OpenSubDiv: add support for an OpenGL evaluator

This evaluator is used in order to evaluate subdivision at render time, allowing for
faster renders of meshes with a subdivision surface modifier placed at the last
position in the modifier list.

When evaluating the subsurf modifier, we detect whether we can delegate evaluation
to the draw code. If so, the subdivision is first evaluated on the GPU using our own
custom evaluator (only the coarse data needs to be initially sent to the GPU), then,
buffers for the final `MeshBufferCache` are filled on the GPU using a set of
compute shaders. However, some buffers are still filled on the CPU side, if doing so
on the GPU is impractical (e.g. the line adjacency buffer used for x-ray, whose
logic is hardly GPU compatible).

This is done at the mesh buffer extraction level so that the result can be readily used
in the various OpenGL engines, without having to write custom geometry or tesselation
shaders.

We use our own subdivision evaluation shaders, instead of OpenSubDiv's vanilla one, in
order to control the data layout, and interpolation. For example, we store vertex colors
as compressed 16-bit integers, while OpenSubDiv's default evaluator only work for float
types.

In order to still access the modified geometry on the CPU side, for use in modifiers
or transform operators, a dedicated wrapper type is added `MESH_WRAPPER_TYPE_SUBD`.
Subdivision will be lazily evaluated via `BKE_object_get_evaluated_mesh` which will
create such a wrapper if possible. If the final subdivision surface is not needed on
the CPU side, `BKE_object_get_evaluated_mesh_no_subsurf` should be used.

Enabling or disabling GPU subdivision can be done through the user preferences (under
Viewport -> Subdivision).

See patch description for benchmarks.

Reviewed By: campbellbarton, jbakker, fclem, brecht, #eevee_viewport

Differential Revision: https://developer.blender.org/D12406

1 files changed, 176 insertions, 0 deletions

diff --git a/source/blender/draw/intern/shaders/common_subdiv_lib.glsl b/source/blender/draw/intern/shaders/common_subdiv_lib.glsl
new file mode 100644
index 00000000000..005561964b8
--- /dev/null
+++ b/source/blender/draw/intern/shaders/common_subdiv_lib.glsl
@@ -0,0 +1,176 @@
+
+layout(local_size_x = 64, local_size_y = 1, local_size_z = 1) in;
+
+/* Uniform block for #DRWSubivUboStorage. */
+layout(std140) uniform shader_data
+{
+  /* Offsets in the buffers data where the source and destination data start. */
+  int src_offset;
+  int dst_offset;
+
+  /* Parameters for the DRWPatchMap. */
+  int min_patch_face;
+  int max_patch_face;
+  int max_depth;
+  int patches_are_triangular;
+
+  /* Coarse topology information. */
+  int coarse_poly_count;
+  uint edge_loose_offset;
+
+  /* Subdiv topology information. */
+  uint num_subdiv_loops;
+
+  /* Subdivision settings. */
+  bool optimal_display;
+
+  /* Sculpt data. */
+  bool has_sculpt_mask;
+
+  /* Masks for the extra coarse face data. */
+  uint coarse_face_select_mask;
+  uint coarse_face_smooth_mask;
+  uint coarse_face_active_mask;
+  uint coarse_face_loopstart_mask;
+
+  /* Total number of elements to process. */
+  uint total_dispatch_size;
+};
+
+uint get_global_invocation_index()
+{
+  uint invocations_per_row = gl_WorkGroupSize.x * gl_NumWorkGroups.x;
+  return gl_GlobalInvocationID.x + gl_GlobalInvocationID.y * invocations_per_row;
+}
+
+/* Structure for #CompressedPatchCoord. */
+struct BlenderPatchCoord {
+  int patch_index;
+  uint encoded_uv;
+};
+
+vec2 decode_uv(uint encoded_uv)
+{
+  float u = float((encoded_uv >> 16) & 0xFFFFu) / 65535.0;
+  float v = float(encoded_uv & 0xFFFFu) / 65535.0;
+  return vec2(u, v);
+}
+
+/* This structure is a carbon copy of OpenSubDiv's PatchTable::PatchHandle. */
+struct PatchHandle {
+  int array_index;
+  int patch_index;
+  int vertex_index;
+};
+
+/* This structure is a carbon copy of OpenSubDiv's PatchCoord. */
+struct PatchCoord {
+  int array_index;
+  int patch_index;
+  int vertex_index;
+  float u;
+  float v;
+};
+
+/* This structure is a carbon copy of OpenSubDiv's PatchCoord.QuadNode.
+ * Each child is a bitfield. */
+struct QuadNode {
+  uvec4 child;
+};
+
+bool is_set(uint i)
+{
+  /* QuadNode.Child.isSet is the first bit of the bitfield. */
+  return (i & 0x1u) != 0;
+}
+
+bool is_leaf(uint i)
+{
+  /* QuadNode.Child.isLeaf is the second bit of the bitfield. */
+  return (i & 0x2u) != 0;
+}
+
+uint get_index(uint i)
+{
+  /* QuadNode.Child.index is made of the remaining bits. */
+  return (i >> 2) & 0x3FFFFFFFu;
+}
+
+/* Duplicate of #PosNorLoop from the mesh extract CPU code.
+ * We do not use a vec3 for the position as it will be padded to a vec4 which is incompatible with
+ * the format.  */
+struct PosNorLoop {
+  float x, y, z;
+  /* TODO(kevindietrich) : figure how to compress properly as GLSL does not have char/short types,
+   * bit operations get tricky. */
+  float nx, ny, nz;
+  float flag;
+};
+
+vec3 get_vertex_pos(PosNorLoop vertex_data)
+{
+  return vec3(vertex_data.x, vertex_data.y, vertex_data.z);
+}
+
+vec3 get_vertex_nor(PosNorLoop vertex_data)
+{
+  return vec3(vertex_data.nx, vertex_data.ny, vertex_data.nz);
+}
+
+void set_vertex_pos(inout PosNorLoop vertex_data, vec3 pos)
+{
+  vertex_data.x = pos.x;
+  vertex_data.y = pos.y;
+  vertex_data.z = pos.z;
+}
+
+void set_vertex_nor(inout PosNorLoop vertex_data, vec3 nor, uint flag)
+{
+  vertex_data.nx = nor.x;
+  vertex_data.ny = nor.y;
+  vertex_data.nz = nor.z;
+  vertex_data.flag = float(flag);
+}
+
+/* Set the vertex normal but preserve the existing flag. This is for when we compute manually the
+ * vertex normals when we cannot use the limit surface, in which case the flag and the normal are
+ * set by two separate compute pass. */
+void set_vertex_nor(inout PosNorLoop vertex_data, vec3 nor)
+{
+  set_vertex_nor(vertex_data, nor, 0);
+}
+
+#define ORIGINDEX_NONE -1
+
+#ifdef SUBDIV_POLYGON_OFFSET
+layout(std430, binding = 0) readonly buffer inputSubdivPolygonOffset
+{
+  uint subdiv_polygon_offset[];
+};
+
+/* Given the index of the subdivision quad, return the index of the corresponding coarse polygon.
+ * This uses subdiv_polygon_offset and since it is a growing list of offsets, we can use binary
+ * search to locate the right index. */
+uint coarse_polygon_index_from_subdiv_quad_index(uint subdiv_quad_index, uint coarse_poly_count)
+{
+  uint first = 0;
+  uint last = coarse_poly_count;
+
+  while (first != last) {
+    uint middle = (first + last) / 2;
+
+    if (subdiv_polygon_offset[middle] < subdiv_quad_index) {
+      first = middle + 1;
+    }
+    else {
+      last = middle;
+    }
+  }
+
+  if (subdiv_polygon_offset[first] == subdiv_quad_index) {
+    return first;
+  }
+
+  return first - 1;
+}
+#endif