walkero
/
openjpeg
1
0
Fork 0
Code Issues Pull Requests Projects Releases 1 Wiki Activity

IDWT 5x3: fix bug in AVX2 implementation (#953, #957)

This commit is contained in:
Even Rouault 2017-06-30 00:03:05 +02:00
parent 6239ed7be4
commit 8fa405ee15
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/lib/openjp2/dwt.c
View File

@ -685,9 +685,9 @@ static void opj_idwt53_v_cas0_mcols_SSE2_OR_AVX2(
s1n_0 = LOADU(in_even + ((len - 1) / 2) * stride); s1n_0 = LOADU(in_even + ((len - 1) / 2) * stride);
/* tmp_len_minus_1 = s1n - ((d1n + 1) >> 1); */ /* tmp_len_minus_1 = s1n - ((d1n + 1) >> 1); */
tmp_len_minus_1 = SUB(s1n_0, SAR(ADD3(d1n_0, d1n_0, two), 2)); tmp_len_minus_1 = SUB(s1n_0, SAR(ADD3(d1n_0, d1n_0, two), 2));
STORE(tmp + 8 * (len - 1), tmp_len_minus_1); STORE(tmp + PARALLEL_COLS_53 * (len - 1), tmp_len_minus_1);
/* d1n + ((s0n + tmp_len_minus_1) >> 1) */ /* d1n + ((s0n + tmp_len_minus_1) >> 1) */
STORE(tmp + 8 * (len - 2), STORE(tmp + PARALLEL_COLS_53 * (len - 2),
ADD(d1n_0, SAR(ADD(s0n_0, tmp_len_minus_1), 1))); ADD(d1n_0, SAR(ADD(s0n_0, tmp_len_minus_1), 1)));
s1n_1 = LOADU(in_even + ((len - 1) / 2) * stride + VREG_INT_COUNT); s1n_1 = LOADU(in_even + ((len - 1) / 2) * stride + VREG_INT_COUNT);