I have the following code intended to execute a linear search through an array using streaming SIMD extensions in c++:
#include <iostream>
#include <emmintrin.h>
using namespace std;
bool sse2_search_array(int* arr, int size, int key) {
int iterations;
if (size % 16 == 0) {
iterations = size / 16;
}
else {
iterations = size / 16 + 1;
}
__m128i* arr_ = reinterpret_cast<__m128i*>(arr); /*Cast to corresponding int type for 128 bit registers. Each __m128i
occupies 8 bits, so 16 integers can be processed simultaneously.*/
__declspec(align(16)) int key_arr[16];
fill_n(key_arr, 16, key); /*fill key array with 16 keys (for SSE comparisons)*/
__m128i* key_arr_ = reinterpret_cast<__m128i*>(key_arr);
int result;
/*Actual search begins here.*/
for (int i = 0; i < iterations; i++, arr_++) {
result = _mm_movemask_epi8(_mm_cmpeq_epi8( *key_arr_, *arr_)); /*Comparison of 2 16 bit arrays simultaneously.*/
cout << "result: " << result << endl;
if (result != 0) { return true; }
}
return false;
}
int main() {
__declspec(align(16)) int example_array[16] = { 1, 2, 3, 4, 5, 6, 7, 8, 9, 0, 1, 2, 3, 4, 5, 6 };
cout << "found: " << sse2_search_array(example_array, 16, 128);
cin.get();
}
It works but the example in the main function should return false since 128 is not in example_array
, but sse2_search_array
seems to always return true, and the value of result
in the example is 1110111011101110b or 61166 and that does not make sense to me because I'm expecting it to be 0. So can somebody tell me what the problem is and how I can fix it? I am not very experienced with c++ and know very little of SSE.
Aucun commentaire:
Enregistrer un commentaire