Rock and animated-prop shaders

ForzaTechShadersApplies toFH6

uber_rock_array (197 compiled permutations, bin.zip _0x00000022) and uber_rock (_0x00000023) draw rural retaining walls and stone tiles, coastal boulders and alpine rock faces. Their compiled pixel shaders (all 12 permutations met in the rural, coastal and mountain test boxes) were interpreted on real merged constants and real textures and match the formulas below to ≤ 3e-7 albedo, 3e-8 roughness and 1e-7 normal where no high-tiling uv is involved. Observed The larger residuals on the micro and mid layers (up to 3.4e-4 albedo, 4.7e-3 normal at a camera about 1,900 m from the origin) are float32 uv rounding at tilings 9 to 20; with those tilings set to 1 near the origin they fall to ≤ 2.6e-5 / 2.4e-6 / 2.5e-5. Observed

Counts come from a recorded census of the Tokyo track Brio and its rural and coastal test boxes; no longer re-checkable. Observed

Permutations met

Family	Permutation	Materials in the test boxes
`uber_rock_array`	`_Array_Micro_Spline`	rural retaining walls
	`_Array_Micro_Up_Modulate_USU`, `_Array_Up_Modulate_USU` (+ `_Moss`)	10 and 7 (+1 and +1)
`uber_rock`	`_diff1_simple_Modulate_Spline`	rural `_LOD4` and `_PLOD`
	`_diff1_simple_Modulate_Upshader`, `_diff1_simple_Upshader`, `_diff1_simple_Frost`	15, 7, 9
	`_Micro2_Modulate_UpShader`, `_Micro1_Modulate`, `_Micro1`	3, 1, 1

All rows Observed. Each of the 64 rock materials matches exactly one permutation. Texture indices sit at material dword 88 + the TXMP slot (uber_rock_array) and 76 + the slot (uber_rock). “TextureArray” is not a GPU texture array: the material binds five ordinary textures MidDiffuse_A..E and MidNormal_A..E and picks two per pixel; the only Texture2DArray in these shaders is the terrain clipmap. Observed INSTANCE_VARIATION x, y, z are the placement record bytes +0x34, +0x35, +0x36 divided by 255; COLOR is the vertex colour (.g up mask, .b and .a micro weights). The draw scale S is the largest column length of the draw’s 3 × 3 (cb9[0..2]; instanced: rows × scale); uber_rock’s micro uses the per-axis column lengths. Observed Frost has no effect on the G-buffer outputs. Observed

Shared pieces

UpMask:   a = n.y · COLOR.g,  b = sat(2·COLOR.g − 1) · N.y       (n = the layer normal so far, N = vertex normal; game Y up)
          sm = sat((b − UpMaskSmoothMinAngle/90) / ((UpMaskSmoothMaxAngle − UpMaskSmoothMinAngle)/90))
          up = UpMaskOnOffOnOff ? sat((a − UpMaskMinAngle/90 + sm·(b − a)) / ((UpMaskMaxAngle − UpMaskMinAngle)/90)) : 0
World layer normal (Up textures): n = normalize(X·x − B·y + N·z), X = (1,0,0), B = cross(N, X)
Modulate: uvM = (dot(P − c0, A)·sA + .5, dot(P − c0, B)·sB + .5) (ModulateConstants frame; P absolute)
          M = ModulateTexture(uvM); t = 0.5 + TM_Max·IV.x·(M − 0.5); o = overlay(c, t)
          c' = o + (c − o)·sat(0.01·|WORLDPOS| + ViewDist_offset)                 (fades out with camera distance)
Tide:     e = WORLDPOS.y + cb2[8].y + TideMarkAOEffect·AO', h = frac(IV.y)·1000 + 1.15
          satf = 1 + (1 − sat((e − h)/Blend))·SaturationAmount, dark = max(sat((e + DarkenOffset − h)/Blend), DarkenAmount)
          L = 0.33(r + g + b), k = L·(0.27, 0.6, 0.75);  c' = (satf·(c − k) + k)·dark

B is unnormalised in uber_rock and normalised in uber_rock_array. The tide-mark height comes from the placement record byte +0x35 (IV.y). AO' in uber_rock_array is the traced layer AO; in uber_rock it is 1 + cb2[92].x·(sat(AO)^|cb2[92].w| − 1), a run-time AO curve, and TideMarkAOEffect is non-zero on every census material (1, 1.5, 3.8, 4.36, 7), so uber_rock’s water line depends on that run-time global. Observed Hand-placed (.pvs) rock instances have no tide byte. Observed

