Resolve exact SetSurface blend, alpha-test, and fog state once during extraction and preserve it through recipe-10 prepared payloads, Wb/EnvCell command data, Vulkan pipelines, push constants, and both ordinary/atmospheric one-pass shaders. Preserve AP-240 Wb pure-Clip immediate opaque/A2C while EnvCell uses retail premultiplied Clip; detail-off routing remains unchanged. Correct AP-232 and register the remaining detail-off state divergence.
Mutation first failures (all restored):
- raw Add->SRCALPHA/ONE: WalkStaticStreamPopulatorTests.ImmediateBuildingDetail_UsesExactResolvedSetSurfaceState line 1278, expected wb-mesh-raw-additive-1x.
- inverse-add->alpha-add: same test line 1278, expected wb-mesh-inverse-additive-1x.
- remove Env inverse: EnvCellAlphaDrawSourceTests.DetailOn_EveryEnvCellFamilyDrawsOnceInPlaceWithAuthoredOpacity line 132, expected envcell-inverse.
- raw IsAdditive precedence: same test line 132, Translucent|Clip|Additive expected envcell-alpha.
- Wb paletted ParamB=0: OrderPreservingSubmitterTests line 333, expected 0.392156869.
- Wb DDS ParamB=0.05: same test line 333, expected 0.784313738.
- disable Alpha+Clip test: WalkStaticStreamPopulatorTests line 1286, expected 0.784313738.
- always fog raw Add: same test line 1287, expected no-fog true.
- disable fog non-Add: same test line 1287, expected no-fog false.
- X=a*qA: RetailDetailTextureContractTests line 261, shared squared-alpha substring absent.
- X includes base alpha: same test line 262, forbidden baseTexel.a present.
- CLIP uses 0.05: EnvCellAlphaDrawSourceTests.ClipShaders_UseGreaterEqualForThePerRangeReference line 367.
- second detail draw: EnvCell detail-on line 131, collection contained 2 draws.
- straight-alpha substitute: Wb immediate line 1278, expected wb-mesh-additive-1x.
- omit ordered detail arm: OrderPreservingSubmitterTests line 318, expected (77,3.5), got (0,0).
- omit atmospheric combine: RetailDetailTextureContractTests line 260, shared include absent.
- drop serialized opacity: ObjectMeshDataSerializerTests line 292, expected opacity bits, got 1.0.
- stale detail arm: atmospheric adjacency line 402, expected slot 0, got 77.
- per-frame surface map: EnvCell warmed allocation line 285, expected 0 B, got 204800 B.
Replace the building and EnvCell detail replay with retail's exact single-pass stage result, including authored surface opacity, squared detail alpha, final-alpha clipping, and the original subset pipeline/order. Arm the ordered walk command in place to close#471, delete the replay pipelines/shaders, and advance prepared content to recipe 9.
Mutation witnesses (each restored before commit):
- X=a*qA: RetailDetailTextureContractTests.BothShaderFamiliesUseTheSharedOnePassSourceAndDebugPrecedesDetailSample line 174, missing materialAlpha * detail.a * detail.a.
- X*=base alpha: same test line 175, forbidden baseTexel.a found.
- CLIP against base alpha: EnvCellAlphaDrawSourceTests.ClipShaders_UseGreaterEqualForThePerRangeReference line 260, final-X conditional missing.
- second detail draw: EnvCellAlphaDrawSourceTests.DetailOn_EveryEnvCellFamilyDrawsOnceInPlaceWithAuthoredOpacity line 105, Assert.Single saw 2 MDI calls.
- straight-alpha substitution: WalkStaticStreamPopulatorTests.ImmediateBuildingDetail_RetainsOriginalFramebufferFamily line 1244, Additive first failed (only wb-mesh-alpha-1x recorded; InvAlpha also failed).
- omit ordered arm: OrderPreservingSubmitterTests.PrepareThenDraw_OrdinaryBuildingClipBuildingOrdinary_ArmsOnePassInPlace line 305, expected (77,3.5), got (0,0).
- omit atmospheric combine: RetailDetailTextureContractTests.BothShaderFamiliesUseTheSharedOnePassSourceAndDebugPrecedesDetailSample line 173, atmospheric shared include missing.
- drop serialized opacity: ObjectMeshDataSerializerTests.SurfaceOpacity_RoundTripsBitExactlyAndDeterministically line 288, first reported 0.5 bits 1056964608 vs 1065353216.
- stale detail arm: ordered adjacency test line 307, expected following ordinary (0,0), got (77,3.5).
