diff options
Diffstat (limited to 'src/video_core/renderer_opengl')
| -rw-r--r-- | src/video_core/renderer_opengl/gl_rasterizer.cpp | 56 | ||||
| -rw-r--r-- | src/video_core/renderer_opengl/gl_rasterizer.h | 9 | ||||
| -rw-r--r-- | src/video_core/renderer_opengl/gl_shader_gen.cpp | 115 | ||||
| -rw-r--r-- | src/video_core/renderer_opengl/gl_state.cpp | 13 | ||||
| -rw-r--r-- | src/video_core/renderer_opengl/gl_state.h | 3 |
5 files changed, 144 insertions, 52 deletions
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp index 1c6c15a58..7e09e4712 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.cpp +++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp @@ -28,6 +28,9 @@ MICROPROFILE_DEFINE(OpenGL_Blits, "OpenGL", "Blits", MP_RGB(100, 100, 255)); MICROPROFILE_DEFINE(OpenGL_CacheManagement, "OpenGL", "Cache Mgmt", MP_RGB(100, 255, 100)); RasterizerOpenGL::RasterizerOpenGL() : shader_dirty(true) { + // Clipping plane 0 is always enabled for PICA fixed clip plane z <= 0 + state.clip_distance[0] = true; + // Create sampler objects for (size_t i = 0; i < texture_samplers.size(); ++i) { texture_samplers[i].Create(); @@ -166,6 +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(); @@ -232,13 +237,24 @@ void RasterizerOpenGL::DrawTriangles() { glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, color_surface != nullptr ? color_surface->texture.handle : 0, 0); - glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_TEXTURE_2D, - depth_surface != nullptr ? depth_surface->texture.handle : 0, 0); - bool has_stencil = - regs.framebuffer.framebuffer.depth_format == Pica::FramebufferRegs::DepthFormat::D24S8; - glFramebufferTexture2D( - GL_DRAW_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, GL_TEXTURE_2D, - (has_stencil && depth_surface != nullptr) ? depth_surface->texture.handle : 0, 0); + if (depth_surface != nullptr) { + if (regs.framebuffer.framebuffer.depth_format == + Pica::FramebufferRegs::DepthFormat::D24S8) { + // attach both depth and stencil + glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, + depth_surface->texture.handle, 0); + } else { + // attach depth + glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_TEXTURE_2D, + depth_surface->texture.handle, 0); + // clear stencil attachment + glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, 0); + } + } else { + // clear both depth and stencil attachment + glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, + 0); + } // Sync the viewport // These registers hold half-width and half-height, so must be multiplied by 2 @@ -398,6 +414,18 @@ 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(); @@ -1277,6 +1305,20 @@ 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; diff --git a/src/video_core/renderer_opengl/gl_rasterizer.h b/src/video_core/renderer_opengl/gl_rasterizer.h index 78e218efe..46c62961c 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.h +++ b/src/video_core/renderer_opengl/gl_rasterizer.h @@ -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(); diff --git a/src/video_core/renderer_opengl/gl_shader_gen.cpp b/src/video_core/renderer_opengl/gl_shader_gen.cpp index bb192affd..9fe183944 100644 --- a/src/video_core/renderer_opengl/gl_shader_gen.cpp +++ b/src/video_core/renderer_opengl/gl_shader_gen.cpp @@ -8,6 +8,7 @@ #include "common/assert.h" #include "common/bit_field.h" #include "common/logging/log.h" +#include "core/core.h" #include "video_core/regs_framebuffer.h" #include "video_core/regs_lighting.h" #include "video_core/regs_rasterizer.h" @@ -24,6 +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; @@ -525,11 +562,12 @@ static void WriteLighting(std::string& out, const PicaShaderConfig& config) { "float geo_factor = 1.0;\n"; // Compute fragment normals and tangents - const std::string pertubation = - "2.0 * (" + SampleTexture(config, lighting.bump_selector) + ").rgb - 1.0"; + auto Perturbation = [&]() { + return "2.0 * (" + SampleTexture(config, lighting.bump_selector) + ").rgb - 1.0"; + }; if (lighting.bump_mode == LightingRegs::LightingBumpMode::NormalMap) { // Bump mapping is enabled using a normal map - out += "vec3 surface_normal = " + pertubation + ";\n"; + out += "vec3 surface_normal = " + Perturbation() + ";\n"; // Recompute Z-component of perturbation if 'renorm' is enabled, this provides a higher // precision result @@ -543,7 +581,7 @@ static void WriteLighting(std::string& out, const PicaShaderConfig& config) { out += "vec3 surface_tangent = vec3(1.0, 0.0, 0.0);\n"; } else if (lighting.bump_mode == LightingRegs::LightingBumpMode::TangentMap) { // Bump mapping is enabled using a tangent map - out += "vec3 surface_tangent = " + pertubation + ";\n"; + out += "vec3 surface_tangent = " + Perturbation() + ";\n"; // Mathematically, recomputing Z-component of the tangent vector won't affect the relevant // computation below, which is also confirmed on 3DS. So we don't bother recomputing here // even if 'renorm' is enabled. @@ -593,8 +631,8 @@ static void WriteLighting(std::string& out, const PicaShaderConfig& config) { // Note: even if the normal vector is modified by normal map, which is not the // normal of the tangent plane anymore, the half angle vector is still projected // using the modified normal vector. - std::string half_angle_proj = "normalize(half_vector) - normal / dot(normal, " - "normal) * dot(normal, normalize(half_vector))"; + std::string half_angle_proj = + "normalize(half_vector) - normal * dot(normal, normalize(half_vector))"; // Note: the half angle vector projection is confirmed not normalized before the dot // product. The result is in fact not cos(phi) as the name suggested. index = "dot(" + half_angle_proj + ", tangent)"; @@ -749,7 +787,8 @@ static void WriteLighting(std::string& out, const PicaShaderConfig& config) { } // Fresnel - if (lighting.lut_fr.enable && + // Note: only the last entry in the light slots applies the Fresnel factor + if (light_index == lighting.src_num - 1 && lighting.lut_fr.enable && LightingRegs::IsLightingSamplerSupported(lighting.config, LightingRegs::LightingSampler::Fresnel)) { // Lookup fresnel LUT value @@ -758,17 +797,17 @@ static void WriteLighting(std::string& out, const PicaShaderConfig& config) { lighting.lut_fr.type, lighting.lut_fr.abs_input); value = "(" + std::to_string(lighting.lut_fr.scale) + " * " + value + ")"; - // Enabled for difffuse lighting alpha component + // Enabled for diffuse lighting alpha component if (lighting.fresnel_selector == LightingRegs::LightingFresnelSelector::PrimaryAlpha || lighting.fresnel_selector == LightingRegs::LightingFresnelSelector::Both) { - out += "diffuse_sum.a *= " + value + ";\n"; + out += "diffuse_sum.a = " + value + ";\n"; } // Enabled for the specular lighting alpha component if (lighting.fresnel_selector == LightingRegs::LightingFresnelSelector::SecondaryAlpha || lighting.fresnel_selector == LightingRegs::LightingFresnelSelector::Both) { - out += "specular_sum.a *= " + value + ";\n"; + out += "specular_sum.a = " + value + ";\n"; } } @@ -1007,8 +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]; @@ -1020,36 +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; @@ -1058,7 +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); @@ -1111,7 +1122,10 @@ vec4 secondary_fragment_color = vec4(0.0); "gl_FragCoord.y < scissor_y2)) discard;\n"; } - out += "float z_over_w = 1.0 - gl_FragCoord.z * 2.0;\n"; + // After perspective divide, OpenGL transform z_over_w from [-1, 1] to [near, far]. Here we use + // default near = 0 and far = 1, and undo the transformation to get the original z_over_w, then + // do our own transformation according to PICA specification. + out += "float z_over_w = 2.0 * gl_FragCoord.z - 1.0;\n"; out += "float depth = z_over_w * depth_scale + depth_offset;\n"; if (state.depthmap_enable == RasterizerRegs::DepthBuffering::WBuffering) { out += "depth /= gl_FragCoord.w;\n"; @@ -1151,6 +1165,11 @@ vec4 secondary_fragment_color = vec4(0.0); // Blend the fog out += "last_tex_env_out.rgb = mix(fog_color.rgb, last_tex_env_out.rgb, fog_factor);\n"; + } else if (state.fog_mode == TexturingRegs::FogMode::Gas) { + Core::Telemetry().AddField(Telemetry::FieldType::Session, "VideoCore_Pica_UseGasMode", + true); + LOG_CRITICAL(Render_OpenGL, "Unimplemented gas mode"); + UNIMPLEMENTED(); } out += "gl_FragDepth = depth;\n"; @@ -1186,6 +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; @@ -1194,7 +1219,9 @@ void main() { texcoord0_w = vert_texcoord0_w; normquat = vert_normquat; view = vert_view; - gl_Position = vec4(vert_position.x, vert_position.y, -vert_position.z, vert_position.w); + gl_Position = vert_position; + gl_ClipDistance[0] = -vert_position.z; // fixed PICA clipping plane z <= 0 + gl_ClipDistance[1] = dot(clip_coef, vert_position); } )"; diff --git a/src/video_core/renderer_opengl/gl_state.cpp b/src/video_core/renderer_opengl/gl_state.cpp index bc9d34b84..5770ae08f 100644 --- a/src/video_core/renderer_opengl/gl_state.cpp +++ b/src/video_core/renderer_opengl/gl_state.cpp @@ -68,6 +68,8 @@ OpenGLState::OpenGLState() { draw.vertex_buffer = 0; draw.uniform_buffer = 0; draw.shader_program = 0; + + clip_distance = {}; } void OpenGLState::Apply() const { @@ -261,6 +263,17 @@ void OpenGLState::Apply() const { glUseProgram(draw.shader_program); } + // Clip distance + for (size_t i = 0; i < clip_distance.size(); ++i) { + if (clip_distance[i] != cur_state.clip_distance[i]) { + if (clip_distance[i]) { + glEnable(GL_CLIP_DISTANCE0 + static_cast<GLenum>(i)); + } else { + glDisable(GL_CLIP_DISTANCE0 + static_cast<GLenum>(i)); + } + } + } + cur_state = *this; } diff --git a/src/video_core/renderer_opengl/gl_state.h b/src/video_core/renderer_opengl/gl_state.h index 745a74479..437fe34c4 100644 --- a/src/video_core/renderer_opengl/gl_state.h +++ b/src/video_core/renderer_opengl/gl_state.h @@ -4,6 +4,7 @@ #pragma once +#include <array> #include <glad/glad.h> namespace TextureUnits { @@ -123,6 +124,8 @@ public: GLuint shader_program; // GL_CURRENT_PROGRAM } draw; + std::array<bool, 2> clip_distance; // GL_CLIP_DISTANCE + OpenGLState(); /// Get the currently active OpenGL state |