hashirama/uvg266
1
0
Fork 0
mirror of https://github.com/ultravideo/uvg266.git synced 2024-11-27 11:24:05 +00:00
Code Issues Actions Projects Releases Wiki Activity

Use unsigned min() to correctly clip -32768

Browse source 
If a coeff happens to be -32768 (0x8000), its 16-bit abs() is also
0x8000. It should ultimately be clipped to 3, so interpret absolute
values as unsigned instead to make that happen.
This commit is contained in:
Pauli Oikkonen 2020-02-04 20:28:14 +02:00
parent b93b90c0d7
commit 2abd733199
1 changed files with 2 additions and 2 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

4
src/strategies/avx2/quant-avx2.c
View file

@ -817,14 +817,14 @@ static uint32_t fast_coeff_cost_avx2(const coeff_t *coeff, int32_t width, int32_
for (int i = 0; i < width * width; i += 32) {
__m256i curr_lo = _mm256_loadu_si256 ((const __m256i *)(coeff + i));
__m256i curr_abs_lo = _mm256_abs_epi16 (curr_lo);
__m256i curr_max3_lo = _mm256_min_epi16 (curr_abs_lo, threes);
__m256i curr_max3_lo = _mm256_min_epu16 (curr_abs_lo, threes);
// 4x4 blocks only have 16 coeffs, so handle them separately
__m256i curr_max3_hi;
if (width >= 8) {
__m256i curr_hi = _mm256_loadu_si256 ((const __m256i *)(coeff + i + 16));
__m256i curr_abs_hi = _mm256_abs_epi16 (curr_hi);
curr_max3_hi = _mm256_min_epi16 (curr_abs_hi, threes);
curr_max3_hi = _mm256_min_epu16 (curr_abs_hi, threes);
curr_max3_hi = _mm256_slli_epi16 (curr_max3_hi, 8);
} else {
// Set MSBs for high bytes if they're meaningless, so shuffles will