On Wed, Nov 2, 2022 at 2:04 PM Michael Niedermayer <mich...@niedermayer.cc>
wrote:
> On Sun, Oct 30, 2022 at 05:32:35PM -0700, mindm...@gmail.com wrote:
> > From: Mark Reid <mindm...@gmail.com>
> >
> > ---
> > libswscale/output.c | 92 ++++++++++++++++++++++++
> > libswscale/swscale_unscaled.c | 4 +-
> > libswscale/tests/floatimg_cmp.c | 4 +-
> > libswscale/utils.c | 16 +++--
> > libswscale/yuv2rgb.c | 2 +
> > tests/ref/fate/filter-pixdesc-rgbaf32be | 1 +
> > tests/ref/fate/filter-pixdesc-rgbaf32le | 1 +
> > tests/ref/fate/filter-pixdesc-rgbf32be | 1 +
> > tests/ref/fate/filter-pixdesc-rgbf32le | 1 +
> > tests/ref/fate/filter-pixfmts-copy | 4 ++
> > tests/ref/fate/filter-pixfmts-crop | 4 ++
> > tests/ref/fate/filter-pixfmts-field | 4 ++
> > tests/ref/fate/filter-pixfmts-fieldorder | 4 ++
> > tests/ref/fate/filter-pixfmts-hflip | 4 ++
> > tests/ref/fate/filter-pixfmts-il | 4 ++
> > tests/ref/fate/filter-pixfmts-null | 4 ++
> > tests/ref/fate/filter-pixfmts-scale | 4 ++
> > tests/ref/fate/filter-pixfmts-transpose | 4 ++
> > tests/ref/fate/filter-pixfmts-vflip | 4 ++
> > tests/ref/fate/sws-floatimg-cmp | 16 +++++
> > 20 files changed, 170 insertions(+), 8 deletions(-)
> > create mode 100644 tests/ref/fate/filter-pixdesc-rgbaf32be
> > create mode 100644 tests/ref/fate/filter-pixdesc-rgbaf32le
> > create mode 100644 tests/ref/fate/filter-pixdesc-rgbf32be
> > create mode 100644 tests/ref/fate/filter-pixdesc-rgbf32le
> >
> > diff --git a/libswscale/output.c b/libswscale/output.c
> > index 0e1c1225a0..e2ec9cbdf5 100644
> > --- a/libswscale/output.c
> > +++ b/libswscale/output.c
> > @@ -2474,6 +2474,92 @@ yuv2gbrpf32_full_X_c(SwsContext *c, const int16_t
> *lumFilter,
> > }
> > }
> >
> > +static void
> > +yuv2rgbaf32_full_X_c(SwsContext *c, const int16_t *lumFilter,
> > + const int16_t **lumSrcx, int lumFilterSize,
> > + const int16_t *chrFilter, const int16_t **chrUSrcx,
> > + const int16_t **chrVSrcx, int chrFilterSize,
> > + const int16_t **alpSrcx, uint8_t *dest,
> > + int dstW, int y)
> > +{
> > + const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(c->dstFormat);
> > + int i;
> > + int alpha = desc->flags & AV_PIX_FMT_FLAG_ALPHA;
> > + int hasAlpha = alpha && alpSrcx;
> > + int pixelStep = alpha ? 4 : 3;
> > + uint32_t *dest32 = (uint32_t*)dest;
> > + const int32_t **lumSrc = (const int32_t**)lumSrcx;
> > + const int32_t **chrUSrc = (const int32_t**)chrUSrcx;
> > + const int32_t **chrVSrc = (const int32_t**)chrVSrcx;
> > + const int32_t **alpSrc = (const int32_t**)alpSrcx;
> > + static const float float_mult = 1.0f / 65535.0f;
> > + uint32_t a = av_float2int(1.0f);
> > +
> > + for (i = 0; i < dstW; i++) {
> > + int j;
> > + int Y = -0x40000000;
> > + int U = -(128 << 23);
> > + int V = -(128 << 23);
> > + int R, G, B, A;
> > +
> > + for (j = 0; j < lumFilterSize; j++)
> > + Y += lumSrc[j][i] * (unsigned)lumFilter[j];
> > +
> > + for (j = 0; j < chrFilterSize; j++) {
> > + U += chrUSrc[j][i] * (unsigned)chrFilter[j];
> > + V += chrVSrc[j][i] * (unsigned)chrFilter[j];
> > + }
> > +
> > + Y >>= 14;
> > + Y += 0x10000;
> > + U >>= 14;
> > + V >>= 14;
> > +
> > + if (hasAlpha) {
> > + A = -0x40000000;
> > +
> > + for (j = 0; j < lumFilterSize; j++)
> > + A += alpSrc[j][i] * (unsigned)lumFilter[j];
> > +
> > + A >>= 1;
> > + A += 0x20002000;
> > + a = av_float2int(float_mult * (float)(av_clip_uintp2(A, 30)
> >> 14));
> > + }
> > +
> > + Y -= c->yuv2rgb_y_offset;
> > + Y *= c->yuv2rgb_y_coeff;
> > + Y += 1 << 13;
> > + R = V * c->yuv2rgb_v2r_coeff;
> > + G = V * c->yuv2rgb_v2g_coeff + U * c->yuv2rgb_u2g_coeff;
> > + B = U * c->yuv2rgb_u2b_coeff;
> > +
>
> > + R = av_clip_uintp2(Y + R, 30);
> > + G = av_clip_uintp2(Y + G, 30);
> > + B = av_clip_uintp2(Y + B, 30);
>
> these additions can overflow i think given sufficiently "bad" input
> especially with the bt2020 matrix
> ive posted a proposed solution for the rgba64 / gbrp16/32f cases
> something similar can be done here
>
> thx
>
>
Make sense. I'll take a look and submit a new version with the fix.
I also have some yuv to float16 patches coming that I will incorporate the
fix into too!
> [...]
> --
> Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB
>
> Opposition brings concord. Out of discord comes the fairest harmony.
> -- Heraclitus
> _______________________________________________
> 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".
>
_______________________________________________
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".
