From the series: Coder Summit 2018
Zhen Wang, MathWorks
Embedded Coder® generates native SIMD instructions starting in R2018a. This support offers 10x or more code efficiency improvements by leveraging the parallelism afforded by embedded hardware.
Embedded Coder native SIMD code generation is a major improvement upon previous SIMD code generation, which used code wrappers. The native SIMD code is faster and of higher qualtiy, and appears like professional handwritten SIMD code.
In some examples, featuring Intel® AVX and SSE for Windows® and Linux®, the code executes 16x faster with SIMD instructions.
This presentation is from Coder Summit Talks, which are live recordings of developers and engineers debuting their best MATLAB and Simulink Coder™ optimizations and examples at an annual technical interchange.