cuda_byte_perm_imp

uint32_t _bperm(const uint32_t a, const uint32_t b, const uint32_t s) {

	unsigned char input[8], selector[4];

	input[0] = a & 0xFF;// x<7:0>
	input[1] = ( a >> 8 ) & 0xFF; //x<15:8>
	input[2] = ( a >> 16 ) & 0xFF; //x<23:16>
	input[3] = ( a >> 24 ) & 0xFF; //x<31:24>

	input[4] = b & 0xFF;// y<7:0>
	input[5] = ( b >> 8 ) & 0xFF; //y<15:8>
	input[6] = ( b >> 16 ) & 0xFF; //y<23:16>
	input[7] = ( b >> 24 ) & 0xFF; //y<31:24>

	selector[0] = s & 0x07;				// s<2 : 0>
	selector[1] = ( s >> 4 ) & 0x07;	// s<6 : 4>
	selector[2] = ( s >> 8 ) & 0x07;	//s<10 : 8>
	selector[3] = ( s >> 12 ) & 0x07;	//s<14 : 12>
	
	uint32_t v  = input[selector[3]];
	v = (v << 8) + input[selector[2]];
	v = (v << 8) + input[selector[1]];
	v = (v << 8) + input[selector[0]];

	return v;
}