Optimize text rendering for long strings

main.cpp

@@ -76,7 +76,7 @@ struct debug_clock_t
 #define TIME_BLOCK(name) debug_clock_t _name_##timer(#name, __COUNTER__);
 
 struct charglyph_t {
-  uint32_t tex;       // handle to glyph texture
+  int32_t tex;        // index of character in tex array
   glm::ivec2 size;    // 2d dimensions
   glm::ivec2 bearing; // offset from baseline to left/top of glyph
   uint32_t advance;   // offset to next glyph
@@ -109,10 +109,16 @@ struct selection_t {
   int32_t nselected;
 };
 
+const int32_t ARRAY_LIMIT = 400;
+
 charglyph_t glyphmap[128];
 uint32_t text_program;
 uint32_t textvao, textvbo;
 uint32_t hex_program;
+uint32_t texture_array;
+glm::mat4 transforms[ARRAY_LIMIT];
+int32_t letter_map[ARRAY_LIMIT];
+
 
 // error processing
 enum err_type {
@@ -238,16 +244,24 @@ void render_text(const char *text, float x, float y, float scale, glm::vec3 colo
 {
     // TODO: Group draw calls together instead of doing 1 draw call/character. Use an atlas with GL_TEXTURE_ARRAY
     // Take tips from https://www.youtube.com/watch?v=S0PyZKX4lyI
+    scale *= 48.0f/256;
 
     // activate corresponding render state
     glUseProgram(text_program);
     glUniform3f(glGetUniformLocation(text_program, "text_color"), color.x, color.y, color.z);
     glActiveTexture(GL_TEXTURE0);
+    glBindTexture(GL_TEXTURE_2D_ARRAY, texture_array);
     glBindVertexArray(textvao);
+    glBindBuffer(GL_ARRAY_BUFFER, textvbo);
 
+    int idx = 0;
     // iterate through 128 ASCI characters
     for (const char *p = text; *p != '\0'; ++p)
     {
+        // Avoid drawing more than 400 chars at a time
+        if (idx == ARRAY_LIMIT - 1)
+            break;
+
         char c = *p;
         charglyph_t charglyph = glyphmap[(size_t)c];
 
@@ -258,31 +272,19 @@ void render_text(const char *text, float x, float y, float scale, glm::vec3 colo
         }
 
         float xpos = x + charglyph.bearing.x * scale;
-        float ypos = y - (charglyph.size.y - charglyph.bearing.y) * scale;
+        float ypos = y - (256 - charglyph.bearing.y) * scale;
 
-        glm::mat4 transform = glm::translate(glm::mat4(1.0f), glm::vec3(xpos, ypos, 0)) * glm::scale(glm::mat4(1.0f), glm::vec3(charglyph.size.x * scale, charglyph.size.y * scale, 0));
+        transforms[idx] = glm::translate(glm::mat4(1.0f), glm::vec3(xpos, ypos, 0)) * glm::scale(glm::mat4(1.0f), glm::vec3(256 * scale, 256 * scale, 0));
-        glUniformMatrix4fv(glGetUniformLocation(text_program, "transform"), 1, GL_FALSE, &transform[0][0]);
+        letter_map[idx] = charglyph.tex;
 
-        // update VBO for each character
-        // float vertices[4][4] = {
-        //   { xpos,     ypos + h,   0.0f, 0.0f },
-        //   { xpos,     ypos,       0.0f, 1.0f },
-        //   { xpos + w, ypos,       1.0f, 1.0f },
-        //   { xpos + w, ypos + h,   1.0f, 0.0f }
-        // };
-        // render glyph texture over quad
-        glBindTexture(GL_TEXTURE_2D, charglyph.tex);
-        // update content of VBO memory
-        glBindBuffer(GL_ARRAY_BUFFER, textvbo);
-        //glBufferSubData(GL_ARRAY_BUFFER, 0, sizeof(vertices), vertices);
-
-        // render quad
-        glEnable(GL_CULL_FACE);
-        glDrawArraysInstanced(GL_TRIANGLE_FAN, 0, 4, 1);
-        glDisable(GL_CULL_FACE);
         // now advance cursors for next glyph (note that advance is number of 1/64 pixels)
         x += (charglyph.advance >> 6) * scale; // bitshift by 6 to get value in pixels (2^6 = 64)
+        ++idx;
     }
+    glUniformMatrix4fv(glGetUniformLocation(text_program, "transforms"), idx, GL_FALSE, &transforms[0][0][0]);
+    glUniform1iv(glGetUniformLocation(text_program, "letter_map"), idx, &letter_map[0]);
+    glDrawArraysInstanced(GL_TRIANGLE_FAN, 0, 4, idx);
+
     glBindBuffer(GL_ARRAY_BUFFER, 0);
     glBindVertexArray(0);
     glBindTexture(GL_TEXTURE_2D, 0);
@@ -413,10 +415,15 @@ int main([[maybe_unused]]int argc, [[maybe_unused]]char **argv)
         MAIN_ERROR(ERR_FONT_INIT);
     }
 
