Optimizing C Code with Neon Intrinsics

Neon is the implementation of the Advanced SIMD extension to the Arm architecture.

All processors compliant with the Armv8-A architecture (for example, the Cortex-A76 or Cortex-A57) include Neon. In the programmer's view, Neon provides an additional 32 128-bit registers with instructions that operate on 8, 16, 32, or 64 bit lanes within these registers.

arm_neon.h

#include <arm_neon.h> must appear before the use of any Neon intrinsics.

The mul in the function name is a hint that this intrinsic uses the MUL instruction. Based on the types of the arguments and return value, sixteen bytes of signed integers, we might guess this intrinsic maps to the following instruction:

MUL Vd.16B, Vn.16B, Vm.16B

So this function multiplies corresponding elements of a and b and returns the result. Checking the definition shows this is indeed true.

1. Changing the arrays, but not the function? In this case padding the arrays with zeros would be the simplest option, but this padding may have to be accounted for in other functions.
2. Changing the function, but not the arrays? One way would be to use the Neon deinterleave for every whole multiple of sixteen values, and then use the C deinterleave for the remainder.

• float64_t is a scalar type which is a 64-bit floating-point type.
• poly64x2_t is a vector type of two 64-bit polynomial scalars.
• int8x8x3_t is a vector array type of three vectors of eight 8-bit signed integers.