Begin work on cloth (again)

1 files changed, 53 insertions, 67 deletions

diff --git a/Scripts/make_dfg_lut.py b/Scripts/make_dfg_lut.py
index d58f3cf..4e16c99 100755
--- a/Scripts/make_dfg_lut.py
+++ b/Scripts/make_dfg_lut.py
@@ -73,45 +73,54 @@ def G_Cloth_L(x, a, b, c, d, e):
 
 
 @numba.njit(cache=True)
-def G_Cloth(roughness, LoH):
+def Lambda_Cloth(roughness, cos_theta):
     a0, a1 = 25.3245, 21.5473
     b0, b1 = 3.32435, 3.82987
     c0, c1 = 0.16801, 0.19823
     d0, d1 = -1.27393, -1.97760
     e0, e1 = -4.85967, -4.32054
 
-    one_minus_r = 1.0 - roughness
-    one_minus_r_sq = one_minus_r * one_minus_r
-    one_minus_LoH = 1.0 - LoH
+    # Matches shader: interpolator = r^2 blends toward rough (a1) column
+    r_sq = roughness * roughness
 
     lambda_val = 0.0
-    if LoH < 0.5:
-        L0 = G_Cloth_L(LoH, a0, b0, c0, d0, e0)
-        L1 = G_Cloth_L(LoH, a1, b1, c1, d1, e1)
-        L = lerp(L0, L1, one_minus_r_sq)
+    if cos_theta < 0.5:
+        L0 = G_Cloth_L(cos_theta, a0, b0, c0, d0, e0)
+        L1 = G_Cloth_L(cos_theta, a1, b1, c1, d1, e1)
+        L = lerp(L0, L1, r_sq)
         lambda_val = math.exp(L)
     else:
         L_05_0 = G_Cloth_L(0.5, a0, b0, c0, d0, e0)
         L_05_1 = G_Cloth_L(0.5, a1, b1, c1, d1, e1)
-        L_05 = lerp(L_05_0, L_05_1, one_minus_r_sq)
+        L_05 = lerp(L_05_0, L_05_1, r_sq)
 
-        L_LoH_0 = G_Cloth_L(one_minus_LoH, a0, b0, c0, d0, e0)
-        L_LoH_1 = G_Cloth_L(one_minus_LoH, a1, b1, c1, d1, e1)
-        L_LoH = lerp(L_LoH_0, L_LoH_1, one_minus_r_sq)
+        one_minus_cos = 1.0 - cos_theta
+        L_c_0 = G_Cloth_L(one_minus_cos, a0, b0, c0, d0, e0)
+        L_c_1 = G_Cloth_L(one_minus_cos, a1, b1, c1, d1, e1)
+        L_c = lerp(L_c_0, L_c_1, r_sq)
 
-        lambda_val = math.exp(2.0 * L_05 - L_LoH)
+        lambda_val = math.exp(2.0 * L_05 - L_c)
 
     # Apply terminator softening (equation 4)
-    return pow(lambda_val, 1.0 + 2.0 * pow(one_minus_LoH, 8.0))
+    return pow(lambda_val, 1.0 + 2.0 * pow(1.0 - cos_theta, 8.0))
 
 
 @numba.njit(cache=True)
-def integrate_brdf_jitted(roughness, NoV, brdf_type, num_samples):
+def V_Cloth(roughness, NoL, NoV):
+    # Height-correlated Smith: G2 / (4 * NoL * NoV)
+    lambda_l = Lambda_Cloth(roughness, NoL)
+    lambda_v = Lambda_Cloth(roughness, NoV)
+    return 1.0 / ((1.0 + lambda_l + lambda_v) * 4.0 * NoL * NoV + 1e-6)
+
+
+@numba.njit(cache=True)
+def integrate_brdf_jitted(roughness, NoV, num_samples):
     V_x = math.sqrt(1.0 - NoV * NoV)
     V_y = 0.0
     V_z = NoV
 
