Hands-On High Performance Programming with Qt 5

Additional instruction sets

As already mentioned, to increase processor performance chip, manufacturers started to add more sophisticated instructions that can either vectorize computations or execute algorithms that hitherto had to be implemented in application code.

The SIMD or vector instructions can be used to parallelize a scalar computation by executing calculations on several scalar values in parallel. For that, we have to load several float values in two sets of SIMD registers and then apply an operation to all of them at once. Intel processors introduced SIMD in a series of extensions named, namely the following:

• Streaming SIMD Extension (SSE): Uses 128-bit registers and is available in several versions from SSE, over SSE2 to ...