Try to use avx_hs() - slow and requires alignment 4096 bits < (l.size*l.siz...

AlexeyAB

2018-08-08 a284a7da8d1facbf984a22302665a2a50295a687

 src/network.c

@@ -862,6 +862,8 @@
            if (l->xnor) {
                //printf("\n %d \n", j);
                size_t ldb_align = 256; // 256bit for AVX2
                if (l->size*l->size*l->c > 4096) ldb_align = 4096;

                binary_transpose_align_weights(l, ldb_align);
            }
        }