moonpower/P3DS-test
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

rework the lut system

Browse Source
This commit is contained in:
Samuliak
2024-10-31 09:25:01 +01:00
parent 90420160f2
commit 158be432fc
9 changed files with 273 additions and 152 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
src/host_shaders/metal_copy_to_lut_texture.metal
View File

@@ -4,6 +4,6 @@ using namespace metal;
constant ushort lutTextureWidth [[function_constant(0)]];
// The copy is done in a vertex shader instead of a compute kernel, since dispatching compute would require ending the render pass
vertex void vertexCopyToLutTexture(uint vid [[vertex_id]], texture2d<float, access::write> out [[texture(0)]], constant float2* data [[buffer(0)]], constant uint& arrayOffset [[buffer(1)]]) {
vertex void vertexCopyToLutTexture(uint vid [[vertex_id]], texture2d<float, access::write> out [[texture(0)]], device float2* data [[buffer(0)]], constant uint& arrayOffset [[buffer(1)]]) {
out.write(float4(data[vid], 0.0, 0.0), uint2(vid % lutTextureWidth, arrayOffset + vid / lutTextureWidth));
}

58
src/host_shaders/metal_shaders.metal
View File

@@ -406,13 +406,11 @@ uint4 performLogicOpU(LogicOp logicOp, uint4 s, uint4 d) {
#define RG_LUT 5u
#define RR_LUT 6u
#define FOG_INDEX 24
float lutLookup(texture2d<float> texLut, uint lut, uint index) {
return texLut.read(uint2(index, lut)).r;
float lutLookup(texture2d_array<float> texLut, uint slice, uint lut, uint index) {
return texLut.read(uint2(index, lut), slice).r;
}
float lightLutLookup(thread Globals& globals, thread DrawVertexOut& in, constant PicaRegs& picaRegs, texture2d<float> texLut, uint environment_id, uint lut_id, uint light_id, float3 light_vector, float3 half_vector) {
float lightLutLookup(thread Globals& globals, thread DrawVertexOut& in, constant PicaRegs& picaRegs, texture2d_array<float> texLut, uint slice, uint environment_id, uint lut_id, uint light_id, float3 light_vector, float3 half_vector) {
uint lut_index;
int bit_in_config1;
if (lut_id == SP_LUT) {
@@ -498,12 +496,12 @@ float lightLutLookup(thread Globals& globals, thread DrawVertexOut& in, constant
delta = abs(delta);
}
int index = int(clamp(floor(delta * 255.0), 0.f, 255.f));
return lutLookup(texLut, lut_index, index) * scale;
return lutLookup(texLut, slice, lut_index, index) * scale;
} else {
// Range is [-1, 1] so we need to map it to [0, 1]
int index = int(clamp(floor(delta * 128.0), -128.f, 127.f));
if (index < 0) index += 256;
return lutLookup(texLut, lut_index, index) * scale;
return lutLookup(texLut, slice, lut_index, index) * scale;
}
}
@@ -515,7 +513,7 @@ float3 regToColor(uint reg) {
}
// Implements the following algorthm: https://mathb.in/26766
void calcLighting(thread Globals& globals, thread DrawVertexOut& in, constant PicaRegs& picaRegs, texture2d<float> texLut, sampler linearSampler, thread float4& primaryColor, thread float4& secondaryColor) {
void calcLighting(thread Globals& globals, thread DrawVertexOut& in, constant PicaRegs& picaRegs, texture2d_array<float> texLut, uint slice, sampler linearSampler, thread float4& primaryColor, thread float4& secondaryColor) {
// Quaternions describe a transformation from surface-local space to eye space.
