On 05/01/2011 09:03 AM, Reinhard Tartler wrote:
> +static void encode_rgb48_10bit(AVCodecContext *avctx, const AVPicture *pic,
> uint8_t *dst)
> +{
> + DPXContext *s = avctx->priv_data;
> + const uint8_t *src = pic->data[0];
> + int x, y;
> +
> + for (y = 0; y < avctx->height; y++) {
> + for (x = 0; x < avctx->width; x++) {
> + int value;
> + if ((avctx->pix_fmt & 1)) {
> + value = ((AV_RB16(src + 6*x + 4) & 0xFFC0) >> 4)
> + | ((AV_RB16(src + 6*x + 2) & 0xFFC0) << 6)
> + | ((AV_RB16(src + 6*x + 0) & 0xFFC0) << 16);
> + } else {
> + value = ((AV_RL16(src + 6*x + 4) & 0xFFC0) >> 4)
> + | ((AV_RL16(src + 6*x + 2) & 0xFFC0) << 6)
> + | ((AV_RL16(src + 6*x + 0) & 0xFFC0) << 16);
> + }
> + write32(dst, value);
> + dst += 4;
> + }
> + src += pic->linesize[0];
> + }
> +}
It might be faster here to put the "if (pix_fmt&1)" check outside the
loops. If gcc does something sane with it that might not be necessary
though. Also, (avctx->pix_fmt & 1) is not really obvious here. Either
it should check for RGB48BE or there should be a comment.
Rest of the encoder looks fine to me.
-Justin
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel
