Abstract:Data transfers are essential in today's computing systems as latency and complex memory access patterns are increasingly challenging to manage. Direct memory access engines (DMAEs) are critically needed to transfer data independently of the processing elements, hiding latency and achieving high throughput even for complex access patterns to high-latency memory. With the prevalence of heterogeneous systems, DMAEs must operate efficiently in increasingly diverse environments. This work proposes a modular and highly configurable open-source DMAE architecture called intelligent DMA (iDMA), split into three parts that can be composed and customized independently. The front-end implements the control plane binding to the surrounding system. The mid-end accelerates complex data transfer patterns such as multi-dimensional transfers, scattering, or gathering. The back-end interfaces with the on-chip communication fabric (data plane). We assess the efficiency of iDMA in various instantiations: In high-performance systems, we achieve speedups of up to 15.8x with only 1 % additional area compared to a base system without a DMAE. We achieve an area reduction of 10 % while improving ML inference performance by 23 % in ultra-low-energy edge AI systems over an existing DMAE solution. We provide area, timing, latency, and performance characterization to guide its instantiation in various systems.

A Instruction and Task Parallel DMA Interface in DSP

Design of an Intelligent Dma System Architecture with Data Pre-Processing

Design and Implementation of DMA Transfers in WISHBONE Interface.

Logical Architecture Design of Integrated Electronic System with High Real-Time and Reliability in ZDPS-2 Pico-Satellite

DMA-based High-Speed Parallel Acquisition System of Digital Signals

Design of the DDR SDRAM Controller Integrated by the Data Intensive Computing Architecture

Design and implementation of an advanced DMA controller on AMBA-based SoC

Design Of An Enhanced Dma With Data Pre-Computing For Digital Image Processing

Design and Implementation of a Memory Architecture in Dsp for Wireless Communication

DMA Data Transferring on PCI Bus Based on DSP

Design and Implementation of an AHB-based High-Speed Memory Access Layer

YHFT-QDSP:High-Performance Heterogeneous Multi-Core DSP

A High-performance, Energy-efficient Modular DMA Engine Architecture

A Design of DSP External Memory Interface

Design and Implementation of Communication Between HOST and DSP Based on PCI-X Core

The Floppy Driver Control System Based on DSP

Architecture Design of Simultaneous Multithreading VLIW DSP

Implementation of an Instruction Dispatch Unit Applied to Digital Signal Processors with VLIW Architecture

The Design of Infterface Between PCI Bus and DSP Based on Dual-port RAM

Design of real-time data acquisition and processing system based on DMA way

MCS-DMA:An Optimization Design of Memory Controller for DMA Transfers in SoC