PlatformCore: Change shader order

Move xBRZ shader to the end of the shader stack, making it
an output shader. This necessitated changes to the rendering loop and
an addition of another shader uniform.

Author: GranMinigun

src/platform/core/include/platform/device/ogl_video_device.hpp

@@ -30,6 +30,7 @@ struct OGLVideoDevice : VideoDevice {
   void UpdateTextures();
   void CreateShaderPrograms();
   void ReleaseShaderPrograms();
+  void UpdateShaderUniforms();
 
   auto CompileShader(
     GLenum type,

src/platform/core/src/device/ogl_video_device.cpp

@@ -40,7 +40,7 @@ OGLVideoDevice::~OGLVideoDevice() {
   glDeleteVertexArrays(1, &quad_vao);
   glDeleteBuffers(1, &quad_vbo);
   glDeleteFramebuffers(1, &fbo);
-  glDeleteTextures(4, texture);
+  glDeleteTextures(3, texture);
 }
 
 void OGLVideoDevice::Initialize() {
@@ -59,7 +59,7 @@ void OGLVideoDevice::Initialize() {
 
   // Create three textures and a framebuffer for postprocessing.
   glGenFramebuffers(1, &fbo);
-  glGenTextures(4, texture);
+  glGenTextures(3, texture);
 
   glEnable(GL_FRAMEBUFFER_SRGB);
 
@@ -106,20 +106,6 @@ void OGLVideoDevice::CreateShaderPrograms() {
 
   ReleaseShaderPrograms();
 
-  // xBRZ freescale upsampling filter (two passes)
-  if(video.filter == Video::Filter::xBRZ) {
-    auto [success0, program0] = CompileProgram(xbrz0_vert, xbrz0_frag);
-    auto [success1, program1] = CompileProgram(xbrz1_vert, xbrz1_frag);
-    
-    if(success0 && success1) {
-      programs.push_back(program0);
-      programs.push_back(program1);
-    } else {
-      if(success0) glDeleteProgram(program0);
-      if(success1) glDeleteProgram(program1);
-    }
-  }
-
   // Color correction pass
   switch(video.color) {
     case Video::Color::higan: {
@@ -147,29 +133,56 @@ void OGLVideoDevice::CreateShaderPrograms() {
   }
 
   // Output pass (final)
-  auto [success, program] = CompileProgram(output_vert, output_frag);
-  if(success) {
-    programs.push_back(program);
+  switch (video.filter) {
+    // xBRZ freescale upsampling filter (two passes)
+    case Video::Filter::xBRZ: {
+      auto [success0, program0] = CompileProgram(xbrz0_vert, xbrz0_frag);
+      auto [success1, program1] = CompileProgram(xbrz1_vert, xbrz1_frag);
+
+      if(success0 && success1) {
+        programs.push_back(program0);
+        programs.push_back(program1);
+      } else {
+        if(success0) glDeleteProgram(program0);
+        if(success1) glDeleteProgram(program1);
+      }
+      break;
+    }
+    // Plain linear/nearest output.
+    default: {
+      auto [success, program] = CompileProgram(output_vert, output_frag);
+      if(success) {
+        programs.push_back(program);
+      }
+      break;
+    }
   }
 
+  UpdateShaderUniforms();
+}
+
+void OGLVideoDevice::UpdateShaderUniforms() {
   // Set constant shader uniforms.
-  for(auto program : programs) {
+  for(const auto program : programs) {
     glUseProgram(program);
 
-    auto screen_map = glGetUniformLocation(program, "u_screen_map");
+    const auto screen_map = glGetUniformLocation(program, "u_screen_map");
     if(screen_map != -1) {
       glUniform1i(screen_map, 0);
     }
 
-    auto history_map = glGetUniformLocation(program, "u_history_map");
+    const auto history_map = glGetUniformLocation(program, "u_history_map");
     if(history_map != -1) {
       glUniform1i(history_map, 1);
     }
 
-    auto source_map = glGetUniformLocation(program, "u_source_map");
-    if(source_map != -1) {
-      glUniform1i(source_map, 2);
+    const auto info_map = glGetUniformLocation(program, "u_info_map");
+    if(info_map != -1) {
+      glUniform1i(info_map, 2);
     }
+
+    const auto output_size = glGetUniformLocation(program, "u_output_size");
+    glUniform2f(output_size, view_width, view_height);
   }
 }
 
