Iteration on the sky rendering pipeline to restore stars/moon visibility
at night and fix washed-out grey daytime clouds. Key fixes:
* sky.frag: disable fog-mix on sky meshes. Retail's keyframe FogEnd
(0..400m at midnight, up to 2400m during day) is calibrated for
terrain; sky meshes are authored at radii 1050-14271m which sits
past FogEnd universally, causing every sky pixel to saturate to
fogColor (dark navy). Stars, moon, dome texture all got
obliterated. The horizon-glow trade-off is noted in the shader
comment; research item to find retail's sky-specific fog range
later.
* SkyRenderer + sky.frag: promote rep.Luminosity into uEmissive so the
vertex lighting saturates properly for bright keyframes. Retail's
FUN_0059da60 non-luminous path writes rep.Luminosity into
material.Emissive via the cache +0x3c slot; we were instead using
it as a post-fragment multiply which could only dim, never brighten.
Net effect: daytime clouds now render saturated white, dome dims
correctly at night (rep.Luminosity=0.11 → Emissive=0.11), stars
and moon unchanged.
* terrain.vert: MIN_FACTOR 0.08 -> 0.0 per retail FUN_00532440 decompile
(DAT_00796344 ambient-floor = 0.0). Back-lit terrain now falls to
pure ambient rather than getting an 8% sun floor.
New research / tooling (no runtime impact):
* docs/research/2026-04-24-lambert-brightness-split.md — retail's
ambient-brightness formula pinned from PE .rdata read + live
RetailTimeProbe capture: effAmbBright = AmbBright + |sunDir| * 0.2
where scale constant 0x0079a1e8 = 0.2f exactly.
* docs/research/2026-04-23-lightning-real.md — research note on the
dat-baked PhysicsScript-driven lightning path (Rainy DayGroup has
explicit PES-triggered flash SkyObjects with 5ms time windows).
* Corrections stapled to sky-decompile-hunt-{B,C}.md: DAT_00842778 is
DirColor, DAT_0084277c is AmbColor (the hunt docs had the swap
backwards).
* tools/RetailTimeProbe/Program.cs: extended with pid=NNNN selector,
sky global probe (DirColor/AmbColor/AmbBright/sunDir/cache.amb),
and the 0x0079a1e8 scale-factor readout.
* tools/SkyObjectInspect/: throwaway dat-inspector built by the Opus
deep-dive agent. Identified GfxObj 0x010015EF as the stars layer
(A8R8G8B8 128x128 texture, 4% bright-pixel ratio).
* src/AcDream.App/Rendering/TextureCache.cs: per-texture alpha
histogram dump under ACDREAM_DUMP_SKY=1 for diagnosing "are the
clouds decoded with proper alpha" type questions.
README: rewrite to reflect current state (playable pre-alpha rendering
Dereth with animated characters, day-night cycle, weather, etc.)
instead of the stale "Phase 0 dat inventory only" description.
All 742 tests green.
19 KiB
Sky Decompile Hunt B — D3D Render-State Signature Trace
Date: 2026-04-23 Hunter: Hunt Agent B (render-state signatures) Status: SIGNIFICANT FINDINGS — but NOT a "celestial-body iteration draw loop"
⚠ 2026-04-24 correction: Any occurrences in this doc that call
DAT_00842778the "ambient" colour are backwards.DAT_00842778= DirColor (directional / sun),DAT_0084277c= AmbColor,DAT_00842780= AmbBright. Cross-verified againstSkyTimeOfDay.UnpackandFUN_00501600's output mapping. Full re-analysis:docs/research/2026-04-24-lambert-brightness-split.md.
TL;DR
The retail acclient does NOT appear to have a classical "sky dome + iterate celestial meshes with per-mesh blend swaps" render function. Instead, sky is implemented as:
- A per-frame keyframe sampler (
FUN_005062e0) that interpolates the current RegionDesc keyframe (sun angle, sun color, ambient color, fog color/near/far) and stashes results into globals. - Per-mesh ambient color pushed via
_DAT_008682bc/c0/c4(RGB floats, no D3DRS_AMBIENT call). - D3D fog state (D3DRS_FOGCOLOR/FOGSTART/FOGEND) set via
FUN_005a41b0— this is the ONLY sky-keyframe → D3D state write path. - A "pre-world" pass (
FUN_00507a50(0)) that disables Zwrite and iterates weather volume objects (fog/rain/snow, NOT sun/moon/stars).
No "celestial body draw loop" found. No D3DRS_AMBIENT call found. No matrix-translation-zero (camera-anchor) found. No lightning RNG found. No huge far-plane constant found at the sky call site. This strongly suggests retail AC is far simpler than modern sky demos assume — the sun/moon/stars are either baked into the RegionDesc scene geometry (treated as regular objects that happen to be positioned far away) or do not exist as discrete draw calls at all, with "sun color" being expressed purely via per-vertex lighting on world meshes.
D3D Vtable Map
All state writes go through pIDirect3DDevice at (param_1 + 0x468),
vtable offset 0xe4 = SetRenderState, offset 0xd4 = SetTextureStageState,
offset 0xc4 = SetMaterial, offset 0x1c (of a PARENT wrapper object at
DAT_0086734c) = SetPerspective/SetFrustum.
Discovered D3D state wrappers in chunk_005A0000.c:
| Wrapper | Addr | D3DRS | Meaning | Cache offset |
|---|---|---|---|---|
| FUN_005a3ba0 | 0x005A3BA0 | 0x1b=27 | ALPHABLENDENABLE | +0x475 |
| FUN_005a3be0 | 0x005A3BE0 | 0x0f=15 | ALPHATESTENABLE | +0x476 |
| FUN_005a3c20 | 0x005A3C20 | 0x19=25 | ALPHAFUNC | +0x478 |
| FUN_005a3c60 | 0x005A3C60 | 0x18=24 | ALPHAREF | +0x47c |
| FUN_005a3ca0 | 0x005A3CA0 | 0x13=19 src, 0x14=20 dst, 0xab=171 op | blend state | +0x4f0/+0x49c |
| FUN_005a3d80 | 0x005A3D80 | 0x16=22 | CULLMODE | +0x488 |
| FUN_005a3dc0 | 0x005A3DC0 | 0xaf=175 | (fog-gated) | +0x48c |
| FUN_005a3e00 | 0x005A3E00 | 7 | ZENABLE | none (stateless) |
| FUN_005a3e20 | 0x005A3E20 | 0x17=23 (ZFUNC) + 0xe=14 (ZWRITE) | combined ZFunc+ZWrite | +0x494/+0x498 |
| FUN_005a3eb0 | 0x005A3EB0 | 0x8b=139 | AMBIENT | +0x4a0 |
| FUN_005a3ef0 | 0x005A3EF0 | 0x91=145 | COLORVERTEX | +0x4a8 |
| FUN_005a3f40 | 0x005A3F40 | 0x93=147 | ? | +0x4a4 |
| FUN_005a3f90 | 0x005A3F90 | 0x1c=28 | FOGENABLE | +0x4bc |
| FUN_005a4080 | 0x005A4080 | 0x22=34 (FOGCOLOR), 0x24=36 (FOGSTART), 0x25=37 (FOGEND) | fog triple | +0x4ac/+0x4c0/+0x4c4 |
| FUN_005a41b0 | 0x005A41B0 | wrapper → FUN_005a4080 | SetFog(color, start, end) | — |
| FUN_005a41f0 | 0x005A41F0 | 0x89=137 | LIGHTING | +0x4c8 |
| FUN_005a4280 | 0x005A4280 | 0x92=146 | LOCALVIEWER | +0x4d8 |
| FUN_005a4350 | 0x005A4350 | 0x3c=60 | ? | +0x4e0 |
| FUN_005a4390 | 0x005A4390 | 0x08 | SHADEMODE | +0x4e4 |
| FUN_005a4420 | 0x005A4420 | stages | SetTextureStageState blend args | — |
| FUN_005a4010 | 0x005A4010 | internal flag at +0x4be → re-calls FUN_005a3f90 | fog-master-enable gate | +0x4be |
Device-reset-to-default function: FUN_005a10f0 at 0x005A10F0, lines
687-740 of chunk_005A0000.c. This is NOT the sky render; it is called
once on device init and full-state rebuild.
