Claypool

Courses

Publications

Students

Projects

Service

Downloads

Misc


WPI (Worcester Polytechnic Institute)

Computer Science Department
------------------------------------------

NVIDIA Performance Testing for Emulation of the Grace CPU

Gong Fan, Muyun He, and Shundong Li

Advisor: Mark Claypool

Major Qualifying Project
Computer Science Department, WPI
Term B 2021

Sponsored by: NVidia

Abstract

NVIDIA is developing a new CPU and needs to test and compare the performance of competitors' chips against their simulated CPU. However, testing is cumbersome and time-consuming with current test suites, especially on simulated hardware. Our project is to create a new tool that consolidates other test suites for easier testing for their simulated CPU. We developed a series of Python scripts, defined a database solution for data storage, and created notebooks on Tableau for data visualization. The tester uses a Python script (shell) to run all tests in one place and upload the results to a database, and the developer views and compares the results in Tableau. For ease of use, testers can mount this software on any Linux machine, which simplifies multiple running benchmarks at once and compares results.


Download


[WPI Homepage] [CS Homepage] [Mark Claypool's Homepage]