-    FT_Set_Pixel_Sizes(face, 0, 48);
+    FT_Set_Pixel_Sizes(face, 256, 256);
 
     glPixelStorei(GL_UNPACK_ALIGNMENT, 1); // disable byte-alignment restriction
 
+    glGenTextures(1, &texture_array);
+    glActiveTexture(GL_TEXTURE0);
+    glBindTexture(GL_TEXTURE_2D_ARRAY, texture_array);
+    glTexImage3D(GL_TEXTURE_2D_ARRAY, 0, GL_R8, 256, 256, 128, 0, GL_RED, GL_UNSIGNED_BYTE, 0);
+
     for (unsigned char c = 0; c < 128; c++)
     {
         // load character glyph
@@ -424,29 +431,21 @@ int main([[maybe_unused]]int argc, [[maybe_unused]]char **argv)
         {
             MAIN_ERROR(ERR_CHAR_LOAD);
         }
-        // generate texture
+
-        uint32_t texture;
+        glTexSubImage3D(GL_TEXTURE_2D_ARRAY,0, 0, 0, int(c),
-        glGenTextures(1, &texture);
-        glBindTexture(GL_TEXTURE_2D, texture);
-        glTexImage2D(
-                     GL_TEXTURE_2D,
-                     0,
-                     GL_RED,
                         face->glyph->bitmap.width,
                         face->glyph->bitmap.rows,
-                     0,
+                        1, GL_RED, GL_UNSIGNED_BYTE,
-                     GL_RED,
+                        face->glyph->bitmap.buffer);
-                     GL_UNSIGNED_BYTE,
+
-                     face->glyph->bitmap.buffer
-                     );
         // set texture options
-        glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
+        glTexParameteri(GL_TEXTURE_2D_ARRAY, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
-        glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
+        glTexParameteri(GL_TEXTURE_2D_ARRAY, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
-        glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
+        glTexParameteri(GL_TEXTURE_2D_ARRAY, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
-        glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
+        glTexParameteri(GL_TEXTURE_2D_ARRAY, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
         // now store character for later use
         charglyph_t charglyph = {
-          texture,
+            int(c),
             glm::ivec2(face->glyph->bitmap.width, face->glyph->bitmap.rows),
             glm::ivec2(face->glyph->bitmap_left, face->glyph->bitmap_top),
             (uint32_t)face->glyph->advance.x
@@ -460,6 +459,12 @@ int main([[maybe_unused]]int argc, [[maybe_unused]]char **argv)
     // text initialization
     text_program = make_gl_program("shaders/text.vert", "shaders/text.frag");
 
+    for (int i = 0; i < ARRAY_LIMIT; ++i)
+    {
+        letter_map[i] = 0;
+        transforms[i] = glm::mat4(1.0f);
+    }
+
     static GLfloat vertex_data[] = {
         0.0f, 1.0f,
         0.0f, 0.0f,

shaders/text.frag

@@ -1,12 +1,18 @@
 #version 330 core
-in vec2 texuv;
+
 out vec4 color;
 
-uniform sampler2D text;
+in VS_OUT {
+    vec2 texuv;
+    flat int index;
+} fs_in;
+
+uniform sampler2DArray text;
+uniform int letter_map[400];
 uniform vec3 text_color;
 
 void main()
 {
-    vec4 sampled = vec4(1.0, 1.0, 1.0, texture(text, texuv).r);
+    vec4 sampled = vec4(1.0, 1.0, 1.0, texture(text, vec3(fs_in.texuv.xy, letter_map[fs_in.index])).r);
     color = vec4(text_color, 1.0) * sampled;
 }

shaders/text.vert

@@ -1,13 +1,17 @@
 #version 330 core
 layout (location = 0) in vec2 vertex; // <vec2 pos>
-out vec2 texuv;
 
-uniform mat4 transform;
+out VS_OUT {
+    vec2 texuv;
+    flat int index;
+} vs_out;
+
+uniform mat4 transforms[400];
 uniform mat4 projection;
 
 void main()
 {
-    gl_Position = projection * transform * vec4(vertex.xy, 0.0, 1.0);
+    gl_Position = projection * transforms[gl_InstanceID] * vec4(vertex.xy, 0.0, 1.0);
-    texuv = vertex.xy;
+    vs_out.index = gl_InstanceID;
-    texuv.y=1.0f-texuv.y;
+    vs_out.texuv = vec2(vertex.x, 1.0f - vertex.y);
 }