// In surface-local space, by definition (and up to permutation) the normal vector is (0,0,1),
// the tangent vector is (1,0,0), and the bitangent vector is (0,1,0).
@@ -566,10 +564,10 @@ void calcLighting(thread Globals& globals, thread DrawVertexOut& in, constant Pi
globals.GPUREG_LIGHTi_CONFIG = picaRegs.read(0x0149u + (lightId << 4u));
float lightDistance;
float3 lightPosition = normalize(float3(
float3 lightPosition = float3(
decodeFP(extract_bits(GPUREG_LIGHTi_VECTOR_LOW, 0, 16), 5u, 10u), decodeFP(extract_bits(GPUREG_LIGHTi_VECTOR_LOW, 16, 16), 5u, 10u),
decodeFP(extract_bits(GPUREG_LIGHTi_VECTOR_HIGH, 0, 16), 5u, 10u)
));
);
// Positional Light
if (extract_bits(globals.GPUREG_LIGHTi_CONFIG, 0, 1) == 0u) {
@@ -613,23 +611,23 @@ void calcLighting(thread Globals& globals, thread DrawVertexOut& in, constant Pi
float delta = lightDistance * distanceAttenuationScale + distanceAttenuationBias;
delta = clamp(delta, 0.0, 1.0);
int index = int(clamp(floor(delta * 255.0), 0.0, 255.0));
distanceAttenuation = lutLookup(texLut, 16u + lightId, index);
distanceAttenuation = lutLookup(texLut, slice, 16u + lightId, index);
}
float spotlightAttenuation = lightLutLookup(globals, in, picaRegs, texLut, environmentId, SP_LUT, lightId, lightVector, halfVector);
float specular0Distribution = lightLutLookup(globals, in, picaRegs, texLut, environmentId, D0_LUT, lightId, lightVector, halfVector);
float specular1Distribution = lightLutLookup(globals, in, picaRegs, texLut, environmentId, D1_LUT, lightId, lightVector, halfVector);
float spotlightAttenuation = lightLutLookup(globals, in, picaRegs, texLut, slice, environmentId, SP_LUT, lightId, lightVector, halfVector);
float specular0Distribution = lightLutLookup(globals, in, picaRegs, texLut, slice, environmentId, D0_LUT, lightId, lightVector, halfVector);
float specular1Distribution = lightLutLookup(globals, in, picaRegs, texLut, slice, environmentId, D1_LUT, lightId, lightVector, halfVector);
float3 reflectedColor;
reflectedColor.r = lightLutLookup(globals, in, picaRegs, texLut, environmentId, RR_LUT, lightId, lightVector, halfVector);
reflectedColor.r = lightLutLookup(globals, in, picaRegs, texLut, slice, environmentId, RR_LUT, lightId, lightVector, halfVector);
if (isSamplerEnabled(environmentId, RG_LUT)) {
reflectedColor.g = lightLutLookup(globals, in, picaRegs, texLut, environmentId, RG_LUT, lightId, lightVector, halfVector);
reflectedColor.g = lightLutLookup(globals, in, picaRegs, texLut, slice, environmentId, RG_LUT, lightId, lightVector, halfVector);
} else {
reflectedColor.g = reflectedColor.r;
}
if (isSamplerEnabled(environmentId, RB_LUT)) {
reflectedColor.b = lightLutLookup(globals, in, picaRegs, texLut, environmentId, RB_LUT, lightId, lightVector, halfVector);
reflectedColor.b = lightLutLookup(globals, in, picaRegs, texLut, slice, environmentId, RB_LUT, lightId, lightVector, halfVector);
} else {
reflectedColor.b = reflectedColor.r;
}
@@ -655,7 +653,7 @@ void calcLighting(thread Globals& globals, thread DrawVertexOut& in, constant Pi
float fresnelFactor;
if (fresnelOutput1 == 1u || fresnelOutput2 == 1u) {
fresnelFactor = lightLutLookup(globals, in, picaRegs, texLut, environmentId, FR_LUT, lightId, lightVector, halfVector);
fresnelFactor = lightLutLookup(globals, in, picaRegs, texLut, slice, environmentId, FR_LUT, lightId, lightVector, halfVector);
}
if (fresnelOutput1 == 1u) {
@@ -676,9 +674,7 @@ float4 performLogicOp(LogicOp logicOp, float4 s, float4 d) {
return as_type<float4>(performLogicOpU(logicOp, as_type<uint4>(s), as_type<uint4>(d)));
}
fragment float4 fragmentDraw(DrawVertexOut in [[stage_in]], float4 prevColor [[color(0)]], constant PicaRegs& picaRegs [[buffer(0)]], constant FragTEV& tev [[buffer(1)]], constant LogicOp& logicOp [[buffer(2)]],
texture2d<float> tex0 [[texture(0)]], texture2d<float> tex1 [[texture(1)]], texture2d<float> tex2 [[texture(2)]], texture2d<float> texLut [[texture(3)]],
sampler samplr0 [[sampler(0)]], sampler samplr1 [[sampler(1)]], sampler samplr2 [[sampler(2)]], sampler linearSampler [[sampler(3)]]) {
fragment float4 fragmentDraw(DrawVertexOut in [[stage_in]], float4 prevColor [[color(0)]], constant PicaRegs& picaRegs [[buffer(0)]], constant FragTEV& tev [[buffer(1)]], constant LogicOp& logicOp [[buffer(2)]], constant uint2& lutSlices [[buffer(3)]], texture2d<float> tex0 [[texture(0)]], texture2d<float> tex1 [[texture(1)]], texture2d<float> tex2 [[texture(2)]], texture2d_array<float> texLightingLut [[texture(3)]], texture1d_array<float> texFogLut [[texture(4)]], sampler samplr0 [[sampler(0)]], sampler samplr1 [[sampler(1)]], sampler samplr2 [[sampler(2)]], sampler linearSampler [[sampler(3)]]) {
Globals globals;
// HACK
@@ -689,7 +685,7 @@ fragment float4 fragmentDraw(DrawVertexOut in [[stage_in]], float4 prevColor [[c
globals.tevSources[0] = in.color;
if (lightingEnabled) {
calcLighting(globals, in, picaRegs, texLut, linearSampler, globals.tevSources[1], globals.tevSources[2]);
calcLighting(globals, in, picaRegs, texLightingLut, lutSlices.x, linearSampler, globals.tevSources[1], globals.tevSources[2]);
} else {
globals.tevSources[1] = float4(0.0);
globals.tevSources[2] = float4(0.0);
@@ -729,13 +725,13 @@ fragment float4 fragmentDraw(DrawVertexOut in [[stage_in]], float4 prevColor [[c
bool enable_fog = (textureEnvUpdateBuffer & 7u) == 5u;
if (enable_fog) {
bool flip_depth = (textureEnvUpdateBuffer & (1u << 16)) != 0u;
float fog_index = flip_depth ? 1.0 - in.position.z : in.position.z;
fog_index *= 128.0;
float clamped_index = clamp(floor(fog_index), 0.0, 127.0);
float delta = fog_index - clamped_index;
float2 value = texLut.read(uint2(clamped_index, FOG_INDEX)).rg;
float fog_factor = clamp(value.r + value.g * delta, 0.0, 1.0);
bool flipDepth = (textureEnvUpdateBuffer & (1u << 16)) != 0u;
float fogIndex = flipDepth ? 1.0 - in.position.z : in.position.z;
fogIndex *= 128.0;
float clampedIndex = clamp(floor(fogIndex), 0.0, 127.0);
float delta = fogIndex - clampedIndex;
float2 value = texFogLut.read(clampedIndex, lutSlices.y).rg;
float fogFactor = clamp(value.r + value.g * delta, 0.0, 1.0);
uint GPUREG_FOG_COLOR = picaRegs.read(0x00E1u);
@@ -743,9 +739,9 @@ fragment float4 fragmentDraw(DrawVertexOut in [[stage_in]], float4 prevColor [[c
float r = (GPUREG_FOG_COLOR & 0xFFu) / 255.0;
float g = ((GPUREG_FOG_COLOR >> 8) & 0xFFu) / 255.0;
float b = ((GPUREG_FOG_COLOR >> 16) & 0xFFu) / 255.0;
float3 fog_color = float3(r, g, b);
float3 fogColor = float3(r, g, b);
color.rgb = mix(fog_color, color.rgb, fog_factor);
color.rgb = mix(fogColor, color.rgb, fogFactor);
}
// Perform alpha test