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

2 files changed, 27 insertions, 33 deletions

diff --git a/source/blender/nodes/shader/nodes/node_shader_tex_environment.c b/source/blender/nodes/shader/nodes/node_shader_tex_environment.c
index 72ad5581050..6c380efe0b2 100644
--- a/source/blender/nodes/shader/nodes/node_shader_tex_environment.c
+++ b/source/blender/nodes/shader/nodes/node_shader_tex_environment.c
@@ -88,7 +88,7 @@ static int node_shader_gpu_tex_environment(GPUMaterial *mat,
              "node_tex_environment_equirectangular",
              in[0].link,
              GPU_constant(&clamp_size),
-             GPU_image(ima, iuser, 0),
+             GPU_image(ima, iuser),
              &in[0].link);
   }
   else {
@@ -103,7 +103,7 @@ static int node_shader_gpu_tex_environment(GPUMaterial *mat,
       GPU_link(mat,
                "node_tex_image_linear_no_mip",
                in[0].link,
-               GPU_image(ima, iuser, 0),
+               GPU_image(ima, iuser),
                &out[0].link,
                &outalpha);
       break;
@@ -111,17 +111,13 @@ static int node_shader_gpu_tex_environment(GPUMaterial *mat,
       GPU_link(mat,
                "node_tex_image_nearest",
                in[0].link,
-               GPU_image(ima, iuser, 0),
+               GPU_image(ima, iuser),
                &out[0].link,
                &outalpha);
       break;
     default:
-      GPU_link(mat,
-               "node_tex_image_cubic",
-               in[0].link,
-               GPU_image(ima, iuser, 0),
-               &out[0].link,
-               &outalpha);
+      GPU_link(
+          mat, "node_tex_image_cubic", in[0].link, GPU_image(ima, iuser), &out[0].link, &outalpha);
       break;
   }
 
diff --git a/source/blender/nodes/shader/nodes/node_shader_tex_image.c b/source/blender/nodes/shader/nodes/node_shader_tex_image.c
index 34a5e323490..781fd1eb579 100644
--- a/source/blender/nodes/shader/nodes/node_shader_tex_image.c
+++ b/source/blender/nodes/shader/nodes/node_shader_tex_image.c
@@ -130,21 +130,19 @@ static int node_shader_gpu_tex_image(GPUMaterial *mat,
   node_shader_gpu_tex_mapping(mat, node, in, out);
 
   if (ima->source == IMA_SRC_TILED) {
-    GPUNodeLink *map;
-    GPU_link(mat, "node_tex_tile_map", in[0].link, &out[0].link, &map);
-    /* This is not exactly great, but if we want to support different sizes per
-     * tile and older hardware, which rules out better methods like texture arrays. */
-    LISTBASE_FOREACH (ImageTile *, tile, &ima->tiles) {
-      float tile_number = tile->tile_number;
-      GPU_link(mat,
-               names_tiled[tex->interpolation],
-               map,
-               GPU_uniform(&tile_number),
-               GPU_image(ima, iuser, tile->tile_number),
-               out[0].link,
-               &out[0].link,
-               &out[1].link);
-    }
+    /* The tiled shader needs both the tile array itself as well as the mapping from tile to array
+     * position. Which of these to allocate is automatically decided based on the shader argument
+     * type, so here the first GPU_image(ima, iuser) will resolve to the array and the second to
+     * the mapping since the third argument in the shader has type sampler2DArray while
+     * the fourth is sampler1DArray.
+     */
+    GPU_stack_link(mat,
+                   node,
+                   names_tiled[tex->interpolation],
+                   in,
+                   out,
+                   GPU_image(ima, iuser),
+                   GPU_image(ima, iuser));
   }
   else {
     switch (tex->projection) {
@@ -159,20 +157,20 @@ static int node_shader_gpu_tex_image(GPUMaterial *mat,
           GPU_link(mat, "set_rgb", *texco, &input_coords);
         }
         if (do_texco_extend) {
-          GPU_link(mat, "point_texco_clamp", *texco, GPU_image(ima, iuser, 0), texco);
+          GPU_link(mat, "point_texco_clamp", *texco, GPU_image(ima, iuser), texco);
         }
-        GPU_stack_link(mat, node, gpu_node_name, in, out, GPU_image(ima, iuser, 0));
+        GPU_stack_link(mat, node, gpu_node_name, in, out, GPU_image(ima, iuser));
         break;
 
       case SHD_PROJ_BOX:
         vnor = GPU_builtin(GPU_WORLD_NORMAL);
         ob_mat = GPU_builtin(GPU_OBJECT_MATRIX);
         blend = GPU_uniform(&tex->projection_blend);
-        gpu_image = GPU_image(ima, iuser, 0);
+        gpu_image = GPU_image(ima, iuser);
 
         /* equivalent to normal_world_to_object */
         GPU_link(mat, "normal_transform_transposed_m4v3", vnor, ob_mat, &norm);
-        GPU_link(mat, gpu_node_name, *texco, norm, GPU_image(ima, iuser, 0), &col1, &col2, &col3);
+        GPU_link(mat, gpu_node_name, *texco, norm, GPU_image(ima, iuser), &col1, &col2, &col3);
         GPU_stack_link(
             mat, node, "node_tex_image_box", in, out, norm, col1, col2, col3, gpu_image, blend);
         break;
@@ -186,9 +184,9 @@ static int node_shader_gpu_tex_image(GPUMaterial *mat,
           GPU_link(mat, "set_rgb", *texco, &input_coords);
         }
         if (do_texco_extend) {
-          GPU_link(mat, "point_texco_clamp", *texco, GPU_image(ima, iuser, 0), texco);
+          GPU_link(mat, "point_texco_clamp", *texco, GPU_image(ima, iuser), texco);
         }
-        GPU_stack_link(mat, node, gpu_node_name, in, out, GPU_image(ima, iuser, 0));
+        GPU_stack_link(mat, node, gpu_node_name, in, out, GPU_image(ima, iuser));
         break;
 
       case SHD_PROJ_TUBE:
@@ -200,9 +198,9 @@ static int node_shader_gpu_tex_image(GPUMaterial *mat,
           GPU_link(mat, "set_rgb", *texco, &input_coords);
         }
         if (do_texco_extend) {
-          GPU_link(mat, "point_texco_clamp", *texco, GPU_image(ima, iuser, 0), texco);
+          GPU_link(mat, "point_texco_clamp", *texco, GPU_image(ima, iuser), texco);
         }
-        GPU_stack_link(mat, node, gpu_node_name, in, out, GPU_image(ima, iuser, 0));
+        GPU_stack_link(mat, node, gpu_node_name, in, out, GPU_image(ima, iuser));
         break;
     }
 
@@ -210,7 +208,7 @@ static int node_shader_gpu_tex_image(GPUMaterial *mat,
       if (do_texco_clip) {
         gpu_node_name = names_clip[tex->interpolation];
         in[0].link = input_coords;
-        GPU_stack_link(mat, node, gpu_node_name, in, out, GPU_image(ima, iuser, 0), out[0].link);
+        GPU_stack_link(mat, node, gpu_node_name, in, out, GPU_image(ima, iuser), out[0].link);
       }
     }
   }