Writing Fast Java Code with the Vector API - JEP Café 18

Explore the Vector API in Java to significantly accelerate computations using SIMD capabilities of modern CPUs. Dive into parallel computing on SIMD machines, learn how to leverage the Java Vector API for enhanced performance, and discover techniques to structure code for optimal utilization of these powerful features. Gain insights into vector creation, manipulation, and parallel operations such as summing, computing norms, averaging, filtering, and reducing. Examine practical examples, patterns, and performance metrics to master fast Java code writing using the Vector API.

Syllabus

Intro
Introducing the SIMD features of your CPU
Scalar parallel computing based on concurrency
Vector parallel computing based on SIMD machine
Shape and species of a Vector
Creating vectors from arrays to sum them
Loading any array in a vector using masks
Avoiding masking when it is not supported
Parallel cross-lanes and lane-wise operations
Computing the norm of a vector in parallel
Computing the average of vector components in parallel
Filtering and compressing a vector in parallel
Reducing a vector in parallel
Wrapping up parallel computations using vectors
Examples, patterns and performances
Outro