VSCode ARM Neon configuration

Arm Neon is an advanced single instruction multiple data (SIMD) architecture extension for the Arm Cortex-A and Arm Cortex-R series of processors with capabilities that vastly improve use cases on mobile devices, such as multimedia encoding/decoding, user interface, 2D/3D graphics and gaming.

It essentially uses 128-bit NEON SIMD registers, which means that if you operate 32-bit floating-point numbers, you can operate 4 at the same time (variables can be defined: float32x4_t); if you operate 16-bit integers (short), you can operate 8 at the same time ( Variables can be defined: int16x8_t); and if 8-bit integers are operated, 16 can be operated at the same time (variables can be defined: int8x16_t).

Prerequisites

• CMake3.13+
• VSCode1.62+
• NDK r23b+
• ADB

Please refer to this article for VSCode and NDK configuration.

Include header files

First, include Neon header file in your source code file.

#include <arm_neon.h>

Add Neon codes

The following is a simple example of assigning values to memory using the Neon instruction set.

void arm_neon_test()
{
    const int sz = 1000000;
    const float val = 1.0f;
    float *arr = new float[sz];
    float32x4_t _v = vdupq_n_f32(val);
    
    int i = 0;
    for(; i < sz; i+=4)
    {
        vst1q_f32(arr + i, _v);
    }
    for(; i < sz; i++)
    {
        arr[i] = val;
    }

    delete[] arr;
}

Build & Run

NDK natively supports NEON, no need to link other libraries, compile and run directly refering to the previous script.