Skip to content

feat(weave/android): hand the v4 2D overlay to the DP (2D under the lens) #1845

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
dfattal merged 1 commit into from
Oct 7, 2026
Merged

feat(weave/android): hand the v4 2D overlay to the DP (2D under the lens) #1845

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
22 changes: 22 additions & 0 deletions src/xrt/compositor/multi/comp_multi_private.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -804,6 +804,20 @@ struct multi_compositor
uint32_t overlay_refusals;
//! @}

/*!
* @name 2D under the lens (ADR-027 Amendment, XR_DXR_weave v14/v15)
* Android twin of the desktop-Linux fields: per-submit overlay hints set by
* comp_multi_weave_android_set_overlay_hints right before each submit
* (never sticky), forwarded to the DP with the overlay when its
* set_overlay_2d slot takes it. @c overlay_in_dp is the last verdict so the
* WARN fires on a change only.
* @{
*/
bool overlay_unchanged;
float overlay_filter_strength;
uint8_t overlay_in_dp; //!< 0 = no overlay seen yet, 1 = runtime blend, 2 = DP composites
//! @}

//! @name Arch-C weave satellite (#1277 P0) — service-presented weave
//! `debug.dxr.weave_satellite=1`: the woven output is blitted onto a
//! SERVICE-owned full-panel overlay surface (the #558 machinery) and
Expand Down Expand Up @@ -1500,6 +1514,14 @@ comp_multi_weave_android_satellite_clear(struct multi_compositor *mc);
*/
bool
comp_multi_weave_android_satellite_presented(struct multi_compositor *mc);

/*!
* XR_DXR_weave v14/v15 overlay hints for the NEXT comp_multi_weave_submit only
* (Android twin of comp_multi_weave_linux_set_overlay_hints). @p filter_strength
* outside [0, 1] (incl. NaN) means "not set by the app": the DP's own default.
*/
void
comp_multi_weave_android_set_overlay_hints(struct xrt_compositor *xc, bool overlay_unchanged, float filter_strength);
#endif

bool
Expand Down
64 changes: 63 additions & 1 deletion src/xrt/compositor/multi/comp_multi_weave_android.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1439,6 +1439,22 @@ comp_multi_weave_android_satellite_clear(struct multi_compositor *mc)
U_LOG_W("weave satellite(#1277): overlay CLEARED on weave-idle (stale-frame guard)");
}

void
comp_multi_weave_android_set_overlay_hints(struct xrt_compositor *xc, bool overlay_unchanged, float filter_strength)
{
struct multi_compositor *mc = multi_compositor(xc);
if (mc == NULL || mc->msc == NULL) {
return;
}
weave_ensure_mutex(mc);
os_mutex_lock(&mc->weave.mutex);
mc->weave.overlay_unchanged = overlay_unchanged;
// Out of range (incl. NaN) = not set by the app: the DP's own default.
mc->weave.overlay_filter_strength =
(filter_strength >= 0.0f && filter_strength <= 1.0f) ? filter_strength : -1.0f;
os_mutex_unlock(&mc->weave.mutex);
}