- per-frame surface map: EnvCellAlphaDrawSourceTests.ProductionWholeLeaf_WarmedScanSubmitRhiAndFilteredReplayDoNotAllocate line 178, expected 0 B, got 147456 B.
Verification before commit: shader compiler 23/23; focused App 213/213; Content 75/75; Core Wb 10/10; launcher migration 6/6; Release solution build 0 warnings / 0 errors; git diff --check clean.
Exclude deferred EnvCell subsets from the opaque turn, preserve exact CLIP/ALPHA fixed-state groups through leaf replay, and use retail's row-3 override, blend, depth, and texture-class alpha references. Keep the existing building-detail sentinel distinct from the two CLIP references.
Bound rejected source payload, replace no-op allocation proofs with actual EnvCell and particle RHI paths, rebuild checked-in SPIR-V, and correct AP-238/AP-240 plus the S4-c2 evidence record.
Gates: Release 0W/0E; hermetic 16735/0/0; InstalledDat 255 pass/10 known fail/1 skip; shaders 32/32; focused 239/239; allocation 2/2 at 0 B. The evidence/comment repair re-ran Release 0W/0E, shaders 32/32, affected 37/37, and allocation 2/2; mesh_detail.vert.spv remained byte-identical at SHA-256 5346247ab7d606046943e19b28888c814e08dc6cb27cd9750096ac055457eb57.
Mutation proof, with each mutation restored after its named first failure:
1. Restoring the opaque predicate to !IsAdditive fails WholeLeaf_MixedCellDrawsOpaqueAtTurnThenClipAndAlphaAtDrain first at draw count: expected 1, actual 3.
2. Selecting _alphaPipeline for CLIP fails WholeLeaf_ClipDrainBindsExactStateAndTextureClassReference first at the bind sequence: expected [envcell-clip], actual [envcell-alpha].
3. Disabling CLIP depth write fails that production Theory first at Assert.True(clipPipeline.Depth.Write): expected true, actual false.
4. Swapping palette/DDS references fails the DDS row first: expected 0.784313738, actual 0.392156869; the palette row reports the inverse.
5. Mutating mesh_modern.frag from < to <= fails ClipShaders_UseGreaterEqualForThePerRangeReference first at Assert.Contains("if (color.a < alphaCutoff) discard;"): the required source spelling is absent.
6. Restoring row-3 OverrideClipmap=true fails the real-interface clip Theory first at Assert.False: expected false, actual true.
7. Deleting failed-append rollback fails the flush/end/abort rejection-storm rows first at the bounded pending-count assertion: expected 0, actual 9000.
8. Selecting _transparentDetailPipeline for a ClipMap detail contribution fails the leaf detail pin first at the bind sequence: expected second bind envcell-retail-detail-clip, actual envcell-retail-detail-alpha.
9. Resetting detail ParamB to zero fails the same detail pin first at the second pushed reference: expected 0.784313738, actual 0.
10. Classifying CLIP with exact mask equality excludes legal 0x09 and fails WholeLeaf_PositiveStippleClipMaskUsesClipPipelineAndDdsReference first at pipeline: expected envcell-clip, actual envcell-alpha.
11. Mapping the new blend to SRC_ALPHA/INVSRCALPHA fails AllRetailBlendModesAreRepresentable first at the tuple: expected (One, OneMinusSrcAlpha), actual (SrcAlpha, OneMinusSrcAlpha).
12. Restoring mesh_detail.vert's uParamB > 0.5 category predicate fails ClipShaders_UseGreaterEqualForThePerRangeReference first because vDetailCategory = uParamB == 1.0 is absent.
13. Treating every positive detail uParamB as a cutoff fails that source pin first because isRetailClipReference(uParamB) ? uParamB : 0.05 is absent.
14. Adding arbitrary 0.5 as an accepted reference to either mesh_modern.frag or mesh_detail.frag fails that source pin first at Assert.DoesNotContain("value - 0.5"); both mutations were run and reversed independently.
Retail: D3DPolyRender::SetSurface @ 0x0059c4d0; paired binary @ 0x0059c72a, 0x0059c747/0x0059c74f, 0x0059c821, 0x0059c838, 0x0059c866.
Round-1 three-lens FAIL at d60ca4ea0 found: the K2 fix was right in
production but nothing pinned the draw side (restoring the pre-fix gate
left every lane green); a stale "still ends an active scissor" comment
asserted a mechanism the same round had already deleted; the sky.vert
comment claimed retail clips the sky when it draws unclipped; the K6
pin ran through ClipFrameAssembler.Assemble, which has zero production
callers, so it proved nothing about the producer that actually runs.
