Advantages of Vectorization and the Effects of Data Size
This video demonstrates the increases in speed and decreases in loop trip counts to be gained by vectorizing code, as well as examining how data type sizes can affect these results and be tweaked to maximize the gains of vectorization.
The demonstration is achieved by comparing the trip counts and run times of the scalar and vectorized versions of a pair of loops using floats and doubles, to show both the massive improvement gained by vectorization and the consequences of using more bytes than you need. It also shows the viewer how to use certain aspects of Advisor’s survey report to gain useful diagnostic information about the details of how their loops vectorized.
Product and Performance Information
Performance varies by use, configuration and other factors. Learn more at www.Intel.com/PerformanceIndex.