Task scheduler: Optimize subsequent pushing bunch of tasks

The idea is to accumulate all new tasks in a thread local queue
first without doing any thread synchronization (aka, locks and
conditional variables) and move those tasks to a scheduler queue
once they are all ready. This way we avoid per-task-pool lock
and only have one lock per bunch of tasks.

This is particularly handy when scheduling new dependency graph
node children. Brings FPS of cached simulation from the linked
below file from ~30 to ~50.

See documentation for BLI_task_pool_delayed_push_{begin, end}
and for TaskThreadLocalStorage::do_delayed_push.

Fixes T50027: Rigidbody playback and simulation performance regression with new depsgraph

Thanks Bastien for the review!

1 files changed, 7 insertions, 0 deletions

diff --git a/source/blender/blenlib/BLI_task.h b/source/blender/blenlib/BLI_task.h
index c3c587275e1..721327d26a8 100644
--- a/source/blender/blenlib/BLI_task.h
+++ b/source/blender/blenlib/BLI_task.h
@@ -106,6 +106,13 @@ void *BLI_task_pool_userdata(TaskPool *pool);
 /* optional mutex to use from run function */
 ThreadMutex *BLI_task_pool_user_mutex(TaskPool *pool);
 
+/* Delayed push, use that to reduce thread overhead by accumulating
+ * all new tasks into local queue first and pushing it to scheduler
+ * from within a single mutex lock.
+ */
+void BLI_task_pool_delayed_push_begin(TaskPool *pool, int thread_id);
+void BLI_task_pool_delayed_push_end(TaskPool *pool, int thread_id);
+
 /* Parallel for routines */
 typedef void (*TaskParallelRangeFunc)(void *userdata, const int iter);
 typedef void (*TaskParallelRangeFuncEx)(void *userdata, void *userdata_chunk, const int iter, const int thread_id);