summaryrefslogtreecommitdiffstats
path: root/src/video_core/swrasterizer
diff options
context:
space:
mode:
Diffstat (limited to 'src/video_core/swrasterizer')
-rw-r--r--src/video_core/swrasterizer/clipper.cpp31
-rw-r--r--src/video_core/swrasterizer/lighting.cpp46
-rw-r--r--src/video_core/swrasterizer/lighting.h3
-rw-r--r--src/video_core/swrasterizer/rasterizer.cpp4
4 files changed, 63 insertions, 21 deletions
diff --git a/src/video_core/swrasterizer/clipper.cpp b/src/video_core/swrasterizer/clipper.cpp
index cdbc71502..a52129eb7 100644
--- a/src/video_core/swrasterizer/clipper.cpp
+++ b/src/video_core/swrasterizer/clipper.cpp
@@ -31,7 +31,7 @@ public:
: coeffs(coeffs), bias(bias) {}
bool IsInside(const Vertex& vertex) const {
- return Math::Dot(vertex.pos + bias, coeffs) <= float24::FromFloat32(0);
+ return Math::Dot(vertex.pos + bias, coeffs) >= float24::FromFloat32(0);
}
bool IsOutSide(const Vertex& vertex) const {
@@ -116,19 +116,18 @@ void ProcessTriangle(const OutputVertex& v0, const OutputVertex& v1, const Outpu
static const float24 f0 = float24::FromFloat32(0.0);
static const float24 f1 = float24::FromFloat32(1.0);
static const std::array<ClippingEdge, 7> clipping_edges = {{
- {Math::MakeVec(f1, f0, f0, -f1)}, // x = +w
- {Math::MakeVec(-f1, f0, f0, -f1)}, // x = -w
- {Math::MakeVec(f0, f1, f0, -f1)}, // y = +w
- {Math::MakeVec(f0, -f1, f0, -f1)}, // y = -w
- {Math::MakeVec(f0, f0, f1, f0)}, // z = 0
- {Math::MakeVec(f0, f0, -f1, -f1)}, // z = -w
- {Math::MakeVec(f0, f0, f0, -f1), Math::Vec4<float24>(f0, f0, f0, EPSILON)}, // w = EPSILON
+ {Math::MakeVec(-f1, f0, f0, f1)}, // x = +w
+ {Math::MakeVec(f1, f0, f0, f1)}, // x = -w
+ {Math::MakeVec(f0, -f1, f0, f1)}, // y = +w
+ {Math::MakeVec(f0, f1, f0, f1)}, // y = -w
+ {Math::MakeVec(f0, f0, -f1, f0)}, // z = 0
+ {Math::MakeVec(f0, f0, f1, f1)}, // z = -w
+ {Math::MakeVec(f0, f0, f0, f1), Math::Vec4<float24>(f0, f0, f0, EPSILON)}, // w = EPSILON
}};
// Simple implementation of the Sutherland-Hodgman clipping algorithm.
// TODO: Make this less inefficient (currently lots of useless buffering overhead happens here)
- for (auto edge : clipping_edges) {
-
+ auto Clip = [&](const ClippingEdge& edge) {
std::swap(input_list, output_list);
output_list->clear();
@@ -147,12 +146,24 @@ void ProcessTriangle(const OutputVertex& v0, const OutputVertex& v1, const Outpu
}
reference_vertex = &vertex;
}
+ };
+
+ for (auto edge : clipping_edges) {
+ Clip(edge);
// Need to have at least a full triangle to continue...
if (output_list->size() < 3)
return;
}
+ if (g_state.regs.rasterizer.clip_enable) {
+ ClippingEdge custom_edge{g_state.regs.rasterizer.GetClipCoef()};
+ Clip(custom_edge);
+
+ if (output_list->size() < 3)
+ return;
+ }
+
InitScreenCoordinates((*output_list)[0]);
InitScreenCoordinates((*output_list)[1]);
diff --git a/src/video_core/swrasterizer/lighting.cpp b/src/video_core/swrasterizer/lighting.cpp
index 39a3e396d..5fa748611 100644
--- a/src/video_core/swrasterizer/lighting.cpp
+++ b/src/video_core/swrasterizer/lighting.cpp
@@ -22,18 +22,37 @@ static float LookupLightingLut(const Pica::State::Lighting& lighting, size_t lut
std::tuple<Math::Vec4<u8>, Math::Vec4<u8>> ComputeFragmentsColors(
const Pica::LightingRegs& lighting, const Pica::State::Lighting& lighting_state,
- const Math::Quaternion<float>& normquat, const Math::Vec3<float>& view) {
+ const Math::Quaternion<float>& normquat, const Math::Vec3<float>& view,
+ const Math::Vec4<u8> (&texture_color)[4]) {
- // TODO(Subv): Bump mapping
- Math::Vec3<float> surface_normal = {0.0f, 0.0f, 1.0f};
+ Math::Vec3<float> surface_normal;
+ Math::Vec3<float> surface_tangent;
if (lighting.config0.bump_mode != LightingRegs::LightingBumpMode::None) {
- LOG_CRITICAL(HW_GPU, "unimplemented bump mapping");
- UNIMPLEMENTED();
+ Math::Vec3<float> perturbation =
+ texture_color[lighting.config0.bump_selector].xyz().Cast<float>() / 127.5f -
+ Math::MakeVec(1.0f, 1.0f, 1.0f);
+ if (lighting.config0.bump_mode == LightingRegs::LightingBumpMode::NormalMap) {
+ if (!lighting.config0.disable_bump_renorm) {
+ const float z_square = 1 - perturbation.xy().Length2();