This round closes L1-L9.
L1 (BLOCKING) — draw-side weather-gate pin. Added
RetailPViewPassExecutorTests.DrawLandscapeDynamicsPhase_GatesDrawWeatherOnceOnWalkDriverWeatherTurnFired:
reads DrawLandscapeDynamicsPhase's compiled call graph and asserts (a)
the call immediately before DrawWeatherOnce is
WalkFrameDriver.get_WeatherTurnFired and (b) exactly one
brfalse/brfalse.s branch sits between that call and the draw, jumping
forward past it. Three mutations, each shown failing:
M1 (restore `if (clipAssembly.OutsideViewSlices.Length != 0)`):
Assert.Equal() Failure: Values differ
Expected: typeof(AcDream.App.Rendering.Walk.WalkFrameDriver)
Actual: typeof(AcDream.App.Rendering.ClipFrameAssembly)
M2 (drop the gate, unconditional call):
Assert.Equal() Failure: Values differ
Expected: typeof(AcDream.App.Rendering.Walk.WalkFrameDriver)
Actual: typeof(AcDream.App.Rendering.RetailPViewPassExecutor)
M3 (invert to `if (!walkDriver.WeatherTurnFired)`):
Assert.Single() Failure: The collection did not contain any matching items
Expected: (predicate expression)
Collection: [CompiledBranch { Offset = 7, OpCode = brfalse.s, TargetOffset = 17 }, CompiledBranch { Offset = 24, OpCode = brtrue.s, TargetOffset = 33 }]
Corrected WalkFrameDriverTranscriptTests.cs's doc comment (the
Collect_WeatherTurnFiredMatchesThePrintedOcLineExactly block): it pins
the print half and the flag only, and now names the real draw-side pin
instead of falsely claiming to double as one.
L2 (MAJOR) — scissor stack deleted, K4 comment corrected. Deleted
IWorldPassSurface.BeginScissor/EndScissor and their RhiWorldPassSurface
bodies, the scissor call inside ClearInteriorDepth, NdcScissorRect.cs +
NdcScissorRectTests.cs (BeginScissor had zero remaining callers, and
RhiWorldPassSurface.BeginScissor was NdcScissorRect.ToPixels' only
production caller). This left RhiWorldPassSurface's 4th constructor
parameter (IRetailPViewFramebufferSource) and RetailPViewPassExecutor.cs's
RetailPViewFramebufferSize/IRetailPViewFramebufferSource/
SilkRetailPViewFramebufferSource types entirely dead (framebuffer size
was needed only for the scissor's NDC-to-pixel conversion) — removed
them and their one call site in FrameRootComposition.cs. Rewrote
RetailPViewPassExecutor.cs's K4 comment: it no longer asserts a live
scissor mechanism; it states that VulkanGpuPassEncoder's constructor
sets the full-attachment scissor once, at pass begin
(VulkanGpuPassEncoder.cs:87), and nothing narrows it after that. KEPT
per the round-1 verdict: ScissorFallbacks, OutsidePlaneCount,
OutsideViewNdcAabb, HasOutsideView, OutdoorVisible, OutdoorSlot,
ClipViewSlice.NdcAabb (the >8-edge zero-plane slice case the punch fans
still consume) and the VulkanViewportMapping.ScissorToVulkan pass-begin
path (IGpuPassEncoder.SetScissor stays). Also deleted
ClipFrameAssembly.TerrainMode/TerrainScissorNdcAabb and their writers
in ClipFrameAssembler.cs (both Assemble and ReassembleOutsideViewFromWalk);
OutsidePlaneCount's formula reduces to `outsideHasScissorFallback ? 0 :
outsideMaxPlaneCount` without needing the deleted TerrainClipMode
comparison (a scissor-fallback slice can only exist when outdoorVisible
is already true, so the three original cases collapse identically).
Deleted the two "terrain=" / "outMode=" diagnostic reads in
WorldRenderDiagnostics.cs (:345, :409) and the now-vestigial
EmitClipRouteProbe "ubo: n=..." segment that read ClipFrame's deleted
terrain bytes.
DEVIATION from L2's literal text: kept the TerrainClipMode enum type
itself — WorldSceneRenderer.cs's flat-world safety path (explicitly
"out of this chunk's scope" per K4's own round-1 comment) still uses it
for an unrelated "did the flat terrain draw" diagnostic flag, with its
own WorldSceneDiagnosticsController/WorldRenderDiagnostics/test
consumers. Deleting the type would require rewriting files outside
every round's declared file list. Reworded its doc comment to state
this plainly.
