citra-emu
/
citra
Archived
1
0
Fork 0

VideoCore: Extract swrast-specific data from OutputVertex

This commit is contained in:
Yuri Kunde Schlesner 2016-12-18 23:43:37 -08:00
parent 8ed9f9d49f
commit dcdffabfe6
5 changed files with 64 additions and 58 deletions

View File

@ -18,6 +18,8 @@
#include "video_core/rasterizer.h" #include "video_core/rasterizer.h"
#include "video_core/shader/shader.h" #include "video_core/shader/shader.h"
using Pica::Rasterizer::Vertex;
namespace Pica { namespace Pica {
namespace Clipper { namespace Clipper {
@ -29,20 +31,20 @@ public:
float24::FromFloat32(0), float24::FromFloat32(0))) float24::FromFloat32(0), float24::FromFloat32(0)))
: coeffs(coeffs), bias(bias) {} : coeffs(coeffs), bias(bias) {}
bool IsInside(const OutputVertex& vertex) const { 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 OutputVertex& vertex) const { bool IsOutSide(const Vertex& vertex) const {
return !IsInside(vertex); return !IsInside(vertex);
} }
OutputVertex GetIntersection(const OutputVertex& v0, const OutputVertex& v1) const { Vertex GetIntersection(const Vertex& v0, const Vertex& v1) const {
float24 dp = Math::Dot(v0.pos + bias, coeffs); float24 dp = Math::Dot(v0.pos + bias, coeffs);
float24 dp_prev = Math::Dot(v1.pos + bias, coeffs); float24 dp_prev = Math::Dot(v1.pos + bias, coeffs);
float24 factor = dp_prev / (dp_prev - dp); float24 factor = dp_prev / (dp_prev - dp);
return OutputVertex::Lerp(factor, v0, v1); return Vertex::Lerp(factor, v0, v1);
} }
private: private:
@ -51,7 +53,7 @@ private:
Math::Vec4<float24> bias; Math::Vec4<float24> bias;
}; };
static void InitScreenCoordinates(OutputVertex& vtx) { static void InitScreenCoordinates(Vertex& vtx) {
struct { struct {
float24 halfsize_x; float24 halfsize_x;
float24 offset_x; float24 offset_x;
@ -91,8 +93,8 @@ void ProcessTriangle(const OutputVertex& v0, const OutputVertex& v1, const Outpu
// introduces at most 1 new vertex to the polygon. Since we start with a triangle and have a // introduces at most 1 new vertex to the polygon. Since we start with a triangle and have a
// fixed 6 clipping planes, the maximum number of vertices of the clipped polygon is 3 + 6 = 9. // fixed 6 clipping planes, the maximum number of vertices of the clipped polygon is 3 + 6 = 9.
static const size_t MAX_VERTICES = 9; static const size_t MAX_VERTICES = 9;
static_vector<OutputVertex, MAX_VERTICES> buffer_a = {v0, v1, v2}; static_vector<Vertex, MAX_VERTICES> buffer_a = {v0, v1, v2};
static_vector<OutputVertex, MAX_VERTICES> buffer_b; static_vector<Vertex, MAX_VERTICES> buffer_b;
auto* output_list = &buffer_a; auto* output_list = &buffer_a;
auto* input_list = &buffer_b; auto* input_list = &buffer_b;
@ -123,7 +125,7 @@ void ProcessTriangle(const OutputVertex& v0, const OutputVertex& v1, const Outpu
std::swap(input_list, output_list); std::swap(input_list, output_list);
output_list->clear(); output_list->clear();
const OutputVertex* reference_vertex = &input_list->back(); const Vertex* reference_vertex = &input_list->back();
for (const auto& vertex : *input_list) { for (const auto& vertex : *input_list) {
// NOTE: This algorithm changes vertex order in some cases! // NOTE: This algorithm changes vertex order in some cases!
@ -148,9 +150,9 @@ void ProcessTriangle(const OutputVertex& v0, const OutputVertex& v1, const Outpu
InitScreenCoordinates((*output_list)[1]); InitScreenCoordinates((*output_list)[1]);
for (size_t i = 0; i < output_list->size() - 2; i++) { for (size_t i = 0; i < output_list->size() - 2; i++) {
OutputVertex& vtx0 = (*output_list)[0]; Vertex& vtx0 = (*output_list)[0];
OutputVertex& vtx1 = (*output_list)[i + 1]; Vertex& vtx1 = (*output_list)[i + 1];
OutputVertex& vtx2 = (*output_list)[i + 2]; Vertex& vtx2 = (*output_list)[i + 2];
InitScreenCoordinates(vtx2); InitScreenCoordinates(vtx2);

