vl: Apply luma key filter before CSC conversion

Apply the luma key filter to the YCbCr values during the CSC conversion
    in video buffer shader. The initial values of max and min luma are set
    to opposite values to disable the filter initially and will be set when
    enabling it.

    Add extra parmeters min and max luma for the luma key filter in
    vl_compositor_set_csc_matrix in va, xvmc. Setting them
    to opposite value 1.f and 0.f respectively won't effect the CSC
    conversion

    v2: -Squash 1,2 and 3 into one patch to avoid breaking build of
        other components. (Christian)
        -use ureg_swizzle. (Christian)
        -change name of the variables. (Christian)

    v3: -Squash all patches in one to avoid breaking of build. (Emil)
        -wrap functions properly. (Emil)
        -use 0.0f and 1.0f instead of 0.f and 1.f respectively. (Emil)

    v4: -Divide it in two patches one which introduces the functionality
	 and assigs dummy values to the changed functions and second which
	 implements the lumakey filter. (Christian)
	-use ureg_scalar instead ureg_swizzle. (Christian)

Signed-off-by: Nayan Deshmukh <nayan26deshmukh@gmail.com>
Reviewed-by: Christian König <christian.koenig@amd.com>

src/gallium/auxiliary/vl/vl_compositor.c

@@ -132,8 +132,10 @@ create_frag_shader_video_buffer(struct vl_compositor *c)
    struct ureg_src tc;
    struct ureg_src csc[3];
    struct ureg_src sampler[3];
+   struct ureg_src lumakey;
    struct ureg_dst texel;
    struct ureg_dst fragment;
+   struct ureg_dst temp[2];
    unsigned i;
 
    shader = ureg_create(PIPE_SHADER_FRAGMENT);
@@ -145,6 +147,11 @@ create_frag_shader_video_buffer(struct vl_compositor *c)
       csc[i] = ureg_DECL_constant(shader, i);
       sampler[i] = ureg_DECL_sampler(shader, i);
    }
+
+   for (i = 0; i < 2; ++i)
+      temp[i] = ureg_DECL_temporary(shader);
+
+   lumakey = ureg_DECL_constant(shader, 3);
    texel = ureg_DECL_temporary(shader);
    fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0);
 
@@ -160,7 +167,17 @@ create_frag_shader_video_buffer(struct vl_compositor *c)
    for (i = 0; i < 3; ++i)
       ureg_DP4(shader, ureg_writemask(fragment, TGSI_WRITEMASK_X << i), csc[i], ureg_src(texel));
 
-   ureg_MOV(shader, ureg_writemask(fragment, TGSI_WRITEMASK_W), ureg_imm1f(shader, 1.0f));
+   ureg_MOV(shader, ureg_writemask(temp[0], TGSI_WRITEMASK_W),
+            ureg_scalar(ureg_src(texel), TGSI_SWIZZLE_Z));
+   ureg_SLE(shader, ureg_writemask(temp[1],TGSI_WRITEMASK_W),
+            ureg_src(temp[0]), ureg_scalar(lumakey, TGSI_SWIZZLE_X));
+   ureg_SGT(shader, ureg_writemask(temp[0],TGSI_WRITEMASK_W),
+            ureg_src(temp[0]), ureg_scalar(lumakey, TGSI_SWIZZLE_Y));
+   ureg_MAX(shader, ureg_writemask(fragment, TGSI_WRITEMASK_W),
+            ureg_src(temp[0]), ureg_src(temp[1]));
+
+   for (i = 0; i < 2; ++i)
+       ureg_release_temporary(shader, temp[i]);
 
    ureg_release_temporary(shader, texel);
    ureg_END(shader);
@@ -852,20 +869,23 @@ vl_compositor_cleanup(struct vl_compositor *c)
 }
 
 void
-vl_compositor_set_csc_matrix(struct vl_compositor_state *s, vl_csc_matrix const *matrix)
+vl_compositor_set_csc_matrix(struct vl_compositor_state *s,
+                             vl_csc_matrix const *matrix,
+                             float luma_min, float luma_max)
 {
    struct pipe_transfer *buf_transfer;
 
    assert(s);
 
-   memcpy
-   (
-      pipe_buffer_map(s->pipe, s->csc_matrix,
-                      PIPE_TRANSFER_WRITE | PIPE_TRANSFER_DISCARD_RANGE,
-                      &buf_transfer),
-      matrix,
-      sizeof(vl_csc_matrix)
-   );
+   float *ptr = pipe_buffer_map(s->pipe, s->csc_matrix,
+                               PIPE_TRANSFER_WRITE | PIPE_TRANSFER_DISCARD_RANGE,
+                               &buf_transfer);
+
+   memcpy(ptr, matrix, sizeof(vl_csc_matrix));
+
+   ptr += sizeof(vl_csc_matrix)/sizeof(float);
+   ptr[0] = luma_min;
+   ptr[1] = luma_max;
 
    pipe_buffer_unmap(s->pipe, buf_transfer);
 }
@@ -1142,13 +1162,13 @@ vl_compositor_init_state(struct vl_compositor_state *s, struct pipe_context *pip
       pipe->screen,
       PIPE_BIND_CONSTANT_BUFFER,
       PIPE_USAGE_DEFAULT,
-      sizeof(csc_matrix)
+      sizeof(csc_matrix) + 2*sizeof(float)
    );
 
