Hi Dmitry,
W dniu 19.03.2023 o 00:27, Dmitry Osipenko pisze:
On 3/19/23 02:23, Daniel Almeida wrote:
Hi Andrzej, one thing I noticed while looking at this:
+struct hantro_vp8_entropy {
+ /* TODO: int32_t? */
+ int32_t coeff_prob[4][8][3][11];
+ int32_t coeff_prob_old[4][8][3][11];
+ int32_t mv_prob[2][19];
+ int32_t mv_prob_old[2][19];
+ int32_t y_mode_prob[4];
+ int32_t uv_mode_prob[3];
+};
In VP9, probabilities are coded in the 0..255 range, so if you want to
save space here, you can go with uint8_t.
BTW, there is a mix of int32_t and s32 in the code, the s32 is a correct
choice. Similar for u8. Checkpatch --strict should warn about it all.
Thanks. Something for a series proper (not RFC) indeed.
Andrzej