canvas renderer: parallel tiling pass and sorting pass

This commit is contained in:
martinfouilleul 2023-02-05 21:03:16 +01:00
parent e59f2b152b
commit fd836c00dd
12 changed files with 352 additions and 380 deletions

2
.gitignore vendored
View File

@ -8,3 +8,5 @@ bin/*
*.ilk
*.vs
*.obj
src/gles_canvas_shaders.h

View File

@ -1,7 +1,7 @@
if not exist bin mkdir bin
set gles_shaders=src\gles_canvas_shaders\gles_canvas_fragment.glsl src\gles_canvas_shaders\gles_canvas_vertex.glsl src\gles_canvas_shaders\gles_canvas_tile.glsl
set gles_shaders=src\gles_canvas_shaders\gles_canvas_blit_vertex.glsl src\gles_canvas_shaders\gles_canvas_blit_fragment.glsl src\gles_canvas_shaders\gles_canvas_clear_counters.glsl src\gles_canvas_shaders\gles_canvas_tile.glsl src\gles_canvas_shaders\gles_canvas_sort.glsl src\gles_canvas_shaders\gles_canvas_draw.glsl
call python scripts\embed_text.py %gles_shaders% --output src\gles_canvas_shaders.h
set INCLUDES=/I src /I src/util /I src/platform /I ext /I ext/angle_headers

View File

@ -183,6 +183,7 @@ int main()
mg_ellipse_fill(x+70, y+50, 30, 50);
// text
//*
mg_set_color_rgba(0, 0, 1, 1);
mg_set_font(font);
mg_set_font_size(12);
@ -193,12 +194,21 @@ int main()
frameTime,
1./frameTime);
mg_text_outlines(text);
//*/
mg_fill();
//*
printf("Milepost vector graphics test program (frame time = %fs, fps = %f)...\n",
frameTime,
1./frameTime);
//*/
/*
mg_set_color_rgba(1, 1, 0, 1);
mg_rectangle_fill(8, 8, 100, 50);
*/
mg_flush();
mg_surface_present(surface);

View File

@ -22,8 +22,13 @@ typedef struct mg_gles_canvas_backend
GLint indexBuffer;
GLint tileCounterBuffer;
GLint tileArrayBuffer;
GLint clearCounterProgram;
GLint tileProgram;
GLint sortProgram;
GLint drawProgram;
GLint blitProgram;
GLint outTexture;
char* indexMapping;
char* vertexMapping;
@ -170,8 +175,16 @@ void mg_gles_canvas_draw_batch(mg_canvas_backend* interface, u32 vertexCount, u3
//TODO: ensure there's enough space in tile buffer
//NOTE: first clear counters
/*
glUseProgram(backend->clearCounterProgram);
glBindBufferBase(GL_SHADER_STORAGE_BUFFER, 0, backend->tileCounterBuffer);
glDispatchCompute(tileCountX*tileCountY, 1, 1);
*/
//NOTE: we first distribute triangles into tiles:
/*
glUseProgram(backend->tileProgram);
glBindBufferBase(GL_SHADER_STORAGE_BUFFER, 0, backend->vertexBuffer);
glBindBufferBase(GL_SHADER_STORAGE_BUFFER, 1, backend->indexBuffer);
glBindBufferBase(GL_SHADER_STORAGE_BUFFER, 2, backend->tileCounterBuffer);
@ -182,26 +195,55 @@ void mg_gles_canvas_draw_batch(mg_canvas_backend* interface, u32 vertexCount, u3
glUniform1ui(2, tileSize);
glUniform1ui(3, tileArraySize);
glDispatchCompute(tileCountX, tileCountY, 1);
glDispatchCompute(indexCount/3, 1, 1);
*/
//NOTE: next we sort triangles in each tile
/*
glUseProgram(backend->sortProgram);
glBindBufferBase(GL_SHADER_STORAGE_BUFFER, 0, backend->vertexBuffer);
glBindBufferBase(GL_SHADER_STORAGE_BUFFER, 1, backend->indexBuffer);
glBindBufferBase(GL_SHADER_STORAGE_BUFFER, 2, backend->tileCounterBuffer);
glBindBufferBase(GL_SHADER_STORAGE_BUFFER, 3, backend->tileArrayBuffer);
glUniform1ui(0, indexCount);
glUniform2ui(1, tileCountX, tileCountY);
glUniform1ui(2, tileSize);
glUniform1ui(3, tileArraySize);
glDispatchCompute(tileCountX * tileCountY, 1, 1);
*/
//TODO: then we fire the fragment shader that will select only triangles in its tile
// glMemoryBarrier(GL_SHADER_STORAGE_BARRIER_BIT);
glUseProgram(backend->drawProgram);
glBindBuffer(GL_ARRAY_BUFFER, backend->dummyVertexBuffer);
glBindBufferBase(GL_SHADER_STORAGE_BUFFER, 0, backend->vertexBuffer);
glBindBufferBase(GL_SHADER_STORAGE_BUFFER, 1, backend->indexBuffer);
/* glBindBufferBase(GL_SHADER_STORAGE_BUFFER, 1, backend->indexBuffer);
glBindBufferBase(GL_SHADER_STORAGE_BUFFER, 2, backend->tileCounterBuffer);
glBindBufferBase(GL_SHADER_STORAGE_BUFFER, 3, backend->tileArrayBuffer);
*/
/*
glBindImageTexture(0, backend->outTexture, 0, GL_FALSE, 0, GL_WRITE_ONLY, GL_RGBA8);
glUniform1ui(0, indexCount);
glUniform2ui(1, tileCountX, tileCountY);
glUniform1ui(2, tileSize);
glUniform1ui(3, tileArraySize);
*/
glDispatchCompute(tileCountX, tileCountY, 1);
//NOTE: now blit out texture to surface
/*
glUseProgram(backend->blitProgram);
glBindBuffer(GL_ARRAY_BUFFER, backend->dummyVertexBuffer);
glActiveTexture(GL_TEXTURE0);
glBindTexture(GL_TEXTURE_2D, backend->outTexture);
glUniform1i(0, 0);
glDrawArrays(GL_TRIANGLES, 0, 6);
*/
mg_gles_canvas_update_vertex_layout(backend);
}
@ -217,7 +259,7 @@ void mg_gles_canvas_atlas_upload(mg_canvas_backend* interface, mp_rect rect, u8*
//TODO
}
void compile_shader(GLuint shader, const char* source)
static void mg_gles_compile_shader(GLuint shader, const char* source)
{
glShaderSource(shader, 1, &source, 0);
glCompileShader(shader);
@ -284,12 +326,40 @@ mg_canvas_backend* mg_gles_canvas_create(mg_surface surface)
glGenBuffers(1, &backend->dummyVertexBuffer);
glBindBuffer(GL_ARRAY_BUFFER, backend->dummyVertexBuffer);
mp_rect frame = mg_surface_get_frame(backend->surface);
glGenTextures(1, &backend->outTexture);
