]> git.sesse.net Git - ffmpeg/blobdiff - libavfilter/vf_colorchannelmixer.c
Merge commit '0801853e640624537db386727b36fa97aa6258e7'
[ffmpeg] / libavfilter / vf_colorchannelmixer.c
index 2e068fa7947c7c9be3cff1e7b3e516bca05533b5..3a9cd37b78aeef29d9acfc1c4a99fae8b2d07ebc 100644 (file)
@@ -19,6 +19,7 @@
  */
 
 #include "libavutil/opt.h"
+#include "libavutil/pixdesc.h"
 #include "avfilter.h"
 #include "drawutils.h"
 #include "formats.h"
 #define B 2
 #define A 3
 
+typedef struct ThreadData {
+    AVFrame *in, *out;
+} ThreadData;
+
 typedef struct ColorChannelMixerContext {
     const AVClass *class;
     double rr, rg, rb, ra;
@@ -42,6 +47,8 @@ typedef struct ColorChannelMixerContext {
     int *buffer;
 
     uint8_t rgba_map[4];
+
+    int (*filter_slice)(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs);
 } ColorChannelMixerContext;
 
 #define OFFSET(x) offsetof(ColorChannelMixerContext, x)
@@ -78,6 +85,12 @@ static int query_formats(AVFilterContext *ctx)
         AV_PIX_FMT_RGB0,   AV_PIX_FMT_BGR0,
         AV_PIX_FMT_RGB48,  AV_PIX_FMT_BGR48,
         AV_PIX_FMT_RGBA64, AV_PIX_FMT_BGRA64,
+        AV_PIX_FMT_GBRP,   AV_PIX_FMT_GBRAP,
+        AV_PIX_FMT_GBRP9,
+        AV_PIX_FMT_GBRP10, AV_PIX_FMT_GBRAP10,
+        AV_PIX_FMT_GBRP12, AV_PIX_FMT_GBRAP12,
+        AV_PIX_FMT_GBRP14,
+        AV_PIX_FMT_GBRP16, AV_PIX_FMT_GBRAP16,
         AV_PIX_FMT_NONE
     };
 
@@ -87,25 +100,319 @@ static int query_formats(AVFilterContext *ctx)
     return ff_set_common_formats(ctx, fmts_list);
 }
 
+static av_always_inline int filter_slice_rgba_planar(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs,
+                                                     int have_alpha)
+{
+    ColorChannelMixerContext *s = ctx->priv;
+    ThreadData *td = arg;
+    AVFrame *in = td->in;
+    AVFrame *out = td->out;
+    const int slice_start = (out->height * jobnr) / nb_jobs;
+    const int slice_end = (out->height * (jobnr+1)) / nb_jobs;
+    const uint8_t *srcg = in->data[0] + slice_start * in->linesize[0];
+    const uint8_t *srcb = in->data[1] + slice_start * in->linesize[1];
+    const uint8_t *srcr = in->data[2] + slice_start * in->linesize[2];
+    const uint8_t *srca = in->data[3] + slice_start * in->linesize[3];
+    uint8_t *dstg = out->data[0] + slice_start * out->linesize[0];
+    uint8_t *dstb = out->data[1] + slice_start * out->linesize[1];
+    uint8_t *dstr = out->data[2] + slice_start * out->linesize[2];
+    uint8_t *dsta = out->data[3] + slice_start * out->linesize[3];
+    int i, j;
+
+    for (i = slice_start; i < slice_end; i++) {
+        for (j = 0; j < out->width; j++) {
+            const uint8_t rin = srcr[j];
+            const uint8_t gin = srcg[j];
+            const uint8_t bin = srcb[j];
+            const uint8_t ain = srca[j];
+
+            dstr[j] = av_clip_uint8(s->lut[R][R][rin] +
+                                    s->lut[R][G][gin] +
+                                    s->lut[R][B][bin] +
+                                    (have_alpha == 1 ? s->lut[R][A][ain] : 0));
+            dstg[j] = av_clip_uint8(s->lut[G][R][rin] +
+                                    s->lut[G][G][gin] +
+                                    s->lut[G][B][bin] +
+                                    (have_alpha == 1 ? s->lut[G][A][ain] : 0));
+            dstb[j] = av_clip_uint8(s->lut[B][R][rin] +
+                                    s->lut[B][G][gin] +
+                                    s->lut[B][B][bin] +
+                                    (have_alpha == 1 ? s->lut[B][A][ain] : 0));
+            if (have_alpha == 1) {
+                dsta[j] = av_clip_uint8(s->lut[A][R][rin] +
+                                        s->lut[A][G][gin] +