@@ -233,65 +246,59 @@ void OGLVideoDevice::SetViewport(int x, int y, int width, int height) {
   view_y = y;
   view_width  = width;
   view_height = height;
+
+  UpdateShaderUniforms();
 }
 
 void OGLVideoDevice::SetDefaultFBO(GLuint fbo) {
   default_fbo = fbo;
 }
 
 void OGLVideoDevice::Draw(u32* buffer) {
-  int target = 0;
+  // Bind LCD history map
+  glActiveTexture(GL_TEXTURE1);
+  glBindTexture(GL_TEXTURE_2D, texture[2]);
 
   // Update and bind LCD screen texture
   glActiveTexture(GL_TEXTURE0);
-  glBindTexture(GL_TEXTURE_2D, texture[3]);
+  glBindTexture(GL_TEXTURE_2D, texture[0]);
   glTexSubImage2D(
     GL_TEXTURE_2D, 0, 0, 0, 240, 160, GL_BGRA, GL_UNSIGNED_BYTE, buffer
   );
 
-  // Bind LCD history map
-  glActiveTexture(GL_TEXTURE1);
-  glBindTexture(GL_TEXTURE_2D, texture[2]);
-
-  // Bind LCD source map
-  glActiveTexture(GL_TEXTURE2);
-  glBindTexture(GL_TEXTURE_2D, texture[3]);
-
-  auto program_count = programs.size();
-
   glViewport(0, 0, 240, 160);
   glBindVertexArray(quad_vao);
+  glBindFramebuffer(GL_FRAMEBUFFER, fbo);
 
-  if(program_count <= 2) {
-    target = 2;
-  }
+  for(auto pass_iter = programs.begin(); pass_iter != programs.end(); ++pass_iter) {
+    glUseProgram(*pass_iter);
 
-  for(int i = 0; i < program_count; i++) {
-    glUseProgram(programs[i]);
+    if (pass_iter != programs.begin()) {
+      // Swap input and output textures.
+      std::swap(texture[0], texture[1]);
+    }
+    if (pass_iter == std::prev(programs.end())) {
+      if (config->video.filter == Video::Filter::xBRZ) {
+        glActiveTexture(GL_TEXTURE2);
+        glBindTexture(GL_TEXTURE_2D, texture[0]);
+        glActiveTexture(GL_TEXTURE0);
+        glBindTexture(GL_TEXTURE_2D, texture[1]);
+        glFramebufferTexture2D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, texture[1], 0);
+      } else {
+        glBindTexture(GL_TEXTURE_2D, texture[0]);
+      }
+      glActiveTexture(GL_TEXTURE1);
+      glReadBuffer(GL_COLOR_ATTACHMENT0);
+      glCopyTexSubImage2D(GL_TEXTURE_2D, 0, 0, 0, 0, 0, 240, 160);
 
-    if(i == program_count - 1) {
-      glViewport(view_x, view_y, view_width, view_height);
       glBindFramebuffer(GL_FRAMEBUFFER, default_fbo);
+      glViewport(view_x, view_y, view_width, view_height);
     } else {
-      glBindFramebuffer(GL_FRAMEBUFFER, fbo);
-      glFramebufferTexture2D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, texture[target], 0);
+      glBindTexture(GL_TEXTURE_2D, texture[0]);
+      glFramebufferTexture2D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, texture[1], 0);
     }
 
     glDrawArrays(GL_TRIANGLES, 0, 6);
-
-    // Output of the current pass is the input for the next pass.
-    glActiveTexture(GL_TEXTURE0);
-    glBindTexture(GL_TEXTURE_2D, texture[target]);
-
-    if(i == program_count - 3) {
-      /* The next pass is the next-to-last pass, before we render to screen.
-       * Render that pass into a separate texture, so that it can be 
-       * used in the next frame to calculate LCD ghosting.
-       */
-      target = 2;
-    } else {
-      target ^= 1;
-    }
   }
 }
 

src/platform/core/src/device/shader/common.glsl.hpp

@@ -19,4 +19,18 @@ constexpr auto common_vert = R"(
     v_uv = uv;
     gl_Position = vec4(position, 0.0, 1.0);
   }
-)";
+)";
+
+constexpr auto common_vert_flip = R"(
+  #version 330 core
+
+  layout(location = 0) in vec2 position;
+  layout(location = 1) in vec2 uv;
+
+  out vec2 v_uv;
+
+  void main() {
+    v_uv = vec2(uv.x, 1.0 - uv.y);
+    gl_Position = vec4(position, 0.0, 1.0);
+  }
+)";

src/platform/core/src/device/shader/output.glsl.hpp

@@ -9,7 +9,7 @@
 
 #include "device/shader/common.glsl.hpp"
 
