Cleanup: Optimize gl query code path

Currently whenever gl queries are performed for the viewport, a large
1024 byte array is allocated to store the query results (256 of them).

Unfortunately, if any gizmo using a `draw_select` callback is active
(e.g. the transform gizmos), these queries (and allocations) will occur
during every mouse move event.

Change the vector to allow for up to 16 query results before making an
allocation. This provides enough space for every built-in gizmo except
Scale Cage (which needs 27 queries). It also removes unnecessary
allocations from two other related vectors used during query processing.

Differential Revision: https://developer.blender.org/D13784

source/blender/gpu/intern/gpu_query.hh

@@ -11,6 +11,8 @@
 
 namespace blender::gpu {
 
+#define QUERY_MIN_LEN 16
+
 typedef enum GPUQueryType {
   GPU_QUERY_OCCLUSION = 0,
 } GPUQueryType;

source/blender/gpu/intern/gpu_select_sample_query.cc

@@ -37,7 +37,7 @@ struct GPUSelectQueryState {
   /** GPU queries abstraction. Contains an array of queries. */
   QueryPool *queries;
   /** Array holding the id corresponding id to each query. */
-  Vector<uint> *ids;
+  Vector<uint, QUERY_MIN_LEN> *ids;
   /** Cache on initialization. */
   GPUSelectResult *buffer;
   /** The capacity of the `buffer` array. */
@@ -71,7 +71,7 @@ void gpu_select_query_begin(GPUSelectResult *buffer,
   g_query_state.index = 0;
   g_query_state.oldhits = oldhits;
 
-  g_query_state.ids = new Vector<uint>();
+  g_query_state.ids = new Vector<uint, QUERY_MIN_LEN>();
   g_query_state.queries = GPUBackend::get()->querypool_alloc();
   g_query_state.queries->init(GPU_QUERY_OCCLUSION);
 
@@ -149,7 +149,7 @@ uint gpu_select_query_end()
   }
 
   Span<uint> ids = *g_query_state.ids;
-  Vector<uint32_t> result(ids.size());
+  Vector<uint32_t, QUERY_MIN_LEN> result(ids.size());
   g_query_state.queries->get_occlusion_result(result);
 
   for (int i = 0; i < result.size(); i++) {

source/blender/gpu/opengl/gl_query.cc

@@ -37,8 +37,9 @@ void GLQueryPool::begin_query()
   /* TODO: add assert about expected usage. */
   while (query_issued_ >= query_ids_.size()) {
     int64_t prev_size = query_ids_.size();
-    query_ids_.resize(prev_size + QUERY_CHUNCK_LEN);
-    glGenQueries(QUERY_CHUNCK_LEN, &query_ids_[prev_size]);
+    int64_t chunk_size = prev_size == 0 ? query_ids_.capacity() : QUERY_CHUNCK_LEN;
+    query_ids_.resize(prev_size + chunk_size);
+    glGenQueries(chunk_size, &query_ids_[prev_size]);
   }
   glBeginQuery(gl_type_, query_ids_[query_issued_++]);
 }

source/blender/gpu/opengl/gl_query.hh

@@ -18,7 +18,7 @@ namespace blender::gpu {
 class GLQueryPool : public QueryPool {
  private:
   /** Contains queries object handles. */
-  Vector<GLuint> query_ids_;
+  Vector<GLuint, QUERY_MIN_LEN> query_ids_;
   /** Type of this query pool. */
   GPUQueryType type_;
   /** Associated GL type. */