L3 (MAJOR, completes K3) — dead TerrainClip UBO deleted from the sky
and terrain shaders. Deleted the block, both gl_ClipDistance loops, and
the gl_PerVertex redeclaration (nothing else needs it) from sky.vert,
terrain_modern.vert, terrain_atmospheric.vert. Recompiled via
tools/compile-shaders.ps1 (glslc via the Vulkan SDK, managed shaderc
fallback also runs) — 24/24 pairs compiled; only sky.vert.spv,
terrain_atmospheric.vert.spv and terrain_modern.vert.spv changed.
Re-pinned VulkanShaderManifestTests.cs's frozen retail-oracle hashes
for sky.vert.spv (7d67a9e3624d198b370d402b5c12e4ce925bf9b8e646ef5123636a86d5985ab5)
and terrain_modern.vert.spv (8a73d89ef0e51e550327b9ff8c24857e309103b1d491030cf0d4d8594b45068c)
with dated comments, matching the existing re-pin convention.
Deleted WorldFrameSectionBinding.BindTerrainClip and
WorldFrameSections.TerrainClip (+ its Reset) with its two callers
(SkyRenderer.Rhi.cs:258, TerrainModernRenderer.Rhi.cs:259). DEVIATION
from L3's literal text: kept the shared Zeroed(...) helper in
WorldPassScope.cs — it is also called by BindSceneLighting and
BindClipRegions, both of which stay; deleting it would have broken
those two live bindings. Deleted IWorldPassSurface.BindTerrainClip
(no-op interface method + RhiWorldPassSurface body) and
WorldScenePassExecutor.cs's two calls to it (:121, :235).
Deleted ClipFrame's _terrainBytes field, TerrainBytes/TerrainBytesForTest
properties, Reset's Array.Clear(_terrainBytes), and the stale
header/K3 comment paragraphs. DEVIATION from L3's literal text: kept
ClipFrame.TerrainUboBytes and ClipFrame.TerrainClipUboBinding —
PortalDepthMaskRenderer.Rhi.cs (:140, :187) is a live production
consumer of both constants for the KEPT exit-seal/punch-fan clip block,
which was never in scope for deletion (portal_depth.vert's own
TerrainClip UBO declaration is the KEEP block, untouched). Reworded
both constants' doc comments (ClipFrame.cs, VulkanPipelineLayouts.cs's
UniformTerrainClip) to say only the portal-depth clip block uses
binding 2 now, and corrected the two GpuBindingModel.cs comments that
explained why terrain-tiling/sky-params bindings are 3/4 by naming "the
terrain clip block". Updated VulkanShaderDescriptorContractTests.cs's
TerrainVertexShaderDeclaresItsClipBlockInTheUniformSet (renamed
TerrainVertexShaderDeclaresOnlySceneLightingInTheUniformSet):
terrain_modern.vert's uniform set is now {SceneLighting} only. Deleted
ClipFrameLayoutTests.NoClip_TerrainBytes_Count0_AllZeros and its K3
comment; corrected the class doc comment and LayoutConstants_MatchShaderStruct's
"terrain UBO"/binding-contract comments to describe the portal-depth
consumer instead.
REQUIRED L3 fact (declared-but-never-bound dynamic uniform binding):
VulkanFrameBindings's constructor seeds EVERY declared uniform binding
(0..UniformBindingCount-1, including binding 2) with the shared dummy
buffer's range before any renderer runs (VulkanFrameBindings.cs:128-130,
`_arena.SeedUniform(binding, dummy.Handle.Handle, dummyUniformRange)`
in a loop over every binding) — "Every binding is always bound, whether
a renderer uses it or not... unused ones point at a shared dummy range"
(VulkanFrameBindings.cs:27-31). So after this round, when the sky or
terrain pipeline draws, set 1's shared descriptor layout still declares
binding 2 (portal_depth.vert's own declaration keeps
IsDeclaredUniformBinding(2) true), and its descriptor still points at a
valid (dummy) range from that seed — vkCmdBindDescriptorSets stays
legal even though neither shader statically uses binding 2 any more.
No SPIR-V-side change was needed to keep this legal.