    vl_compositor_clear_layers(s);
 
    vl_csc_get_matrix(VL_CSC_COLOR_STANDARD_IDENTITY, NULL, true, &csc_matrix);
-   vl_compositor_set_csc_matrix(s, (const vl_csc_matrix *)&csc_matrix);
+   vl_compositor_set_csc_matrix(s, (const vl_csc_matrix *)&csc_matrix, 1.0f, 0.0f);
 
    return true;
 }

src/gallium/auxiliary/vl/vl_compositor.h

@@ -138,7 +138,9 @@ vl_compositor_init_state(struct vl_compositor_state *state, struct pipe_context
  * set yuv -> rgba conversion matrix
  */
 void
-vl_compositor_set_csc_matrix(struct vl_compositor_state *settings, const vl_csc_matrix *matrix);
+vl_compositor_set_csc_matrix(struct vl_compositor_state *settings,
+                             const vl_csc_matrix *matrix,
+                             float luma_min, float luma_max);
 
 /**
  * reset dirty area, so it's cleared with the clear colour

src/gallium/state_trackers/va/context.c

@@ -159,7 +159,7 @@ VA_DRIVER_INIT_FUNC(VADriverContextP ctx)
    vl_compositor_init_state(&drv->cstate, drv->pipe);
 
    vl_csc_get_matrix(VL_CSC_COLOR_STANDARD_BT_601, NULL, true, &drv->csc);
-   vl_compositor_set_csc_matrix(&drv->cstate, (const vl_csc_matrix *)&drv->csc);
+   vl_compositor_set_csc_matrix(&drv->cstate, (const vl_csc_matrix *)&drv->csc, 1.0f, 0.0f);
    pipe_mutex_init(drv->mutex);
 
    ctx->pDriverData = (void *)drv;

src/gallium/state_trackers/vdpau/mixer.c

@@ -69,7 +69,7 @@ vlVdpVideoMixerCreate(VdpDevice device,
 
    vl_csc_get_matrix(VL_CSC_COLOR_STANDARD_BT_601, NULL, true, &vmixer->csc);
    if (!debug_get_bool_option("G3DVL_NO_CSC", FALSE))
-      vl_compositor_set_csc_matrix(&vmixer->cstate, (const vl_csc_matrix *)&vmixer->csc);
+      vl_compositor_set_csc_matrix(&vmixer->cstate, (const vl_csc_matrix *)&vmixer->csc, 1.0f, 0.0f);
 
    *mixer = vlAddDataHTAB(vmixer);
    if (*mixer == 0) {
@@ -671,7 +671,7 @@ vlVdpVideoMixerSetAttributeValues(VdpVideoMixer mixer,
          else
             memcpy(vmixer->csc, vdp_csc, sizeof(vl_csc_matrix));
          if (!debug_get_bool_option("G3DVL_NO_CSC", FALSE))
-            vl_compositor_set_csc_matrix(&vmixer->cstate, (const vl_csc_matrix *)&vmixer->csc);
+            vl_compositor_set_csc_matrix(&vmixer->cstate, (const vl_csc_matrix *)&vmixer->csc, 1.0f, 0.0f);
          break;
 
       case VDP_VIDEO_MIXER_ATTRIBUTE_NOISE_REDUCTION_LEVEL:

src/gallium/state_trackers/vdpau/output.c

@@ -492,9 +492,9 @@ vlVdpOutputSurfacePutBitsYCbCr(VdpOutputSurface surface,
    if (!csc_matrix) {
       vl_csc_matrix csc;
       vl_csc_get_matrix(VL_CSC_COLOR_STANDARD_BT_601, NULL, 1, &csc);
-      vl_compositor_set_csc_matrix(cstate, (const vl_csc_matrix*)&csc);
+      vl_compositor_set_csc_matrix(cstate, (const vl_csc_matrix*)&csc, 1.0f, 0.0f);
    } else {
-      vl_compositor_set_csc_matrix(cstate, csc_matrix);
+      vl_compositor_set_csc_matrix(cstate, csc_matrix, 1.0f, 0.0f);
    }
 
    vl_compositor_clear_layers(cstate);

src/gallium/state_trackers/xvmc/attributes.c

@@ -110,7 +110,7 @@ Status XvMCSetAttribute(Display *dpy, XvMCContext *context, Atom attribute, int
       context_priv->color_standard,
       &context_priv->procamp, true, &csc
    );
-   vl_compositor_set_csc_matrix(&context_priv->cstate, (const vl_csc_matrix *)&csc);
+   vl_compositor_set_csc_matrix(&context_priv->cstate, (const vl_csc_matrix *)&csc, 1.0f, 0.0f);
 
    XVMC_MSG(XVMC_TRACE, "[XvMC] Set attribute %s to value %d.\n", attr, value);
 

src/gallium/state_trackers/xvmc/context.c

@@ -293,7 +293,7 @@ Status XvMCCreateContext(Display *dpy, XvPortID port, int surface_type_id,
       context_priv->color_standard,
       &context_priv->procamp, true, &csc
    );
-   vl_compositor_set_csc_matrix(&context_priv->cstate, (const vl_csc_matrix *)&csc);
+   vl_compositor_set_csc_matrix(&context_priv->cstate, (const vl_csc_matrix *)&csc, 1.0f, 0.0f);
 
    context_priv->vscreen = vscreen;
    context_priv->pipe = pipe;