Personal tools

Resource Partitioning of Caches

From iis-projects

Revision as of 11:16, 15 November 2022 by Balasr (talk | contribs)
Jump to: navigation, search


Overview

Status: Available

Introduction

We are moving towards more predictable application-class cores. Our in-house designed Ariane [3] (CVA6) RISC-V 64-bit core is Linux capable and features L1 data and instruction caches. On a system integration level we also have a last-level cache (LLC) [4]. The strategy we are using is to make the whole system more predictable is resource partitioning. Specifically, partitioning the caches allows multiple processes or operating systems (through a hypervisor) to share the same caches with guaranteed worst-case performance.

Project

The goal of the project is to investigate and implement resource partitioning schemes for caches and make them controllable through a high-level framework following the idea of Arm MPAM [2]. There are various known schemes such as randomization, cache-coloring and temporal partitioning[1] that can serve as starting point.

  • Get familiar with cache and the controller design
  • Study existing resource partitioning techniques
  • Implement and verify on FPGA
  • Benchmark results

Character

  • 20% Literature Review
  • 60% Hardware Design
  • 20% Verification and Bechmarking

Prerequisites

  • Strong interest in computer architecture and memory systems
  • Experience with digital design in SystemVerilog as taught in VLSI I
  • Preferred: Knowledge or experience with AXI and RISC-V

References

[1] “A Survey of Techniques for Cache Partitioning in Multicore Processors“ https://dl.acm.org/doi/10.1145/3062394
[2] “Supporting Temporal and Spatial Isolation in a Hypervisor for ARM Multicore Platforms” https://retis.sssup.it/~a.biondi/papers/isol-ICIT18.pdf
[4] ”Last-level cache (LLC)” https://github.com/pulp-platform/axi_llc