-constexpr auto output_vert = common_vert;
+constexpr auto output_vert = common_vert_flip;
 
 constexpr auto output_frag = R"(
   #version 330 core
@@ -21,6 +21,6 @@ constexpr auto output_frag = R"(
   uniform sampler2D u_screen_map;
 
   void main() {
-    frag_color = texture(u_screen_map, vec2(v_uv.x, 1.0 - v_uv.y));
+    frag_color = texture(u_screen_map, v_uv);
   }
-)";
+)";

src/platform/core/src/device/shader/xbrz.glsl.hpp

@@ -52,9 +52,9 @@ constexpr auto xbrz0_frag = R"(
 
   in vec2 v_uv;
 
-  uniform sampler2D u_source_map;
+  uniform sampler2D u_screen_map;
 
-  #define u_source_size vec4(240.0, 160.0, 1.0/240.0, 1.0/160.0)
+  #define u_screen_size vec4(240.0, 160.0, 1.0/240.0, 1.0/160.0)
 
   #define BLEND_NONE 0
   #define BLEND_NORMAL 1
@@ -98,7 +98,7 @@ constexpr auto xbrz0_frag = R"(
   #define eq(a,b)  (a == b)
   #define neq(a,b) (a != b)
 
-  #define P(x,y) texture(u_source_map, coord + u_source_size.zw * vec2(x, y)).rgb
+  #define P(x,y) texture(u_screen_map, coord + u_screen_size.zw * vec2(x, y)).rgb
 
 void main() {
   //---------------------------------------
@@ -108,8 +108,8 @@ void main() {
   //                       x|G|H|I|x
   //                       -|x|x|x|-
 
-  vec2 pos = fract(v_uv * u_source_size.xy) - vec2(0.5, 0.5);
-  vec2 coord = v_uv - pos * u_source_size.zw;
+  vec2 pos = fract(v_uv * u_screen_size.xy) - vec2(0.5, 0.5);
+  vec2 coord = v_uv - pos * u_screen_size.zw;
 
   vec3 A = P(-1,-1);
   vec3 B = P( 0,-1);
@@ -268,35 +268,18 @@ frag_color /= 255.0;
 }
 )";
 
-constexpr auto xbrz1_vert = R"(
-  #version 330 core
-
-  layout(location = 0) in vec2 position;
-  layout(location = 1) in vec2 uv;
-
-  out vec2 v_uv;
-  out vec4 u_screen_size;
-
-  uniform sampler2D u_screen_map;
-
-  void main() {
-    v_uv = uv;
-    u_screen_size.xy = textureSize(u_screen_map, 0);
-    u_screen_size.zw = 1.0 / textureSize(u_screen_map, 0);
-    gl_Position = vec4(position, 0.0, 1.0);
-  }
-)";
+constexpr auto xbrz1_vert = common_vert_flip;
 
 constexpr auto xbrz1_frag = R"(
   #version 330 core
 
   layout(location = 0) out vec4 frag_color;
 
   in vec2 v_uv;
-  in vec4 u_screen_size;
 
-  uniform sampler2D u_screen_map; // info texture
-  uniform sampler2D u_source_map; // LCD texture
+  uniform sampler2D u_info_map; // info texture
+  uniform sampler2D u_screen_map; // LCD texture
+  uniform vec2 u_output_size;
 
   #define u_source_size vec4(240.0, 160.0, 1.0/240.0, 1.0/160.0)
 
@@ -342,15 +325,15 @@ constexpr auto xbrz1_frag = R"(
   #define eq(a,b)  (a == b)
   #define neq(a,b) (a != b)
 
-  #define P(x,y) texture(u_source_map, coord + u_source_size.zw * vec2(x, y)).rgb
+  #define P(x,y) texture(u_screen_map, coord + u_source_size.zw * vec2(x, y)).rgb
 
   void main() {
    //---------------------------------------
     // Input Pixel Mapping: -|B|-
     //                      D|E|F
     //                      -|H|-
 
-    vec2 scale = u_screen_size.xy * u_source_size.zw;
+    vec2 scale = u_output_size.xy * u_source_size.zw;
     vec2 pos = fract(v_uv * u_source_size.xy) - vec2(0.5, 0.5);
     vec2 coord = v_uv - pos * u_source_size.zw;
 
@@ -360,7 +343,7 @@ constexpr auto xbrz1_frag = R"(
     vec3 F = P( 1, 0);
     vec3 H = P( 0, 1);
 
-    vec4 info = floor(texture(u_screen_map, coord) * 255.0 + 0.5);
+    vec4 info = floor(texture(u_info_map, coord) * 255.0 + 0.5);
 
     // info Mapping: x|y|
     //               w|z|
@@ -448,4 +431,4 @@ constexpr auto xbrz1_frag = R"(
 
   frag_color = vec4(res, 1.0);
   }
-)";
+)";