Abstract:3D-stacking memory such as High-Bandwidth Memory (HBM) and Hybrid Memory Cube (HMC) provides orders of magnitude more bandwidth and significantly increased channel-level parallelism (CLP) due to its new parallel memory architecture. However, it is challenging to fully exploit the abundant CLP for performance as the bandwidth utilization is highly dependent on address mapping in the memory controller. Unfortunately, CLP is very sensitive to a program’s data access pattern, which is not made available to OS/hardware by existing mechanisms. In this work, we address these challenges with software-defined address mapping (SDAM) that, for the first time, enables user program to obtain a direct control of the low-level memory hardware in a more intelligent and fine-grained manner. In particular, we develop new mechanisms that can effectively communicate a program’s data access properties to the OS and hardware and to use it to control data placement in hardware. To guarantee correctness and reduce overhead in storage and performance, we extend Linux kernel and C-language memory allocators to support multiple address mappings. For advanced system optimization, we develop machine learning methods that can automatically identify access patterns of major variables in a program and cluster these with similar access patterns to reduce the overhead for SDAM. We demonstrate the benefits of our design on real system prototype, comprising (1) a RISC-V processor, near memory accelerators and HBM modules using Xilinx FPGA platform, and (2) modified Linux and glibc. Our evaluation on standard CPU benchmarks and data-intensive benchmarks (for both CPU and accelerators) demonstrates 1.41×, 1.84× speedup on CPU and 2.58× on near memory accelerators in our system with SDAM compared to a baseline system that uses a fixed address mapping.

Application specific parallel memory architecture for software defined radio

On Memory Accelerated Signal Processing within Software Defined Radios

Design and Optimization of the Vector Memory Applying for SDR

Software Programmable Data Allocation in Multi-bank Memory of SIMD Processors.

Parallel Programming and Its Architectures Based on Data Access Separated Algorithm Kernels

A Novel GPP-based Software-Defined Radio Architecture.

SPVA: A Novel Digital Signal Processor Architecture for Software Defined Radio

Design and implementation of parallel multi-access memory interface

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

Epuma: A Unique Memory Access Based Parallel DSP Processor for SDR and CR

An Efficient Vector Memory Unit for SIMD DSP

Epuma Embedded Parallel DSP Processor with Unique Memory Access.

An Efficient SIMD Parallel Memory Structure for Radix-2 FFT Computation

Data Memory Structure and Management Strategy of Reconfigurable System for Radar

Case Study of Efficient Parallel Memory Access Programming for the Embedded Heterogeneous Multicore DSP Architecture Epuma

Software-defined Address Mapping - a Case on 3D Memory.

A Novel Design of High-speed Multi-port Memory Interface for Digital Signal Processor

A Configurable Management Strategy for Parallel Access of Coarse-Grained Reconfigurable Architecture for Radar Processing

Convolutional Decoding on Deep-pipelined SIMD Processor with Flexible Parallel Memory

Efficient Hardware Architecture of Deterministic MPA Decoder for SCMA

Software Defined Radio Architecture Using a Multicasting Network-on-chip