L4 (MAJOR) — one outside-view slice constructor. Extracted
ClipFrameAssembler.AppendOutsideSlice (frame, ViewPolygon,
outsideSlicesList, ref maxPlaneCount, ref hasScissorFallback, ref
scissorFallbacks) — the exact ClipPlaneSet.From/AppendSlot/
ClipViewSlice-construction body — and call it from both Assemble's
outside_view loop and ReassembleOutsideViewFromWalk's outside_view
loop. Rewrote ClipFrameLayoutTests's K6 pin
(ClipViewSlicePlanes_PunchFanPath_EqualsCpuViewPolygonEdgePlanes_ForASyntheticView):
builds a synthetic WalkPortalView from pixel-space points (the
WalkCopyViewTests pattern) via a trivial IWalkRayCaster, runs
BeginWalkFrame(frame, outdoorRoot: false) then
ReassembleOutsideViewFromWalk(assembly, walkView, 640, 480), and reads
assembly.OutsideViewSlices[0].Planes — the exact production pair
RetailPViewPassExecutor.cs's BeginWalkFrame call and
RetailPViewRenderer.cs's ReassembleOutsideViewFromWalk call make.
Mutation (perturbed planes[0].W inside the shared helper, right after
`planes = cps.PlaneArray;`):
edge 0 midpoint (-0.3,0.10000001) should lie ~on its OWN GPU plane; the closest plane was only 0.5251073 away
The pin's own source contains no `Assemble(` call — grep-checked by
extracting the method body and searching it for the literal text; no
match.
L5 — shader comments corrected. sky.vert/terrain_modern.vert/
terrain_atmospheric.vert now state retail draws the sky once
(LScape::draw @0x00506330 -> GameSky::Draw(sky,0) @0x0050633c, before
draw_check_blocks) and the landscape is view-culled per cell, never
GPU-clipped (RenderDeviceD3D::DrawBlock @0x005a17c0). The "Phase W
Stage 4"/"U.3 default" clip narratives are deleted along with the
blocks they described.
L6 — grep sweep. `grep -rn "<token>" src tests docs/architecture`
returns nothing for SetTerrainClip, BeginDoorwayScissor,
BindTerrainClip, TerrainBytes, "active scissor", and NdcScissorRect —
confirmed after this commit (docs/plans and docs/research keep the
historical record, untouched). TerrainClipMode is the one deliberate
exception (see the L2 deviation note above); every one of its 9
remaining hits is a live, non-stale reference (the enum declaration,
WorldSceneRenderer's flat-path local, or their diagnostic/test
plumbing), not a stale mention of a deleted mechanism.
L7 — Issue130DoorwayStripTests. Deleted AnySliceAdmitsScissor,
worstScissorGapPx and its PIN 1 assertion, the header's scissor
sentences, and the scissorGap half of MeasureTopEdgeGap (dropped the
fbW parameter it alone needed). Rewrote the header: the scissor
mechanism is retired (the sky and landscape draw unclipped; aperture
exactness comes from the depth clear, the exit seals and the interior
repaint). The remaining plane-gap half (the canary PIN,
`worstPlaneGapPx <= 1.2f`) still pins something production reads:
AnySliceAdmitsPlanes walks slice.Planes from the SAME ProjectToClip ->
ClipToRegion -> ClipPlaneSet.From pipeline
RetailPViewPassExecutor.DrawWalkPunchFan reads through
clipAssembly.OutsideViewSlices[activeViewIndex].Planes — so the test
was kept, not deleted.
L8 — DrawWalkSky loop-shape pin. Added
DrawWalkSky_RenderSkyCallSiteHasNoEnclosingBackwardBranch (same
backward-branch-span shape as K1's DrawWeatherOnce pin), plus a note in
both pins that CompiledCallGraph.ReadBranches does not decode a
compiled switch jump table, but no C# loop construct compiles to one.
Mutation (wrapped the RenderSky call in `for (int i = 0; i < 2; i++)`):
Assert.DoesNotContain() Failure: Filter matched in collection
Collection: [..., CompiledBranch { Offset = 20, OpCode = brtrue.s, TargetOffset = 25 }, CompiledBranch { Offset = 23, OpCode = br.s, TargetOffset = 66 }, CompiledBranch { Offset = 72, OpCode = blt.s, TargetOffset = 13 }, CompiledBranch { Offset = 80, OpCode = brfalse.s, TargetOffset = 116 }, CompiledBranch { Offset = 88, OpCode = brfalse.s, TargetOffset = 116 }]
(the Offset=72 -> TargetOffset=13 entry is the injected loop's backward
branch spanning the call).
L9 — this message.
Gates: dotnet build 0 warnings / 0 errors. Hermetic filter
(Lane!=InstalledDat&...&Status!=KnownFailure): 6827 passed, 0 failed.
