With [vr] eye_tracked_foveation (on by default on the Steam Frame, off
elsewhere) the runtime asks for XR_EXT_eye_gaze_interaction. When the
system reports an eye tracker, OpenXRInput binds the gaze pose and
locates it for each packet's display time, in the space the eye views
are located in; vr/eye_gaze.h turns it into tangents of each eye's own
view, which AuroraStereoFrame now carries (appended, after the existing
prefix).
Aurora centres the eye's fragment density map on the gaze snapped to a
cell of two map texels (about 3 degrees). Each eye keeps up to 32 maps,
one per cell looked at, so a glance back reuses its map; a new map is
bound once its upload completes, and until then the eye keeps the map
it had. Without a tracked gaze (a blink, no tracker, the setting off)
foveation centres on the forward direction exactly as before: the
forward maps are byte-identical.
Also logs every extension the OpenXR runtime offers at startup, so the
first Steam Frame session shows what SteamVR's Android runtime has.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019HBRGKTE1GnN2ah8gcZKr3
- Updated stereo_frame_worker_smoke.cpp to allow dynamic headset rates and prediction lead time.
- Improved logging to include motion diagnostics and adjusted frame submission logic based on headset frequency.
- Enhanced stereo_interpolation_test.cpp with additional tests for camera motion separation and playback cadence.
- Introduced MkwVRReadSceneView function to read the camera view matrix for improved scene rendering.
- Modified VR first-person logic to support scene view reading and validation.
- Added scene_camera.hpp to encapsulate camera motion handling and inverse view calculations.
- Ensured that the VR integration layer correctly logs motion diagnostics and handles scene playback accurately.
- Implemented DVDReadVrAsset function to read mapped disc paths for VR assets.
- Created HeldItem structure and ReadHeldItem function for managing held items in the game.
- Developed unit tests for ReadHeldItem to ensure correct functionality and edge case handling.
- Added cockpit item data tests to validate model indexing and parsing of archives.
- Added support for hiding model arrays in the rendering pipeline to optimize performance during VR gameplay.
- Introduced new functions in `gx_model_visibility` to manage hidden model arrays and their visibility based on game state.
- Enhanced cockpit recentering logic to ensure accurate seat measurements and eye positioning during gameplay.
- Updated `FirstPersonState` to track cockpit height and forward direction, improving VR experience.
- Added tests for Bullet Bill model visibility and cockpit height adjustments to ensure functionality and stability.
- Refactored existing code to accommodate new features and improve overall code organization.
- New [vr] hand_tracking (default off, Quest only for now): the Quest launcher's Settings > VR and
the headset panel's VR tab, under hand steering. Two XR_EXT_hand_tracking trackers are located
every XR frame: with the controllers held the Quest builds the joints from their touch sensors
(XR_EXT_hand_tracking_data_source's controller source), once they are put down from its
cameras. The trackers exist only while the option and hand steering are on and also serve the
runtime hand mesh; the extensions (plus XR_FB_hand_tracking_aim) are asked for when either is
on at launch.
- Aurora skins the runtime mesh with the joints themselves (tracked pose times inverse bind pose,
no curl, no grip); runtimes with joints but no mesh get a skeleton; non-finite joints put only
that hand back on its grip curl. AuroraCockpitHand carries the 26 seated-frame joints and radii.
- The manifest declares horizonos.permission.HAND_TRACKING (and the deprecated
com.oculus.permission.HAND_TRACKING), both normal permissions with no prompt on a Quest 3, and
oculus.software.handtracking as optional; without it Horizon OS keeps the app controllers-only.
Bare hands then drive khr/simple_controller, so on Android a hand whose squeeze action is
inactive and select active is treated as bare: with the option off it presses nothing but the
menu gesture, is not drawn and feeds no Wii Remote motion.
- Interaction-profile changes and tracker sources are logged. The pure rules (grasp from finger
flexion, bare latch, pinch gate, bare-hand buttons, flick) live in the OpenXR-free
vr/openxr_hand_tracking.h with mkw_vr_hand_tracking_tests; the driving and flick parts are
wired in the next commits. Docs: OPENXR.md, docs/quest-port.md, README.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Introduced a new configuration option for immersive window mode in runtime_config.h.
- Updated the parsing and setting functions to handle the immersive window state.
- Modified the OpenXR backend to support rendering with the immersive window, blending the race view with the surrounding environment.
- Enhanced the settings overlay to allow users to select between immersive, immersive window, and flat screen race views.