View File

@ -307,8 +307,8 @@ MICROPROFILE_DEFINE(GPU_Rasterization, "GPU", "Rasterization", MP_RGB(50, 50, 24
* Helper function for ProcessTriangle with the "reversed" flag to allow for implementing * Helper function for ProcessTriangle with the "reversed" flag to allow for implementing
* culling via recursion. * culling via recursion.
*/ */
static void ProcessTriangleInternal(const Shader::OutputVertex& v0, const Shader::OutputVertex& v1, static void ProcessTriangleInternal(const Vertex& v0, const Vertex& v1, const Vertex& v2,
const Shader::OutputVertex& v2, bool reversed = false) { bool reversed = false) {
const auto& regs = g_state.regs; const auto& regs = g_state.regs;
MICROPROFILE_SCOPE(GPU_Rasterization); MICROPROFILE_SCOPE(GPU_Rasterization);
@ -1276,8 +1276,7 @@ static void ProcessTriangleInternal(const Shader::OutputVertex& v0, const Shader
} }
} }
void ProcessTriangle(const Shader::OutputVertex& v0, const Shader::OutputVertex& v1, void ProcessTriangle(const Vertex& v0, const Vertex& v1, const Vertex& v2) {
const Shader::OutputVertex& v2) {
ProcessTriangleInternal(v0, v1, v2); ProcessTriangleInternal(v0, v1, v2);
} }

View File

@ -4,16 +4,44 @@
#pragma once #pragma once
namespace Pica { #include "video_core/shader/shader.h"
namespace Shader { namespace Pica {
struct OutputVertex;
}
namespace Rasterizer { namespace Rasterizer {
void ProcessTriangle(const Shader::OutputVertex& v0, const Shader::OutputVertex& v1, struct Vertex : Shader::OutputVertex {
const Shader::OutputVertex& v2); Vertex(const OutputVertex& v) : OutputVertex(v) {}
// Attributes used to store intermediate results
// position after perspective divide
Math::Vec3<float24> screenpos;
// Linear interpolation
// factor: 0=this, 1=vtx
void Lerp(float24 factor, const Vertex& vtx) {
pos = pos * factor + vtx.pos * (float24::FromFloat32(1) - factor);
// TODO: Should perform perspective correct interpolation here...
tc0 = tc0 * factor + vtx.tc0 * (float24::FromFloat32(1) - factor);
tc1 = tc1 * factor + vtx.tc1 * (float24::FromFloat32(1) - factor);
tc2 = tc2 * factor + vtx.tc2 * (float24::FromFloat32(1) - factor);
screenpos = screenpos * factor + vtx.screenpos * (float24::FromFloat32(1) - factor);
color = color * factor + vtx.color * (float24::FromFloat32(1) - factor);
}
// Linear interpolation
// factor: 0=v0, 1=v1
static Vertex Lerp(float24 factor, const Vertex& v0, const Vertex& v1) {
Vertex ret = v0;
ret.Lerp(factor, v1);
return ret;
}
};
void ProcessTriangle(const Vertex& v0, const Vertex& v1, const Vertex& v2);
} // namespace Rasterizer } // namespace Rasterizer

View File

@ -26,7 +26,7 @@ OutputVertex OutputVertex::FromAttributeBuffer(const Regs& regs, AttributeBuffer
OutputVertex ret{}; OutputVertex ret{};
std::array<float24, 24> vertex_slots; std::array<float24, 24> vertex_slots;
}; };
static_assert(sizeof(vertex_slots) <= sizeof(ret), "Struct and array have different sizes."); static_assert(sizeof(vertex_slots) == sizeof(ret), "Struct and array have different sizes.");
unsigned int num_attributes = regs.vs_output_total; unsigned int num_attributes = regs.vs_output_total;
ASSERT(num_attributes <= 7); ASSERT(num_attributes <= 7);

View File

@ -28,9 +28,6 @@ struct AttributeBuffer {
}; };
struct OutputVertex { struct OutputVertex {
OutputVertex() = default;
// VS output attributes
Math::Vec4<float24> pos; Math::Vec4<float24> pos;
Math::Vec4<float24> quat; Math::Vec4<float24> quat;
Math::Vec4<float24> color; Math::Vec4<float24> color;
@ -42,42 +39,22 @@ struct OutputVertex {
INSERT_PADDING_WORDS(1); INSERT_PADDING_WORDS(1);
Math::Vec2<float24> tc2; Math::Vec2<float24> tc2;
// Padding for optimal alignment
INSERT_PADDING_WORDS(4);
// Attributes used to store intermediate results
// position after perspective divide
Math::Vec3<float24> screenpos;
INSERT_PADDING_WORDS(1);
// Linear interpolation
// factor: 0=this, 1=vtx
void Lerp(float24 factor, const OutputVertex& vtx) {
pos = pos * factor + vtx.pos * (float24::FromFloat32(1) - factor);
// TODO: Should perform perspective correct interpolation here...
tc0 = tc0 * factor + vtx.tc0 * (float24::FromFloat32(1) - factor);
tc1 = tc1 * factor + vtx.tc1 * (float24::FromFloat32(1) - factor);
tc2 = tc2 * factor + vtx.tc2 * (float24::FromFloat32(1) - factor);
screenpos = screenpos * factor + vtx.screenpos * (float24::FromFloat32(1) - factor);
color = color * factor + vtx.color * (float24::FromFloat32(1) - factor);
}
// Linear interpolation
// factor: 0=v0, 1=v1
static OutputVertex Lerp(float24 factor, const OutputVertex& v0, const OutputVertex& v1) {
OutputVertex ret = v0;
ret.Lerp(factor, v1);
return ret;
}
static OutputVertex FromAttributeBuffer(const Regs& regs, AttributeBuffer& output); static OutputVertex FromAttributeBuffer(const Regs& regs, AttributeBuffer& output);
}; };
#define ASSERT_POS(var, pos) \
static_assert(offsetof(OutputVertex, var) == pos * sizeof(float24), "Semantic at wrong " \
"offset.")
ASSERT_POS(pos, Regs::VSOutputAttributes::POSITION_X);
ASSERT_POS(quat, Regs::VSOutputAttributes::QUATERNION_X);
ASSERT_POS(color, Regs::VSOutputAttributes::COLOR_R);
ASSERT_POS(tc0, Regs::VSOutputAttributes::TEXCOORD0_U);
ASSERT_POS(tc1, Regs::VSOutputAttributes::TEXCOORD1_U);
ASSERT_POS(tc0_w, Regs::VSOutputAttributes::TEXCOORD0_W);
ASSERT_POS(view, Regs::VSOutputAttributes::VIEW_X);
ASSERT_POS(tc2, Regs::VSOutputAttributes::TEXCOORD2_U);
#undef ASSERT_POS
static_assert(std::is_pod<OutputVertex>::value, "Structure is not POD"); static_assert(std::is_pod<OutputVertex>::value, "Structure is not POD");
static_assert(sizeof(OutputVertex) == 32 * sizeof(float), "OutputVertex has invalid size"); static_assert(sizeof(OutputVertex) == 24 * sizeof(float), "OutputVertex has invalid size");
/** /**
* This structure contains the state information that needs to be unique for a shader unit. The 3DS * This structure contains the state information that needs to be unique for a shader unit. The 3DS