Overview / Usage

This program is in development and aims to take command line arguments to run various operations such as vector multiplies, vector adds and matrix operations such as inverse calculation, division and multiplication.

This program will be able to do all these operations and time them through various methods to determine which one is quicker on a given PC.

​

The images in the thumbnail and featured section are just images of the launching of one of many multi thread operations.

Methodology / Approach

Each operation has its own kernel written for the specific method of completing it (Scalar, AVX & DPC++) which can be run to benchmark the devices performance of that specific operation using that method. This is designed to show the performance characteristics of simple mathematical operations on the hardware. This is done by allocating a few buffers of a desired length then specifying how many times to iterate over each buffer and do operations for to attain useful figures for relative performance and overhead.

Technologies Used

AVX, AVX2, AVX-512, DPC++