Single instruction multiple data (SIMD) describes a class of CPU operations that allow to perform the same operation on multiple data at the same time. The counterpart is single instruction single data (SISD). All modern Instruction set architecture (ISA) support SIMD operations.

In the numpy library of python array operations use SIMD to process data simultaniously instead of each value of Array after another. This is also often called a vectorized Operation. Therefore due to SIMD a vectorized operation is faster than an explicit loop.

References

  • Blog post
  • Andy Pavlo Advanced Database Systems