InstalledDat lane against a real DAT directory: 244 passed, 4 failed —
exactly the four known failures (LayoutImporterMediaBearingChildSweepTests
+ LayoutImporterInvisibleSweepTests, both #383; TowerAscentReplayTests,
KnownFailure; WalkTraceConformanceTests.Oh_doorway_still_first_frame_diff,
#458 KnownFailure). No register row added (every deletion here removes
an acdream-only rule; the KEEP items already had their equivalence
pins from round 1). No new flag/probe.
OWED: the lead's ACDREAM_DEVTOOLS=1 four-pose visual self-gate against
the running client is not run by this agent — CLAUDE.md and this
task's own instructions forbid launching the graphical client from
here; that visual/validation-layer pass remains the user's to run
before this round is accepted.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
VM2's live cdb read against the PDB-paired retail client (GUID
9e847e2f-777c-4bd9-886c-22256bb87f32) proved
m_caps.bCanDoSinglePassDetailing = 1 and trysinglepass = 1 on real hardware,
so D3DPolyRender::RenderMeshSubset (0x0059ca10) never falls back to the
two-pass framebuffer blend the earlier #226 port reproduced. Every loaded
CGfxObj sets use_built_mesh = 1 (CGfxObj::InitLoad 0x005346b0), so buildings
and EnvCells always take the single-pass texture-stage combine set up in
D3DPolyRender::SetSurface (0x0059c4d0):
result = lerp(base * diffuse, detail.rgb, detail.a * diffuse.a)
RenderMeshSubset lights opaque built-mesh subsets with
tmpmaterial.Diffuse.a = 1, so on the live Dereth category texture
0x06006D58 (mean rgb 0.165, mean alpha 0.132) the combine works out to
~0.868 * base + 0.022 — a mild darkening, the opposite sign of the fallback
DstColor blend's brightening.
Also removes the invented 10 m / 50 m distance fade. Retail's
ACRender::get_alpha_for_z (0x006b6230) is only evaluated in
D3DPolyRender::DrawPolyInternal (0x0059d7c0, the immediate-polygon path)
and only when the static noFadeDetail (0x00820e38, initialised to 1) is 0 —
unreachable for built meshes. Attenuation is the sampler's linear mip chain
converging to the texture mean, not a scripted ramp.
Changes:
- mesh_detail.vert/.frag: drop vDetailFade and its distance term; add
vDetailOpacity mirroring mesh_modern.vert's InstanceAlphaBuf (binding 7)
read, and output detail.rgb with alpha = detail.a * vDetailOpacity under
the corrected pipeline blend.
- VulkanViewportMapping.BlendFactorsOf / GpuEnums.GpuBlendMode.RetailDetail:
SrcAlpha + OneMinusSrcAlpha instead of DstColor + OneMinusSrcAlpha.
- RetailDetailTextureContract: replaced the distance-fade constants and
FramebufferFactor with Expected(base, detail, opacity) and IsNeutral,
matching the lerp; contract tests cover zero-alpha/zero-opacity no-ops,
the measured darkening on the live category texture, and full-alpha
replacement.
- Regenerated mesh_detail's committed SPIR-V and the shader manifest
(tools/compile-shaders.ps1); no other shader pair changed.
- Docs: #226's pseudocode note, the docs/ISSUES.md #226 entry, and the
retired TS-52 divergence-register row corrected from the two-pass
DESTCOLOR description to the single-pass path and the darkening
expectation, each citing the VM2 cdb note.
Verified: dotnet build AcDream.slnx -c Release (0 warnings, 0 errors);
dotnet test on AcDream.App.Tests and AcDream.Core.Tests (Release, hermetic
lanes) both green.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
VulkanViewportMapping has inverted the front face since V6c, on the standard
argument that rendering with a negative viewport height mirrors framebuffer
space and therefore reverses triangle orientation. The world arm is the first
consumer that culls anything, and it falsified the inversion twice over on one
frame.
Nothing exercised it before now. Every Vulkan consumer through V6i - TextRenderer,
DebugLineRenderer and the bring-up scene - declares Cull = GpuCullMode.None, so
the mapping had never decided a single fragment. That is why a wrong answer
survived four slices and a validation-clean run: an unexercised path.
What the world arm measured, on the same offline scene the GL pixel gate captures.
Terrain is the one single-sided surface acdream draws - FrontFace(Ccw) plus
Cull(Back), matching ACRender::landPolysDraw's per-triangle eye-side predicate -
and under the inversion it vanished completely, 190 multi-draw commands issuing
against 625 loaded landblocks with nothing on screen. Every closed building shell
rendered inside-out in the same frame: the front wall culled and the interior
beams visible through the gap, which is what a back-face-front cull looks like on
geometry that is only nearly convex. Declaring the GL winding verbatim restores
both at once - terrain draws single-sided from above, and the shells close.
