and traversal-tile, a processing-near-memory (PNM) tile optimized for traversal-centric DP workloads, such as DNA sequence alignment. Our hardware-software co-design employs recursive partitioning and reconfigurable windowed bit-parallel logic to ensure exact computation. Results show the matrix tile achieves 42.8x speedup and 392x energy efficiency over the NVIDIA H100 GPU for APSP. For sequence-to-graph alignment, the traversal tile sustains 2.56 million reads/s (short-reads) and 39.3 thousand reads/s (long-reads), outperforming state-of-the-art accelerators by up to 2.56x in throughput. GEN-Graph provides the first scalable, exact solution for general DP dataflows by matching hardware specialization to algorithmic structure.
title GEN-Graph: Heterogeneous PIM Accelerator for General Computational Patterns in Graph-based Dynamic Programming
topic Hardware Architecture
Social and Information Networks
url https://arxiv.org/abs/2604.15361