-    A, B = 0.0, 0.0
+    # R: GGX scale, G: GGX bias, B: cloth DFG
+    std_scale, std_bias, cloth_val = 0.0, 0.0, 0.0
 
     for i in range(num_samples):
         e1, e2 = random.random(), random.random()
@@ -137,47 +146,33 @@ def integrate_brdf_jitted(roughness, NoV, brdf_type, num_samples):
 
         NoL = saturate(L_z)
         NoH = saturate(H_z)
-        NoV_proxy = saturate(V_z) # NoV is V_z
+        NoV_proxy = saturate(V_z)
 
         if NoL > 0:
-            scale, bias = 0.0, 0.0
             # --- Standard BRDF ---
-            if brdf_type == 1:
-                # Note that the D term is present in the numerator and the denominator, so it cancels out.
-                #D = D_GGX(roughness, NoH)
-                G = G_GGXSmith(roughness, NoL, NoV_proxy)
-                Fc_term = pow(1.0 - VoH, 5.0)
-
-                # PDF of GGX Importance Sampling is D * NoH / (4 * VoH).
-                # The full term is (D * G * NoL) / PDF, which simplifies to:
-                # G * NoL * (4 * VoH / NoH).
-                # This can be unstable when NoH is close to zero, so we clamp the denominator.
-                common_term = (G * NoL * 4.0 * VoH) / max(NoH, 1e-5)
-
-                # We are baking the two components of the split-sum approximation for IBL:
-                # reflectance = f0 * scale + bias
-                scale = common_term * (1.0 - Fc_term)
-                bias = common_term * Fc_term
-            # --- Cloth BRDF ---
-            elif brdf_type == 2:
-                # We are importance sampling GGX, so must account for its PDF.
-                D_c = D_Cloth(roughness, NoH)
-                G_c = G_Cloth(roughness, VoH)
+            # D cancels between numerator and PDF.
+            G = G_GGXSmith(roughness, NoL, NoV_proxy)
+            Fc_term = pow(1.0 - VoH, 5.0)
 
-                # PDF = D_GGX(r, NoH) * NoH / (4 * VoH)
-                pdf_ggx = D_GGX(roughness, NoH) * NoH / (4.0 * VoH + 1e-6)
+            # PDF = D_GGX * NoH / (4 * VoH), so (D * G * NoL) / PDF simplifies to:
+            common_term = (G * NoL * 4.0 * VoH) / max(NoH, 1e-5)
 
-                # We must divide by the PDF and multiply by our target distribution and the cosine term.
-                scale = (D_c * G_c * NoL) / (pdf_ggx + 1e-6)
-                bias = 0.0
+            std_scale += common_term * (1.0 - Fc_term)
+            std_bias += common_term * Fc_term
 
-            A += scale
-            B += bias
+            # --- Cloth BRDF ---
+            # Same GGX importance samples, reweighted for cloth D and V.
+            if roughness >= 1e-4:
+                D_c = D_Cloth(roughness, NoH)
+                V_c = V_Cloth(roughness, NoL, NoV_proxy)
+                pdf_ggx = D_GGX(roughness, NoH) * NoH / (4.0 * VoH + 1e-6)
+                cloth_val += (D_c * V_c * NoL) / (pdf_ggx + 1e-6)
 
-    return A / num_samples, B / num_samples
+    inv_n = 1.0 / num_samples
+    return std_scale * inv_n, std_bias * inv_n, cloth_val * inv_n
 
 
-def calculate_pixel(coords, resolution, brdf_type, num_samples):
+def calculate_pixel(coords, resolution, num_samples):
     x, y = coords
     u = (x + 0.5) / resolution
     v = (y + 0.5) / resolution
@@ -186,18 +181,14 @@ def calculate_pixel(coords, resolution, brdf_type, num_samples):
     roughness = saturate(v)
     if NoV < 1e-4: return x, y, 0.0, 0.0, 0.0
 
