c++ - Matrix vectorization using SIMD -
i'm trying vectorized following loop (only inner) :
for (int =0; i<n; i++){ const int line = * width; (int j = 0; j < n; j++){ a[line + j] = 3; } }
but i'm getting following error:
not vectorized: not suitable gather d.38226_277 = *d.38225_276;
i googled around , it's seems compiler not vectorize because, not know data contiguous. i'm sure data ([line + j]) contiguous. how can tell compiler trust me , vectorize loop ?
i'm using gcc 4.7
edit:
i able vetorize loop without change in code using intel compiler.
Comments
Post a Comment