- Implemented GPU rendering logic for the immersive window mask, ensuring correct visual output in various rendering paths.
- Added tests to validate the immersive window functionality and its interaction with existing race view settings.
- Eyes render under a VK_EXT_fragment_density_map: full rate around each eye's forward direction,
2x2 then 4x4 pixel blocks towards the edges ([vr] foveation = off|low|medium|high, default off).
XR_FB_foveation cannot help here: the runtime's maps only shape passes drawing into its
swapchain, and the eyes reach it through a copy.
- aurora-main/patches/dawn/aurora_fdm.inc: Dawn enables the extension only on request and for
dynamic rendering, flags every render pipeline, and chains an immutable RG8 map into any pass
whose first color attachment is a view bound to one (ABI: include/aurora/dawn_fdm_abi.h).
- android/Build-QuestDawn.ps1 builds the pinned Dawn revision with those patches for arm64
(dawn-build CI flags, protobuf off) into a cached package; Build-Quest.ps1 links it
(-StockDawn opts out) and AuroraDawnProvider.cmake enables the ABI from its manifest.
- lib/gfx/foveation.hpp generates the maps (32 px per texel, densities 255/127/63); an eye is
foveated only when single_pass_eyes draws it in one render pass. Menus never are.
- Live level from the headset panel's VR tab and the launcher; the launch decides whether the
device has maps. debug.wiicompiled.foveation and debug.wiicompiled.fdm for A/B.
- Tests: Foveation cases in gx_fifo_tests, mkw_vr_config_tests. Docs: OPENXR.md, quest-port.md.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Added eye_pass_plan.hpp: an eye keeps drawing in the render pass it has open across the frame's
GX copies (only the mono render performs them), skips passes a later clear of the whole color and
depth erases, and splits only for a clear of color alone or depth alone.
- render_stereo_eye follows that plan; the cockpit fallback is drawn inside the last pass instead of
a render pass of its own that loaded the eye back.
- The stencil is now cleared where the eye actually starts, so a cockpit mask drawn in an erased pass
can no longer punch holes in the HUD.
- [vr] single_pass_eyes (default on, live from F10) and debug.wiicompiled.eye_passes 0/1 on the Quest
switch back to one render pass per recorded pass for A/B timing.
- Each new plan shape is logged once ("Eye replay plan: ...").
- Tests: EyePassPlan cases in gx_fifo_tests; stereo_multiplayer_smoke and stereo_frame_worker_smoke
pass on D3D12.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
OpenXR's grip +X is normal to the palm but points away from it on the left hand
and into it on the right, which is exactly what makes both grips carry the same
orientation when the hands hold a wheel symmetrically. The fingers therefore run
along -Y on both hands, and it is the geometry across the palm that mirrors.
Building the right hand's fingers on +Y instead left them pointing at the player
while the real hand faced forward.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The animated copy is registered for the vehicle's whole position array, and a
race frame has 228 primitives that bind it. Suppressing draw merging for all of
them turned the kart's display list back into hundreds of separate draw calls,
each recorded once and replayed in the mono pass and both eyes: on a Quest 3 the
app frame went from 15.7 ms to 21.7 ms with the GPU pinned at 97 %, which past
the 72 Hz deadline reads as 56 FPS against 42.
Merging is unsafe only between draws that resolved the array differently, since
the merged whole renders through the binding of the draw it folds into. So
resolve the replacement once per draw, before the merge test, and merge into a
draw that reached the same decision. Draws of an opponent sharing the asset
merge with each other again too, and with no set registered the test is exactly
what it was before the feature.
Measured at the same place on a Quest 3: 60 FPS against 42, the app frame back
to 13.9 ms, 564 draw calls a frame with 11808 primitives merged away.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The Quest offers XR_FB_hand_tracking_mesh without the app declaring hand
tracking, so a Quest 3 draws the runtime's own hand mesh rather than the
procedural gloves, and its fingers bent backwards on grip: the mesh's
joints point -Z towards the fingertip and +Y out of the back of the hand,
so flexion is negative about the joint's own X, on both hands. Fix and
test by the repository owner; OPENXR.md said the Quest always drew the
gloves, which was wrong.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
On a Quest 3 the gloves' fingers pointed up out of the fist and bent out
of the back of the hand as the grip closed: they were built along the
grip's -Z, which runs up the tube the curled fingers form towards the
thumb, and curled towards -Y, which is backwards on the right hand.
