DRW: Use USHORT for vertex color and upload them in linear color to the GPU

This way we remove the need for the srgb boolean uniform and a lot of code complexity. However, mesh update is going to be a bit slower.

I did not benchmark the performance impact.

This also fix a typo in draw_cache_impl_particles.c and fix hair not using vertex color in workbench.

Reviewed By: jbakker

Differential Revision: https://developer.blender.org/D6610

7 files changed, 25 insertions, 137 deletions

diff --git a/source/blender/draw/intern/draw_cache.c b/source/blender/draw/intern/draw_cache.c
index 4d10e781e79..a0e4af8fc8d 100644
--- a/source/blender/draw/intern/draw_cache.c
+++ b/source/blender/draw/intern/draw_cache.c
@@ -826,25 +826,11 @@ GPUBatch *DRW_cache_object_surface_get(Object *ob)
 
 GPUBatch **DRW_cache_object_surface_material_get(struct Object *ob,
                                                  struct GPUMaterial **gpumat_array,
-                                                 uint gpumat_array_len,
-                                                 char **auto_layer_names,
-                                                 int **auto_layer_is_srgb,
-                                                 int *auto_layer_count)
+                                                 uint gpumat_array_len)
 {
-  if (auto_layer_names != NULL) {
-    *auto_layer_names = NULL;
-    *auto_layer_is_srgb = NULL;
-    *auto_layer_count = 0;
-  }
-
   switch (ob->type) {
     case OB_MESH:
-      return DRW_cache_mesh_surface_shaded_get(ob,
-                                               gpumat_array,
-                                               gpumat_array_len,
-                                               auto_layer_names,
-                                               auto_layer_is_srgb,
-                                               auto_layer_count);
+      return DRW_cache_mesh_surface_shaded_get(ob, gpumat_array, gpumat_array_len);
     case OB_CURVE:
       return DRW_cache_curve_surface_shaded_get(ob, gpumat_array, gpumat_array_len);
     case OB_SURF:
@@ -2733,18 +2719,10 @@ GPUBatch *DRW_cache_mesh_surface_edges_get(Object *ob)
 /* Return list of batches with length equal to max(1, totcol). */
 GPUBatch **DRW_cache_mesh_surface_shaded_get(Object *ob,
                                              struct GPUMaterial **gpumat_array,
-                                             uint gpumat_array_len,
-                                             char **auto_layer_names,
-                                             int **auto_layer_is_srgb,
-                                             int *auto_layer_count)
+                                             uint gpumat_array_len)
 {
   BLI_assert(ob->type == OB_MESH);
-  return DRW_mesh_batch_cache_get_surface_shaded(ob->data,
-                                                 gpumat_array,
-                                                 gpumat_array_len,
-                                                 auto_layer_names,
-                                                 auto_layer_is_srgb,
-                                                 auto_layer_count);
+  return DRW_mesh_batch_cache_get_surface_shaded(ob->data, gpumat_array, gpumat_array_len);
 }
 