+ perturbation.z = std::sqrt(std::max(z_square, 0.0f));
+ }
+ surface_normal = perturbation;
+ surface_tangent = Math::MakeVec(1.0f, 0.0f, 0.0f);
+ } else if (lighting.config0.bump_mode == LightingRegs::LightingBumpMode::TangentMap) {
+ surface_normal = Math::MakeVec(0.0f, 0.0f, 1.0f);
+ surface_tangent = perturbation;
+ } else {
+ LOG_ERROR(HW_GPU, "Unknown bump mode %u", lighting.config0.bump_mode.Value());
+ }
+ } else {
+ surface_normal = Math::MakeVec(0.0f, 0.0f, 1.0f);
+ surface_tangent = Math::MakeVec(1.0f, 0.0f, 0.0f);
}
// Use the normalized the quaternion when performing the rotation
auto normal = Math::QuaternionRotate(normquat, surface_normal);
+ auto tangent = Math::QuaternionRotate(normquat, surface_tangent);
Math::Vec4<float> diffuse_sum = {0.0f, 0.0f, 0.0f, 1.0f};
Math::Vec4<float> specular_sum = {0.0f, 0.0f, 0.0f, 1.0f};
@@ -102,6 +121,16 @@ std::tuple<Math::Vec4<u8>, Math::Vec4<u8>> ComputeFragmentsColors(
result = Math::Dot(light_vector, spot_dir.Cast<float>() / 2047.0f);
break;
}
+ case LightingRegs::LightingLutInput::CP:
+ if (lighting.config0.config == LightingRegs::LightingConfig::Config7) {
+ const Math::Vec3<float> norm_half_vector = half_vector.Normalized();
+ const Math::Vec3<float> half_vector_proj =
+ norm_half_vector - normal * Math::Dot(normal, norm_half_vector);
+ result = Math::Dot(half_vector_proj, tangent);
+ } else {
+ result = 0.0f;
+ }
+ break;
default:
LOG_CRITICAL(HW_GPU, "Unknown lighting LUT input %u\n", static_cast<u32>(input));
UNIMPLEMENTED();
@@ -201,7 +230,8 @@ std::tuple<Math::Vec4<u8>, Math::Vec4<u8>> ComputeFragmentsColors(
d1_lut_value * refl_value * light_config.specular_1.ToVec3f();
// Fresnel
- if (lighting.config1.disable_lut_fr == 0 &&
+ // Note: only the last entry in the light slots applies the Fresnel factor
+ if (light_index == lighting.max_light_index && lighting.config1.disable_lut_fr == 0 &&
LightingRegs::IsLightingSamplerSupported(lighting.config0.config,
LightingRegs::LightingSampler::Fresnel)) {
@@ -213,14 +243,14 @@ std::tuple<Math::Vec4<u8>, Math::Vec4<u8>> ComputeFragmentsColors(
if (lighting.config0.fresnel_selector ==
LightingRegs::LightingFresnelSelector::PrimaryAlpha ||
lighting.config0.fresnel_selector == LightingRegs::LightingFresnelSelector::Both) {
- diffuse_sum.a() *= lut_value;
+ diffuse_sum.a() = lut_value;
}
// Enabled for the specular lighting alpha component
if (lighting.config0.fresnel_selector ==
LightingRegs::LightingFresnelSelector::SecondaryAlpha ||
lighting.config0.fresnel_selector == LightingRegs::LightingFresnelSelector::Both) {
- specular_sum.a() *= lut_value;
+ specular_sum.a() = lut_value;
}
}
diff --git a/src/video_core/swrasterizer/lighting.h b/src/video_core/swrasterizer/lighting.h
index 438dca926..d807a3d94 100644
--- a/src/video_core/swrasterizer/lighting.h
+++ b/src/video_core/swrasterizer/lighting.h
@@ -13,6 +13,7 @@ namespace Pica {
std::tuple<Math::Vec4<u8>, Math::Vec4<u8>> ComputeFragmentsColors(
const Pica::LightingRegs& lighting, const Pica::State::Lighting& lighting_state,
- const Math::Quaternion<float>& normquat, const Math::Vec3<float>& view);
+ const Math::Quaternion<float>& normquat, const Math::Vec3<float>& view,
+ const Math::Vec4<u8> (&texture_color)[4]);
} // namespace Pica
diff --git a/src/video_core/swrasterizer/rasterizer.cpp b/src/video_core/swrasterizer/rasterizer.cpp
index fdc1df199..862135614 100644
--- a/src/video_core/swrasterizer/rasterizer.cpp
+++ b/src/video_core/swrasterizer/rasterizer.cpp
@@ -437,8 +437,8 @@ static void ProcessTriangleInternal(const Vertex& v0, const Vertex& v1, const Ve
GetInterpolatedAttribute(v0.view.y, v1.view.y, v2.view.y).ToFloat32(),
GetInterpolatedAttribute(v0.view.z, v1.view.z, v2.view.z).ToFloat32(),
};
- std::tie(primary_fragment_color, secondary_fragment_color) =
- ComputeFragmentsColors(g_state.regs.lighting, g_state.lighting, normquat, view);
+ std::tie(primary_fragment_color, secondary_fragment_color) = ComputeFragmentsColors(
+ g_state.regs.lighting, g_state.lighting, normquat, view, texture_color);
}
for (unsigned tev_stage_index = 0; tev_stage_index < tev_stages.size();