93     __m256i r_lo[4], r_hi[4];  in vpx_sad64x64x4d_avx2()  local
97     r_lo[0] = _mm256_loadu_si256((const __m256i *)refs[0]);  in vpx_sad64x64x4d_avx2()
99     r_lo[1] = _mm256_loadu_si256((const __m256i *)refs[1]);  in vpx_sad64x64x4d_avx2()
101     r_lo[2] = _mm256_loadu_si256((const __m256i *)refs[2]);  in vpx_sad64x64x4d_avx2()
103     r_lo[3] = _mm256_loadu_si256((const __m256i *)refs[3]);  in vpx_sad64x64x4d_avx2()
107     r_lo[0] = _mm256_sad_epu8(r_lo[0], s_lo);  in vpx_sad64x64x4d_avx2()
108     r_lo[1] = _mm256_sad_epu8(r_lo[1], s_lo);  in vpx_sad64x64x4d_avx2()
109     r_lo[2] = _mm256_sad_epu8(r_lo[2], s_lo);  in vpx_sad64x64x4d_avx2()
110     r_lo[3] = _mm256_sad_epu8(r_lo[3], s_lo);  in vpx_sad64x64x4d_avx2()
117     sums[0] = _mm256_add_epi32(sums[0], r_lo[0]);  in vpx_sad64x64x4d_avx2()
118     sums[1] = _mm256_add_epi32(sums[1], r_lo[1]);  in vpx_sad64x64x4d_avx2()
119     sums[2] = _mm256_add_epi32(sums[2], r_lo[2]);  in vpx_sad64x64x4d_avx2()
120     sums[3] = _mm256_add_epi32(sums[3], r_lo[3]);  in vpx_sad64x64x4d_avx2()