The fingers now run along Y, forwards out of the palm (+Y on the right
hand, -Y on the left, the frame being right-handed), and close towards
+X, the palm's outward normal, with the thumb on the -Z side of both
hands. A test pins the reach, the closing direction and that nothing
bends out of the back of the hand.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
A stereo packet can now carry an AuroraCockpit: tracked hands and, when the
vehicle's own wheel cannot be animated, a synthetic steering wheel or
handlebar, all in metres in the seated frame. Each eye draws it inside the
scene's pass just before the first virtual-screen draw, depth-tested with
the world's own depth mapping (captured from a full-view world draw), so
the kart and track occlude the hands and the 2D layer cannot hide them.
Hands use a runtime-provided hand mesh when one is supplied and a
procedural glove otherwise.
aurora_set_stereo_scene_anchor_scaled lets the sealed frame own its world
scale: each eye's head translation is rescaled from the packet's scale to
the frame's. A non-finite cockpit is dropped with one warning; the frame
still renders.
Ported from heurazy's mario-kart-wii-VR-port (GPL-3.0-or-later).
The VR first-person camera will animate the local vehicle's steering wheel
by handing Aurora a rotated copy of one of the vehicle's position arrays.
A draw picks up the copy only when it binds that exact array with the local
vehicle's position matrix (per-position ownership for indexed-matrix
draws), so opponents sharing the asset keep the original vertices. The copy
takes the same padded upload path as the original and never touches the
shared array cache.
Ported from heurazy's mario-kart-wii-VR-port (GPL-3.0-or-later).
Brings in patchzyy/Wiicompiled main: os_sleep parked-thread fix (#195),
HTTPS Retro WFC payload (#198), macOS build guide (#177), and the
reverse-Z depth fix (#134).
Conflicts were in aurora-main/lib/gfx/common.cpp and lib/gx/shader.cpp,
both from #134, which lands squarely on the VR stereo replay path.
#134 makes UseReversedZ genuinely reversed: the near/far correction now
applies exactly once, inside effective_projection(), instead of being
applied there AND per-vertex in the shader (the double application had
been cancelling out, so "reversed" Z silently behaved like forward Z).
Three pieces of the VR path were built against that old behaviour and
would have broken silently, so they are adapted here:
- shader.cpp exact-screen-depth parked the virtual screen at -0.5*w
specifically so the shader's following negation would land it at
+0.5*w. With that negation gone it now writes +0.5*w directly; keeping
the minus sign would park the screen at NDC -0.5, outside the clip
volume, discarding every 2D/HUD draw.
- stereo_replay.hpp backend_ndc_depth_row re-applied the correction to
the projection it was handed. That projection is effective_projection()
output, which now already carries it, so the function is a pass-through
of the Z row and no longer depends on the reversed-Z setting; the dead
bool parameter is dropped. Re-applying it would invert the virtual
screen's depth ordering, so 2D layers meant to sit on top would lose
the depth test to the ones behind them.
- shader_info.cpp stages the host depth window for that exact-depth path.
It now uses the same reversed-Z remap as upstream's new SetViewport
code, since frag_depth is written directly and has to reproduce the
window the fixed viewport transform would have applied. Restricted
depth windows (how the game forces an element in front of everything)
are exactly the 2D draws the virtual screen carries.
The SetViewport resolution keeps upstream's remap but retains the
ordering/clamp guard our version had: for any ordered guest range the
result is identical to upstream, and it avoids handing WebGPU
minDepth > maxDepth for the swapped pair MKW is known to emit. The VR eye
replay reuses these recorded values, so the guard covers that path too.
Test updates:
- stereo_replay_test now asserts the composed Z row against the staged
projection's own Z row rather than against the helper's output, so it
actually catches a re-introduced double correction (verified: it fails
when the old negation is put back; the previous self-consistent form
passed).
- gx_fifo_test's clearDepthValue expectation followed #134's deliberate
clear_depth_value() inversion, expressed through UseReversedZ rather
than hardcoded. Upstream changed the behaviour without updating this
test, so it fails on upstream/main as-is.
Verified: aurora suite 247 passed with the same 2 failures that already
fail on the pre-merge branch (IndexedPaletteHistoryKeepsAbsoluteVertexSlots,
PacksOneUniformWhenBothHalvesNeedInitialValue - both pre-existing, unrelated
to depth); shader.cpp and common.cpp compile clean; translator suite 577
passed. Not yet validated on-device in VR.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude