Merge pull request #2900 from wwylele/clip-2

PICA: implement custom clip plane
merge-requests/60/head
Yuri Kunde Schlesner 2017-09-16 10:23:00 +07:00 committed by GitHub
commit 699c920991
5 changed files with 116 additions and 46 deletions

@ -5,10 +5,10 @@
#pragma once
#include <array>
#include "common/bit_field.h"
#include "common/common_funcs.h"
#include "common/common_types.h"
#include "video_core/pica_types.h"
namespace Pica {
@ -31,7 +31,17 @@ struct RasterizerRegs {
BitField<0, 24, u32> viewport_size_y;
INSERT_PADDING_WORDS(0x9);
INSERT_PADDING_WORDS(0x3);
BitField<0, 1, u32> clip_enable;
BitField<0, 24, u32> clip_coef[4]; // float24
Math::Vec4<float24> GetClipCoef() const {
return {float24::FromRaw(clip_coef[0]), float24::FromRaw(clip_coef[1]),
float24::FromRaw(clip_coef[2]), float24::FromRaw(clip_coef[3])};
}
INSERT_PADDING_WORDS(0x1);
BitField<0, 24, u32> viewport_depth_range; // float24
BitField<0, 24, u32> viewport_depth_near_plane; // float24

@ -169,6 +169,8 @@ RasterizerOpenGL::RasterizerOpenGL() : shader_dirty(true) {
glTexBuffer(GL_TEXTURE_BUFFER, GL_RGBA32F, proctex_diff_lut_buffer.handle);
// Sync fixed function OpenGL state
SyncClipEnabled();
SyncClipCoef();
SyncCullMode();
SyncBlendEnabled();
SyncBlendFuncs();
@ -401,6 +403,18 @@ void RasterizerOpenGL::NotifyPicaRegisterChanged(u32 id) {
SyncCullMode();
break;
// Clipping plane
case PICA_REG_INDEX(rasterizer.clip_enable):
SyncClipEnabled();
break;
case PICA_REG_INDEX_WORKAROUND(rasterizer.clip_coef[0], 0x48):
case PICA_REG_INDEX_WORKAROUND(rasterizer.clip_coef[1], 0x49):
case PICA_REG_INDEX_WORKAROUND(rasterizer.clip_coef[2], 0x4a):
case PICA_REG_INDEX_WORKAROUND(rasterizer.clip_coef[3], 0x4b):
SyncClipCoef();
break;
// Depth modifiers
case PICA_REG_INDEX(rasterizer.viewport_depth_range):
SyncDepthScale();
@ -1280,6 +1294,20 @@ void RasterizerOpenGL::SetShader() {
}
}
void RasterizerOpenGL::SyncClipEnabled() {
state.clip_distance[1] = Pica::g_state.regs.rasterizer.clip_enable != 0;
}
void RasterizerOpenGL::SyncClipCoef() {
const auto raw_clip_coef = Pica::g_state.regs.rasterizer.GetClipCoef();
const GLvec4 new_clip_coef = {raw_clip_coef.x.ToFloat32(), raw_clip_coef.y.ToFloat32(),
raw_clip_coef.z.ToFloat32(), raw_clip_coef.w.ToFloat32()};
if (new_clip_coef != uniform_block_data.data.clip_coef) {
uniform_block_data.data.clip_coef = new_clip_coef;
uniform_block_data.dirty = true;
}
}
void RasterizerOpenGL::SyncCullMode() {
const auto& regs = Pica::g_state.regs;

@ -151,14 +151,21 @@ private:
LightSrc light_src[8];
alignas(16) GLvec4 const_color[6]; // A vec4 color for each of the six tev stages
alignas(16) GLvec4 tev_combiner_buffer_color;
alignas(16) GLvec4 clip_coef;
};
static_assert(
sizeof(UniformData) == 0x460,
sizeof(UniformData) == 0x470,
"The size of the UniformData structure has changed, update the structure in the shader");
static_assert(sizeof(UniformData) < 16384,
"UniformData structure must be less than 16kb as per the OpenGL spec");
/// Syncs the clip enabled status to match the PICA register
void SyncClipEnabled();
/// Syncs the clip coefficients to match the PICA register
void SyncClipCoef();
/// Sets the OpenGL shader in accordance with the current PICA register state
void SetShader();