+                                        s->lut[A][B][bin] +
+                                        s->lut[A][A][ain]);
+            }
+        }
+
+        srcg += in->linesize[0];
+        srcb += in->linesize[1];
+        srcr += in->linesize[2];
+        srca += in->linesize[3];
+        dstg += out->linesize[0];
+        dstb += out->linesize[1];
+        dstr += out->linesize[2];
+        dsta += out->linesize[3];
+    }
+
+    return 0;
+}
+
+static av_always_inline int filter_slice_rgba16_planar(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs,
+                                                       int have_alpha, int depth)
+{
+    ColorChannelMixerContext *s = ctx->priv;
+    ThreadData *td = arg;
+    AVFrame *in = td->in;
+    AVFrame *out = td->out;
+    const int slice_start = (out->height * jobnr) / nb_jobs;
+    const int slice_end = (out->height * (jobnr+1)) / nb_jobs;
+    const uint16_t *srcg = (const uint16_t *)(in->data[0] + slice_start * in->linesize[0]);
+    const uint16_t *srcb = (const uint16_t *)(in->data[1] + slice_start * in->linesize[1]);
+    const uint16_t *srcr = (const uint16_t *)(in->data[2] + slice_start * in->linesize[2]);
+    const uint16_t *srca = (const uint16_t *)(in->data[3] + slice_start * in->linesize[3]);
+    uint16_t *dstg = (uint16_t *)(out->data[0] + slice_start * out->linesize[0]);
+    uint16_t *dstb = (uint16_t *)(out->data[1] + slice_start * out->linesize[1]);
+    uint16_t *dstr = (uint16_t *)(out->data[2] + slice_start * out->linesize[2]);
+    uint16_t *dsta = (uint16_t *)(out->data[3] + slice_start * out->linesize[3]);
+    int i, j;
+
+    for (i = slice_start; i < slice_end; i++) {
+        for (j = 0; j < out->width; j++) {
+            const uint16_t rin = srcr[j];
+            const uint16_t gin = srcg[j];
+            const uint16_t bin = srcb[j];
+            const uint16_t ain = srca[j];
+
+            dstr[j] = av_clip_uintp2(s->lut[R][R][rin] +
+                                     s->lut[R][G][gin] +
+                                     s->lut[R][B][bin] +
+                                     (have_alpha == 1 ? s->lut[R][A][ain] : 0), depth);
+            dstg[j] = av_clip_uintp2(s->lut[G][R][rin] +
+                                     s->lut[G][G][gin] +
+                                     s->lut[G][B][bin] +
+                                     (have_alpha == 1 ? s->lut[G][A][ain] : 0), depth);
+            dstb[j] = av_clip_uintp2(s->lut[B][R][rin] +
+                                     s->lut[B][G][gin] +
+                                     s->lut[B][B][bin] +
+                                     (have_alpha == 1 ? s->lut[B][A][ain] : 0), depth);
+            if (have_alpha == 1) {
+                dsta[j] = av_clip_uintp2(s->lut[A][R][rin] +
+                                         s->lut[A][G][gin] +
+                                         s->lut[A][B][bin] +
+                                         s->lut[A][A][ain], depth);
+            }
+        }
+
+        srcg += in->linesize[0] / 2;
+        srcb += in->linesize[1] / 2;
+        srcr += in->linesize[2] / 2;
+        srca += in->linesize[3] / 2;
+        dstg += out->linesize[0] / 2;
+        dstb += out->linesize[1] / 2;
+        dstr += out->linesize[2] / 2;
+        dsta += out->linesize[3] / 2;
+    }
+
+    return 0;
+}
+
+static int filter_slice_gbrp(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
+{
+    return filter_slice_rgba_planar(ctx, arg, jobnr, nb_jobs, 0);
+}
+
+static int filter_slice_gbrap(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
+{
+    return filter_slice_rgba_planar(ctx, arg, jobnr, nb_jobs, 1);
+}
+
