[FFmpeg-devel] [PATCH] lavfi/kerndeint: use width instead of linesize.

Clément Bœsch ubitux at gmail.com
Sun Jan 6 08:05:43 CET 2013


On Sun, Jan 06, 2013 at 07:58:25AM +0100, Clément Bœsch wrote:
[...]

Some explanations following.

> From afb2350a13085e8f628415d98c88310a2c554dde Mon Sep 17 00:00:00 2001
> From: =?UTF-8?q?Cl=C3=A9ment=20B=C5=93sch?= <ubitux at gmail.com>
> Date: Sun, 6 Jan 2013 07:23:43 +0100
> Subject: [PATCH] lavfi/kerndeint: fix various width/linesize bugs.
> 
> ---
>  libavfilter/vf_kerndeint.c | 38 +++++++++++++++++++-------------------
>  tests/ref/lavfi/kerndeint  |  2 +-
>  2 files changed, 20 insertions(+), 20 deletions(-)
> 
> diff --git a/libavfilter/vf_kerndeint.c b/libavfilter/vf_kerndeint.c
> index 26f2e49..e7bc74d 100644
> --- a/libavfilter/vf_kerndeint.c
> +++ b/libavfilter/vf_kerndeint.c
> @@ -39,9 +39,9 @@ typedef struct {
>      const AVClass *class;
>      int           frame; ///< frame count, starting from 0
>      int           thresh, map, order, sharp, twoway;
> -    int           vsub;
> +    int           vsub, hsub;
>      uint8_t       *tmp_data [4];  ///< temporary plane data buffer
> -    int           tmp_bwidth[4];  ///< temporary plane byte width
> +    int           tmp_linesize[4];

This is actually a real linesize.

>      int           pixel_step;
>  } KerndeintContext;
>  
> @@ -101,9 +101,10 @@ static int config_props(AVFilterLink *inlink)
>      int ret;
>  
>      kerndeint->vsub = desc->log2_chroma_h;
> -    kerndeint->pixel_step = av_get_bits_per_pixel(desc) >> 3;
> +    kerndeint->hsub = desc->log2_chroma_w;

> +    kerndeint->pixel_step = av_get_padded_bits_per_pixel(desc) >> 3;

av_get_bits_per_pixel() returns 3 for "0rgb" (and all the other packed rgb
with a 0 "channel"), we need 4, just like "rgba".

>  
> -    ret = av_image_alloc(kerndeint->tmp_data, kerndeint->tmp_bwidth,
> +    ret = av_image_alloc(kerndeint->tmp_data, kerndeint->tmp_linesize,
>                            inlink->w, inlink->h, inlink->format, 1);
>      if (ret < 0)
>          return ret;
> @@ -137,8 +138,8 @@ static int filter_frame(AVFilterLink *inlink, AVFilterBufferRef *inpic)
>      uint8_t *dstp, *dstp_saved;
>      const uint8_t *srcp_saved;
>  
> -    int src_linesize, psrc_linesize, dst_linesize, bwidth;
> -    int x, y, plane, val, hi, lo, g, h, n = kerndeint->frame++;
> +    int src_linesize, psrc_linesize, dst_linesize;
> +    int x, y, plane, val, hi, lo, w, h, n = kerndeint->frame++;
>      double valf;
>  
>      const int thresh = kerndeint->thresh;
> @@ -159,27 +160,27 @@ static int filter_frame(AVFilterLink *inlink, AVFilterBufferRef *inpic)
>  
>      for (plane = 0; inpic->data[plane] && plane < 4; plane++) {
>          h = plane == 0 ? inlink->h : inlink->h >> kerndeint->vsub;
> -        bwidth = kerndeint->tmp_bwidth[plane];
> +        w = (inlink->w * kerndeint->pixel_step) >> (plane == 0 ? 0 : kerndeint->hsub);
>  

Here we handle both planar and packed. It also works for the packed
yuyv422.

>          srcp = srcp_saved = inpic->data[plane];
>          src_linesize      = inpic->linesize[plane];
> -        psrc_linesize     = outpic->linesize[plane];
> +        psrc_linesize     = kerndeint->tmp_linesize[plane];

We must use the linesize of our temporary buffer for psrc (which is
tmp_data), since it may mismatch the one from the outpic for various
reasons (crop filter or something).

>          dstp = dstp_saved = outpic->data[plane];
>          dst_linesize      = outpic->linesize[plane];
>          srcp              = srcp_saved + (1 - order) * src_linesize;
>          dstp              = dstp_saved + (1 - order) * dst_linesize;
>  
>          for (y = 0; y < h; y += 2) {
> -            memcpy(dstp, srcp, bwidth);
> +            memcpy(dstp, srcp, w);

Here and following, we use the width, and not the linesize since we don't
want to outread.