glBindTexture(GL_TEXTURE_2D, backend->outTexture);
glTexStorage2D(GL_TEXTURE_2D, 1, GL_RGBA8, frame.w, frame.h);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
//NOTE: create clear program
{
GLuint clearShader = glCreateShader(GL_COMPUTE_SHADER);
backend->clearCounterProgram = glCreateProgram();
mg_gles_compile_shader(clearShader, gles_canvas_clear_counters);
glAttachShader(backend->clearCounterProgram, clearShader);
glLinkProgram(backend->clearCounterProgram);
int status = 0;
glGetProgramiv(backend->clearCounterProgram, GL_LINK_STATUS, &status);
if(!status)
{
char buffer[256];
int size = 0;
glGetProgramInfoLog(backend->clearCounterProgram, 256, &size, buffer);
printf("link error in gl_canvas_clear_counters: %.*s\n", size, buffer);
exit(-1);
}
}
//NOTE: create tile program
{
GLuint tileShader = glCreateShader(GL_COMPUTE_SHADER);
backend->tileProgram = glCreateProgram();
compile_shader(tileShader, gles_canvas_tile);
mg_gles_compile_shader(tileShader, gles_canvas_tile);
glAttachShader(backend->tileProgram, tileShader);
glLinkProgram(backend->tileProgram);
@ -301,20 +371,40 @@ mg_canvas_backend* mg_gles_canvas_create(mg_surface surface)
char buffer[256];
int size = 0;
glGetProgramInfoLog(backend->tileProgram, 256, &size, buffer);
printf("link error: %.*s\n", size, buffer);
printf("link error in gl_canvas_tile: %.*s\n", size, buffer);
exit(-1);
}
}
//NOTE: create sort program
{
GLuint sortShader = glCreateShader(GL_COMPUTE_SHADER);
backend->sortProgram = glCreateProgram();
mg_gles_compile_shader(sortShader, gles_canvas_sort);
glAttachShader(backend->sortProgram, sortShader);
glLinkProgram(backend->sortProgram);
int status = 0;
glGetProgramiv(backend->sortProgram, GL_LINK_STATUS, &status);
if(!status)
{
char buffer[256];
int size = 0;
glGetProgramInfoLog(backend->sortProgram, 256, &size, buffer);
printf("link error gl_canvas_sort: %.*s\n", size, buffer);
exit(-1);
}
}
//NOTE: create draw program
{
GLuint vertexShader = glCreateShader(GL_VERTEX_SHADER);
GLuint fragmentShader = glCreateShader(GL_FRAGMENT_SHADER);
GLuint shader = glCreateShader(GL_COMPUTE_SHADER);
backend->drawProgram = glCreateProgram();
compile_shader(vertexShader, gles_canvas_vertex);
compile_shader(fragmentShader, gles_canvas_fragment);
mg_gles_compile_shader(shader, gles_canvas_draw);
glAttachShader(backend->drawProgram, vertexShader);
glAttachShader(backend->drawProgram, fragmentShader);
glAttachShader(backend->drawProgram, shader);
glLinkProgram(backend->drawProgram);
int status = 0;
@ -324,7 +414,33 @@ mg_canvas_backend* mg_gles_canvas_create(mg_surface surface)
char buffer[256];
int size = 0;
glGetProgramInfoLog(backend->drawProgram, 256, &size, buffer);
printf("link error: %.*s\n", size, buffer);
printf("link error gl_canvas_draw: %.*s\n", size, buffer);
exit(-1);
}
}
//NOTE: create blit program
{
GLuint vertexShader = glCreateShader(GL_VERTEX_SHADER);
GLuint fragmentShader = glCreateShader(GL_FRAGMENT_SHADER);
backend->blitProgram = glCreateProgram();
mg_gles_compile_shader(vertexShader, gles_canvas_blit_vertex);
mg_gles_compile_shader(fragmentShader, gles_canvas_blit_fragment);
glAttachShader(backend->blitProgram, vertexShader);
glAttachShader(backend->blitProgram, fragmentShader);
glLinkProgram(backend->blitProgram);
int status = 0;
glGetProgramiv(backend->blitProgram, GL_LINK_STATUS, &status);
if(!status)
{
char buffer[256];
int size = 0;
glGetProgramInfoLog(backend->blitProgram, 256, &size, buffer);
printf("link error gl_canvas_blit: %.*s\n", size, buffer);
exit(-1);
}
}

View File

@ -1,293 +0,0 @@
/*********************************************************************
*
* file: gles_canvas_shaders.h
* note: string literals auto-generated by embed_text.py
* date: 03/022023
*
**********************************************************************/
#ifndef __GLES_CANVAS_SHADERS_H__
#define __GLES_CANVAS_SHADERS_H__
//NOTE: string imported from src\gles_canvas_shaders\gles_canvas_fragment.glsl
const char* gles_canvas_fragment =
"#version 310 es\n"
"\n"
"precision mediump float;\n"
"layout(std430) buffer;\n"
"\n"
"struct vertex {\n"
" vec2 pos;\n"
" vec4 cubic;\n"
" vec2 uv;\n"
" vec4 color;\n"
" vec4 clip;\n"
" int zIndex;\n"
"};\n"
"\n"
"layout(binding = 0) buffer vertexBufferSSBO {\n"
" vertex elements[];\n"
"} vertexBuffer ;\n"
"\n"
"layout(binding = 1) buffer indexBufferSSBO {\n"
" uint elements[];\n"
"} indexBuffer ;\n"
"\n"
"layout(binding = 2) coherent buffer tileCounterBufferSSBO {\n"
" uint elements[];\n"
"} tileCounterBuffer ;\n"
"\n"
"layout(binding = 3) coherent buffer tileArrayBufferSSBO {\n"
" uint elements[];\n"
"} tileArrayBuffer ;\n"
"\n"
"layout(location = 0) uniform uint indexCount;\n"
"layout(location = 1) uniform uvec2 tileCount;\n"
"layout(location = 2) uniform uint tileSize;\n"
"layout(location = 3) uniform uint tileArraySize;\n"
"\n"
"layout(location = 0) out vec4 fragColor;\n"
"\n"
"\n"
"bool is_top_left(ivec2 a, ivec2 b)\n"
"{\n"
" return( (a.y == b.y && b.x < a.x)\n"
" ||(b.y < a.y));\n"
"}\n"
"\n"
"int orient2d(ivec2 a, ivec2 b, ivec2 p)\n"
"{\n"
" return((b.x-a.x)*(p.y-a.y) - (b.y-a.y)*(p.x-a.x));\n"
"}\n"
"\n"
"void main()\n"
"{\n"
" uvec2 tileCoord = uvec2(gl_FragCoord.xy)/tileSize;\n"
" uint tileIndex = tileCoord.y * tileCount.x + tileCoord.x;\n"
" uint tileCounter = tileCounterBuffer.elements[tileIndex];\n"
"\n"
" const float subPixelFactor = 16.;\n"
" const int sampleCount = 8;\n"
"\n"
" ivec2 centerPoint = ivec2(round(gl_FragCoord.xy * subPixelFactor));\n"
" ivec2 samplePoints[sampleCount] = ivec2[sampleCount](centerPoint + ivec2(1, 3),\n"
" centerPoint + ivec2(-1, -3),\n"
" centerPoint + ivec2(5, -1),\n"
" centerPoint + ivec2(-3, 5),\n"
" centerPoint + ivec2(-5, -5),\n"
" centerPoint + ivec2(-7, 1),\n"
" centerPoint + ivec2(3, -7),\n"
" centerPoint + ivec2(7, 7));\n"
"\n"
"\n"
" //DEBUG\n"
"/*\n"
" if( int(gl_FragCoord.x - 0.5) % 16 == 0\n"
" ||int(gl_FragCoord.y - 0.5) % 16 == 0)\n"
" {\n"
" fragColor = vec4(0, 0, 0, 1);\n"
" }\n"
" else if(tileCounterBuffer.elements[tileIndex] == 2u)\n"
" {\n"
" fragColor = vec4(1, 1, 0, 1);\n"
" }\n"
" else if(tileCounter != 0u)\n"
" {\n"
" fragColor = vec4(0, 1, 0, 1);\n"
" }\n"
" else\n"
" {\n"
" fragColor = vec4(1, 0, 0, 1);\n"
" }\n"
" return;\n"
"//*/\n"
" //----\n"
"\n"
" vec4 sampleColor[sampleCount];\n"
" vec4 currentColor[sampleCount];\n"
" int currentZIndex[sampleCount];\n"
" int flipCount[sampleCount];\n"
"\n"
" for(int i=0; i<sampleCount; i++)\n"
" {\n"
" currentZIndex[i] = -1;\n"
" flipCount[i] = 0;\n"
" sampleColor[i] = vec4(0, 0, 0, 0);\n"
" currentColor[i] = vec4(0, 0, 0, 0);\n"
" }\n"
"\n"