+static int filter_slice_gbrp9(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
+{
+    return filter_slice_rgba16_planar(ctx, arg, jobnr, nb_jobs, 0, 9);
+}
+
+static int filter_slice_gbrp10(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
+{
+    return filter_slice_rgba16_planar(ctx, arg, jobnr, nb_jobs, 0, 10);
+}
+
+static int filter_slice_gbrap10(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
+{
+    return filter_slice_rgba16_planar(ctx, arg, jobnr, nb_jobs, 1, 10);
+}
+
+static int filter_slice_gbrp12(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
+{
+    return filter_slice_rgba16_planar(ctx, arg, jobnr, nb_jobs, 0, 12);
+}
+
+static int filter_slice_gbrap12(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
+{
+    return filter_slice_rgba16_planar(ctx, arg, jobnr, nb_jobs, 1, 12);
+}
+
+static int filter_slice_gbrp14(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
+{
+    return filter_slice_rgba16_planar(ctx, arg, jobnr, nb_jobs, 0, 14);
+}
+
+static int filter_slice_gbrp16(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
+{
+    return filter_slice_rgba16_planar(ctx, arg, jobnr, nb_jobs, 0, 16);
+}
+
+static int filter_slice_gbrap16(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
+{
+    return filter_slice_rgba16_planar(ctx, arg, jobnr, nb_jobs, 1, 16);
+}
+
+static av_always_inline int filter_slice_rgba_packed(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs,
+                                                     int have_alpha, int step)
+{
+    ColorChannelMixerContext *s = ctx->priv;
+    ThreadData *td = arg;
+    AVFrame *in = td->in;
+    AVFrame *out = td->out;
+    const int slice_start = (out->height * jobnr) / nb_jobs;
+    const int slice_end = (out->height * (jobnr+1)) / nb_jobs;
+    const uint8_t roffset = s->rgba_map[R];
+    const uint8_t goffset = s->rgba_map[G];
+    const uint8_t boffset = s->rgba_map[B];
+    const uint8_t aoffset = s->rgba_map[A];
+    const uint8_t *srcrow = in->data[0] + slice_start * in->linesize[0];
+    uint8_t *dstrow = out->data[0] + slice_start * out->linesize[0];
+    int i, j;
+
+    for (i = slice_start; i < slice_end; i++) {
+        const uint8_t *src = srcrow;
+        uint8_t *dst = dstrow;
+
+        for (j = 0; j < out->width * step; j += step) {
+            const uint8_t rin = src[j + roffset];
+            const uint8_t gin = src[j + goffset];
+            const uint8_t bin = src[j + boffset];
+            const uint8_t ain = src[j + aoffset];
+
+            dst[j + roffset] = av_clip_uint8(s->lut[R][R][rin] +
+                                             s->lut[R][G][gin] +
+                                             s->lut[R][B][bin] +
+                                             (have_alpha == 1 ? s->lut[R][A][ain] : 0));
+            dst[j + goffset] = av_clip_uint8(s->lut[G][R][rin] +
+                                             s->lut[G][G][gin] +
+                                             s->lut[G][B][bin] +
+                                             (have_alpha == 1 ? s->lut[G][A][ain] : 0));
+            dst[j + boffset] = av_clip_uint8(s->lut[B][R][rin] +
+                                             s->lut[B][G][gin] +
+                                             s->lut[B][B][bin] +
+                                             (have_alpha == 1 ? s->lut[B][A][ain] : 0));
+            if (have_alpha == 1) {
+                dst[j + aoffset] = av_clip_uint8(s->lut[A][R][rin] +
+                                                 s->lut[A][G][gin] +
+                                                 s->lut[A][B][bin] +
+                                                 s->lut[A][A][ain]);