>              srcp += 2 * src_linesize;
>              dstp += 2 * dst_linesize;
>          }
>  
>          // Copy through the lines that will be missed below.
> -        memcpy(dstp_saved + order            * dst_linesize, srcp_saved + (1 -     order) * src_linesize, bwidth);
> -        memcpy(dstp_saved + (2 + order    )  * dst_linesize, srcp_saved + (3 -     order) * src_linesize, bwidth);
> -        memcpy(dstp_saved + (h - 2 + order)  * dst_linesize, srcp_saved + (h - 1 - order) * src_linesize, bwidth);
> -        memcpy(dstp_saved + (h - 4 + order)  * dst_linesize, srcp_saved + (h - 3 - order) * src_linesize, bwidth);
> +        memcpy(dstp_saved + order            * dst_linesize, srcp_saved + (1 -     order) * src_linesize, w);
> +        memcpy(dstp_saved + (2 + order    )  * dst_linesize, srcp_saved + (3 -     order) * src_linesize, w);
> +        memcpy(dstp_saved + (h - 2 + order)  * dst_linesize, srcp_saved + (h - 1 - order) * src_linesize, w);
> +        memcpy(dstp_saved + (h - 4 + order)  * dst_linesize, srcp_saved + (h - 3 - order) * src_linesize, w);
>  
>          /* For the other field choose adaptively between using the previous field
>             or the interpolant from the current field. */
> @@ -205,18 +206,17 @@ static int filter_frame(AVFilterLink *inlink, AVFilterBufferRef *inpic)
>          dstp   = dstp_saved + 5 * dst_linesize - (1 - order) * dst_linesize;
>  
>          for (y = 5 - (1 - order); y <= h - 5 - (1 - order); y += 2) {
> -            for (x = 0; x < bwidth; x++) {
> +            for (x = 0; x < w; x++) {
>                  if (thresh == 0 || n == 0 ||
>                      (abs((int)prvp[x]  - (int)srcp[x])  > thresh) ||
>                      (abs((int)prvpp[x] - (int)srcpp[x]) > thresh) ||
>                      (abs((int)prvpn[x] - (int)srcpn[x]) > thresh)) {
>                      if (map) {
> -                        g = x & ~3;
> -
>                          if (is_packed_rgb) {
> -                            AV_WB32(dstp + g, 0xffffffff);
> -                            x = g + 3;
> +                            AV_WB32(dstp + x, 0xffffffff);
> +                            x += kerndeint->pixel_step - 1;

We just increment by the pixel step, the minus one is because the next
loop is going to +1.

>                          } else if (inlink->format == AV_PIX_FMT_YUYV422) {
> +                            int g = x & ~3;

Just a move locally because it's not needed elsewhere.

>                              // y <- 235, u <- 128, y <- 235, v <- 128
>                              AV_WB32(dstp + g, 0xeb80eb80);
>                              x = g + 3;
> @@ -287,7 +287,7 @@ static int filter_frame(AVFilterLink *inlink, AVFilterBufferRef *inpic)
>  
>          srcp = inpic->data[plane];
>          dstp = kerndeint->tmp_data[plane];
> -        av_image_copy_plane(dstp, psrc_linesize, srcp, src_linesize, bwidth, h);
> +        av_image_copy_plane(dstp, psrc_linesize, srcp, src_linesize, w, h);
>      }
>  
>      avfilter_unref_buffer(inpic);
> diff --git a/tests/ref/lavfi/kerndeint b/tests/ref/lavfi/kerndeint
> index b03f977..2769319 100644
> --- a/tests/ref/lavfi/kerndeint
> +++ b/tests/ref/lavfi/kerndeint
> @@ -6,5 +6,5 @@ bgr0                364b8bcd1c7a384902077bc7190c5ea3
>  bgra                81ac8315a4c66e363bc6fa3e99d9cd2b
>  rgb0                ae0c2afbc266345c1372276755595105
>  rgba                42a6cc9b815ca0ee69c29db3616ce25e
> -yuv420p             a935cce07c5287b92c6d5220361866ed
> +yuv420p             40ca042814882b0b791cbec38e289702
>  yuyv422             f549c98059ba9ce50e28204256d13b5d

Just one test changed; I believe this is the cause of various FATE
failures.

-- 
Clément B.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 490 bytes
Desc: not available
URL: <http://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20130106/75d28a9d/attachment.asc>


More information about the ffmpeg-devel mailing list