 /* Return list of batches with length equal to max(1, totcol). */
@@ -2895,8 +2873,7 @@ GPUBatch **DRW_cache_curve_surface_shaded_get(Object *ob,
   struct Curve *cu = ob->data;
   struct Mesh *mesh_eval = ob->runtime.mesh_eval;
   if (mesh_eval != NULL) {
-    return DRW_mesh_batch_cache_get_surface_shaded(
-        mesh_eval, gpumat_array, gpumat_array_len, NULL, NULL, NULL);
+    return DRW_mesh_batch_cache_get_surface_shaded(mesh_eval, gpumat_array, gpumat_array_len);
   }
   else {
     return DRW_curve_batch_cache_get_surface_shaded(cu, gpumat_array, gpumat_array_len);
@@ -3036,8 +3013,7 @@ GPUBatch **DRW_cache_text_surface_shaded_get(Object *ob,
     return NULL;
   }
   if (mesh_eval != NULL) {
-    return DRW_mesh_batch_cache_get_surface_shaded(
-        mesh_eval, gpumat_array, gpumat_array_len, NULL, NULL, NULL);
+    return DRW_mesh_batch_cache_get_surface_shaded(mesh_eval, gpumat_array, gpumat_array_len);
   }
   else {
     return DRW_curve_batch_cache_get_surface_shaded(cu, gpumat_array, gpumat_array_len);
@@ -3131,8 +3107,7 @@ GPUBatch **DRW_cache_surf_surface_shaded_get(Object *ob,
   struct Curve *cu = ob->data;
   struct Mesh *mesh_eval = ob->runtime.mesh_eval;
   if (mesh_eval != NULL) {
-    return DRW_mesh_batch_cache_get_surface_shaded(
-        mesh_eval, gpumat_array, gpumat_array_len, NULL, NULL, NULL);
+    return DRW_mesh_batch_cache_get_surface_shaded(mesh_eval, gpumat_array, gpumat_array_len);
   }
   else {
     return DRW_curve_batch_cache_get_surface_shaded(cu, gpumat_array, gpumat_array_len);
diff --git a/source/blender/draw/intern/draw_cache.h b/source/blender/draw/intern/draw_cache.h
index 11564464546..3759654931a 100644
--- a/source/blender/draw/intern/draw_cache.h
+++ b/source/blender/draw/intern/draw_cache.h
@@ -54,10 +54,7 @@ struct GPUBatch *DRW_cache_object_surface_get(struct Object *ob);
 struct GPUBatch *DRW_cache_object_loose_edges_get(struct Object *ob);
 struct GPUBatch **DRW_cache_object_surface_material_get(struct Object *ob,
                                                         struct GPUMaterial **gpumat_array,
-                                                        uint gpumat_array_len,
-                                                        char **auto_layer_names,
-                                                        int **auto_layer_is_srgb,
-                                                        int *auto_layer_count);
+                                                        uint gpumat_array_len);
 struct GPUBatch *DRW_cache_object_face_wireframe_get(struct Object *ob);
 
 /* Empties */
@@ -127,10 +124,7 @@ struct GPUBatch *DRW_cache_mesh_surface_get(struct Object *ob);
 struct GPUBatch *DRW_cache_mesh_surface_edges_get(struct Object *ob);
 struct GPUBatch **DRW_cache_mesh_surface_shaded_get(struct Object *ob,
                                                     struct GPUMaterial **gpumat_array,
-                                                    uint gpumat_array_len,
-                                                    char **auto_layer_names,
-                                                    int **auto_layer_is_srgb,
-                                                    int *auto_layer_count);
+                                                    uint gpumat_array_len);
 struct GPUBatch **DRW_cache_mesh_surface_texpaint_get(struct Object *ob);
 struct GPUBatch *DRW_cache_mesh_surface_texpaint_single_get(struct Object *ob);
 struct GPUBatch *DRW_cache_mesh_surface_vertpaint_get(struct Object *ob);
diff --git a/source/blender/draw/intern/draw_cache_extract.h b/source/blender/draw/intern/draw_cache_extract.h
index 2d10199782b..9228147af44 100644
--- a/source/blender/draw/intern/draw_cache_extract.h
+++ b/source/blender/draw/intern/draw_cache_extract.h
@@ -213,12 +213,6 @@ typedef struct MeshBatchCache {
 
   GPUBatch **surface_per_mat;
 
