micahpearlman

// Really awesome code taken from: http://apangborn.com/2011/05/pixel-processing-using-arm-assembly/
inline static void neon_rgba_to_bgra(unsigned char *src, unsigned char *dst, int numPixels)
{
#ifdef __ARM_NEON__
    int simd_pixels = numPixels & ~7; // round down to nearest 8
    int simd_iterations = simd_pixels >> 3;
    int col;
    if(simd_iterations) { // make sure at least 1 iteration
        __asm__ __volatile__ ("1: \n\t"
                              // structured load of 8 pixels into d0-d3 (64-bit) NEON registers