Commit 08d2f548 authored by Peter de Rivaz's avatar Peter de Rivaz

Fixed idct16x16_10 highbitdepth transform

In the case when there are only non-zero coefficients
in the first 4x4 block a special routine is called.
The highbitdepth optimized version of this routine
examined the wrong positions when deciding whether
to call an assembler or C inverse transform.

Change-Id: I62da663ca11775dadb66e402e42f4a1cb1927893
parent a1b72611
......@@ -4463,11 +4463,11 @@ void vp9_highbd_idct16x16_10_add_sse2(const tran_low_t *input, uint8_t *dest8,
// Find the min & max for the row transform
// Since all non-zero dct coefficients are in upper-left 4x4 area,
// we only need to consider first 4 rows here.
max_input = _mm_max_epi16(inptr[0], inptr[2]);
min_input = _mm_min_epi16(inptr[0], inptr[2]);
max_input = _mm_max_epi16(inptr[0], inptr[1]);
min_input = _mm_min_epi16(inptr[0], inptr[1]);
for (i = 2; i < 4; i++) {
max_input = _mm_max_epi16(max_input, inptr[2*i]);
min_input = _mm_min_epi16(min_input, inptr[2*i]);
max_input = _mm_max_epi16(max_input, inptr[i]);
min_input = _mm_min_epi16(min_input, inptr[i]);
max_input = _mm_cmpgt_epi16(max_input, max);
min_input = _mm_cmplt_epi16(min_input, min);
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment