WHAT’S IN THIS CHAPTER?
When ARM first released its original SIMD extensions, it was a huge success. Finally, single instructions worked on multiple data values accelerating multimedia applications, and enabling ARM cores access to a whole range of multimedia devices. Single instructions operating on multiple data values packed into registers meant that ARM cores could be used in DSP applications, or simply to obtain better performance. Mobile telephones could decode MP3 music using even less power, meaning longer battery life.
NEON is an extension of the original SIMD instruction set and is often referred to as the Advanced SIMD Extensions. It extends the SIMD concept by adding instructions that work on 64-bit registers (D for double word) and 128-bit registers (Q for quad word).
NEON instructions are executed as part of the ARM instruction stream, simplifying development and debugging.
NEON isn’t just about having huge amount of registers. The advantage of SIMD instructions is to execute an operation on several data values packed into a single register in a single instruction, but the data must first be correctly placed into the registers. How exactly is that done?
For example, consider a 24-bit image. There are three channels comprising a total of 24-bits ...