Hi,
On Sat, Dec 24, 2011 at 8:23 AM, Kostya Shishkov
<[email protected]> wrote:
> +#if IVI4_STREAM_ANALYSER
> +static uint8_t has_b_frames = 0;
> +static uint8_t has_transp = 0;
> +static uint8_t uses_tiling = 0;
> +static uint8_t uses_haar = 0;
> +static uint8_t uses_fullpel = 0;
> +#endif
Can those (under ifdef is fine) go in the IVI4DecContext please?
> +struct {
> + InvTransformPtr *inv_trans;
> + DCTransformPtr *dc_trans;
> + int is_2d_trans;
> +} transforms[18] = {
> + { ff_ivi_inverse_haar_8x8, ff_ivi_dc_haar_2d, 1 },
> + { NULL, NULL, 0 }, /* inverse Haar 8x1 */
> + { NULL, NULL, 0 }, /* inverse Haar 1x8 */
> + { ff_ivi_put_pixels_8x8, ff_ivi_put_dc_pixel_8x8, 1 },
> + { ff_ivi_inverse_slant_8x8, ff_ivi_dc_slant_2d, 1 },
> + { ff_ivi_row_slant8, ff_ivi_dc_row_slant, 1 },
> + { ff_ivi_col_slant8, ff_ivi_dc_col_slant, 1 },
> + { NULL, NULL, 0 }, /* inverse DCT 8x8 */
> + { NULL, NULL, 0 }, /* inverse DCT 8x1 */
> + { NULL, NULL, 0 }, /* inverse DCT 1x8 */
> + { NULL, NULL, 0 }, /* inverse Haar 4x4 */
> + { ff_ivi_inverse_slant_4x4, ff_ivi_dc_slant_2d, 1 },
> + { NULL, NULL, 0 }, /* no transform 4x4 */
Later on, you use != NULL as an indicator that it's implemented. No
transform suggests that no IDCT is needed, and that NULL is a correct
value. Which is correct?
> + /* Decode chroma subsampling. We support only 4:4 aka YVU9. */
> + if (get_bits(&ctx->gb, 2)) {
> + av_log(avctx, AV_LOG_ERROR, "Only YVU9 picture format is
> supported!\n");
> + return AVERROR_INVALIDDATA;
> + }
Do we know the meaning of the other combinations? If they're valid,
maybe ask for samples?
> + scan_indx = get_bits(&ctx->gb, 4);
> + if (scan_indx == 15) {
> + av_log(avctx, AV_LOG_ERROR, "Custom scan pattern
> encountered!\n");
> + return AVERROR_INVALIDDATA;
> + }
> + band->scan = scan_index_to_tab[scan_indx];
> +
> + band->quant_mat = get_bits(&ctx->gb, 5);
> + if (band->quant_mat == 31) {
> + av_log(avctx, AV_LOG_ERROR, "Custom quant matrix
> encountered!\n");
> + return AVERROR_INVALIDDATA;
> + }
Ask for samples x2?
Ronald
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel
