edgevoid/armorpaint
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases 1 Wiki Activity

base: raytrace faster

Browse Source
This commit is contained in:
luboslenco
2026-08-26 10:11:25 +02:00
parent a21e0c3f5e
commit 4960ad5c56
8 changed files with 212 additions and 176 deletions
Download Patch File Download Diff File Expand all files Collapse all files
base/shaders/raytrace/raytrace_bake_ao.metal
+9 -2
View File
@@ -25,6 +25,13 @@ struct RayPayload {
constant int SAMPLES = 4;//64;
uint table_byte(texture2d<float, access::read> tex, int i) {
int t = (i & 131071) >> 2;
float4 c = tex.read(uint2(uint(t & 127), uint(t >> 7)), 0);
int ch = i & 3;
return uint((ch == 0 ? c.r : (ch == 1 ? c.g : (ch == 2 ? c.b : c.a))) * 255);
}
float rand(int pixel_i, int pixel_j, int sample_index, int sample_dimension, int frame, texture2d<float, access::read> sobol, texture2d<float, access::read> scramble, texture2d<float, access::read> rank) {
pixel_i += frame * 9;
pixel_j += frame * 11;
@@ -34,12 +41,12 @@ float rand(int pixel_i, int pixel_j, int sample_index, int sample_dimension, int
sample_dimension = sample_dimension & 255;
int i = (sample_dimension + (pixel_i + pixel_j * 128) * 8) & 131071;
int ranked_sample_index = sample_index ^ int(rank.read(uint2(i % 128, uint(i / 128)), 0).r * 255);
int ranked_sample_index = sample_index ^ int(table_byte(rank, i));
int value = int(sobol.read(uint2(ranked_sample_index, sample_dimension), 0).r * 255);
i = (sample_dimension % 8) + (pixel_i + pixel_j * 128) * 8;
value = value ^ int(scramble.read(uint2(i % 128, uint(i / 128)), 0).r * 255);
value = value ^ int(table_byte(scramble, i));
float v = (0.5f + value) / 256.0f;
return v;
base/shaders/raytrace/raytrace_bake_bent.metal
+9 -2
View File
@@ -25,6 +25,13 @@ struct RayPayload {
constant int SAMPLES = 4;//64;
uint table_byte(texture2d<float, access::read> tex, int i) {
int t = (i & 131071) >> 2;
float4 c = tex.read(uint2(uint(t & 127), uint(t >> 7)), 0);
int ch = i & 3;
return uint((ch == 0 ? c.r : (ch == 1 ? c.g : (ch == 2 ? c.b : c.a))) * 255);
}
float rand(int pixel_i, int pixel_j, int sample_index, int sample_dimension, int frame, texture2d<float, access::read> sobol, texture2d<float, access::read> scramble, texture2d<float, access::read> rank) {
pixel_i += frame * 9;
pixel_j += frame * 11;
@@ -34,12 +41,12 @@ float rand(int pixel_i, int pixel_j, int sample_index, int sample_dimension, int
sample_dimension = sample_dimension & 255;
int i = (sample_dimension + (pixel_i + pixel_j * 128) * 8) & 131071;
int ranked_sample_index = sample_index ^ int(rank.read(uint2(i % 128, uint(i / 128)), 0).r * 255);
int ranked_sample_index = sample_index ^ int(table_byte(rank, i));
int value = int(sobol.read(uint2(ranked_sample_index, sample_dimension), 0).r * 255);
i = (sample_dimension % 8) + (pixel_i + pixel_j * 128) * 8;
value = value ^ int(scramble.read(uint2(i % 128, uint(i / 128)), 0).r * 255);
value = value ^ int(table_byte(scramble, i));
float v = (0.5f + value) / 256.0f;
return v;
base/shaders/raytrace/raytrace_bake_light.metal
+9 -2
View File
@@ -33,6 +33,13 @@ float2 equirect(float3 normal, float angle) {
return float2(theta / PI2, phi / PI);
}
uint table_byte(texture2d<float, access::read> tex, int i) {
int t = (i & 131071) >> 2;
float4 c = tex.read(uint2(uint(t & 127), uint(t >> 7)), 0);
int ch = i & 3;
return uint((ch == 0 ? c.r : (ch == 1 ? c.g : (ch == 2 ? c.b : c.a))) * 255);
}
float rand(int pixel_i, int pixel_j, int sample_ndex, int sample_dimension, int frame, texture2d<float, access::read> sobol, texture2d<float, access::read> scramble, texture2d<float, access::read> rank) {
pixel_i += frame * 9;
pixel_j += frame * 11;
@@ -42,12 +49,12 @@ float rand(int pixel_i, int pixel_j, int sample_ndex, int sample_dimension, int
sample_dimension = sample_dimension & 255;
int i = (sample_dimension + (pixel_i + pixel_j * 128) * 8) & 131071;
int ranked_sample_index = sample_ndex ^ int(rank.read(uint2(i % 128, uint(i / 128)), 0).r * 255);
int ranked_sample_index = sample_ndex ^ int(table_byte(rank, i));
int value = int(sobol.read(uint2(ranked_sample_index, sample_dimension), 0).r * 255);
i = (sample_dimension % 8) + (pixel_i + pixel_j * 128) * 8;
value = value ^ int(scramble.read(uint2(i % 128, uint(i / 128)), 0).r * 255);
value = value ^ int(table_byte(scramble, i));
float v = (0.5f + value) / 256.0f;
return v;
base/shaders/raytrace/raytrace_bake_thick.metal
+9 -2
View File
@@ -25,6 +25,13 @@ struct RayPayload {
constant int SAMPLES = 4;//64;
uint table_byte(texture2d<float, access::read> tex, int i) {
int t = (i & 131071) >> 2;
float4 c = tex.read(uint2(uint(t & 127), uint(t >> 7)), 0);
int ch = i & 3;
return uint((ch == 0 ? c.r : (ch == 1 ? c.g : (ch == 2 ? c.b : c.a))) * 255);
}
float rand(int pixel_i, int pixel_j, int sample_index, int sample_dimension, int frame, texture2d<float, access::read> sobol, texture2d<float, access::read> scramble, texture2d<float, access::read> rank) {
pixel_i += frame * 9;
pixel_j += frame * 11;
@@ -34,12 +41,12 @@ float rand(int pixel_i, int pixel_j, int sample_index, int sample_dimension, int
sample_dimension = sample_dimension & 255;
int i = (sample_dimension + (pixel_i + pixel_j * 128) * 8) & 131071;
int ranked_sample_index = sample_index ^ int(rank.read(uint2(i % 128, uint(i / 128)), 0).r * 255);
int ranked_sample_index = sample_index ^ int(table_byte(rank, i));
int value = int(sobol.read(uint2(ranked_sample_index, sample_dimension), 0).r * 255);
i = (sample_dimension % 8) + (pixel_i + pixel_j * 128) * 8;
value = value ^ int(scramble.read(uint2(i % 128, uint(i / 128)), 0).r * 255);
value = value ^ int(table_byte(scramble, i));
float v = (0.5f + value) / 256.0f;
return v;
base/shaders/raytrace/raytrace_brute_core.metal
+44 -42
View File
@@ -78,41 +78,31 @@ float2 equirect(float3 normal, float angle) {
constant int RANK_CACHE_DIMS = 16;
uint rank_at(int base, int k, texture2d<float, access::read> rank) {
int i = (base + k) & 131071;
return uint(rank.read(uint2(i % 128, uint(i / 128)), 0).r * 255);
uint table_word(texture2d<float, access::read> tex, int i) {
int t = (i & 131071) >> 2;
float4 c = tex.read(uint2(uint(t & 127), uint(t >> 7)), 0);
return uint(c.r * 255) | (uint(c.g * 255) << 8) | (uint(c.b * 255) << 16) | (uint(c.a * 255) << 24);
}
uint table_byte(texture2d<float, access::read> tex, int i) {
return (table_word(tex, i) >> ((i & 3) * 8)) & 255;
}
uint2 init_scramble(uint2 tid, int frame, texture2d<float, access::read> scramble) {
int pixel_i = (int(tid.x) + frame * 9) & 127;
int pixel_j = (int(tid.y) + frame * 11) & 127;
int base = (pixel_i + pixel_j * 128) * 8;
uint lo = 0;
uint hi = 0;
for (int k = 0; k < 4; ++k) {
int i = base + k;
lo |= uint(scramble.read(uint2(i % 128, uint(i / 128)), 0).r * 255) << (k * 8);
i += 4;
hi |= uint(scramble.read(uint2(i % 128, uint(i / 128)), 0).r * 255) << (k * 8);
}
return uint2(lo, hi);
return uint2(table_word(scramble, base), table_word(scramble, base + 4));
}
uint4 init_rank(uint2 tid, int frame, texture2d<float, access::read> rank) {
int pixel_i = (int(tid.x) + frame * 9) & 127;
int pixel_j = (int(tid.y) + frame * 11) & 127;
int base = (pixel_i + pixel_j * 128) * 8;
uint r0 = 0;
uint r1 = 0;
uint r2 = 0;
uint r3 = 0;
for (int k = 0; k < 4; ++k) {
r0 |= rank_at(base, k, rank) << (k * 8);
r1 |= rank_at(base, k + 4, rank) << (k * 8);
r2 |= rank_at(base, k + 8, rank) << (k * 8);
r3 |= rank_at(base, k + 12, rank) << (k * 8);
}
return uint4(r0, r1, r2, r3);
return uint4(table_word(rank, base),
table_word(rank, base + 4),
table_word(rank, base + 8),
table_word(rank, base + 12));
}
float rand(int pixel_i, int pixel_j, int sample_index, int sample_dimension, int frame, uint2 scramble_cache, uint4 rank_cache, texture2d<float, access::read> sobol, texture2d<float, access::read> rank) {
@@ -130,8 +120,8 @@ float rand(int pixel_i, int pixel_j, int sample_index, int sample_dimension, int
ranked_sample_index = sample_index ^ int((packed >> ((sample_dimension & 3) * 8)) & 255);
}
else {
int i = (sample_dimension + (pixel_i + pixel_j * 128) * 8) & 131071;
ranked_sample_index = sample_index ^ int(rank.read(uint2(i % 128, uint(i / 128)), 0).r * 255);
ranked_sample_index = sample_index ^
int(table_byte(rank, sample_dimension + (pixel_i + pixel_j * 128) * 8));
}
int value = int(sobol.read(uint2(ranked_sample_index, sample_dimension), 0).r * 255);
@@ -192,8 +182,8 @@ float2 s16_to_f32(uint val) {
return float2(a, b) / 32767.0f;
}
float3 unpack_position(uint posxy, uint posz_norz) {
return float3(s16_to_f32(posxy), s16_to_f32(posz_norz).x);
float3 srgb_to_linear(float3 c) {
return c * (c * (c * 0.305306011 + 0.682171111) + 0.012522878);
}
float3 hit_world_position(ray ray, intersector_t::result_type intersection) {
@@ -256,9 +246,8 @@ float3 surface_specular(const float3 base_color, const float metalness) {
return mix(float3(0.04, 0.04, 0.04), base_color, metalness);
}
uint2 texel_coord(texture2d<float, access::read> tex, float2 tex_coord) {
uint2 size = uint2(tex.get_width(), tex.get_height());
return uint2(fract(tex_coord) * float2(size));
uint2 texel_coord(float2 tex_coord, float2 tex_size) {
return uint2(fract(tex_coord) * tex_size);
}
kernel void raytracingKernel(
@@ -285,6 +274,8 @@ kernel void raytracingKernel(
int frame = int(constant_buffer.eye.w);
uint2 scramble_cache = init_scramble(tid, frame, mytexture_scramble);
uint4 rank_cache = init_rank(tid, frame, mytexture_rank);
float2 tex_size = float2(mytexture0.get_width(), mytexture0.get_height());
float3 accum = float3(0, 0, 0);
for (int j = 0; j < SAMPLES; ++j) {
@@ -365,18 +356,21 @@ kernel void raytracingKernel(
float2 tex_coord = hit_attribute2d(vertex_uvs, barycentrics) * constant_buffer.params.z;
float3 hit = hit_world_position(ray, intersection);
uint2 texel = texel_coord(mytexture0, tex_coord);
uint2 texel = texel_coord(tex_coord, tex_size);
float4 texpaint0 = mytexture0.read(texel, 0);
float2 zw0 = s16_to_f32(a0.poszw);
float2 zw1 = s16_to_f32(a1.poszw);
float2 zw2 = s16_to_f32(a2.poszw);
float3 vertex_positions[3] = {
unpack_position(a0.posxy, a0.poszw),
unpack_position(a1.posxy, a1.poszw),
unpack_position(a2.posxy, a2.poszw)
float3(s16_to_f32(a0.posxy), zw0.x),
float3(s16_to_f32(a1.posxy), zw1.x),
float3(s16_to_f32(a2.posxy), zw2.x)
};
float3 vertex_normals[3] = {
float3(s16_to_f32(a0.nor), s16_to_f32(a0.poszw).y),
float3(s16_to_f32(a1.nor), s16_to_f32(a1.poszw).y),
float3(s16_to_f32(a2.nor), s16_to_f32(a2.poszw).y)
float3(s16_to_f32(a0.nor), zw0.y),
float3(s16_to_f32(a1.nor), zw1.y),
float3(s16_to_f32(a2.nor), zw2.y)
};
float3 n = normalize(hit_attribute(vertex_normals, barycentrics));
@@ -409,7 +403,7 @@ kernel void raytracingKernel(
if (normal_map) {
texpaint1 = mytexture1.read(texel, 0);
}
float3 texcolor = pow(texpaint0.rgb, float3(2.2, 2.2, 2.2));
float3 texcolor = srgb_to_linear(texpaint0.rgb);
#ifdef _TRANSLUCENCY
if (!intersection.triangle_front_facing) {
@@ -473,14 +467,22 @@ kernel void raytracingKernel(
}
}
create_basis(n, tangent, binormal);
float3 wo = -ray.direction;
float ndotv = dot(n, wo);
if (ndotv <= 0.0) {
break;
if (ndotv < 1e-3) {
if (i > 0) {
break;
}
n = normalize(n + wo * (1e-3 - ndotv));
float ndotg = dot(n, ng);
if (ndotg < 1e-3) {
n = normalize(n + ng * (1e-3 - ndotg));
}
ndotv = max(dot(n, wo), 1e-3);
}
create_basis(n, tangent, binormal);
float3 albedo = surface_albedo(texcolor, texpaint2.b);
float3 f0 = surface_specular(texcolor, texpaint2.b);
float3 fresnel = f_schlick(f0, ndotv);
base/shaders/raytrace/raytrace_brute_full.metal
+44 -42
View File
@@ -79,41 +79,31 @@ float2 equirect(float3 normal, float angle) {
constant int RANK_CACHE_DIMS = 16;
uint rank_at(int base, int k, texture2d<float, access::read> rank) {
int i = (base + k) & 131071;
return uint(rank.read(uint2(i % 128, uint(i / 128)), 0).r * 255);
uint table_word(texture2d<float, access::read> tex, int i) {
int t = (i & 131071) >> 2;
float4 c = tex.read(uint2(uint(t & 127), uint(t >> 7)), 0);
return uint(c.r * 255) | (uint(c.g * 255) << 8) | (uint(c.b * 255) << 16) | (uint(c.a * 255) << 24);
}
uint table_byte(texture2d<float, access::read> tex, int i) {
return (table_word(tex, i) >> ((i & 3) * 8)) & 255;
}
uint2 init_scramble(uint2 tid, int frame, texture2d<float, access::read> scramble) {
int pixel_i = (int(tid.x) + frame * 9) & 127;
int pixel_j = (int(tid.y) + frame * 11) & 127;
int base = (pixel_i + pixel_j * 128) * 8;
uint lo = 0;
uint hi = 0;
for (int k = 0; k < 4; ++k) {
int i = base + k;
lo |= uint(scramble.read(uint2(i % 128, uint(i / 128)), 0).r * 255) << (k * 8);
i += 4;
hi |= uint(scramble.read(uint2(i % 128, uint(i / 128)), 0).r * 255) << (k * 8);
}
return uint2(lo, hi);
return uint2(table_word(scramble, base), table_word(scramble, base + 4));
}
uint4 init_rank(uint2 tid, int frame, texture2d<float, access::read> rank) {
int pixel_i = (int(tid.x) + frame * 9) & 127;
int pixel_j = (int(tid.y) + frame * 11) & 127;
int base = (pixel_i + pixel_j * 128) * 8;
uint r0 = 0;
uint r1 = 0;
uint r2 = 0;
uint r3 = 0;
for (int k = 0; k < 4; ++k) {
r0 |= rank_at(base, k, rank) << (k * 8);
r1 |= rank_at(base, k + 4, rank) << (k * 8);
r2 |= rank_at(base, k + 8, rank) << (k * 8);
r3 |= rank_at(base, k + 12, rank) << (k * 8);
}
return uint4(r0, r1, r2, r3);
return uint4(table_word(rank, base),
table_word(rank, base + 4),
table_word(rank, base + 8),
table_word(rank, base + 12));
}
float rand(int pixel_i, int pixel_j, int sample_index, int sample_dimension, int frame, uint2 scramble_cache, uint4 rank_cache, texture2d<float, access::read> sobol, texture2d<float, access::read> rank) {
@@ -131,8 +121,8 @@ float rand(int pixel_i, int pixel_j, int sample_index, int sample_dimension, int
ranked_sample_index = sample_index ^ int((packed >> ((sample_dimension & 3) * 8)) & 255);
}
else {
int i = (sample_dimension + (pixel_i + pixel_j * 128) * 8) & 131071;
ranked_sample_index = sample_index ^ int(rank.read(uint2(i % 128, uint(i / 128)), 0).r * 255);
ranked_sample_index = sample_index ^
int(table_byte(rank, sample_dimension + (pixel_i + pixel_j * 128) * 8));
}
int value = int(sobol.read(uint2(ranked_sample_index, sample_dimension), 0).r * 255);
@@ -193,8 +183,8 @@ float2 s16_to_f32(uint val) {
return float2(a, b) / 32767.0f;
}
float3 unpack_position(uint posxy, uint posz_norz) {
return float3(s16_to_f32(posxy), s16_to_f32(posz_norz).x);
float3 srgb_to_linear(float3 c) {
return c * (c * (c * 0.305306011 + 0.682171111) + 0.012522878);
}
float3 hit_world_position(ray ray, intersector_t::result_type intersection) {
@@ -257,9 +247,8 @@ float3 surface_specular(const float3 base_color, const float metalness) {
return mix(float3(0.04, 0.04, 0.04), base_color, metalness);
}
uint2 texel_coord(texture2d<float, access::read> tex, float2 tex_coord) {
uint2 size = uint2(tex.get_width(), tex.get_height());
return uint2(fract(tex_coord) * float2(size));
uint2 texel_coord(float2 tex_coord, float2 tex_size) {
return uint2(fract(tex_coord) * tex_size);
}
kernel void raytracingKernel(
@@ -286,6 +275,8 @@ kernel void raytracingKernel(
int frame = int(constant_buffer.eye.w);
uint2 scramble_cache = init_scramble(tid, frame, mytexture_scramble);
uint4 rank_cache = init_rank(tid, frame, mytexture_rank);
float2 tex_size = float2(mytexture0.get_width(), mytexture0.get_height());
float3 accum = float3(0, 0, 0);
for (int j = 0; j < SAMPLES; ++j) {
@@ -366,18 +357,21 @@ kernel void raytracingKernel(
float2 tex_coord = hit_attribute2d(vertex_uvs, barycentrics) * constant_buffer.params.z;
float3 hit = hit_world_position(ray, intersection);
uint2 texel = texel_coord(mytexture0, tex_coord);
uint2 texel = texel_coord(tex_coord, tex_size);
float4 texpaint0 = mytexture0.read(texel, 0);
float2 zw0 = s16_to_f32(a0.poszw);
float2 zw1 = s16_to_f32(a1.poszw);
float2 zw2 = s16_to_f32(a2.poszw);
float3 vertex_positions[3] = {
unpack_position(a0.posxy, a0.poszw),
unpack_position(a1.posxy, a1.poszw),
unpack_position(a2.posxy, a2.poszw)
float3(s16_to_f32(a0.posxy), zw0.x),
float3(s16_to_f32(a1.posxy), zw1.x),
float3(s16_to_f32(a2.posxy), zw2.x)
};
float3 vertex_normals[3] = {
float3(s16_to_f32(a0.nor), s16_to_f32(a0.poszw).y),
float3(s16_to_f32(a1.nor), s16_to_f32(a1.poszw).y),
float3(s16_to_f32(a2.nor), s16_to_f32(a2.poszw).y)
float3(s16_to_f32(a0.nor), zw0.y),
float3(s16_to_f32(a1.nor), zw1.y),
float3(s16_to_f32(a2.nor), zw2.y)
};
float3 n = normalize(hit_attribute(vertex_normals, barycentrics));
@@ -410,7 +404,7 @@ kernel void raytracingKernel(
if (normal_map) {
texpaint1 = mytexture1.read(texel, 0);
}
float3 texcolor = pow(texpaint0.rgb, float3(2.2, 2.2, 2.2));
float3 texcolor = srgb_to_linear(texpaint0.rgb);
#ifdef _TRANSLUCENCY
if (!intersection.triangle_front_facing) {
@@ -474,14 +468,22 @@ kernel void raytracingKernel(
}
}
create_basis(n, tangent, binormal);
float3 wo = -ray.direction;
float ndotv = dot(n, wo);
if (ndotv <= 0.0) {
break;
if (ndotv < 1e-3) {
if (i > 0) {
break;
}
n = normalize(n + wo * (1e-3 - ndotv));
float ndotg = dot(n, ng);
if (ndotg < 1e-3) {
n = normalize(n + ng * (1e-3 - ndotg));
}
ndotv = max(dot(n, wo), 1e-3);
}
create_basis(n, tangent, binormal);
float3 albedo = surface_albedo(texcolor, texpaint2.b);
float3 f0 = surface_specular(texcolor, texpaint2.b);
float3 fresnel = f_schlick(f0, ndotv);
base/shaders/raytrace/raytrace_brute_multi_core.metal
+44 -42
View File
@@ -79,41 +79,31 @@ float2 equirect(float3 normal, float angle) {
constant int RANK_CACHE_DIMS = 16;
uint rank_at(int base, int k, texture2d<float, access::read> rank) {
int i = (base + k) & 131071;
return uint(rank.read(uint2(i % 128, uint(i / 128)), 0).r * 255);
uint table_word(texture2d<float, access::read> tex, int i) {
int t = (i & 131071) >> 2;
float4 c = tex.read(uint2(uint(t & 127), uint(t >> 7)), 0);
return uint(c.r * 255) | (uint(c.g * 255) << 8) | (uint(c.b * 255) << 16) | (uint(c.a * 255) << 24);
}
uint table_byte(texture2d<float, access::read> tex, int i) {
return (table_word(tex, i) >> ((i & 3) * 8)) & 255;
}
uint2 init_scramble(uint2 tid, int frame, texture2d<float, access::read> scramble) {
int pixel_i = (int(tid.x) + frame * 9) & 127;
int pixel_j = (int(tid.y) + frame * 11) & 127;
int base = (pixel_i + pixel_j * 128) * 8;
uint lo = 0;
uint hi = 0;
for (int k = 0; k < 4; ++k) {
int i = base + k;
lo |= uint(scramble.read(uint2(i % 128, uint(i / 128)), 0).r * 255) << (k * 8);
i += 4;
hi |= uint(scramble.read(uint2(i % 128, uint(i / 128)), 0).r * 255) << (k * 8);
}
return uint2(lo, hi);
return uint2(table_word(scramble, base), table_word(scramble, base + 4));
}
uint4 init_rank(uint2 tid, int frame, texture2d<float, access::read> rank) {
int pixel_i = (int(tid.x) + frame * 9) & 127;
int pixel_j = (int(tid.y) + frame * 11) & 127;
int base = (pixel_i + pixel_j * 128) * 8;
uint r0 = 0;
uint r1 = 0;
uint r2 = 0;
uint r3 = 0;
for (int k = 0; k < 4; ++k) {
r0 |= rank_at(base, k, rank) << (k * 8);
r1 |= rank_at(base, k + 4, rank) << (k * 8);
r2 |= rank_at(base, k + 8, rank) << (k * 8);
r3 |= rank_at(base, k + 12, rank) << (k * 8);
}
return uint4(r0, r1, r2, r3);
return uint4(table_word(rank, base),
table_word(rank, base + 4),
table_word(rank, base + 8),
table_word(rank, base + 12));
}
float rand(int pixel_i, int pixel_j, int sample_index, int sample_dimension, int frame, uint2 scramble_cache, uint4 rank_cache, texture2d<float, access::read> sobol, texture2d<float, access::read> rank) {
@@ -131,8 +121,8 @@ float rand(int pixel_i, int pixel_j, int sample_index, int sample_dimension, int
ranked_sample_index = sample_index ^ int((packed >> ((sample_dimension & 3) * 8)) & 255);
}
else {
int i = (sample_dimension + (pixel_i + pixel_j * 128) * 8) & 131071;
ranked_sample_index = sample_index ^ int(rank.read(uint2(i % 128, uint(i / 128)), 0).r * 255);
ranked_sample_index = sample_index ^
int(table_byte(rank, sample_dimension + (pixel_i + pixel_j * 128) * 8));
}
int value = int(sobol.read(uint2(ranked_sample_index, sample_dimension), 0).r * 255);
@@ -193,8 +183,8 @@ float2 s16_to_f32(uint val) {
return float2(a, b) / 32767.0f;
}
float3 unpack_position(uint posxy, uint posz_norz) {
return float3(s16_to_f32(posxy), s16_to_f32(posz_norz).x);
float3 srgb_to_linear(float3 c) {
return c * (c * (c * 0.305306011 + 0.682171111) + 0.012522878);
}
float3 hit_world_position(ray ray, intersector_t::result_type intersection) {
@@ -257,9 +247,8 @@ float3 surface_specular(const float3 base_color, const float metalness) {
return mix(float3(0.04, 0.04, 0.04), base_color, metalness);
}
uint2 texel_coord(texture2d<float, access::read> tex, float2 tex_coord) {
uint2 size = uint2(tex.get_width(), tex.get_height());
return uint2(fract(tex_coord) * float2(size));
uint2 texel_coord(float2 tex_coord, float2 tex_size) {
return uint2(fract(tex_coord) * tex_size);
}
kernel void raytracingKernel(
@@ -286,6 +275,8 @@ kernel void raytracingKernel(
int frame = int(constant_buffer.eye.w);
uint2 scramble_cache = init_scramble(tid, frame, mytexture_scramble);
uint4 rank_cache = init_rank(tid, frame, mytexture_rank);
float2 tex_size = float2(mytexture0.get_width(), mytexture0.get_height());
float3 accum = float3(0, 0, 0);
for (int j = 0; j < SAMPLES; ++j) {
@@ -366,18 +357,21 @@ kernel void raytracingKernel(
float2 tex_coord = hit_attribute2d(vertex_uvs, barycentrics) * constant_buffer.params.z;
float3 hit = hit_world_position(ray, intersection);
uint2 texel = texel_coord(mytexture0, tex_coord);
uint2 texel = texel_coord(tex_coord, tex_size);
float4 texpaint0 = mytexture0.read(texel, 0);
float2 zw0 = s16_to_f32(a0.poszw);
float2 zw1 = s16_to_f32(a1.poszw);
float2 zw2 = s16_to_f32(a2.poszw);
float3 vertex_positions[3] = {
