The shared vertex/index arena is the largest single GPU allocation acdream
makes (384 MiB + 128 MiB) and the one the Vulkan backend has the most specific
plan for (campaign doc section 4.3). This slice swaps the resource handle type
underneath it and changes nothing else: the reclaimable-range allocator, the
growth quanta, the budgeted incremental grow-and-copy, the retirement-ledger
gating, the abort ticket, the LRU that drives eviction, and the 896 MiB
dual-generation physical ceiling are all untouched. That is deliberate - those
are the semantics section 4.3 says the Vulkan arena must mirror exactly, so
preserving them is the point of the slice rather than an incidental constraint.
What moved:
- GlobalMeshBuffer's two GL buffer objects became IGpuBuffer, allocated through
IGpuDevice.CreateBuffer with DeviceLocal residency and Vertex-or-Index plus
both transfer usages (the arena is simultaneously a draw source and both ends
of its own migration, which is exactly why GpuBufferUsage is a flags enum).
- UploadMesh's two hand-rolled BufferSubData sites became IGpuBuffer.Upload.
The old code staged indices through GL_COPY_WRITE_BUFFER specifically so an
upload could not mutate whichever VAO a preceding render pass left bound;
Upload stages through a neutral binding point of the backend's choosing, so
that property now comes for free instead of by hand.
- AdvanceMigration's CopyBufferSubData became IGpuBuffer.CopyTo - a device-side
copy, which the Vulkan backend will record as vkCmdCopyBuffer. The live
prefix still never round-trips through system memory.
- BeginMigration/CommitMigration/AbortMigration/Dispose now carry IGpuBuffer in
the migration record and the abort ticket instead of raw uint names, so the
ticket's identity check is a resource identity rather than a number that goes
stale the moment the buffer is deleted.
What deliberately did not move. A VAO has no RHI verb - Vulkan bakes vertex
input into the pipeline - and WbDrawDispatcher, EnvCellRenderer and
ParticleRenderer still bind VAO/VBO/IBO with raw GL until V4c hands them the
pass encoder. So GlobalMeshBuffer keeps its GL handle for the vertex array and
its attribute layout, and VBO/IBO became computed properties that publish the
backing GL name of the buffer the arena now owns as an IGpuBuffer. One private
RequireGlBuffer helper is the single place that reaches through the interface,
and it disappears with those consumers. ObjectMeshManager therefore needed no
upload-path change at all - it reads those same three properties.
Two decisions worth recording.
First, arena deletes do not route through IGpuBuffer.Dispose. The arena already
gates every delete behind its own GpuRetirementLedger and decrements its
physical-capacity accounting in the same retirement stage; Dispose would defer
the physical free through the device queue a second time, so the accounting
would run ahead of real GPU residency and could admit a migration that breaches
the 896 MiB ceiling. GlGpuBuffer gains DeleteRetired for callers that have
already proved flight safety, and GlobalMeshBuffer composes it into a release
whose four stages match TrackedGlResource.CreateRetryableBufferDeletion exactly
- precondition, mutation-with-validation, byte accounting, resource-count
accounting - so a driver failure re-issues only the delete and never
double-counts.
Second, two corrections in the GL backend, both required to keep this port
behaviour-preserving rather than merely compiling. GlGpuBuffer's glBufferData
usage hint now follows residency (DeviceLocal -> StaticDraw), which is what the
arena has always requested; the host-writable rings and texture table keep
DynamicDraw and are unaffected. And a failed allocation now releases the GL
name it had already created - GL_OUT_OF_MEMORY is a real outcome for a 384 MiB
growth destination, and the previous code leaked the name on that path.
Plumbing: the device reaches the arena through WbMeshAdapter and
ObjectMeshManager. Their constructors became internal because IGpuDevice is an
internal type by the pinned contract, matching what V4a did for BitmapFont,
DebugLineRenderer and TextRenderer; both classes stay public and every caller
already lives inside AcDream.App or its InternalsVisibleTo test assemblies. The
unused public GlobalMeshBuffer(GL) convenience constructor is gone - it could
not supply a device and had no callers.
Gates. Release build green with TreatWarningsAsErrors. App tests 3,843 passed /
3 skipped, exactly the slice baseline; complete Release suite 8,906 passed / 5
skipped. Offline pixel gate against 79ee2361: 25 differing pixels of 563,200
(fraction 4.44e-05), against a same-commit control captured immediately
afterwards of 24 - the change is indistinguishable from capture noise and sits
40x under the 0.001 threshold. An earlier gate run was discarded rather than
interpreted: its client log showed real ScrollUp/ScrollDown input reaching the
offline window, which zoomed the camera, and a camera-motion difference is not
a rendering result.
No divergence-register row: this slice changes no retail-facing behaviour.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Five small post-cleanup items from T7 code review:
I1: Removed dead `datDir` parameter from WbMeshAdapter ctor (parameter
was unused after _wbDats removal; ArgumentNullException.ThrowIfNull
was misleading). Updated call sites in GameWindow.cs and
WbMeshAdapterTests.cs.