AMBIENT (D3DRS=0x8b=139) — The Key Negative Finding
I searched every call to vtable 0xe4 with state 0x8b across the
entire 688K-line decompile. Results:
- chunk_005A0000.c:3388 —
...0x8b,0— inside default-init reset - chunk_005A0000.c:2772 —
...0x8b,param_2— inside wrapper FUN_005a3eb0
External callers of FUN_005a3eb0:
chunk_005A0000.c:704: FUN_005a3eb0(0); // default-reset, ambient=black
ONLY ONE CALLER. The sky keyframe sampler (FUN_005062e0) does NOT
call FUN_005a3eb0. D3DRS_AMBIENT is set to 0 once and never changed.
This falsifies the previous research hypothesis (2026-04-22 doc) that
clouds are tinted by a per-keyframe D3DRS_AMBIENT write.
Sky color instead lives in _DAT_008682bc/c0/c4 (three floats, RGB, set
from DAT_00842950/54/58). See below.
Sky Keyframe Sampler — FUN_005062e0 @ 0x005062E0
chunk_00500000.c, lines 6213-6333. Called each frame from the scene
renderer (FUN_00455a50 path, around chunk_00450000.c:4341, 4468).
Stripped pseudocode:
void FUN_005062e0(int worldRenderer) {
if (worldRenderer.layerBits != 0 && worldRenderer.skyEnabled != 0)
{
FUN_00508010(); // advance sky timer
double now = _DAT_008379a8; // game-time seconds
if (_DAT_00842798 <= now && currentRegionDesc != 0) {
_DAT_00842798 = regionDesc[+0x50].nextKeyTime + now; // reschedule
float turbidity = (weatherSys ? weatherSys[+0x48] : _DAT_00795610);
if (_DAT_008427a0 < now) {
// ---- BLOCK 1: sample ambient/sun ----
int ok = FUN_004ff440(turbidity,
&fogDensity, // out float
&sunColor, // out ARGB
sunDirVec, // out float[3]
&ambientColor); // out ARGB
if (ok) {
if (fogDensity < DAT_0084295c) fogDensity = DAT_0084295c;
// ---- WEATHER FOG BLEND (if weather active) ----
if (DAT_008427a9 /* weatherFogEnabled */) {
if (1.0f <= _DAT_008427b8 /* blendProgress */) {
// blend complete; snap to weather values
fogDensity = DAT_008427ac; // weather density
sunColor = DAT_00842788; // weather fog color
} else {
// lerp currentColor → weatherColor per channel (bytes),
// lerp density toward DAT_008427ac
// advance _DAT_008427b8 by _DAT_007c7208 (blend rate)
}
}
FUN_00505f30(fogDensity, sunColor, sunDirVec, ambientColor);
}
_DAT_008427a0 = now + (currentRegionDesc ?
regionDesc[+0x50].keyInterval : _DAT_007c7200);
}
// ---- BLOCK 2: sample fog near/far/color ----
FUN_005a4010(!DAT_0081dbf8); // disable fog if master fog flag clear
if (DAT_0081dbf8) {
FUN_005a3f90(DAT_0081dbf8); // FOGENABLE = true
int ok2 = FUN_004ff480(turbidity,
&fogNear, // out float
&fogFar, // out float
&fogColor);// out ARGB
if (ok2) {
if (DAT_008427a9 /* weatherFogEnabled */) {
// blend fogNear, fogFar, fogColor toward weather values
// identical lerp structure
}
FUN_005a41b0(&fogColor, fogNear, fogFar);
// → SetRenderState(FOGCOLOR, ...), (FOGSTART, ...), (FOGEND, ...)
}
}
}
}
}
This is the ONLY per-frame writer of fog D3D state. Called by
FUN_00455a50 (scene entry) via FUN_00506270 etc.
