Fix T73133: UDIM texture count in Eevee is limited by OpenGL

Based on @fclem's suggestion in D6421, this commit implements support for
storing all tiles of a UDIM texture in a single 2D array texture on the GPU.

Previously, Eevee was binding one OpenGL texture per tile, quickly running
into hardware limits with nontrivial UDIM texture sets.
Workbench meanwhile had no UDIM support at all, as reusing the per-tile
approach would require splitting the mesh by tile as well as texture.

With this commit, both Workbench as well as Eevee now support huge numbers
of tiles, with the eventual limits being GPU memory and ultimately
GL_MAX_ARRAY_TEXTURE_LAYERS, which tends to be in the 1000s on modern GPUs.

Initially my plan was to have one array texture per unique size, but managing
the different textures and keeping everything consistent ended up being way
too complex.

Therefore, we now use a simpler version that allocates a texture that
is large enough to fit the largest tile and then packs all tiles into as many
layers as necessary.

As a result, each UDIM texture only binds two textures (one for the actual
images, one for metadata) regardless of how many tiles are used.

Note that this rolls back per-tile GPUTextures, meaning that we again have
per-Image GPUTextures like we did before the original UDIM commit,
but now with four instead of two types.

Reviewed By: fclem

Differential Revision: https://developer.blender.org/D6456

1 files changed, 20 insertions, 15 deletions

diff --git a/source/blender/draw/engines/workbench/workbench_forward.c b/source/blender/draw/engines/workbench/workbench_forward.c
index 44f43fc7d09..ae001f8d10c 100644
--- a/source/blender/draw/engines/workbench/workbench_forward.c
+++ b/source/blender/draw/engines/workbench/workbench_forward.c
@@ -172,8 +172,10 @@ WORKBENCH_MaterialData *workbench_forward_get_or_create_material_data(WORKBENCH_
     GPUShader *shader = (wpd->shading.color_type == color_type) ?
                             wpd->transparent_accum_sh :
                             wpd->transparent_accum_uniform_sh;
+    const bool is_tiled = (ima && ima->source == IMA_SRC_TILED);
     if (color_type == V3D_SHADING_TEXTURE_COLOR) {
-      shader = wpd->transparent_accum_textured_sh;
+      shader = is_tiled ? wpd->transparent_accum_textured_array_sh :
+                          wpd->transparent_accum_textured_sh;
     }
 
     grp = DRW_shgroup_create(shader, psl->transparent_accum_pass);
@@ -201,7 +203,7 @@ WORKBENCH_MaterialData *workbench_forward_get_or_create_material_data(WORKBENCH_
       DRW_shgroup_uniform_float_copy(grp, "shadowFocus", wpd->shadow_focus);
     }
 
-    workbench_material_shgroup_uniform(wpd, grp, material, ob, false, interp);
+    workbench_material_shgroup_uniform(wpd, grp, material, ob, false, is_tiled, interp);
     material->shgrp = grp;
 
