SIMD Parallel Programming with the Vector API
summary
Covers Java's incubating Vector API (JEP 469) for SIMD parallelism on CPUs (AVX 128/256/512). Explains using a static Species object, looping in steps of species.length, and handling leftovers with hardware masks or a tail loop; shows expected speedups (about 8x on ints with 256-bit AVX) while noting memory/load overhead and the need for JMH benchmarking on target hardware. GitHub contains example JMH benchmarks.
AI-written summary · about
chapters (7)
- 0:00Introduction
- 3:25API Vector overview
- 13:18SIMD basics
- 20:57API Vector operation
- 31:12Vector operations & algorithms
- 40:29Use cases & performance
- 42:39Q&A