-  /* arrays of bool uniform names (and value) that will be use to
-   * set srgb conversion for auto attributes.*/
-  char *auto_layer_names;
-  int *auto_layer_is_srgb;
-  int auto_layer_len;
-
   DRWBatchFlag batch_requested;
   DRWBatchFlag batch_ready;
 
diff --git a/source/blender/draw/intern/draw_cache_extract_mesh.c b/source/blender/draw/intern/draw_cache_extract_mesh.c
index c98494ebdd9..ee0597c6b21 100644
--- a/source/blender/draw/intern/draw_cache_extract_mesh.c
+++ b/source/blender/draw/intern/draw_cache_extract_mesh.c
@@ -1928,7 +1928,7 @@ static void *extract_vcol_init(const MeshRenderData *mr, void *buf)
       GPU_vertformat_safe_attrib_name(layer_name, attr_safe_name, GPU_MAX_SAFE_ATTRIB_NAME);
 
       BLI_snprintf(attr_name, sizeof(attr_name), "c%s", attr_safe_name);
-      GPU_vertformat_attr_add(&format, attr_name, GPU_COMP_U8, 4, GPU_FETCH_INT_TO_FLOAT_UNIT);
+      GPU_vertformat_attr_add(&format, attr_name, GPU_COMP_U16, 4, GPU_FETCH_INT_TO_FLOAT_UNIT);
 
       if (i == CustomData_get_render_layer(cd_ldata, CD_MLOOPCOL)) {
         GPU_vertformat_alias_add(&format, "c");
@@ -1948,12 +1948,20 @@ static void *extract_vcol_init(const MeshRenderData *mr, void *buf)
   GPU_vertbuf_init_with_format(vbo, &format);
   GPU_vertbuf_data_alloc(vbo, mr->loop_len);
 
-  MLoopCol *vcol_data = (MLoopCol *)vbo->data;
+  typedef struct gpuMeshVcol {
+    ushort r, g, b, a;
+  } gpuMeshVcol;
+
+  gpuMeshVcol *vcol_data = (gpuMeshVcol *)vbo->data;
   for (int i = 0; i < 8; i++) {
     if (vcol_layers & (1 << i)) {
-      void *layer_data = CustomData_get_layer_n(cd_ldata, CD_MLOOPCOL, i);
-      memcpy(vcol_data, layer_data, sizeof(*vcol_data) * mr->loop_len);
-      vcol_data += mr->loop_len;
+      MLoopCol *mcol = (MLoopCol *)CustomData_get_layer_n(cd_ldata, CD_MLOOPCOL, i);
+      for (int l = 0; l < mr->loop_len; l++, mcol++, vcol_data++) {
+        vcol_data->r = unit_float_to_ushort_clamp(BLI_color_from_srgb_table[mcol->r]);
+        vcol_data->g = unit_float_to_ushort_clamp(BLI_color_from_srgb_table[mcol->g]);
+        vcol_data->b = unit_float_to_ushort_clamp(BLI_color_from_srgb_table[mcol->b]);
+        vcol_data->a = unit_float_to_ushort_clamp(mcol->a * (1.0f / 255.0f));
+      }
     }
   }
   return NULL;
diff --git a/source/blender/draw/intern/draw_cache_impl.h b/source/blender/draw/intern/draw_cache_impl.h
index 8cb318bd0bb..bf056d7444d 100644
--- a/source/blender/draw/intern/draw_cache_impl.h
+++ b/source/blender/draw/intern/draw_cache_impl.h
@@ -127,10 +127,7 @@ struct GPUBatch *DRW_mesh_batch_cache_get_surface(struct Mesh *me);
 struct GPUBatch *DRW_mesh_batch_cache_get_surface_edges(struct Mesh *me);
 struct GPUBatch **DRW_mesh_batch_cache_get_surface_shaded(struct Mesh *me,
                                                           struct GPUMaterial **gpumat_array,
-                                                          uint gpumat_array_len,
-                                                          char **auto_layer_names,
-                                                          int **auto_layer_is_srgb,
-                                                          int *auto_layer_count);
+                                                          uint gpumat_array_len);
 struct GPUBatch **DRW_mesh_batch_cache_get_surface_texpaint(struct Mesh *me);
 struct GPUBatch *DRW_mesh_batch_cache_get_surface_texpaint_single(struct Mesh *me);
 struct GPUBatch *DRW_mesh_batch_cache_get_surface_vertpaint(struct Mesh *me);
diff --git a/source/blender/draw/intern/draw_cache_impl_mesh.c b/source/blender/draw/intern/draw_cache_impl_mesh.c
index aadcc2a939a..458029c2580 100644
--- a/source/blender/draw/intern/draw_cache_impl_mesh.c
+++ b/source/blender/draw/intern/draw_cache_impl_mesh.c
@@ -230,68 +230,6 @@ static DRW_MeshCDMask mesh_cd_calc_used_gpu_layers(const Mesh *me,
   return cd_used;
 }
 
