Mesa (main): gallium/u_vbuf: add a fast path to skip refcounting for uploaded user buffers

[GitLab Mirror]

Module: Mesa
Branch: main
Commit: b8d59e68beb27b1308d726c2e515b23e87e47f79
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=b8d59e68beb27b1308d726c2e515b23e87e47f79

Author: Marek Olšák <marek.olsak at amd.com>
Date:   Tue May 25 14:55:42 2021 -0400

gallium/u_vbuf: add a fast path to skip refcounting for uploaded user buffers

This improves performance by 23% with radeonsi and Ryzen 3900X running
the game torcs.

All vertex buffers must be user buffers to get this.

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz at gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10992>

---

 src/gallium/auxiliary/util/u_vbuf.c | 21 +++++++++++++++++++--
 1 file changed, 19 insertions(+), 2 deletions(-)

diff --git a/src/gallium/auxiliary/util/u_vbuf.c b/src/gallium/auxiliary/util/u_vbuf.c
index af6f56e4ac3..02a6ba00300 100644
--- a/src/gallium/auxiliary/util/u_vbuf.c
+++ b/src/gallium/auxiliary/util/u_vbuf.c
@@ -1271,8 +1271,25 @@ static void u_vbuf_set_driver_vertex_buffers(struct u_vbuf *mgr)
    start_slot = ffs(mgr->dirty_real_vb_mask) - 1;
    count = util_last_bit(mgr->dirty_real_vb_mask >> start_slot);
 
-   pipe->set_vertex_buffers(pipe, start_slot, count, 0, false,
-                            mgr->real_vertex_buffer + start_slot);
+   if (mgr->dirty_real_vb_mask == mgr->enabled_vb_mask &&
+       mgr->dirty_real_vb_mask == mgr->user_vb_mask) {
+      /* Fast path that allows us to transfer the VBO references to the driver
+       * to skip atomic reference counting there. These are freshly uploaded
+       * user buffers that can be discarded after this call.
+       */
+      pipe->set_vertex_buffers(pipe, start_slot, count, 0, true,
+                               mgr->real_vertex_buffer + start_slot);
+
+      /* We don't own the VBO references now. Set them to NULL. */
+      for (unsigned i = 0; i < count; i++) {
+         assert(!mgr->real_vertex_buffer[start_slot + i].is_user_buffer);
+         mgr->real_vertex_buffer[start_slot + i].buffer.resource = NULL;
+      }
+   } else {
+      /* Slow path where we have to keep VBO references. */
+      pipe->set_vertex_buffers(pipe, start_slot, count, 0, false,
+                               mgr->real_vertex_buffer + start_slot);
+   }
    mgr->dirty_real_vb_mask = 0;
 }