-    r, g = 0.0, 0.0
-    if brdf_type == 1: # standard
-        r, g = integrate_brdf_jitted(roughness, NoV, 1, num_samples)
-    elif brdf_type == 2: # cloth
-        if roughness < 1e-4: return x, y, 0.0, 0.0, 0.0
-        r, g = integrate_brdf_jitted(roughness, NoV, 2, num_samples)
+    std_scale, std_bias, cloth = integrate_brdf_jitted(roughness, NoV, num_samples)
 
-    return x, y, r, g, 0.0
+    # R: GGX scale, G: GGX bias, B: cloth DFG
+    return x, y, std_scale, std_bias, cloth
 
 
-def generate_exr(resolution, output_filename, brdf_type, num_samples, num_workers):
-    print(f"Generating {resolution}x{resolution} EXR '{output_filename}' ({num_samples} samples/pixel) using {num_workers} workers.")
+def generate_exr(resolution, output_filename, num_samples, num_workers):
+    print(f"Generating {resolution}x{resolution} EXR '{output_filename}' (R=GGX scale, G=GGX bias, B=cloth) ({num_samples} samples/pixel) using {num_workers} workers.")
     header = OpenEXR.Header(resolution, resolution)
     pt = Imath.PixelType(Imath.PixelType.FLOAT)
     header['channels'] = { 'R': Imath.Channel(pt), 'G': Imath.Channel(pt), 'B': Imath.Channel(pt) }
@@ -205,7 +196,7 @@ def generate_exr(resolution, output_filename, brdf_type, num_samples, num_worker
     pixel_data = np.zeros((resolution, resolution, 3), dtype=np.float32)
 
     coords_to_process = [(x, y) for y in range(resolution) for x in range(resolution)]
-    worker_func = partial(calculate_pixel, resolution=resolution, brdf_type=brdf_type, num_samples=num_samples)
+    worker_func = partial(calculate_pixel, resolution=resolution, num_samples=num_samples)
 
     processed_count = 0
     total_pixels = len(coords_to_process)
@@ -241,15 +232,13 @@ def generate_exr(resolution, output_filename, brdf_type, num_samples, num_worker
         raise RuntimeError(f"Failed to write EXR file '{output_filename}': {e}")
 
 def main():
-    parser = argparse.ArgumentParser(description='Generate DFG LUT EXR images for PBR.')
-    parser.add_argument('-t', '--type', choices=['standard', 'cloth'], default='standard',
-                        help='Type of DFG texture to generate (default: standard)')
+    parser = argparse.ArgumentParser(description='Generate packed DFG LUT (R=GGX scale, G=cloth, B=GGX bias).')
     parser.add_argument('-r', '--resolution', type=int, default=512,
                         help='Resolution of the square EXR image (default: 512)')
     parser.add_argument('-s', '--samples', type=int, default=8192,
                         help='Number of samples per pixel for integration (default: 8192)')
-    parser.add_argument('-o', '--output',
-                        help='Output filename (default: dfg_standard.exr or dfg_cloth.exr)')
+    parser.add_argument('-o', '--output', default='dfg.exr',
+                        help='Output filename (default: dfg.exr)')
     parser.add_argument('-j', '--workers', type=int, default=os.cpu_count(),
                         help=f'Number of worker processes (default: {os.cpu_count()})')
 
@@ -259,11 +248,8 @@ def main():
         print("Error: Resolution must be a positive integer")
         return 1
 
-    brdf_type = 1 if args.type == 'standard' else 2
-    output_filename = args.output if args.output else f'dfg_{args.type}.exr'
-
     try:
-        generate_exr(args.resolution, output_filename, brdf_type, args.samples, args.workers)
+        generate_exr(args.resolution, args.output, args.samples, args.workers)
     except Exception as e:
         print(f"Error: {e}")
         return 1