Block-1 Keyframe Layout — FUN_00501600 @ 0x00501600
chunk_00500000.c, lines 1151-1232. Decoded keyframe struct:
| Offset | Field | Type | Notes |
|---|---|---|---|
| +0x04 | SunMagnitude | float | lerp → param_4 (radius) |
| +0x08 | SunYAngle | float | degrees; * (π/180) gives radians |
| +0x0c | SunZAngle | float | degrees; * (π/180) gives radians |
| +0x10–0x12 | AmbientColor | byte[3] | R,G,B (packed via byte-shuffles) |
| +0x14 | FogDensity | float | lerp → param_2 |
| +0x18–0x1a | SunColor | byte[3] | R,G,B |
From chunk_00500000.c:1190-1216:
// out sunDirVec (param_4):
fVar9 = lerp(nextKey[+0x04], prevKey[+0x04], param_1); // magnitude
float sinAngY = lerp(nextKey[+0x08], prevKey[+0x08]) * DEG2RAD;
float sinAngZ = lerp(nextKey[+0x0c], prevKey[+0x0c]) * DEG2RAD;
float cosY = cos(sinAngY);
float sinY = sin(sinAngY);
float cosZ = cos(sinAngZ);
float sinZ = sin(sinAngZ);
param_4[0] = fVar9 * sinZ * cosY;
param_4[1] = fVar9 * cosZ * cosY;
param_4[2] = fVar9 * sinY;
// out sunColor ARGB (param_3): from bytes at +0x18,0x19,0x1a
// out ambientColor ARGB (param_5): from bytes at +0x10,0x11,0x12
(The repeated FUN_005df4c4() calls with no args are Ghidra's artifact
for a union/reinterpret-cast byte shuffle — they are NOT an RNG and NOT
a lightning modulator.)
Block-2 Keyframe Layout — FUN_00501860 @ 0x00501860
chunk_00500000.c, lines 1236-1268. Second block of the keyframe struct:
| Offset | Field | Type | Notes |
|---|---|---|---|
| +0x20 | FogNear | float | |
| +0x24 | FogFar | float | |
| +0x28–0x2a | FogColor | byte[3] | R,G,B |
Apply-Block-1 — FUN_00505f30 @ 0x00505F30
chunk_00500000.c, lines 6026-6092. Stores the sampled values into global
cache and into the _DAT_008682bc/c0/c4 "per-vertex ambient" slot:
DAT_00842780 = fogDensity; // param_2
DAT_0084277c = sunColor; // param_3 (ARGB)
DAT_00842950 = sunDir[0]; // param_4[0]
DAT_00842954 = sunDir[1];
DAT_00842958 = sunDir[2];
DAT_00842778 = ambientColor; // param_5 (ARGB)
_DAT_008682c8 = sunDir[0]; // duplicated
_DAT_008682d0 = sunDir[2];
_DAT_008682cc = sunDir[1];
FUN_00451a60(&_something, ambientColor); // unpacks ARGB → 3 floats
_DAT_008682bc = unpacked R; // RGB ambient floats for mesh lighting
_DAT_008682c0 = unpacked G;
_DAT_008682c4 = unpacked B;
DAT_008682d4 = 0; // ambient dirty flag
_DAT_008682bc/c0/c4 is the THING the sky tints the world with.
It is consumed per-vertex by mesh-draw, not by a D3D state write.
Candidate "Sky Pass" — FUN_00507a50 @ 0x00507A50
chunk_00500000.c, lines 7250-7299.
void FUN_00507a50(weatherMgr, phase) {
if ((FUN_00451ec0() || phase == 0)) { // skyEnabled or pass-0
weatherMgr.renderingFlag = 1;
char oldAlpha = FUN_005a1560();
FUN_005a3f90(DAT_008427a9 != '\0'); // FOGENABLE ← weatherFog flag
float oldFar = DAT_0081fc98;
FUN_0054bf30(DAT_0081fc98 * _DAT_007c6f14); // multiply far plane
FUN_005a3e20(8, 0); // ZFUNC=ALWAYS, ZWRITE=0
if (phase == 0) {
// iterate weather volume objects (rain/snow/fog cells)
for (uint i = 0; i < weatherMgr.count; ++i) {
int obj = *(int*)(weatherMgr.pArray + i*4);
uint flags = *(uint*)(weatherMgr.pFlags + i*4);
if (obj && !(flags & 1) &&
(DAT_0081dbf9 || !(flags & 4)) &&
(!DAT_008427a9 || !FUN_005a1560() || !(flags_byte & 2)))
{
FUN_00511720(obj); // scene-graph walk (update per-frame)
FUN_00511760(obj); // scene-graph walk (draw)
}
}
} else if (DAT_0081dbf9) {
// phase 1: just let the device callback at +0x64 run
(**(code **)(*DAT_00870340 + 0x64))(weatherMgr[+0x28]);
}
FUN_0054bf30(oldFar); // restore far plane
FUN_005a3e20(4, 1); // ZFUNC=LESSEQUAL, ZWRITE=1
FUN_005a3f90(oldAlpha); // restore FOGENABLE
weatherMgr.renderingFlag = 0;
}
}
Called by FUN_00506d90 (chunk_00500000.c:6683):
FUN_00507a50(0)= phase-0 (sky-ish volume draw, before terrain/entity)FUN_00507a50(1)= phase-1 (post-draw overlay, gated by weather flag)
The DAT_0081fc98 * _DAT_007c6f14 line IS the far-plane multiplier —
however _DAT_007c6f14 here is ambiguous, used elsewhere as a small
rotation coefficient (chunk_005E0000.c:258 etc). This deserves a
follow-up: check if that DAT is a configurable sky-far-plane multiplier
or a recycled constant.