-static void mesh_cd_extract_auto_layers_names_and_srgb(Mesh *me,
-                                                       DRW_MeshCDMask cd_used,
-                                                       char **r_auto_layers_names,
-                                                       int **r_auto_layers_srgb,
-                                                       int *r_auto_layers_len)
-{
-  const Mesh *me_final = (me->edit_mesh) ? me->edit_mesh->mesh_eval_final : me;
-  const CustomData *cd_ldata = &me_final->ldata;
-
-  int uv_len_used = count_bits_i(cd_used.uv);
-  int vcol_len_used = count_bits_i(cd_used.vcol);
-  int uv_len = CustomData_number_of_layers(cd_ldata, CD_MLOOPUV);
-  int vcol_len = CustomData_number_of_layers(cd_ldata, CD_MLOOPCOL);
-
-  uint auto_names_len = 32 * (uv_len_used + vcol_len_used);
-  uint auto_ofs = 0;
-  /* Allocate max, resize later. */
-  char *auto_names = MEM_callocN(sizeof(char) * auto_names_len, __func__);
-  int *auto_is_srgb = MEM_callocN(sizeof(int) * (uv_len_used + vcol_len_used), __func__);
-
-  for (int i = 0; i < uv_len; i++) {
-    if ((cd_used.uv & (1 << i)) != 0) {
-      const char *name = CustomData_get_layer_name(cd_ldata, CD_MLOOPUV, i);
-      char safe_name[GPU_MAX_SAFE_ATTRIB_NAME];
-      GPU_vertformat_safe_attrib_name(name, safe_name, GPU_MAX_SAFE_ATTRIB_NAME);
-      auto_ofs += BLI_snprintf_rlen(
-          auto_names + auto_ofs, auto_names_len - auto_ofs, "ba%s", safe_name);
-      /* +1 to include '\0' terminator. */
-      auto_ofs += 1;
-    }
-  }
-
-  uint auto_is_srgb_ofs = uv_len_used;
-  for (int i = 0; i < vcol_len; i++) {
-    if ((cd_used.vcol & (1 << i)) != 0) {
-      const char *name = CustomData_get_layer_name(cd_ldata, CD_MLOOPCOL, i);
-      /* We only do vcols that are not overridden by a uv layer with same name. */
-      if (CustomData_get_named_layer_index(cd_ldata, CD_MLOOPUV, name) == -1) {
-        char safe_name[GPU_MAX_SAFE_ATTRIB_NAME];
-        GPU_vertformat_safe_attrib_name(name, safe_name, GPU_MAX_SAFE_ATTRIB_NAME);
-        auto_ofs += BLI_snprintf_rlen(
-            auto_names + auto_ofs, auto_names_len - auto_ofs, "ba%s", safe_name);
-        /* +1 to include '\0' terminator. */
-        auto_ofs += 1;
-        auto_is_srgb[auto_is_srgb_ofs] = true;
-        auto_is_srgb_ofs++;
-      }
-    }
-  }
-
-  auto_names = MEM_reallocN(auto_names, sizeof(char) * auto_ofs);
-  auto_is_srgb = MEM_reallocN(auto_is_srgb, sizeof(int) * auto_is_srgb_ofs);
-
-  /* WATCH: May have been referenced somewhere before freeing. */
-  MEM_SAFE_FREE(*r_auto_layers_names);
-  MEM_SAFE_FREE(*r_auto_layers_srgb);
-
-  *r_auto_layers_names = auto_names;
-  *r_auto_layers_srgb = auto_is_srgb;
-  *r_auto_layers_len = auto_is_srgb_ofs;
-}
-
 /** \} */
 