I2: Updated stale GameWindow.cs comment that still described
WbMeshAdapter as opening its own dat handles. Now reflects Phase O
state: shared DatCollection via DatCollectionAdapter.
I3: Documented thread-safety contract on RenderStateCache (render-thread
only — required for the mutable-static GL sentinel pattern).
M1: Added comment on IDatReaderWriter's write-path methods noting they
are preserved for verbatim compatibility but unused in acdream.
M3: Added comment on Chorizite.Core PackageReference in Core.csproj
explaining the previously-transitive dependency.
Also excluded SplitFormulaDivergenceTest.cs from the test build via
<Compile Remove>: this N.5b one-time data-collection test referenced
WorldBuilder.Shared types directly; after Phase O-T7 dropped that
project reference it no longer compiles. The sweep data it produced
already informed the N.5b Path-C decision and the file is retained
in the tree for historical reference.
Build green; tests green (1146 + 8 pre-existing failures baseline
maintained).
Spec: docs/superpowers/specs/2026-05-21-phase-o-dat-path-unification-design.md
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
WbDrawDispatcher draws all entities through WB's ObjectRenderData
(VAO/VBO per GfxObj, per-batch IBO) using acdream's TextureCache for
texture resolution. Two-pass rendering (opaque+ClipMap, then
translucent) matching the existing InstancedMeshRenderer pattern.
Per-entity single-instance drawing for N.4 simplicity — true
instancing grouping deferred to N.6.
Atlas-tier entities: mesh from WB, texture from TextureCache via
batch SurfaceId. Per-instance-tier entities: AnimatedEntityState
drives part overrides + hidden-parts, palette/surface overrides
resolve through TextureCache's composite-key caches.
Side-table population (Task 23 folded in): WbMeshAdapter now takes
DatCollection and populates AcSurfaceMetadataTable on first
IncrementRefCount per GfxObj. The side-table provides TranslucencyKind
(critical for ClipMap alpha-test on vegetation) plus Luminosity,
Diffuse, SurfOpacity, NeedsUvRepeat, DisableFog for sky-pass and
lighting.
GameWindow wiring: when WbFoundationFlag is enabled, WbDrawDispatcher
draws everything and InstancedMeshRenderer is skipped. Flag-off path
is unchanged.
Matrix composition: restPose * animOverride * entityWorld, matching
the spec. Three MatrixCompositionTests verify the contract.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Without this, ObjectMeshManager.StagedMeshData and
OpenGLGraphicsDevice._glThreadQueue grow unbounded as background
workers prep mesh data + queue GL actions. Visual stress test of
flag-on at radius 7 showed real FPS drop and rising frame latency
from this leak.
Tick() drains both queues:
1. _graphicsDevice.ProcessGLQueue() applies pending GL state.
2. Loop _meshManager.StagedMeshData.TryDequeue -> UploadMeshData
to materialize VAO/VBO/IBO for each prepared mesh.
Wired into GameWindow's render loop before draw work begins.
No-op when adapter is uninitialized or disposed.
Pattern matches WB's reference ObjectRenderManagerBase.ProcessUploads
without the prioritization heuristics (we're not yet drawing the
results — Task 22's WbDrawDispatcher will add prioritization when
visual budget matters).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
WbMeshAdapter now actually constructs the WB pipeline:
- OpenGLGraphicsDevice(gl, logger, DebugRenderSettings)
- DefaultDatReaderWriter(datDir) — opens its own file handles for now
(memory cost ~50-100MB of duplicate index caches, acceptable for
foundation work per plan Adjustment 1)
- ObjectMeshManager(graphicsDevice, dats, NullLogger)
InstancedMeshRenderer.EnsureUploaded routes through the adapter when
ACDREAM_USE_WB_FOUNDATION=1 is set; uses a WbManagedSentinel entry
in the local cache to mark "this GfxObj lives in WB now". CollectGroups
skips sentinel entries; both Draw passes skip them; Dispose skips them
(no GL resources to free — ObjectMeshManager owns those). Task 22's
WbDrawDispatcher will eventually draw WB-managed objects. With flag
off, behavior is byte-identical to before.
WbMeshAdapter constructor signature changed from (GL, DatCollection,
Logger) to (GL, string datDir, Logger). Updated tests to use
CreateUninitialized() for behavior tests and single null-GL guard test
for constructor validation. GameWindow updated to pass _datDir and to
wire _wbMeshAdapter into InstancedMeshRenderer.
AcDream.App.csproj gets direct ProjectReferences to WorldBuilder.Shared
and Chorizite.OpenGLSDLBackend — project refs are not transitive in
.NET, so AcDream.App must list them explicitly even though AcDream.Core
already references them.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Stub adapter that validates constructor args and exposes the public
shape (IncrementRefCount / DecrementRefCount / GetRenderData / Dispose).
Real ObjectMeshManager init is deferred to Task 9 — for now methods
no-op so call sites can wire the adapter without behavioral effect.
IWbMeshAdapter interface enables mocking in subsequent tasks
(LandblockSpawnAdapter tests in Task 11 need it).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
