A Dynamic Allocation Scheme for Adaptive Shared-Memory Mapping on Kilo-core RV Clusters for Attention-Based Model Deployment

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Wang, Bowen, Bertuletti, Marco, Zhang, Yichao, Jung, Victor J. B., Benini, Luca
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866913971893698560
author Wang, Bowen
Bertuletti, Marco
Zhang, Yichao
Jung, Victor J. B.
Benini, Luca
author_facet Wang, Bowen
Bertuletti, Marco
Zhang, Yichao
Jung, Victor J. B.
Benini, Luca
contents Attention-based models demand flexible hardware to manage diverse kernels with varying arithmetic intensities and memory access patterns. Large clusters with shared L1 memory, a common architectural pattern, struggle to fully utilize their processing elements (PEs) when scaled up due to reduced throughput in the hierarchical PE-to-L1 intra-cluster interconnect. This paper presents Dynamic Allocation Scheme (DAS), a runtime programmable address remapping hardware unit coupled with a unified memory allocator, designed to minimize data access contention of PEs onto the multi-banked L1. We evaluated DAS on an aggressively scaled-up 1024-PE RISC-V cluster with Non-Uniform Memory Access (NUMA) PE-to-L1 interconnect to demonstrate its potential for improving data locality in large parallel machine learning workloads. For a Vision Transformer (ViT)-L/16 model, each encoder layer executes in 5.67 ms, achieving a 1.94x speedup over the fixed word-level interleaved baseline with 0.81 PE utilization. Implemented in 12nm FinFET technology, DAS incurs <0.1 % area overhead.
format Preprint
id arxiv_https___arxiv_org_abs_2508_01180
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle A Dynamic Allocation Scheme for Adaptive Shared-Memory Mapping on Kilo-core RV Clusters for Attention-Based Model Deployment
Wang, Bowen
Bertuletti, Marco
Zhang, Yichao
Jung, Victor J. B.
Benini, Luca
Hardware Architecture
Attention-based models demand flexible hardware to manage diverse kernels with varying arithmetic intensities and memory access patterns. Large clusters with shared L1 memory, a common architectural pattern, struggle to fully utilize their processing elements (PEs) when scaled up due to reduced throughput in the hierarchical PE-to-L1 intra-cluster interconnect. This paper presents Dynamic Allocation Scheme (DAS), a runtime programmable address remapping hardware unit coupled with a unified memory allocator, designed to minimize data access contention of PEs onto the multi-banked L1. We evaluated DAS on an aggressively scaled-up 1024-PE RISC-V cluster with Non-Uniform Memory Access (NUMA) PE-to-L1 interconnect to demonstrate its potential for improving data locality in large parallel machine learning workloads. For a Vision Transformer (ViT)-L/16 model, each encoder layer executes in 5.67 ms, achieving a 1.94x speedup over the fixed word-level interleaved baseline with 0.81 PE utilization. Implemented in 12nm FinFET technology, DAS incurs <0.1 % area overhead.
title A Dynamic Allocation Scheme for Adaptive Shared-Memory Mapping on Kilo-core RV Clusters for Attention-Based Model Deployment
topic Hardware Architecture
url https://arxiv.org/abs/2508.01180