+            } else if (have_alpha == -1 && in != out)
+                dst[j + aoffset] = 0;
+        }
+
+        srcrow += in->linesize[0];
+        dstrow += out->linesize[0];
+    }
+
+    return 0;
+}
+
+static av_always_inline int filter_slice_rgba16_packed(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs,
+                                                       int have_alpha, int step)
+{
+    ColorChannelMixerContext *s = ctx->priv;
+    ThreadData *td = arg;
+    AVFrame *in = td->in;
+    AVFrame *out = td->out;
+    const int slice_start = (out->height * jobnr) / nb_jobs;
+    const int slice_end = (out->height * (jobnr+1)) / nb_jobs;
+    const uint8_t roffset = s->rgba_map[R];
+    const uint8_t goffset = s->rgba_map[G];
+    const uint8_t boffset = s->rgba_map[B];
+    const uint8_t aoffset = s->rgba_map[A];
+    const uint8_t *srcrow = in->data[0] + slice_start * in->linesize[0];
+    uint8_t *dstrow = out->data[0] + slice_start * out->linesize[0];
+    int i, j;
+
+    for (i = slice_start; i < slice_end; i++) {
+        const uint16_t *src = (const uint16_t *)srcrow;
+        uint16_t *dst = (uint16_t *)dstrow;
+
+        for (j = 0; j < out->width * step; j += step) {
+            const uint16_t rin = src[j + roffset];
+            const uint16_t gin = src[j + goffset];
+            const uint16_t bin = src[j + boffset];
+            const uint16_t ain = src[j + aoffset];
+
+            dst[j + roffset] = av_clip_uint16(s->lut[R][R][rin] +
+                                              s->lut[R][G][gin] +
+                                              s->lut[R][B][bin] +
+                                              (have_alpha == 1 ? s->lut[R][A][ain] : 0));
+            dst[j + goffset] = av_clip_uint16(s->lut[G][R][rin] +
+                                              s->lut[G][G][gin] +
+                                              s->lut[G][B][bin] +
+                                              (have_alpha == 1 ? s->lut[G][A][ain] : 0));
+            dst[j + boffset] = av_clip_uint16(s->lut[B][R][rin] +
+                                              s->lut[B][G][gin] +
+                                              s->lut[B][B][bin] +
+                                              (have_alpha == 1 ? s->lut[B][A][ain] : 0));
+            if (have_alpha == 1) {
+                dst[j + aoffset] = av_clip_uint16(s->lut[A][R][rin] +
+                                                  s->lut[A][G][gin] +
+                                                  s->lut[A][B][bin] +
+                                                  s->lut[A][A][ain]);
+            }
+        }
+
+        srcrow += in->linesize[0];
+        dstrow += out->linesize[0];
+    }
+
+    return 0;
+}
+
+static int filter_slice_rgba64(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
+{
+    return filter_slice_rgba16_packed(ctx, arg, jobnr, nb_jobs, 1, 4);
+}
+
+static int filter_slice_rgb48(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
+{
+    return filter_slice_rgba16_packed(ctx, arg, jobnr, nb_jobs, 0, 3);
+}
+
+static int filter_slice_rgba(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
+{
+    return filter_slice_rgba_packed(ctx, arg, jobnr, nb_jobs, 1, 4);
+}
+
+static int filter_slice_rgb24(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
+{
+    return filter_slice_rgba_packed(ctx, arg, jobnr, nb_jobs, 0, 3);
+}
+
+static int filter_slice_rgb0(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
+{
+    return filter_slice_rgba_packed(ctx, arg, jobnr, nb_jobs, -1, 4);
+}
+
 static int config_output(AVFilterLink *outlink)
 {
     AVFilterContext *ctx = outlink->src;
     ColorChannelMixerContext *s = ctx->priv;