No matrix-translation-zero (camera-anchor) is present in this
function. The scene graph objects iterated via FUN_00511720/60 are
rain/snow/fog-shaft volumes — they are positioned in world space and
expected to follow the camera via normal scene-graph propagation.
No per-mesh blend swap (alpha for clouds vs additive for sun) is present. Blend mode is set once for the whole pass.
Fog Write Path — SUMMARY
The complete trail:
FUN_00505f30stores sun/ambient globals (no D3D call).FUN_005062e0lerps current-keyframe fog → weather-fog if needed.FUN_005a41b0(&fogColor, fogNear, fogFar)→FUN_005a4080(pFogColor_as_ARGB, near, far)→- Writes D3DRS_FOGCOLOR=0x22, D3DRS_FOGSTART=0x24, D3DRS_FOGEND=0x25.
Fog ENABLE is wrapped by FUN_005a4010 (master gate at +0x4be) +
FUN_005a3f90 (actual D3DRS_FOGENABLE=0x1c write). Both driven by
the DAT_0081dbf8 master-fog flag.
D3DRS_FOGTABLEMODE=0x23, FOGVERTEXMODE=0x8c, FOGDENSITY=0x26 —
these are only set once in the default-init (FUN_005a10f0) and never
per-frame. Retail uses linear fog (FOGSTART/FOGEND), not exponential
(FOGDENSITY).
Lightning Flash — NOT FOUND
Searched for:
FUN_005df4c4(the RNG/byte-shuffle) — all hits are byte-packing for ARGB color interpolation, not random modulation._DAT_008427b8— it's a weather-fog-blend progress (0 → 1 ramp), not a per-frame random value.- Time-based sine/cos on small periods — none found coupled to sky color.
- Any RNG call near fog/ambient writes — none found.
Lightning is likely a separate system not found in this hunt, OR does not exist in retail and is purely decorative in modern ports.
Matrix-Anchor (Camera-Centered Sky) — NOT FOUND
Searched for patterns like:
*(undefined4*)(mat + 0x30) = 0;SetTransform(D3DTS_VIEW=2, ...)via any vtable offset- A matrix copy that zeroes three consecutive 4-byte fields
Nothing found anchored to the sky pass. The view matrix is NOT rewritten with zero translation before the sky draw. This is consistent with the conclusion that there is no discrete "sky dome" — the weather/fog volume objects follow the camera by being placed in camera-relative world position by their parent scene-graph node.
Huge Far-Plane Constants — NOT FOUND
Searched for:
- 0x49742400 (1e6f) — no hits
- 0x47c35000 (1e5f) — no hits
- 0x4b189680 (1e7f) — no hits
The only far-plane modulator is DAT_0081fc98 * _DAT_007c6f14 at
chunk_00500000.c:7272. Without knowing _DAT_007c6f14's numeric value
from a debugger or disassembly of .rdata, we can't tell if this is
1 (sky-far-plane multiplier) or ~1 (no-op). A follow-up disassembly of the address
0x007c6f14should nail this down.
D3D State Inventory — Sky-Relevant Per-Frame Writes
Only four D3D states are written per-frame in the sky-adjacent code path:
| State | Hex | Name | Writer | Source of value |
|---|---|---|---|---|
| 34 | 0x22 | FOGCOLOR | FUN_005a4080 |
current keyframe+weather lerp |
| 36 | 0x24 | FOGSTART | FUN_005a4080 |
current keyframe+weather lerp |
| 37 | 0x25 | FOGEND | FUN_005a4080 |
current keyframe+weather lerp |
| 28 | 0x1c | FOGENABLE | FUN_005a3f90 |
DAT_0081dbf8 / weather flag |
Plus, inside FUN_00507a50 (phase-0 weather-volume draw only):
| State | Hex | Name | Value |
|---|---|---|---|
| 23+14 | 0x17+0x0e | ZFUNC + ZWRITE | ALWAYS + OFF → restore LESSEQUAL + ON |
Gaps / Unresolved
- The classical sky dome / celestial body draw — NOT FOUND. Either retail literally does not have one (sky is just a clear-color glimpse plus fog between camera and terrain), or it's hidden inside the RegionDesc.Scene entries and rendered as a regular object through the normal scene-graph path.