" for(uint tileArrayIndex=0u; tileArrayIndex < tileCounter; tileArrayIndex++)\n"
" {\n"
" uint triangleIndex = tileArrayBuffer.elements[tileArraySize * tileIndex + tileArrayIndex];\n"
"\n"
" uint i0 = indexBuffer.elements[triangleIndex];\n"
" uint i1 = indexBuffer.elements[triangleIndex+1u];\n"
" uint i2 = indexBuffer.elements[triangleIndex+2u];\n"
"\n"
" ivec2 p0 = ivec2(vertexBuffer.elements[i0].pos * subPixelFactor + vec2(0.5, 0.5));\n"
" ivec2 p1 = ivec2(vertexBuffer.elements[i1].pos * subPixelFactor + vec2(0.5, 0.5));\n"
" ivec2 p2 = ivec2(vertexBuffer.elements[i2].pos * subPixelFactor + vec2(0.5, 0.5));\n"
"\n"
" int zIndex = vertexBuffer.elements[i0].zIndex;\n"
" vec4 color = vertexBuffer.elements[i0].color;\n"
"\n"
" //NOTE(martin): reorder triangle counter-clockwise and compute bias for each edge\n"
" int cw = (p1 - p0).x*(p2 - p0).y - (p1 - p0).y*(p2 - p0).x;\n"
" if(cw < 0)\n"
" {\n"
" uint tmpIndex = i1;\n"
" i1 = i2;\n"
" i2 = tmpIndex;\n"
"\n"
" ivec2 tmpPoint = p1;\n"
" p1 = p2;\n"
" p2 = tmpPoint;\n"
" }\n"
"\n"
" vec4 cubic0 = vertexBuffer.elements[i0].cubic;\n"
" vec4 cubic1 = vertexBuffer.elements[i1].cubic;\n"
" vec4 cubic2 = vertexBuffer.elements[i2].cubic;\n"
"\n"
" int bias0 = is_top_left(p1, p2) ? 0 : -1;\n"
" int bias1 = is_top_left(p2, p0) ? 0 : -1;\n"
" int bias2 = is_top_left(p0, p1) ? 0 : -1;\n"
"\n"
" for(int sampleIndex = 0; sampleIndex < sampleCount; sampleIndex++)\n"
" {\n"
" ivec2 samplePoint = samplePoints[sampleIndex];\n"
"\n"
" int w0 = orient2d(p1, p2, samplePoint);\n"
" int w1 = orient2d(p2, p0, samplePoint);\n"
" int w2 = orient2d(p0, p1, samplePoint);\n"
"\n"
" if((w0+bias0) >= 0 && (w1+bias1) >= 0 && (w2+bias2) >= 0)\n"
" {\n"
" vec4 cubic = (cubic0*float(w0) + cubic1*float(w1) + cubic2*float(w2))/(float(w0)+float(w1)+float(w2));\n"
"\n"
" float eps = 0.0001;\n"
" if(cubic.w*(cubic.x*cubic.x*cubic.x - cubic.y*cubic.z) <= eps)\n"
" {\n"
" if(zIndex == currentZIndex[sampleIndex])\n"
" {\n"
" flipCount[sampleIndex]++;\n"
" }\n"
" else\n"
" {\n"
" if((flipCount[sampleIndex] & 0x01) != 0)\n"
" {\n"
" sampleColor[sampleIndex] = currentColor[sampleIndex];\n"
" }\n"
" currentColor[sampleIndex] = sampleColor[sampleIndex]*(1.-color.a) + color.a*color;\n"
" currentZIndex[sampleIndex] = zIndex;\n"
" flipCount[sampleIndex] = 1;\n"
" }\n"
" }\n"
" }\n"
" }\n"
" }\n"
" vec4 pixelColor = vec4(0);\n"
" for(int sampleIndex = 0; sampleIndex < sampleCount; sampleIndex++)\n"
" {\n"
" if((flipCount[sampleIndex] & 0x01) != 0)\n"
" {\n"
" sampleColor[sampleIndex] = currentColor[sampleIndex];\n"
" }\n"
" pixelColor += sampleColor[sampleIndex];\n"
" }\n"
" fragColor = pixelColor/float(sampleCount);\n"
"}\n";
//NOTE: string imported from src\gles_canvas_shaders\gles_canvas_vertex.glsl
const char* gles_canvas_vertex =
"#version 310 es\n"
"\n"
"precision mediump float;\n"
"\n"
"void main()\n"
"{\n"
" float x = float(((uint(gl_VertexID) + 2u) / 3u)%2u);\n"
" float y = float(((uint(gl_VertexID) + 1u) / 3u)%2u);\n"
"\n"
" gl_Position = vec4(-1.0f + x*2.0f, -1.0f+y*2.0f, 0.0f, 1.0f);\n"
"}\n";
