slow_on_gcc49.c

// __attribute__((aligned(16))) float input[n], weight[n];
float dot_sse(float *input, float *weight, int n)
{
	__attribute__((aligned(16))) float mm[4] = {0};
	__m128 x, u;
	int pk_lp = (n & 0xfffffffc);
	int i;
	float sum = 0;
	u = _mm_setzero_ps();
	for (i = 0; i < pk_lp; i += 4) {
		u = _mm_add_ps(u, _mm_mul_ps(*(const __m128*)&input[i], *(const __m128*)&weight[i]));
	}
	_mm_store_ps(mm, u);
	sum = mm[0] + mm[1] + mm[2] + mm[3];
	for (i = pk_lp; i < n; ++i) {
		sum += input[i] * weight[i];
	}
	return sum;
}