hashirama/uvg266
1
0
Fork 0
mirror of https://github.com/ultravideo/uvg266.git synced 2024-11-27 19:24:06 +00:00
Code Issues Actions Projects Releases Wiki Activity

some more modifications to sao_edge_ddistortion_avx2 to prevent memory failures

Browse source
This commit is contained in:
Reima Hyvönen 2019-03-29 14:38:53 +02:00 committed by Pauli Oikkonen
parent 3496a57f7a
commit afbb6fb960
1 changed files with 3 additions and 10 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

11
src/strategies/avx2/sao-avx2.c
View file

@ -95,8 +95,6 @@ static int sao_edge_ddistortion_avx2(const kvz_pixel *orig_data,
__m256i tmp_offset_epi32;
__m256i tmp1_vec_epi32;
__m256i tmp2_vec_epi32;
__m256i zeros_epi32 = _mm256_setzero_si256();
__m256i offset_zeros_epi32;
for (y = 1; y < block_height - 1; ++y) {
for (x = 1; x < block_width - 8; x+=8) {
@ -113,9 +111,6 @@ static int sao_edge_ddistortion_avx2(const kvz_pixel *orig_data,
tmp_offset_epi32 = _mm256_permutevar8x32_epi32(offsets_epi32, v_cat_epi32);
offset_zeros_epi32 = _mm256_cmpeq_epi32(zeros_epi32, tmp_offset_epi32);
// (diff - offset) * (diff - offset)
tmp1_vec_epi32 = _mm256_mul_epi32(_mm256_sub_epi32(tmp_diff_epi32, tmp_offset_epi32), _mm256_sub_epi32(tmp_diff_epi32, tmp_offset_epi32));
@ -125,7 +120,7 @@ static int sao_edge_ddistortion_avx2(const kvz_pixel *orig_data,
// Offset is applied to reconstruction, so it is subtracted from diff.
// sum += (diff - offset) * (diff - offset) - diff * diff;
tmp_sum_epi32 = _mm256_add_epi32(tmp_sum_epi32, _mm256_andnot_si256(offset_zeros_epi32, _mm256_sub_epi32(tmp1_vec_epi32, tmp2_vec_epi32)));
tmp_sum_epi32 = _mm256_add_epi32(tmp_sum_epi32, _mm256_sub_epi32(tmp1_vec_epi32, tmp2_vec_epi32));
}
bool use_6_elements = block_width - x - 1 == 6;
@ -151,8 +146,6 @@ static int sao_edge_ddistortion_avx2(const kvz_pixel *orig_data,
tmp_offset_epi32 = _mm256_permutevar8x32_epi32(offsets_epi32, v_cat_epi32);
offset_zeros_epi32 = _mm256_cmpeq_epi32(zeros_epi32, tmp_offset_epi32);
// (diff - offset) * (diff - offset)
tmp1_vec_epi32 = _mm256_mul_epi32(_mm256_sub_epi32(tmp_diff_epi32, tmp_offset_epi32), _mm256_sub_epi32(tmp_diff_epi32, tmp_offset_epi32));
@ -162,7 +155,7 @@ static int sao_edge_ddistortion_avx2(const kvz_pixel *orig_data,
// Offset is applied to reconstruction, so it is subtracted from diff.
// sum += (diff - offset) * (diff - offset) - diff * diff;
tmp_sum_epi32 = _mm256_add_epi32(tmp_sum_epi32, _mm256_andnot_si256(offset_zeros_epi32, _mm256_sub_epi32(tmp1_vec_epi32, tmp2_vec_epi32)));
tmp_sum_epi32 = _mm256_add_epi32(tmp_sum_epi32, _mm256_sub_epi32(tmp1_vec_epi32, tmp2_vec_epi32));
tmp_sum_epi32 = _mm256_hadd_epi32(tmp_sum_epi32, tmp_sum_epi32);
tmp_sum_epi32 = _mm256_hadd_epi32(tmp_sum_epi32, tmp_sum_epi32);