bool
comp_multi_weave_android_satellite_presented(struct multi_compositor *mc)
{
Expand Down Expand Up @@ -2329,6 +2345,52 @@ comp_multi_weave_submit(struct xrt_compositor *xc,
// completion contract for the whole frame.
// Kill-switch: debug.xrt.DXR_ANDROID_WEAVE_SPLIT=0 restores the old
// single-submission behaviour.
// 2D under the lens (ADR-027 Amendment): offer the v4 overlay to the DP so
// it composites the layer INSIDE the weave and band-limits it for the lens,
// instead of the post-weave blend below (which aliases per eye wherever the
// lens is on). Only when the layer is exactly the process_atlas target size
// (the slot's contract; the satellite physical-rect and v6 N-view targets
// are not window-sized). The layer is put in SHADER_READ_ONLY here, BEFORE
// the self-submit split, so it is ready when a self-submitting DP (CNSDK)
// samples it during process_atlas. A DP that declines leaves the post-weave
// blend to do exactly what it did before. Per submit (#1835): the import
// is cached by AHB pointer, so a submit that chains no overlay must not
// hand the DP the last one.
struct xrt_display_processor_vk *dp_vk = (struct xrt_display_processor_vk *)mc->weave.dp;
bool ov_in_dp = false;
if (mc->weave.overlay_image != VK_NULL_HANDLE && mc->weave.overlay_this_submit &&
mc->weave.overlay_w == mc->weave.out_w &&
mc->weave.overlay_h == mc->weave.out_h && xrt_display_processor_vk_supports_overlay_2d(dp_vk)) {
VkImageMemoryBarrier ov_to_read = {
.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
.srcAccessMask = VK_ACCESS_MEMORY_WRITE_BIT,
.dstAccessMask = VK_ACCESS_SHADER_READ_BIT,
.oldLayout = mc->weave.overlay_first_use ? VK_IMAGE_LAYOUT_UNDEFINED
: VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL,
.newLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL,
.image = mc->weave.overlay_image,
.subresourceRange = range,
};
mc->weave.overlay_first_use = false;
vk->vkCmdPipelineBarrier(cmd, VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT,
0, 0, NULL, 0, NULL, 1, &ov_to_read);
// v15 strength first: the D3D11 / Linux order.
xrt_display_processor_vk_set_overlay_2d_filter_strength(dp_vk, mc->weave.overlay_filter_strength);
ov_in_dp = xrt_display_processor_vk_set_overlay_2d(
dp_vk, mc->weave.overlay_view, (VkFormat_XDP)WEAVE_VK_FORMAT, mc->weave.overlay_w,
mc->weave.overlay_h, XRT_ATLAS_ENCODING_ENCODED, mc->weave.overlay_unchanged);
}
if (mc->weave.overlay_image != VK_NULL_HANDLE && mc->weave.overlay_this_submit) {
const uint8_t verdict = ov_in_dp ? 2u : 1u;
if (verdict != mc->weave.overlay_in_dp) {
mc->weave.overlay_in_dp = verdict;
U_LOG_W("weave(#1036) v4: 2D overlay %ux%u (target %ux%u) -> %s", mc->weave.overlay_w,
mc->weave.overlay_h, mc->weave.out_w, mc->weave.out_h,
ov_in_dp ? "composited by the DP inside the weave (2D under the lens)"
: "runtime post-weave premul-over blend");
}
}

bool weave_split = weave_split_enabled() && xrt_display_processor_is_self_submitting(mc->weave.dp);
if (weave_split) {
if (vk->vkEndCommandBuffer(cmd) != VK_SUCCESS) {
Expand Down Expand Up @@ -2406,7 +2468,7 @@ comp_multi_weave_submit(struct xrt_compositor *xc,
// blend, so crisp 2D lands on top of the interlaced 3D at screen depth.
// The overlay is NOT woven — it is drawn after process_atlas onto the same
// output attachment.
if (mc->weave.overlay_image != VK_NULL_HANDLE && mc->weave.overlay_this_submit) {
if (mc->weave.overlay_image != VK_NULL_HANDLE && mc->weave.overlay_this_submit && !ov_in_dp) {
bool blend_ready = mc->weave.overlay_blend_initialized;
if (!blend_ready) {
blend_ready = vk_local2d_composite_init(&mc->weave.overlay_blend, vk, WEAVE_VK_FORMAT,
Expand Down
4 changes: 4 additions & 0 deletions src/xrt/ipc/server/ipc_server_handler.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6727,6 +6727,10 @@ ipc_handle_weave_submit(volatile struct ipc_client_state *ics,
comp_multi_weave_linux_set_mono_in_2d(ics->xc, args->mono_in_2d != 0);
// v14/v15: per submit, read only when the DP composites the overlay.
comp_multi_weave_linux_set_overlay_hints(ics->xc, args->overlay_unchanged != 0, args->overlay_filter_strength);
#elif defined(XRT_OS_ANDROID)
// v14/v15: same per-submit hints for the Android weave (2D under the lens).
comp_multi_weave_android_set_overlay_hints(ics->xc, args->overlay_unchanged != 0,
args->overlay_filter_strength);
#endif
bool ok = comp_multi_weave_submit( //
ics->xc, handles[0], //
Expand Down
Loading