+    const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(outlink->format);
+    const int depth = desc->comp[0].depth;
     int i, j, size, *buffer;
 
     ff_fill_rgba_map(s->rgba_map, outlink->format);
 
-    switch (outlink->format) {
-    case AV_PIX_FMT_RGB48:
-    case AV_PIX_FMT_BGR48:
-    case AV_PIX_FMT_RGBA64:
-    case AV_PIX_FMT_BGRA64:
-        size = 65536;
-        break;
-    default:
-        size = 256;
-    }
-
+    size = 1 << depth;
     s->buffer = buffer = av_malloc(16 * size * sizeof(*s->buffer));
     if (!s->buffer)
         return AVERROR(ENOMEM);
@@ -136,192 +443,92 @@ static int config_output(AVFilterLink *outlink)
         s->lut[A][A][i] = lrint(i * s->aa);
     }
 
-    return 0;
-}
-
-static int filter_frame(AVFilterLink *inlink, AVFrame *in)
-{
-    AVFilterContext *ctx = inlink->dst;
-    ColorChannelMixerContext *s = ctx->priv;
-    AVFilterLink *outlink = ctx->outputs[0];
-    const uint8_t roffset = s->rgba_map[R];
-    const uint8_t goffset = s->rgba_map[G];
-    const uint8_t boffset = s->rgba_map[B];
-    const uint8_t aoffset = s->rgba_map[A];
-    const uint8_t *srcrow = in->data[0];
-    uint8_t *dstrow;
-    AVFrame *out;
-    int i, j;
-
-    if (av_frame_is_writable(in)) {
-        out = in;
-    } else {
-        out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
-        if (!out) {
-            av_frame_free(&in);
-            return AVERROR(ENOMEM);
-        }
-        av_frame_copy_props(out, in);
-    }
-
-    dstrow = out->data[0];
     switch (outlink->format) {
     case AV_PIX_FMT_BGR24:
     case AV_PIX_FMT_RGB24:
-        for (i = 0; i < outlink->h; i++) {
-            const uint8_t *src = srcrow;
-            uint8_t *dst = dstrow;
-
-            for (j = 0; j < outlink->w * 3; j += 3) {
-                const uint8_t rin = src[j + roffset];
-                const uint8_t gin = src[j + goffset];
-                const uint8_t bin = src[j + boffset];
-
-                dst[j + roffset] = av_clip_uint8(s->lut[R][R][rin] +
-                                                 s->lut[R][G][gin] +
-                                                 s->lut[R][B][bin]);
-                dst[j + goffset] = av_clip_uint8(s->lut[G][R][rin] +
-                                                 s->lut[G][G][gin] +
-                                                 s->lut[G][B][bin]);
-                dst[j + boffset] = av_clip_uint8(s->lut[B][R][rin] +
-                                                 s->lut[B][G][gin] +
-                                                 s->lut[B][B][bin]);
-            }
-
-            srcrow += in->linesize[0];
-            dstrow += out->linesize[0];
-        }
+        s->filter_slice = filter_slice_rgb24;
         break;
     case AV_PIX_FMT_0BGR:
     case AV_PIX_FMT_0RGB:
     case AV_PIX_FMT_BGR0:
     case AV_PIX_FMT_RGB0:
-        for (i = 0; i < outlink->h; i++) {
-            const uint8_t *src = srcrow;
-            uint8_t *dst = dstrow;
-
-            for (j = 0; j < outlink->w * 4; j += 4) {
-                const uint8_t rin = src[j + roffset];
-                const uint8_t gin = src[j + goffset];
-                const uint8_t bin = src[j + boffset];
-
-                dst[j + roffset] = av_clip_uint8(s->lut[R][R][rin] +
-                                                 s->lut[R][G][gin] +
-                                                 s->lut[R][B][bin]);
-                dst[j + goffset] = av_clip_uint8(s->lut[G][R][rin] +
-                                                 s->lut[G][G][gin] +
-                                                 s->lut[G][B][bin]);
-                dst[j + boffset] = av_clip_uint8(s->lut[B][R][rin] +
-                                                 s->lut[B][G][gin] +
-                                                 s->lut[B][B][bin]);
-                if (in != out)
-                    dst[j + aoffset] = 0;
-            }
-
-            srcrow += in->linesize[0];
-            dstrow += out->linesize[0];
-        }
+        s->filter_slice = filter_slice_rgb0;
         break;
     case AV_PIX_FMT_ABGR:
     case AV_PIX_FMT_ARGB:
     case AV_PIX_FMT_BGRA:
     case AV_PIX_FMT_RGBA:
-        for (i = 0; i < outlink->h; i++) {
-            const uint8_t *src = srcrow;
-            uint8_t *dst = dstrow;
-
-            for (j = 0; j < outlink->w * 4; j += 4) {
-                const uint8_t rin = src[j + roffset];
-                const uint8_t gin = src[j + goffset];
-                const uint8_t bin = src[j + boffset];
-                const uint8_t ain = src[j + aoffset];
-
-                dst[j + roffset] = av_clip_uint8(s->lut[R][R][rin] +
-                                                 s->lut[R][G][gin] +
-                                                 s->lut[R][B][bin] +
-                                                 s->lut[R][A][ain]);
-                dst[j + goffset] = av_clip_uint8(s->lut[G][R][rin] +
-                                                 s->lut[G][G][gin] +
-                                                 s->lut[G][B][bin] +
-                                                 s->lut[G][A][ain]);
-                dst[j + boffset] = av_clip_uint8(s->lut[B][R][rin] +
-                                                 s->lut[B][G][gin] +
-                                                 s->lut[B][B][bin] +
-                                                 s->lut[B][A][ain]);
-                dst[j + aoffset] = av_clip_uint8(s->lut[A][R][rin] +
-                                                 s->lut[A][G][gin] +
-                                                 s->lut[A][B][bin] +
-                                                 s->lut[A][A][ain]);
-            }
-
-            srcrow += in->linesize[0];
-            dstrow += out->linesize[0];
-        }
+        s->filter_slice = filter_slice_rgba;
         break;
     case AV_PIX_FMT_BGR48:
     case AV_PIX_FMT_RGB48:
-        for (i = 0; i < outlink->h; i++) {
-            const uint16_t *src = (const uint16_t *)srcrow;
-            uint16_t *dst = (uint16_t *)dstrow;
-
-            for (j = 0; j < outlink->w * 3; j += 3) {
-                const uint16_t rin = src[j + roffset];
-                const uint16_t gin = src[j + goffset];
-                const uint16_t bin = src[j + boffset];
-
-                dst[j + roffset] = av_clip_uint16(s->lut[R][R][rin] +
-                                                  s->lut[R][G][gin] +
-                                                  s->lut[R][B][bin]);
-                dst[j + goffset] = av_clip_uint16(s->lut[G][R][rin] +
-                                                  s->lut[G][G][gin] +
-                                                  s->lut[G][B][bin]);
-                dst[j + boffset] = av_clip_uint16(s->lut[B][R][rin] +
-                                                  s->lut[B][G][gin] +
-                                                  s->lut[B][B][bin]);
-            }
-
-            srcrow += in->linesize[0];
-            dstrow += out->linesize[0];
-        }
+        s->filter_slice = filter_slice_rgb48;
         break;
     case AV_PIX_FMT_BGRA64:
     case AV_PIX_FMT_RGBA64:
-        for (i = 0; i < outlink->h; i++) {
-            const uint16_t *src = (const uint16_t *)srcrow;
-            uint16_t *dst = (uint16_t *)dstrow;
-
-            for (j = 0; j < outlink->w * 4; j += 4) {
