High-performance register file design for general-purpose graphics processing unit with multi-issue logic

GPGPU requires large capacity and high performance register files to save running state and achieve seamless thread bundle switching to improve parallel performance. However, the area and power consumption of register files increase rapidly with the number of access ports, which makes it difficult for register files to adopt more radical multi-port designs. For multi-issue logic GPGPU, the limited number of ports leads to a significant increase in the probability of read and write conflicts, which severely limits the improvement of GPGPU performance. In this paper, a register file design for multi-issue logic is presented. Different from the existing single-port design, the proposed register file separates the operand read and write logic and distributes the registers interleaved to solve the register conflict from two aspects. Moreover, the improved operand collector supports parallel execution, which can fully leverage the performance of the register file. The effectiveness of the proposed method has been validated by applying GPGPU instances on an FPGA platform and designing test cases.

Authors

Institutions

Publication Details

Published
2025-05-09
DOI
https://doi.org/10.1117/12.3059520
Primary Topic
Parallel Computing and Optimization Techniques
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
article

High-performance register file design for general-purpose graphics processing unit with multi-issue logic

Lele Li, Chaofei Wei, Guofeng Zhao, Fan‐Yi Meng et al.
Parallel Computing and Optimization Techniques
article

High-performance register file design for general-purpose graphics processing unit with multi-issue logic

Lele Li, Chaofei Wei, Guofeng Zhao, Fan‐Yi Meng, Gao Chen, Xinxin Zhao, Shuai Wang, Gang Qin, Kai Jiang
article en

Abstract

GPGPU requires large capacity and high performance register files to save running state and achieve seamless thread bundle switching to improve parallel performance. However, the area and power consumption of register files increase rapidly with the number of access ports, which makes it difficult for register files to adopt more radical multi-port designs. For multi-issue logic GPGPU, the limited number of ports leads to a significant increase in the probability of read and write conflicts, which severely limits the improvement of GPGPU performance. In this paper, a register file design for multi-issue logic is presented. Different from the existing single-port design, the proposed register file separates the operand read and write logic and distributes the registers interleaved to solve the register conflict from two aspects. Moreover, the improved operand collector supports parallel execution, which can fully leverage the performance of the register file. The effectiveness of the proposed method has been validated by applying GPGPU instances on an FPGA platform and designing test cases.

Inspur (China) (CN)
Openalex Percentile: Top 92%
Parallel Computing and Optimization Techniques
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.