     /* Depth */
@@ -226,16 +228,17 @@ WORKBENCH_MaterialData *workbench_forward_get_or_create_material_data(WORKBENCH_
 static GPUShader *ensure_forward_accum_shaders(WORKBENCH_PrivateData *wpd,
                                                bool is_uniform_color,
                                                bool is_hair,
+                                               bool is_tiled,
                                                const WORKBENCH_ColorOverride color_override,
                                                eGPUShaderConfig sh_cfg)
 {
   WORKBENCH_FORWARD_Shaders *sh_data = &e_data.sh_data[sh_cfg];
   int index = workbench_material_get_accum_shader_index(
-      wpd, is_uniform_color, is_hair, color_override);
+      wpd, is_uniform_color, is_hair, is_tiled, color_override);
   if (sh_data->transparent_accum_sh_cache[index] == NULL) {
     const GPUShaderConfigData *sh_cfg_data = &GPU_shader_cfg_data[sh_cfg];
     char *defines = workbench_material_build_defines(
-        wpd, is_uniform_color, is_hair, color_override);
+        wpd, is_uniform_color, is_hair, is_tiled, color_override);
     char *transparent_accum_vert = workbench_build_forward_vert(is_hair);
     char *transparent_accum_frag = workbench_build_forward_transparent_accum_frag();
     sh_data->transparent_accum_sh_cache[index] = GPU_shader_create_from_arrays({
@@ -255,7 +258,7 @@ static GPUShader *ensure_forward_composite_shaders(WORKBENCH_PrivateData *wpd)
   int index = OBJECT_OUTLINE_ENABLED(wpd) ? 1 : 0;
   if (e_data.composite_sh_cache[index] == NULL) {
     char *defines = workbench_material_build_defines(
-        wpd, false, false, WORKBENCH_COLOR_OVERRIDE_OFF);
+        wpd, false, false, false, WORKBENCH_COLOR_OVERRIDE_OFF);
     char *composite_frag = workbench_build_forward_composite_frag();
     e_data.composite_sh_cache[index] = DRW_shader_create_fullscreen(composite_frag, defines);
     MEM_freeN(composite_frag);
@@ -268,17 +271,19 @@ void workbench_forward_choose_shaders(WORKBENCH_PrivateData *wpd, eGPUShaderConf
 {
   wpd->composite_sh = ensure_forward_composite_shaders(wpd);
   wpd->transparent_accum_sh = ensure_forward_accum_shaders(
-      wpd, false, false, WORKBENCH_COLOR_OVERRIDE_OFF, sh_cfg);
+      wpd, false, false, false, WORKBENCH_COLOR_OVERRIDE_OFF, sh_cfg);
   wpd->transparent_accum_hair_sh = ensure_forward_accum_shaders(
-      wpd, false, true, WORKBENCH_COLOR_OVERRIDE_OFF, sh_cfg);
+      wpd, false, true, false, WORKBENCH_COLOR_OVERRIDE_OFF, sh_cfg);
   wpd->transparent_accum_uniform_sh = ensure_forward_accum_shaders(
-      wpd, true, false, WORKBENCH_COLOR_OVERRIDE_OFF, sh_cfg);
+      wpd, true, false, false, WORKBENCH_COLOR_OVERRIDE_OFF, sh_cfg);
   wpd->transparent_accum_uniform_hair_sh = ensure_forward_accum_shaders(
-      wpd, true, true, WORKBENCH_COLOR_OVERRIDE_OFF, sh_cfg);
+      wpd, true, true, false, WORKBENCH_COLOR_OVERRIDE_OFF, sh_cfg);
   wpd->transparent_accum_textured_sh = ensure_forward_accum_shaders(
-      wpd, false, false, WORKBENCH_COLOR_OVERRIDE_TEXTURE, sh_cfg);
+      wpd, false, false, false, WORKBENCH_COLOR_OVERRIDE_TEXTURE, sh_cfg);
+  wpd->transparent_accum_textured_array_sh = ensure_forward_accum_shaders(
+      wpd, false, false, true, WORKBENCH_COLOR_OVERRIDE_TEXTURE, sh_cfg);
   wpd->transparent_accum_vertex_sh = ensure_forward_accum_shaders(
-      wpd, false, false, WORKBENCH_COLOR_OVERRIDE_VERTEX, sh_cfg);
+      wpd, false, false, false, WORKBENCH_COLOR_OVERRIDE_VERTEX, sh_cfg);
 }
 
 void workbench_forward_outline_shaders_ensure(WORKBENCH_PrivateData *wpd, eGPUShaderConfig sh_cfg)
@@ -288,11 +293,11 @@ void workbench_forward_outline_shaders_ensure(WORKBENCH_PrivateData *wpd, eGPUSh
   if (sh_data->object_outline_sh == NULL) {
     const GPUShaderConfigData *sh_cfg_data = &GPU_shader_cfg_data[sh_cfg];
     char *defines = workbench_material_build_defines(
-        wpd, false, false, WORKBENCH_COLOR_OVERRIDE_OFF);
+        wpd, false, false, false, WORKBENCH_COLOR_OVERRIDE_OFF);
     char *defines_texture = workbench_material_build_defines(
-        wpd, true, false, WORKBENCH_COLOR_OVERRIDE_OFF);
+        wpd, true, false, false, WORKBENCH_COLOR_OVERRIDE_OFF);
     char *defines_hair = workbench_material_build_defines(
-        wpd, false, true, WORKBENCH_COLOR_OVERRIDE_OFF);
+        wpd, false, true, false, WORKBENCH_COLOR_OVERRIDE_OFF);
     char *forward_vert = workbench_build_forward_vert(false);
     char *forward_frag = workbench_build_forward_outline_frag();
     char *forward_hair_vert = workbench_build_forward_vert(true);
@@ -533,7 +538,7 @@ static void workbench_forward_cache_populate_particles(WORKBENCH_Data *vedata, O
       DRWShadingGroup *shgrp = DRW_shgroup_hair_create(
           ob, psys, md, psl->transparent_accum_pass, shader);
       DRW_shgroup_uniform_block(shgrp, "world_block", wpd->world_ubo);
-      workbench_material_shgroup_uniform(wpd, shgrp, material, ob, false, interp);
+      workbench_material_shgroup_uniform(wpd, shgrp, material, ob, false, false, interp);
       DRW_shgroup_uniform_vec4(shgrp, "viewvecs[0]", (float *)wpd->viewvecs, 3);
       /* Hairs have lots of layer and can rapidly become the most prominent surface.
        * So lower their alpha artificially. */