Two independent surfaces, one change, and the correction is the identity mapping.
Recorded here rather than worked around in the renderers, because a renderer that
compensates for its backend is exactly the shape this file exists to prevent: the
contract says renderers speak GL and the backend translates, and the backend was
translating wrongly.
The viewport flip itself is untouched and still correct - it is what puts
GL-authored geometry the right way up with no shader or matrix change. What goes
is the claim that a winding inversion has to travel with it. The scissor's
explicit flip is a separate correction with a separate justification and is
likewise untouched.
The test suite says so now rather than describing the old behaviour: the
pass-through is asserted directly, and the exact-inverses test becomes a
travels-alone test, so a later change that reintroduces the inversion fails here
first and on any single-sided surface second.
Gates. Release build green. App tests 4,112 passed / 3 skipped, the unchanged
baseline. GL offline pixel gate unaffected by construction - this file has no GL
arm - and measured with the world arm in commit 2.
No divergence-register row: this corrects a backend translation error rather than
introducing a deviation from retail.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The last of V6's three commits, and the one that makes the backend render.
Plan sections: 4.5 (pipelines and the persisted cache), 4.6 (shaders and the
committed .spv), 4.7 and 3.3 (clip space, the Y flip and winding), 4.9 and 4.10
(swapchain format and the scissor convention), 4.11 (the probe shader V5
deferred), 5.4 (Target: null means the swapchain image, literally).
WHAT RUNS. ACDREAM_RENDER_BACKEND=vulkan now renders a real scene through the
whole RHI on the RX 9070 XT: 60,000-plus frames per twelve-second run, 4x MSAA
resolving into a B8G8R8A8_UNORM swapchain, GPU timer scopes resolving, a
screenshot taken through IGpuDevice.CaptureBackbuffer, and a clean
CloseMainWindow exit with the allocator reporting three device-memory objects.
WHAT IT DRAWS, AND WHY IT IS NOT THE GAME. V6's milestone is "a full game frame
on Vulkan" and on this branch that cannot be the game's own frame. V4c and V4d
are parked by 5.5.5 so the world renderers are still raw GL; and the two
renderers that DO speak the RHI - TextRenderer and DebugLineRenderer, ported at
V4a - both throw for any device that is not a GlGpuDevice, because their loose
uniforms and their classic texture-unit sprite binding have no home in the
pinned contract yet. Converting them is a V4-class change with its own GL pixel
gate, outside this slice's file list.
So the backend is exercised through the contract by a scene of our own, and it
is not a toy. It uses a device-local mesh arena filled through the staging ring,
instance and batch data written straight into mapped ring memory, an offscreen
render target whose colour is registered into the global texture table and
sampled by a later pass, a BC1 texture with a CPU-built mip chain beside an
uncompressed one with a vkCmdBlitImage chain, one multi-draw-indirect covering
five quads with gl_DrawID selecting per-draw batch data, a second pipeline with
line-list topology bound mid-pass, dynamic cull/front-face/depth-write, push
constants, timer scopes, and an MSAA colour attachment resolving into the
swapchain image.
ORIENTATION, BY INSPECTION. Slice V5's screenshot was a uniform clear and its
orientation was right "by construction" - which a uniform clear cannot show. The
scene is therefore deliberately asymmetric in both axes: a quadrant card that is
red top-left, green top-right, blue bottom-left and white bottom-right, four
differently tinted markers at four different corners, and an open L of lines
whose short stub rises at its right end. The captured PNG reads correctly in
every one of those, including a miniature of the same card in the bottom-right
whose own quadrants are also the right way up. The negative viewport height, the
front-face inversion and the capture path agree.
THE SHADER TOOLCHAIN, AND WHAT IT FOUND. tools/compile-shaders.ps1 drives
tools/ShaderCompiler, a small out-of-solution .NET tool over Silk.NET.Shaderc -
the same shaderc glslc is built on, through the already-pinned Silk 2.23.0
family. glslc is preferred when a Vulkan SDK is present and reported when it is;
neither this machine nor CI has one, and requiring a 500 MB manual install
between a contributor and a working checkout is not a reasonable price for a
build step. The GLSL sources stay the single source of truth: the Vulkan dialect
arrives as a preamble injected after the #version line - ACDREAM_UBO_SET becomes
"set = 1,", the texture table becomes a set-2 descriptor array with a required
nonuniformEXT accessor, and the shared 96-byte push block is declared with each
loose uniform name defined onto its member. The only edits to a shader BODY are
mechanical and dialect-level: dropping default-block uniform declarations, which
Vulkan GLSL has no such thing as, and assigning explicit varying locations BY
NAME across a pair, because ordinal assignment would look identical today and
silently swap varyings the first time an author reordered a line.
