Add opt-in async OVRTX rendering

.github/test-subsets/postmerge-rendering.toml

@@ -31,6 +31,8 @@ rendering-correctness-kitless-legacy = [
   "source/isaaclab_tasks/test/core/test_rendering_cartpole_kitless.py::test_rendering_cartpole_kitless[legacy-ovphysx-ovrtx-albedo]",
   "source/isaaclab_tasks/test/core/test_rendering_cartpole_kitless.py::test_rendering_cartpole_kitless[legacy-ovphysx-ovrtx-rgb]",
   "source/isaaclab_tasks/test/core/test_rendering_cartpole_kitless.py::test_rendering_cartpole_kitless[legacy-ovphysx-ovrtx-semantic_segmentation]",
+  "source/isaaclab_tasks/test/core/test_rendering_cartpole_kitless.py::test_rendering_cartpole_kitless_async[legacy-newton-ovrtx-rgb]",
+  "source/isaaclab_tasks/test/core/test_rendering_cartpole_kitless.py::test_rendering_cartpole_kitless_async[legacy-ovphysx-ovrtx-rgb]",
 ]
 
 rendering-correctness-kitless-ovstage = [
@@ -39,4 +41,6 @@ rendering-correctness-kitless-ovstage = [
   "source/isaaclab_tasks/test/core/test_rendering_cartpole_kitless.py::test_rendering_cartpole_kitless[ovstage-ovphysx-ovrtx-albedo]",
   "source/isaaclab_tasks/test/core/test_rendering_cartpole_kitless.py::test_rendering_cartpole_kitless[ovstage-ovphysx-ovrtx-rgb]",
   "source/isaaclab_tasks/test/core/test_rendering_cartpole_kitless.py::test_rendering_cartpole_kitless[ovstage-ovphysx-ovrtx-semantic_segmentation]",
+  "source/isaaclab_tasks/test/core/test_rendering_cartpole_kitless.py::test_rendering_cartpole_kitless_async[ovstage-newton-ovrtx-rgb]",
+  "source/isaaclab_tasks/test/core/test_rendering_cartpole_kitless.py::test_rendering_cartpole_kitless_async[ovstage-ovphysx-ovrtx-rgb]",
 ]

docs/source/testing/benchmarks.rst

@@ -78,6 +78,25 @@ JSON. ``schema`` writes the stable, typed JSON bundle used for programmatic
 comparison. With multiple formatters, their filenames include ``_summary`` and
 ``_schema`` respectively.
 
+OVRTX camera tasks can compare synchronous against asynchronous (one-frame-latency) rendering
+through a Hydra override on the camera's renderer configuration (use ``False`` to force
+synchronous):
+
+.. code-block:: bash
+
+   uv run isaaclab benchmark runtime \
+       --task Isaac-Cartpole-Camera-Direct \
+       --num_envs 256 \
+       --warmup_frames 30 \
+       --num_frames 200 \
+       --benchmark_formatter schema,omniperf \
+       --output_path ./results/ovrtx_async \
+       physics=newton_mjwarp renderer=ovrtx presets=rgb \
+       env.tiled_camera.renderer_cfg.async_rendering=True
+
+``OVRTX_ASYNC_RENDERING=1`` in the environment toggles the same path for any task, which avoids
+naming a camera that a given task may not define.
+
 Warm-Up
 ~~~~~~~
 

source/isaaclab_ov/changelog.d/pv-ovrtx-async-slot-buffers.minor.rst

@@ -0,0 +1,17 @@
+Added
+^^^^^
+
+* Added an opt-in asynchronous OVRTX render path controlled by
+  :attr:`~isaaclab_ov.renderers.ovrtx_renderer_cfg.OVRTXRendererCfg.async_rendering`
+  (default ``False``). When enabled, rendering overlaps simulation and Python work and camera
+  outputs arrive one frame later, improving throughput. The first frame after (re)initialization
+  is consumed immediately so the first camera read returns a valid frame; later frames are
+  pipelined. Available on both the legacy and ovstage scene-ownership paths; under ovstage each
+  scene write first drains any render still in flight, since OVRTX reads the stage's storage in
+  place.
+* Added the ``OVRTX_ASYNC_RENDERING`` environment variable to override
+  :attr:`~isaaclab_ov.renderers.ovrtx_renderer_cfg.OVRTXRendererCfg.async_rendering` for tests.
+* Added the ``OVRTX_NUM_BUFFERS`` environment variable to configure the render queue depth: the
+  number of asynchronous renders kept in flight (default ``2``, values below ``2`` are clamped).
+  Larger values overlap more simulation with rendering at the cost of extra frames of camera
+  latency.