unpack_position(a0.posxy, a0.poszw),
unpack_position(a1.posxy, a1.poszw),
unpack_position(a2.posxy, a2.poszw)
float3(s16_to_f32(a0.posxy), zw0.x),
float3(s16_to_f32(a1.posxy), zw1.x),
float3(s16_to_f32(a2.posxy), zw2.x)
};
float3 vertex_normals[3] = {
float3(s16_to_f32(a0.nor), s16_to_f32(a0.poszw).y),
float3(s16_to_f32(a1.nor), s16_to_f32(a1.poszw).y),
float3(s16_to_f32(a2.nor), s16_to_f32(a2.poszw).y)
float3(s16_to_f32(a0.nor), zw0.y),
float3(s16_to_f32(a1.nor), zw1.y),
float3(s16_to_f32(a2.nor), zw2.y)
};
float3 n = normalize(hit_attribute(vertex_normals, barycentrics));
@@ -410,7 +404,7 @@ kernel void raytracingKernel(
if (normal_map) {
texpaint1 = mytexture1.read(texel, 0);
}
float3 texcolor = pow(texpaint0.rgb, float3(2.2, 2.2, 2.2));
float3 texcolor = srgb_to_linear(texpaint0.rgb);
#ifdef _TRANSLUCENCY
if (!intersection.triangle_front_facing) {
@@ -474,14 +468,22 @@ kernel void raytracingKernel(
}
}
create_basis(n, tangent, binormal);
float3 wo = -ray.direction;
float ndotv = dot(n, wo);
if (ndotv <= 0.0) {
break;
if (ndotv < 1e-3) {
if (i > 0) {
break;
}
n = normalize(n + wo * (1e-3 - ndotv));
float ndotg = dot(n, ng);
if (ndotg < 1e-3) {
n = normalize(n + ng * (1e-3 - ndotg));
}
ndotv = max(dot(n, wo), 1e-3);
}
create_basis(n, tangent, binormal);
float3 albedo = surface_albedo(texcolor, texpaint2.b);
float3 f0 = surface_specular(texcolor, texpaint2.b);
float3 fresnel = f_schlick(f0, ndotv);
base/shaders/raytrace/raytrace_brute_multi_full.metal
+44 -42
View File
@@ -80,41 +80,31 @@ float2 equirect(float3 normal, float angle) {
constant int RANK_CACHE_DIMS = 16;
uint rank_at(int base, int k, texture2d<float, access::read> rank) {
int i = (base + k) & 131071;
return uint(rank.read(uint2(i % 128, uint(i / 128)), 0).r * 255);
uint table_word(texture2d<float, access::read> tex, int i) {
int t = (i & 131071) >> 2;
float4 c = tex.read(uint2(uint(t & 127), uint(t >> 7)), 0);
return uint(c.r * 255) | (uint(c.g * 255) << 8) | (uint(c.b * 255) << 16) | (uint(c.a * 255) << 24);
}
uint table_byte(texture2d<float, access::read> tex, int i) {
return (table_word(tex, i) >> ((i & 3) * 8)) & 255;
}
uint2 init_scramble(uint2 tid, int frame, texture2d<float, access::read> scramble) {
int pixel_i = (int(tid.x) + frame * 9) & 127;
int pixel_j = (int(tid.y) + frame * 11) & 127;
int base = (pixel_i + pixel_j * 128) * 8;
uint lo = 0;
uint hi = 0;
for (int k = 0; k < 4; ++k) {
int i = base + k;
lo |= uint(scramble.read(uint2(i % 128, uint(i / 128)), 0).r * 255) << (k * 8);
i += 4;
hi |= uint(scramble.read(uint2(i % 128, uint(i / 128)), 0).r * 255) << (k * 8);
}
return uint2(lo, hi);
return uint2(table_word(scramble, base), table_word(scramble, base + 4));
}
uint4 init_rank(uint2 tid, int frame, texture2d<float, access::read> rank) {
int pixel_i = (int(tid.x) + frame * 9) & 127;
int pixel_j = (int(tid.y) + frame * 11) & 127;
int base = (pixel_i + pixel_j * 128) * 8;
uint r0 = 0;
uint r1 = 0;
uint r2 = 0;
uint r3 = 0;
for (int k = 0; k < 4; ++k) {
r0 |= rank_at(base, k, rank) << (k * 8);
r1 |= rank_at(base, k + 4, rank) << (k * 8);
r2 |= rank_at(base, k + 8, rank) << (k * 8);
r3 |= rank_at(base, k + 12, rank) << (k * 8);
}
return uint4(r0, r1, r2, r3);
return uint4(table_word(rank, base),
table_word(rank, base + 4),
table_word(rank, base + 8),
table_word(rank, base + 12));
}
float rand(int pixel_i, int pixel_j, int sample_index, int sample_dimension, int frame, uint2 scramble_cache, uint4 rank_cache, texture2d<float, access::read> sobol, texture2d<float, access::read> rank) {
@@ -132,8 +122,8 @@ float rand(int pixel_i, int pixel_j, int sample_index, int sample_dimension, int
