#include <assert.h>


#include <input.h>
#include <vector.h>

#include <tracem/ints.h>

_Static_assert(sizeof(input_t)*INPUT_VECTOR_STEP == sizeof(v128_t), "invalid vector size mul");

#define INPUT_VPROC_STEP 15

__attribute__((pure))
u64 vproc(usize len, const input_t input[const restrict len])
{
	static const v128_t SINGLE = { .as.i16 = { -1, -1, -1, -1, -1, -1, -1, -1 } };
	static const v128_t ONES   = { .as.i16 = {  1,  1,  1,  1,  1,  1,  1,  1 } };
#if DEBUG
	assert(len % INPUT_VPROC_STEP == 0);
#endif

#define LAST(j) input[i +  (j)]
#define NEXT(j) input[i + ((j) * 2)]
	__m128i end = _mm_setzero_si128();
	for(usize i = 0;i<len;i+=INPUT_VPROC_STEP)
	{
		// load the vectors
		v128_t prev = { .as.u16 = {
				input[i+0],
				input[i+2],

				input[i+4],
				input[i+6],

				input[i+8],
				input[i+10],

				input[i+12],
				input[i+14],
			}
		};
		v128_t next = { .as.u16 = {
				input[i+1],
				input[i+3],

				input[i+5],
				input[i+7],

				input[i+9],
				input[i+11],

				input[i+13],
				input[i+15],
			}
		};
		//__m128i res = _mm_subs_epi16(prev.mm, next.mm);
		__m128i pltn = _mm_cmplt_epi16(prev.mm, next.mm); // prev < next
		pltn = _mm_sign_epi16(pltn, SINGLE.mm);
		pltn = _mm_sign_epi16(ONES.mm, pltn);
		end = _mm_add_epi16(end, pltn);
		//__m128i pr =  _mm_sign_epi16(FULL_MASK.mm, res);
	}
	
	v16_u128n _ev = ((union v128)end).as.u16;
	return (u64) (    _ev[0] + _ev[1] 
			+ _ev[2] + _ev[3]
			+ _ev[4] + _ev[5]
			+ _ev[6] + _ev[7] );
}