meta: Support 16x MSAA in the multisample scaled blit shader

v2: Fix the x_scale in the shader. Remove the doubts in the commit
    message.
Reviewed-by: Anuj Phogat <[email protected]>

3 files changed, 35 insertions, 10 deletions

diff --git a/src/mesa/drivers/common/meta.h b/src/mesa/drivers/common/meta.h
index 23fa209905d..d742eaa9f67 100644
--- a/src/mesa/drivers/common/meta.h
+++ b/src/mesa/drivers/common/meta.h
@@ -285,9 +285,11 @@ enum blit_msaa_shader {
    BLIT_2X_MSAA_SHADER_2D_MULTISAMPLE_SCALED_RESOLVE,
    BLIT_4X_MSAA_SHADER_2D_MULTISAMPLE_SCALED_RESOLVE,
    BLIT_8X_MSAA_SHADER_2D_MULTISAMPLE_SCALED_RESOLVE,
+   BLIT_16X_MSAA_SHADER_2D_MULTISAMPLE_SCALED_RESOLVE,
    BLIT_2X_MSAA_SHADER_2D_MULTISAMPLE_ARRAY_SCALED_RESOLVE,
    BLIT_4X_MSAA_SHADER_2D_MULTISAMPLE_ARRAY_SCALED_RESOLVE,
    BLIT_8X_MSAA_SHADER_2D_MULTISAMPLE_ARRAY_SCALED_RESOLVE,
+   BLIT_16X_MSAA_SHADER_2D_MULTISAMPLE_ARRAY_SCALED_RESOLVE,
    BLIT_MSAA_SHADER_COUNT,
 };
 
diff --git a/src/mesa/drivers/common/meta_blit.c b/src/mesa/drivers/common/meta_blit.c
index 5972a5af0c9..b92c2e2f22b 100644
--- a/src/mesa/drivers/common/meta_blit.c
+++ b/src/mesa/drivers/common/meta_blit.c
@@ -72,20 +72,25 @@ setup_glsl_msaa_blit_scaled_shader(struct gl_context *ctx,
    char *sample_map_expr = rzalloc_size(mem_ctx, 1);
    char *texel_fetch_macro = rzalloc_size(mem_ctx, 1);
    const char *sampler_array_suffix = "";
-   float y_scale;
+   float x_scale, y_scale;
    enum blit_msaa_shader shader_index;
 
    assert(src_rb);
    samples = MAX2(src_rb->NumSamples, 1);
-   y_scale = samples * 0.5;
+
+   if (samples == 16)
+      x_scale = 4.0;
+   else
+      x_scale = 2.0;
+   y_scale = samples / x_scale;
 
    /* We expect only power of 2 samples in source multisample buffer. */
    assert(samples > 0 && _mesa_is_pow_two(samples));
    while (samples >> (shader_offset + 1)) {
       shader_offset++;
    }
-   /* Update the assert if we plan to support more than 8X MSAA. */
-   assert(shader_offset > 0 && shader_offset < 4);
+   /* Update the assert if we plan to support more than 16X MSAA. */
+   assert(shader_offset > 0 && shader_offset <= 4);
 
    assert(target == GL_TEXTURE_2D_MULTISAMPLE ||
           target == GL_TEXTURE_2D_MULTISAMPLE_ARRAY);
@@ -129,6 +134,10 @@ setup_glsl_msaa_blit_scaled_shader(struct gl_context *ctx,
       sample_number =  "sample_map[int(2 * fract(coord.x) + 8 * fract(coord.y))]";
       sample_map = ctx->Const.SampleMap8x;
       break;
+   case 16:
+      sample_number =  "sample_map[int(4 * fract(coord.x) + 16 * fract(coord.y))]";
+      sample_map = ctx->Const.SampleMap16x;
+      break;
    default:
       sample_number = NULL;
       sample_map = NULL;
@@ -184,9 +193,9 @@ setup_glsl_msaa_blit_scaled_shader(struct gl_context *ctx,
                                "{\n"
                                "%s"
                                "   vec2 interp;\n"
-                               "   const vec2 scale = vec2(2.0f, %ff);\n"
-                               "   const vec2 scale_inv = vec2(0.5f, %ff);\n"
-                               "   const vec2 s_0_offset = vec2(0.25f, %ff);\n"
+                               "   const vec2 scale = vec2(%ff, %ff);\n"
+                               "   const vec2 scale_inv = vec2(%ff, %ff);\n"
+                               "   const vec2 s_0_offset = vec2(%ff, %ff);\n"
                                "   vec2 s_0_coord, s_1_coord, s_2_coord, s_3_coord;\n"
                                "   vec4 s_0_color, s_1_color, s_2_color, s_3_color;\n"
                                "   vec4 x_0_color, x_1_color;\n"
@@ -219,9 +228,9 @@ setup_glsl_msaa_blit_scaled_shader(struct gl_context *ctx,
                                "}\n",
                                sampler_array_suffix,
                                sample_map_expr,
-                               y_scale,
-                               1.0f / y_scale,
-                               1.0f / samples,
+                               x_scale, y_scale,
+                               1.0f / x_scale, 1.0f / y_scale,
+                               0.5f / x_scale, 0.5f / y_scale,
                                texel_fetch_macro);
 
    _mesa_meta_compile_and_link_program(ctx, vs_source, fs_source, name,
diff --git a/src/mesa/drivers/dri/i965/gen6_multisample_state.c b/src/mesa/drivers/dri/i965/gen6_multisample_state.c
index 49c6ebabfba..8eb620de56b 100644
--- a/src/mesa/drivers/dri/i965/gen6_multisample_state.c
+++ b/src/mesa/drivers/dri/i965/gen6_multisample_state.c
@@ -91,6 +91,17 @@ gen6_get_sample_position(struct gl_context *ctx,
  *           | 6 | 7 |                      | 7 | 1 |
  *           ---------                      ---------
  *
+ * 16X MSAA sample index layout  16x MSAA sample number layout
+ *         -----------------            -----------------
+ *         | 0 | 1 | 2 | 3 |            |15 |10 | 9 | 7 |
+ *         -----------------            -----------------
+ *         | 4 | 5 | 6 | 7 |            | 4 | 1 | 3 |13 |
+ *         -----------------            -----------------
+ *         | 8 | 9 |10 |11 |            |12 | 2 | 0 | 6 |
+ *         -----------------            -----------------
+ *         |12 |13 |14 |15 |            |11 | 8 | 5 |14 |
+ *         -----------------            -----------------
+ *
  * A sample map is used to map sample indices to sample numbers.
  */
 void
@@ -99,10 +110,13 @@ gen6_set_sample_maps(struct gl_context *ctx)
    uint8_t map_2x[2] = {0, 1};
    uint8_t map_4x[4] = {0, 1, 2, 3};
    uint8_t map_8x[8] = {5, 2, 4, 6, 0, 3, 7, 1};
+   uint8_t map_16x[16] = { 15, 10, 9, 7, 4, 1, 3, 13,
+                           12, 2, 0, 6, 11, 8, 5, 14 };
 
    memcpy(ctx->Const.SampleMap2x, map_2x, sizeof(map_2x));
    memcpy(ctx->Const.SampleMap4x, map_4x, sizeof(map_4x));
    memcpy(ctx->Const.SampleMap8x, map_8x, sizeof(map_8x));
+   memcpy(ctx->Const.SampleMap16x, map_16x, sizeof(map_16x));
 }
 
 /**