ranked_sample_index = sample_index ^ int((packed >> ((sample_dimension & 3) * 8)) & 255);
}
else {
int i = (sample_dimension + (pixel_i + pixel_j * 128) * 8) & 131071;
ranked_sample_index = sample_index ^ int(rank.read(uint2(i % 128, uint(i / 128)), 0).r * 255);
ranked_sample_index = sample_index ^
int(table_byte(rank, sample_dimension + (pixel_i + pixel_j * 128) * 8));
}
int value = int(sobol.read(uint2(ranked_sample_index, sample_dimension), 0).r * 255);
@@ -194,8 +184,8 @@ float2 s16_to_f32(uint val) {
return float2(a, b) / 32767.0f;
}
float3 unpack_position(uint posxy, uint posz_norz) {
return float3(s16_to_f32(posxy), s16_to_f32(posz_norz).x);
float3 srgb_to_linear(float3 c) {
return c * (c * (c * 0.305306011 + 0.682171111) + 0.012522878);
}
float3 hit_world_position(ray ray, intersector_t::result_type intersection) {
@@ -258,9 +248,8 @@ float3 surface_specular(const float3 base_color, const float metalness) {
return mix(float3(0.04, 0.04, 0.04), base_color, metalness);
}
uint2 texel_coord(texture2d<float, access::read> tex, float2 tex_coord) {
uint2 size = uint2(tex.get_width(), tex.get_height());
return uint2(fract(tex_coord) * float2(size));
uint2 texel_coord(float2 tex_coord, float2 tex_size) {
return uint2(fract(tex_coord) * tex_size);
}
kernel void raytracingKernel(
@@ -287,6 +276,8 @@ kernel void raytracingKernel(
int frame = int(constant_buffer.eye.w);
uint2 scramble_cache = init_scramble(tid, frame, mytexture_scramble);
uint4 rank_cache = init_rank(tid, frame, mytexture_rank);
float2 tex_size = float2(mytexture0.get_width(), mytexture0.get_height());
float3 accum = float3(0, 0, 0);
for (int j = 0; j < SAMPLES; ++j) {
@@ -367,18 +358,21 @@ kernel void raytracingKernel(
float2 tex_coord = hit_attribute2d(vertex_uvs, barycentrics) * constant_buffer.params.z;
float3 hit = hit_world_position(ray, intersection);
uint2 texel = texel_coord(mytexture0, tex_coord);
uint2 texel = texel_coord(tex_coord, tex_size);
float4 texpaint0 = mytexture0.read(texel, 0);
float2 zw0 = s16_to_f32(a0.poszw);
float2 zw1 = s16_to_f32(a1.poszw);
float2 zw2 = s16_to_f32(a2.poszw);
float3 vertex_positions[3] = {
unpack_position(a0.posxy, a0.poszw),
unpack_position(a1.posxy, a1.poszw),
unpack_position(a2.posxy, a2.poszw)
float3(s16_to_f32(a0.posxy), zw0.x),
float3(s16_to_f32(a1.posxy), zw1.x),
float3(s16_to_f32(a2.posxy), zw2.x)
};
float3 vertex_normals[3] = {
float3(s16_to_f32(a0.nor), s16_to_f32(a0.poszw).y),
float3(s16_to_f32(a1.nor), s16_to_f32(a1.poszw).y),
float3(s16_to_f32(a2.nor), s16_to_f32(a2.poszw).y)
float3(s16_to_f32(a0.nor), zw0.y),
float3(s16_to_f32(a1.nor), zw1.y),
float3(s16_to_f32(a2.nor), zw2.y)
};
float3 n = normalize(hit_attribute(vertex_normals, barycentrics));
@@ -411,7 +405,7 @@ kernel void raytracingKernel(
if (normal_map) {
texpaint1 = mytexture1.read(texel, 0);
}
float3 texcolor = pow(texpaint0.rgb, float3(2.2, 2.2, 2.2));
float3 texcolor = srgb_to_linear(texpaint0.rgb);
#ifdef _TRANSLUCENCY
if (!intersection.triangle_front_facing) {
@@ -475,14 +469,22 @@ kernel void raytracingKernel(
}
}
create_basis(n, tangent, binormal);
float3 wo = -ray.direction;
float ndotv = dot(n, wo);
if (ndotv <= 0.0) {
break;
if (ndotv < 1e-3) {
if (i > 0) {
break;
}
n = normalize(n + wo * (1e-3 - ndotv));
float ndotg = dot(n, ng);
if (ndotg < 1e-3) {
n = normalize(n + ng * (1e-3 - ndotg));
}
ndotv = max(dot(n, wo), 1e-3);
}
create_basis(n, tangent, binormal);
float3 albedo = surface_albedo(texcolor, texpaint2.b);
float3 f0 = surface_specular(texcolor, texpaint2.b);
float3 fresnel = f_schlick(f0, ndotv);