 /* ---------------------------------------------------------------------- */
@@ -492,8 +430,6 @@ static void mesh_batch_cache_discard_shaded_tri(MeshBatchCache *cache)
   mesh_cd_layers_type_clear(&cache->cd_used);
 
   MEM_SAFE_FREE(cache->surface_per_mat);
-  MEM_SAFE_FREE(cache->auto_layer_names);
-  MEM_SAFE_FREE(cache->auto_layer_is_srgb);
 
   cache->mat_len = 0;
 }
@@ -771,10 +707,7 @@ GPUBatch *DRW_mesh_batch_cache_get_edit_mesh_analysis(Mesh *me)
 
 GPUBatch **DRW_mesh_batch_cache_get_surface_shaded(Mesh *me,
                                                    struct GPUMaterial **gpumat_array,
-                                                   uint gpumat_array_len,
-                                                   char **auto_layer_names,
-                                                   int **auto_layer_is_srgb,
-                                                   int *auto_layer_count)
+                                                   uint gpumat_array_len)
 {
   MeshBatchCache *cache = mesh_batch_cache_get(me);
   DRW_MeshCDMask cd_needed = mesh_cd_calc_used_gpu_layers(me, gpumat_array, gpumat_array_len);
@@ -783,21 +716,8 @@ GPUBatch **DRW_mesh_batch_cache_get_surface_shaded(Mesh *me,
 
   mesh_cd_layers_type_merge(&cache->cd_needed, cd_needed);
 
-  if (!mesh_cd_layers_type_overlap(cache->cd_used, cd_needed)) {
-    mesh_cd_extract_auto_layers_names_and_srgb(me,
-                                               cache->cd_needed,
-                                               &cache->auto_layer_names,
-                                               &cache->auto_layer_is_srgb,
-                                               &cache->auto_layer_len);
-  }
-
   mesh_batch_cache_add_request(cache, MBC_SURF_PER_MAT);
 
-  if (auto_layer_names) {
-    *auto_layer_names = cache->auto_layer_names;
-    *auto_layer_is_srgb = cache->auto_layer_is_srgb;
-    *auto_layer_count = cache->auto_layer_len;
-  }
   for (int i = 0; i < cache->mat_len; i++) {
     DRW_batch_request(&cache->surface_per_mat[i]);
   }
diff --git a/source/blender/draw/intern/draw_cache_impl_particles.c b/source/blender/draw/intern/draw_cache_impl_particles.c
index 7de6ee2b3b1..f4057927d55 100644
--- a/source/blender/draw/intern/draw_cache_impl_particles.c
+++ b/source/blender/draw/intern/draw_cache_impl_particles.c
@@ -1199,7 +1199,7 @@ static void particle_batch_cache_ensure_pos_and_seg(PTCacheEdit *edit,
       GPU_vertformat_safe_attrib_name(name, attr_safe_name, GPU_MAX_SAFE_ATTRIB_NAME);
 
       BLI_snprintf(uuid, sizeof(uuid), "c%s", attr_safe_name);
-      col_id[i] = GPU_vertformat_attr_add(&format, uuid, GPU_COMP_F32, 2, GPU_FETCH_FLOAT);
+      col_id[i] = GPU_vertformat_attr_add(&format, uuid, GPU_COMP_U16, 4, GPU_FETCH_FLOAT);
 
       if (i == active_col) {
         GPU_vertformat_alias_add(&format, "c");