An Evaluation of Memory Sharing Performance for Heterogeneous Embedded SoCs with Many-Core Accelerators
Open Access
- 8 February 2015
- conference paper
- Published by Association for Computing Machinery (ACM)
- p. 1-9
Abstract
Today's systems-on-chip (SoCs) more and more conform to the models envisioned by the Heterogeneous System Architecture (HSA) foundation in which massively parallel, programmable many-core accelerators (PMCAs) not only cooperate but also coherently share memory with a powerful, multi-core host processor. Allowing direct access to system memory from both sides greatly simplifies application development, but it increases the potential interference to the memory system due to the PMCA. In this work, we evaluate the impact of a PMCA's memory traffic on the host performance using the Xilinx Zynq-7000 SoC. This platform features a dual-core ARM Cortex-A9 CPU, as well as a field-programmable gate array (FPGA), which we use to model a PMCA. Synthetic workload, real benchmarks from the MiBench and ALPBench suites, and collaborative workloads all show that the interference generated by the PMCA can significantly reduce the memory bandwidth seen by the host (on average up to 25 % for host applications).Keywords
Funding Information
- European Commission Directorate-General for Research and Innovation (611016)
This publication has 11 references indexed in Scilit:
- Fast Matrix Covering in All Programmable Systems-on-ChipElektronika ir Elektrotechnika, 2014
- Architectural support for address translation on GPUs2014
- Supporting x86-64 address translation for 100s of GPU lanes2014
- Energy and performance exploration of accelerator coherency port using Xilinx ZYNQ2013
- Improving the programmability of STHORM-based heterogeneous systems with offload-enabled OpenMP2013
- Platform 2012, a many-core computing accelerator for embedded SoCs2012
- The tradeoffs of fused memory hierarchies in heterogeneous computing architectures2012
- On the Efficacy of a Fused CPU+GPU Processor (or APU) for Parallel Computing2011
- Studying the impact of hardware prefetching and bandwidth partitioning in chip-multiprocessors2011
- SRP: Symbiotic Resource Partitioning of the Memory Hierarchy in CMPsLecture Notes in Computer Science, 2010