_DAT_007c6f14— used as a far-plane multiplier in the sky pass, but cross-referenced use in chunks 0x5E0000 suggests it's a unit-scale float (< 10). Needs raw.rdatadump to confirm its value.- Is
_DAT_008682bc/c0/c4consumed by D3D material or per-vertex color? The writers are clear; the readers need a separate hunt (likely in chunk_00590000.c or the mesh-draw path atFUN_0054c9c0). - Lightning flash — not found in the keyframe/fog paths. If it exists, it's in a separate code path (maybe tied to weather presets 6 "thunderstorm" — DAT_008427a9=1 case in chunk_00550000.c:11886).
Files / Addresses cited (all absolute paths)
C:\Users\erikn\source\repos\acdream\docs\research\decompiled\chunk_005A0000.c- lines 687-740:
FUN_005a10f0device reset (AMBIENT=0 here, only caller) - lines 2606-3035: D3D state wrapper zoo (all vtable-0xe4 writers)
- lines 2868-2907:
FUN_005a4080= SetFog triple - lines 2911-2921:
FUN_005a41b0= SetFog wrapper - lines 3346-3400: default state-init defaults
- lines 3860-3970:
FUN_005a5950world-scene top-level renderer
- lines 687-740:
C:\Users\erikn\source\repos\acdream\docs\research\decompiled\chunk_00500000.c- lines 1151-1232:
FUN_00501600Block-1 keyframe sampler - lines 1236-1268:
FUN_00501860Block-2 (fog) keyframe sampler - lines 6026-6092:
FUN_00505f30apply-keyframe-1 to globals - lines 6213-6333:
FUN_005062e0per-frame sky+fog update (MAIN) - lines 6683-6708:
FUN_00506d90scene renderer (calls sky phase 0/1) - lines 7250-7299:
FUN_00507a50the phase-0 weather-volume draw
- lines 1151-1232:
C:\Users\erikn\source\repos\acdream\docs\research\decompiled\chunk_00450000.c- lines 608-622:
FUN_00451a60ARGB → 3-float unpack - lines 4300-4479: per-landblock ambient setup (
_DAT_008682bcwriters)
- lines 608-622:
C:\Users\erikn\source\repos\acdream\docs\research\decompiled\chunk_00550000.c- lines 11835-12016:
FUN_0055eb40weather-preset-id → fog constants (0=clear, 1=light rain, 2=medium, 3=heavy, 4=blizzard, 5=?, 6=thunderstorm)
- lines 11835-12016:
C:\Users\erikn\source\repos\acdream\docs\research\decompiled\chunk_004F0000.c- lines 10692-10720:
FUN_004ff440/FUN_004ff480keyframe trampolines
- lines 10692-10720:
Recommendation for acdream port
Given these findings, our C# SkyRenderer should:
- Stop expecting a celestial-body draw loop in the decompile — Hunt A will likely confirm via the Region-loader angle.
- Port the keyframe sampler faithfully from
FUN_00501600/ 1860 — the struct layout above is the ground truth for RegionDesc keyframe blobs. - Apply the sun direction and colors to world-mesh vertex lighting,
not to a D3DRS_AMBIENT equivalent. The values live in
_DAT_008682bc/c0/c4and are consumed by the mesh pipeline. - Set fog via SetFog(color, start, end) + FogEnable per frame from
the Block-2 sample. This matches
FUN_005a41b0. - Do NOT add a matrix translation-zero — retail does not anchor the sky dome, because retail does not appear to HAVE a sky dome draw in the D3D state sense. If acdream renders a dome mesh, it's a modern addition.
- Tint clouds (if rendered) via the per-mesh ambient global, not
via the current uniform-one approach. The previous research doc's
conclusion on "clouds multiply by ambient" is likely correct in
spirit — just know that "ambient" here means
_DAT_008682bc/c0/c4, not D3DRS_AMBIENT.