Run over the eight production pairs, exactly one thing happened: none of them
compiled, and every failure is a specific source-level fact belonging to a
renderer-port slice that has not landed. debug_line needs uView/uProjection
converged into one uViewProjection - two matrices are 128 bytes and the shared
block is 96. mesh_modern and particle still pass a uvec2 bindless handle as a
varying, which is V4t's GpuTextureSlot retype. sky has ten loose uniforms and
wants a UBO. ui_text needs uScreenSize/uUseTexture/uTex. particle_mesh needs
uTextureIndex to become uTextureIndexA. terrain_modern needs V4d-1's matrix
convergence. mesh is the legacy pair with no RHI consumer at all. That inventory
is committed as shaders.manifest.json, with each source's SHA-256 and the
compiler's own message, and a test re-hashes it so an edited shader that never
got recompiled fails a build rather than shipping a stale binary.
vk_probe is the pair that does compile, and it is the shader 4.11 already asked
for: V5 recorded "build one real pipeline from the committed .spv" as its single
deliberate deviation because no toolchain existed. It is Vulkan-dialect only and
no GL renderer draws with it, so it forks nothing; it retires when the ported
world renderers become the backend's own proof.
DESCRIPTORS. Sets 0 and 1 are DYNAMIC buffer descriptors bound per flight slot,
so a per-draw range change costs a dynamic offset in vkCmdBindDescriptorSets
rather than a vkUpdateDescriptorSets in the hot path - which is what keeps 4.4's
zero-writes-per-frame property true for buffers as well as for textures. Ten
dynamic storage descriptors is above Vulkan's guaranteed minimum of four, so it
is a real requirement rather than a free choice, it fails loudly at layout
creation on a device that cannot serve it, and V9's lavapipe row must confirm
it. Unused bindings point at a shared dummy range so there is ONE set layout and
one pipeline layout; that is why binding a second pipeline mid-pass costs
nothing and disturbs neither the descriptors nor the push constants.
THE ONE MAPPING FUNCTION. VulkanViewportMapping holds the whole coordinate
reconciliation: negative viewport height, the front-face inversion that pairs
with it, and - separately - the scissor flip, which the viewport sign does NOT
perform. The V3 audit flagged that as a concrete V6 acceptance item and it is
the subtle one: vkCmdSetScissor is always top-left-origin, NdcScissorRect emits
GL bottom-left rectangles, and getting it wrong clips a doorway aperture from
the wrong edge in a scene that has one. Clip space needs nothing, as 4.7
concluded: the cameras already build [0,1]-convention projections.
CONTRACT GAP, RECORDED NOT PAPERED OVER. GpuPipelineDescription cannot name its
colour-attachment format, and Vulkan bakes that into a pipeline. Offscreen
targets therefore adopt the swapchain's B8G8R8A8_UNORM rather than a literal
RGBA order - invisible above the API, because an image is sampled through its
format's component mapping and the one CPU readback swizzles explicitly. The
honest fix is a colour-format field added in a reviewed contract commit, exactly
as GpuBlendMode.InverseAlpha and GpuVertexFormat.UByte4UInt were added when V4c
and V4d met the same wall. It is documented at
VulkanTextureFormatMapping.CanonicalColorAttachmentFormat.
The pipeline cache is persisted to the cache directory and validated by its
32-byte header against this device's vendor, device and cache UUID before use.
Drivers are required to ignore incompatible blobs, but "required to" is a poor
foundation for something that runs before anything else in the process, and the
check costs 32 bytes of comparison. Two consecutive launches report "cold" then
"reused".
Gates: Release build clean; App suite 4056 passed / 3 skipped (4037 at V6b plus
19 new); offline pixel gate PASS at a differing fraction of 5.15e-05 with a
same-commit control immediately after it at 2.84e-05 - 29 and 16 pixels of
563,200, the same class of ambient variation the campaign's 15-23 band records,
and roughly 19x under the 0.001 threshold on a commit that changes no GL code
path.
Validation layers could not be run: this machine has no Vulkan SDK, no
HKLM\SOFTWARE\Khronos\Vulkan\ExplicitLayers key, no VK_LAYER_PATH and no
VkLayer_khronos_validation.json anywhere on disk. Plan 7 already requires one
validation-clean run at V7; it needs the SDK installed first and is reported
rather than assumed here.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Erik