uber_rock_array

Mac = MacroDiffuse(uv0), MN = MacroNormal(uv0), MAO = MacroAO(uv0).x
v = Mid_Variation_Mask(uv0).x;  i = clamp(floor(4v), 0, 3), j = i + 1        (layers A..E)
s = smoothstep(sat((4v − i − lo)/(hi − lo))),  lo/hi = 0.5 ∓ (1 − Array_Optimize)/2
uvm = MidUV_Tiling·uv1·(ScaleMidUVsOnOff ? S : 1);  D_i, D_j = MidDiffuse_[i|j](uvm), N_i, N_j = MidNormal_[i|j](uvm)
b = sat((s − lo2 + HeightScale·(N_j.a − N_i.a))/(hi2 − lo2)),  lo2/hi2 = 0.5 ∓ MidBlendSoftness/2
mid = D_i + (D_j − D_i)·b;  (x, y) = 2·lerp(N_i.rg, N_j.rg, b) − 1, z = sqrt(sat(1 − x² − y²)), nmid = normalize(T1·x − B1n·y + N·z)
(x', y') = xy of normalize(dot(nmid, T0), dot(nmid, B0n), dot(nmid, N));  (mx, my) = 2·MN.rg − 1
vv = normalize(x' + k·mx, y' − k·my, 1), k = NormalScale_MacroMidUpTextureSet_MacroMidUpBlendedTextureSet
nmd = normalize(T0·vv.x + B0n·vv.y + N·vv.z)
c1 = Mac + (sat(ov)^2.2 − Mac)·sat(normalStrength0),  g(x) = x^(1/2.2), ov = g(mid) > .5 ? 2·g(Mac)·g(mid) : 1 − 2(1 − g(Mac))(1 − g(mid))
r1 = clamp(mid.a + MacroRoughnessStrength·(2·rm − 1), 0, 1),  rm = Mac.a + (sat(1 − 2(1 − g(Mac.a))(1 − g(mid.a)))^2.2 − Mac.a)·sat(normalStrength0)
moss (_Moss): mm = (1 − s) [i = 0];  AO base = MAO + (N_i.b − MAO)·mm  (else MAO)
Up (UpBlend + UseSimpleUp): UD, EX, UN = UpSimpleDiffuse/Extras/Normal at the absolute world (X, Z) (1 m repeat)
     up0 = UpMask(n = nmd);  kh = UpMaskMegaHeightStrength·EX.b;  w = sat((up0 − (0.5 − kh))/(2kh))
     c = lerp(c1, UD, w), r = lerp(r1, EX.r, w), n = (1 − w)·nmd + w·nUp (world layer normal, UN), porosity lerp to EX.g
     AO' = lerp(1 + (1 − w)(AObase − 1), UN.b, w)
  (no Up: c = c1, r = r1, n = nmd, AO' = 1 + (AObase − 1)(1 − sat((UpMaskMegaHeightStrength − 0.5)/(2·UpMaskMegaHeightStrength))))
Micro:
  _Array_Micro_Spline: uvμ = MicroUVTiling·uv1·S; mA, mB = MicroNormalA/BTexture(uvμ), m = min(v, 1), I = MicroIntensity
       md = lerp(mA.a, mB.a, m), mr = lerp(mA.b, mB.b, m), micro (x, y, z) = lerp of the standard decodes
  Up permutations: mB = MicroNormalBTexture(uvμ) only, I = MicroIntensity·((1 − up0)(1 − COLOR.b) + COLOR.b)
  c += (overlay_c(md) − c)·I,  overlay_c(d) = c > .5 ? 1 − 2(1 − d)(1 − c) : 2dc;  r += I·MicroRoughnessStrength·(2·mr − 1)
  bb = normalize(dot(n, T1), dot(n, B1n), dot(n, N));  d = 1 + I(mz − 1), p = I·mx, q = −I·my
  cm = normalize((d + q)·bb.x + p·bb.z, (d + p)·bb.y + q·bb.z, d·bb.z − q·bb.y − p·bb.x)
  t = bb + (cm − bb)·sat(|N·V|)  (V toward the camera: view dependent);  n = normalize(T1·t.x + B1n·t.y + N·t.z)
Modulate (_Modulate): colour with M.rgb, roughness with M.a, both weighted 1 − max(up0, mm)
then tide(c, AO'), clamp(c, 0, 1), times the run-time AO factor K;  roughness written unclamped;  metal 0;  emissive 0

UpMaskMegaHeightStrength is 0 on 16 of 19 coastal and mountain array materials; there w = sat((up0 − 0.5)/0) is the GPU’s hard step (+inf → 1, −inf → 0, NaN → 0). Observed