@ -25,6 +25,42 @@ using TevStageConfig = TexturingRegs::TevStageConfig;
namespace GLShader {
static const std::string UniformBlockDef = R"(
#define NUM_TEV_STAGES 6
#define NUM_LIGHTS 8
struct LightSrc {
vec3 specular_0;
vec3 specular_1;
vec3 diffuse;
vec3 ambient;
vec3 position;
vec3 spot_direction;
float dist_atten_bias;
float dist_atten_scale;
};
layout (std140) uniform shader_data {
vec2 framebuffer_scale;
int alphatest_ref;
float depth_scale;
float depth_offset;
int scissor_x1;
int scissor_y1;
int scissor_x2;
int scissor_y2;
vec3 fog_color;
vec2 proctex_noise_f;
vec2 proctex_noise_a;
vec2 proctex_noise_p;
vec3 lighting_global_ambient;
LightSrc light_src[NUM_LIGHTS];
vec4 const_color[NUM_TEV_STAGES];
vec4 tev_combiner_buffer_color;
vec4 clip_coef;
};
)";
PicaShaderConfig PicaShaderConfig::BuildFromRegs(const Pica::Regs& regs) {
PicaShaderConfig res;
@ -1010,8 +1046,6 @@ std::string GenerateFragmentShader(const PicaShaderConfig& config) {
std::string out = R"(
#version 330 core
#define NUM_TEV_STAGES 6
#define NUM_LIGHTS 8
in vec4 primary_color;
in vec2 texcoord[3];
@ -1023,36 +1057,6 @@ in vec4 gl_FragCoord;
out vec4 color;
struct LightSrc {
vec3 specular_0;
vec3 specular_1;
vec3 diffuse;
vec3 ambient;
vec3 position;
vec3 spot_direction;
float dist_atten_bias;
float dist_atten_scale;
};
layout (std140) uniform shader_data {
vec2 framebuffer_scale;
int alphatest_ref;
float depth_scale;
float depth_offset;
int scissor_x1;
int scissor_y1;
int scissor_x2;
int scissor_y2;
vec3 fog_color;
vec2 proctex_noise_f;
vec2 proctex_noise_a;
vec2 proctex_noise_p;
vec3 lighting_global_ambient;
LightSrc light_src[NUM_LIGHTS];
vec4 const_color[NUM_TEV_STAGES];
vec4 tev_combiner_buffer_color;
};
uniform sampler2D tex[3];
uniform samplerBuffer lighting_lut;
uniform samplerBuffer fog_lut;
@ -1061,7 +1065,11 @@ uniform samplerBuffer proctex_color_map;
uniform samplerBuffer proctex_alpha_map;
uniform samplerBuffer proctex_lut;
uniform samplerBuffer proctex_diff_lut;
)";
out += UniformBlockDef;
out += R"(
// Rotate the vector v by the quaternion q
vec3 quaternion_rotate(vec4 q, vec3 v) {
return v + 2.0 * cross(q.xyz, cross(q.xyz, v) + q.w * v);
@ -1197,6 +1205,12 @@ out float texcoord0_w;
out vec4 normquat;
out vec3 view;
)";
out += UniformBlockDef;
out += R"(
void main() {
primary_color = vert_color;
texcoord[0] = vert_texcoord0;
@ -1207,7 +1221,7 @@ void main() {
view = vert_view;
gl_Position = vert_position;
gl_ClipDistance[0] = -vert_position.z; // fixed PICA clipping plane z <= 0
// TODO (wwylele): calculate gl_ClipDistance[1] from user-defined clipping plane
gl_ClipDistance[1] = dot(clip_coef, vert_position);
}
)";

@ -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]);