IndecisiveTurtle:
> From: IndecisiveTurtle <geoste...@gmail.com>
>
> Performance wise, encoding a 1080p 1-minute video is performed in about 2.5
> minutes with the cpu encoder running on my Ryzen 5 4600H, while it takes
> about 30 seconds on my NVIDIA GTX 1650
>
> Haar shader has a subgroup optimized variant that applies when configured
> wavelet depth allows it
> ---
> +
> +void put_vc2_ue_uint(inout PutBitContext pb, uint val)
> +{
> + int pbits = 0, topbit = 1, maxval = 1, bits = 0;
> + if (val == 0)
> + {
> + put_bits(pb, 1, 1);
> + return;
> + }
> + val++;
> +
> + while (val > maxval)
> + {
> + topbit <<= 1;
> + bits++;
> + maxval <<= 1;
> + maxval |= 1;
> + }
> +
> + for (int i = 0; i < bits; i++)
> + {
> + topbit >>= 1;
> + pbits <<= 2;
> + if ((val & topbit) != 0)
> + pbits |= 1;
> + }
> +
> + put_bits(pb, bits * 2 + 1, (pbits << 1) | 1);
> +}
> +
You are still using the old and inefficient way to write VC-2
exponential coded integers. Improving this gave a nice speed boost to
the software encoder.
- Andreas
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".