-                const uint16_t rin = src[j + roffset];
-                const uint16_t gin = src[j + goffset];
-                const uint16_t bin = src[j + boffset];
-                const uint16_t ain = src[j + aoffset];
-
-                dst[j + roffset] = av_clip_uint16(s->lut[R][R][rin] +
-                                                  s->lut[R][G][gin] +
-                                                  s->lut[R][B][bin] +
-                                                  s->lut[R][A][ain]);
-                dst[j + goffset] = av_clip_uint16(s->lut[G][R][rin] +
-                                                  s->lut[G][G][gin] +
-                                                  s->lut[G][B][bin] +
-                                                  s->lut[G][A][ain]);
-                dst[j + boffset] = av_clip_uint16(s->lut[B][R][rin] +
-                                                  s->lut[B][G][gin] +
-                                                  s->lut[B][B][bin] +
-                                                  s->lut[B][A][ain]);
-                dst[j + aoffset] = av_clip_uint16(s->lut[A][R][rin] +
-                                                  s->lut[A][G][gin] +
-                                                  s->lut[A][B][bin] +
-                                                  s->lut[A][A][ain]);
-            }
+        s->filter_slice = filter_slice_rgba64;
+        break;
+    case AV_PIX_FMT_GBRP:
+        s->filter_slice = filter_slice_gbrp;
+        break;
+    case AV_PIX_FMT_GBRAP:
+        s->filter_slice = filter_slice_gbrap;
+        break;
+    case AV_PIX_FMT_GBRP9:
+        s->filter_slice = filter_slice_gbrp9;
+        break;
+    case AV_PIX_FMT_GBRP10:
+        s->filter_slice = filter_slice_gbrp10;
+        break;
+    case AV_PIX_FMT_GBRAP10:
+        s->filter_slice = filter_slice_gbrap10;
+        break;
+    case AV_PIX_FMT_GBRP12:
+        s->filter_slice = filter_slice_gbrp12;
+        break;
+    case AV_PIX_FMT_GBRAP12:
+        s->filter_slice = filter_slice_gbrap12;
+        break;
+    case AV_PIX_FMT_GBRP14:
+        s->filter_slice = filter_slice_gbrp14;
+        break;
+    case AV_PIX_FMT_GBRP16:
+        s->filter_slice = filter_slice_gbrp16;
+        break;
+    case AV_PIX_FMT_GBRAP16:
+        s->filter_slice = filter_slice_gbrap16;
+        break;
+    }
+
+    return 0;
+}
+
+static int filter_frame(AVFilterLink *inlink, AVFrame *in)
+{
+    AVFilterContext *ctx = inlink->dst;
+    ColorChannelMixerContext *s = ctx->priv;
+    AVFilterLink *outlink = ctx->outputs[0];
+    ThreadData td;
+    AVFrame *out;
 
-            srcrow += in->linesize[0];
-            dstrow += out->linesize[0];
+    if (av_frame_is_writable(in)) {
+        out = in;
+    } else {
+        out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
+        if (!out) {
+            av_frame_free(&in);
+            return AVERROR(ENOMEM);
         }
+        av_frame_copy_props(out, in);
     }
 
+    td.in = in;
+    td.out = out;
+    ctx->internal->execute(ctx, s->filter_slice, &td, NULL, FFMIN(outlink->h, ff_filter_get_nb_threads(ctx)));
+
     if (in != out)
         av_frame_free(&in);
-    return ff_filter_frame(ctx->outputs[0], out);
+    return ff_filter_frame(outlink, out);
 }
 
 static av_cold void uninit(AVFilterContext *ctx)
@@ -358,5 +565,5 @@ AVFilter ff_vf_colorchannelmixer = {
     .query_formats = query_formats,
     .inputs        = colorchannelmixer_inputs,
     .outputs       = colorchannelmixer_outputs,
-    .flags         = AVFILTER_FLAG_SUPPORT_TIMELINE_GENERIC,
+    .flags         = AVFILTER_FLAG_SUPPORT_TIMELINE_GENERIC | AVFILTER_FLAG_SLICE_THREADS,
 };