uber_rock

D = DiffuseA(uv0), AO = AOA(uv0).x, nb = normalize(T·x − B·y + N·z) of NormalA(uv0) (B = cross(N, T)·w, unnormalised)
Up (_Upshader): C = DiffuseC / NormalC / AOC at 0.2·world (X, Z); up = UpMask(n = nb)
     c = lerp(D.rgb, C.rgb, up), n = normalize(lerp(nb, nUp, up)), r = lerp(D.a, C.a, up), AO = lerp(AOA, AOC, up)
  (else up = the constant mask, c = D.rgb, n = nb, r = D.a)
Micro (_Micro1: M1 only, cw = 1; _Micro2: M1 + M2, cw = COLOR.a); cz = COLOR.b:
     M1 = MicroNormal(MicroUVTiling·uv1·cs), M2 = Micro2Normal(Micro2UVTiling·uv1·cs)   (cs = column scales)
     dd = M2.a + cw((M1.a − .5)cz + .5 − M2.a);  c += (dd − .5)(1 − up)·sat((0.875 + 0.125·cs.z)·lerp(MicroDiffuse02Strength, MicroDiffuse01Strength, cw))
     mr = M2.b + cw((M1.b − .5)cz + .5 − M2.b);  mr += (.5 − mr)·up;  r = overlay(r, mr) (condition on r)
     t = normalize(dot(n, T), dot(n, Bn), dot(n, N));  mx = lerp(M2.x, M1.x·cz, cw), my = −lerp(M2.y, M1.y·cz, cw)
     f = _Micro2 ? max((1 − up)·cz, 1 − cw) : max((1 − up)·cz, 0)
     n = T·t2.x + Bn·t2.y + N·t2.z,  t2 = normalize(t.x + f·mx·(1 − up), t.y + f·my·(1 − up), t.z)
Modulate (_Modulate): colour only
albedo = lerp(tide(clamp(c_mod, 0, 1)), c_premodulate, up) · K;  roughness r;  metal 0

The MegaSkirt blend into the terrain megatexture near the ground (run-time terrain height and clipmap, cb7[29..51]; every uber_rock_array permutation above and uber_rock _Spline and _Micro1*) cannot be reproduced from files. The AO-into-albedo factor K is run time. Unknown

uber_asset_animated

Coastal boats AMB_CST_BOAT_0x_A_Main use _Animate_AnimOscillate and *_Alpha and TEX_CST_DECAL_COMM_A use _AlphaTest_AnimOscillate (buckets 0x19 and 0x38, G-buffer pass). Texture slot 0 AOTexture, 1 DiffuseATexture, 2 NormalATexture, 3 RoughnessMetalnessATexture; register r7.x is WetRoughnessScale: Observed

wf = sat(N.y + 1);  D = DiffuseATexture(uv0);  RM = RoughnessMetalnessATexture(uv0)
k  = lerp(1 + wf(cb7[3].x − 1), 1 + wf(cb7[3].z − 1), sat(1 − RM.r));  L = dot(D.rgb, (0.3, 0.59, 0.11))
RT0.rgb = ((1 + (1 − k)·0.2)(D.rgb·k − L) + L) · (1 + K(max(max(cb2[43].x, ao), max(cb2[44].x, ao)) − 1)),  ao = AOTexture(uv1).r
RT0.w = RM.g;  RT1.w = RM.r + WetRoughnessScale(RM.r^(1 + wf(cb7[3].y − 1)) − RM.r);  RT2 = standard NormalATexture(uv0)
RT1.xyz seed 0;  _AlphaTest: discard D.a < cb2[128].x

cb7[3] is a run-time weather global; with cb7[3] = (1, 1, 1) every weather term is the identity. Observed The vertex shader translates and rotates the mesh by triangle waves of game time cb2[9].x with per-draw speeds and amplitudes. Observed The boats’ pixel shader was interpreted on the real meshes and textures: albedo ≤ 6.4e-8, roughness ≤ 4.7e-8, normal ≤ 1e-7, identical discard decisions. Observed

anim_ledflipbookpanel

_Standard draws the festival LED screen prp_fst_display_01_a (bucket 0x19); texture indices sit in the material rows from register 7 (r7.x FlipbookTexture, r7.y MaskTexture): Observed

M = MaskTexture(uv1 · MaskUVTiling).r      (BC4 SNORM: signed)
frame = fmod(FramesPerSecond·TimeDilationAmount·cb2[9].x + OffsetMultiplier·trunc(uv0.x), X·Y),  X, Y = int(XFrames), int(YFrames)
col = trunc(fmod(frame, X)), row = trunc(frame / X);  uvF = (frac(uv0.x)/(NumVariations·X) + col/X, frac(uv0.y)/Y + row/Y)
c = FlipbookTexture(uvF).rgb · Brightness · M;   RT0.rgb = c × the AO term (AO input 1);  RT1.xyz seed = c·g(d)
RT0.w = 0;  RT1.w = lerp(RoughnessBoard, RoughnessLED, M);  RT2 = vertex normal;  opaque

c is not clamped by the shader (Brightness 10; negative between the LED dots from the signed mask); the render target format is not traced. The shader interpreted at game times 0, 0.07, 37.25 and 1,234.5 s matches the formulas (albedo ≤ 4.8e-7, roughness ≤ 2.9e-8, emissive ≤ 5.8e-8). Observed