//NOTE: string imported from src\gles_canvas_shaders\gles_canvas_tile.glsl
const char* gles_canvas_tile =
"#version 310 es\n"
"layout(local_size_x = 1, local_size_y = 1, local_size_z = 1) in;\n"
"\n"
"precision mediump float;\n"
"layout(std430) buffer;\n"
"\n"
"struct vertex {\n"
" vec2 pos;\n"
" vec4 cubic;\n"
" vec2 uv;\n"
" vec4 color;\n"
" vec4 clip;\n"
" int zIndex;\n"
"};\n"
"\n"
"layout(binding = 0) buffer vertexBufferSSBO {\n"
" vertex elements[];\n"
"} vertexBuffer ;\n"
"\n"
"layout(binding = 1) buffer indexBufferSSBO {\n"
" uint elements[];\n"
"} indexBuffer ;\n"
"\n"
"layout(binding = 2) coherent buffer tileCounterBufferSSBO {\n"
" uint elements[];\n"
"} tileCounterBuffer ;\n"
"\n"
"layout(binding = 3) coherent buffer tileArrayBufferSSBO {\n"
" uint elements[];\n"
"} tileArrayBuffer ;\n"
"\n"
"layout(location = 0) uniform uint indexCount;\n"
"layout(location = 1) uniform uvec2 tileCount;\n"
"layout(location = 2) uniform uint tileSize;\n"
"layout(location = 3) uniform uint tileArraySize;\n"
"\n"
"void main()\n"
"{\n"
" uint tileIndex = gl_WorkGroupID.y * tileCount.x + gl_WorkGroupID.x;\n"
" uint tileArrayOffset = tileArraySize * tileIndex;\n"
"\n"
" vec4 tileBox = vec4(gl_WorkGroupID.x * tileSize,\n"
" gl_WorkGroupID.y * tileSize,\n"
" gl_WorkGroupID.x * tileSize + tileSize,\n"
" gl_WorkGroupID.y * tileSize + tileSize);\n"
"\n"
" uint tileCounter = 0u;\n"
"\n"
" for(uint triangleIndex = 0u; triangleIndex < indexCount; triangleIndex += 3u)\n"
" {\n"
" uint i0 = indexBuffer.elements[triangleIndex];\n"
" uint i1 = indexBuffer.elements[triangleIndex+1u];\n"
" uint i2 = indexBuffer.elements[triangleIndex+2u];\n"
"\n"
" vec2 p0 = vertexBuffer.elements[i0].pos;\n"
" vec2 p1 = vertexBuffer.elements[i1].pos;\n"
" vec2 p2 = vertexBuffer.elements[i2].pos;\n"
"\n"
" vec4 bbox = vec4(min(min(p0.x, p1.x), p2.x),\n"
" min(min(p0.y, p1.y), p2.y),\n"
" max(max(p0.x, p1.x), p2.x),\n"
" max(max(p0.y, p1.y), p2.y));\n"
"\n"
" if(!( bbox.x > tileBox.z\n"
" || bbox.z < tileBox.x\n"
" || bbox.y > tileBox.w\n"
" || bbox.w < tileBox.y))\n"
" {\n"
" tileArrayBuffer.elements[tileArrayOffset + tileCounter] = triangleIndex;\n"
" tileCounter++;\n"
" }\n"
" }\n"
" tileCounterBuffer.elements[tileIndex] = tileCounter;\n"
"\n"
"// tileCounterBuffer.elements[tileIndex] = 1u;\n"
"}\n";
#endif // __GLES_CANVAS_SHADERS_H__

View File

@ -0,0 +1,13 @@
#version 310 es
precision mediump float;
in vec2 uv;
out vec4 fragColor;
layout(location=0) uniform sampler2D tex;
void main()
{
fragColor = texture(tex, uv);
}

View File

@ -2,10 +2,13 @@
precision mediump float;
out vec2 uv;
void main()
{
float x = float(((uint(gl_VertexID) + 2u) / 3u)%2u);
float y = float(((uint(gl_VertexID) + 1u) / 3u)%2u);
gl_Position = vec4(-1.0f + x*2.0f, -1.0f+y*2.0f, 0.0f, 1.0f);
uv = vec2(x, y);
}

View File

@ -0,0 +1,15 @@
#version 310 es
layout(local_size_x = 1, local_size_y = 1, local_size_z = 1) in;
precision mediump float;
layout(std430) buffer;
layout(binding = 0) coherent restrict writeonly buffer tileCounterBufferSSBO {
uint elements[];
} tileCounterBuffer ;
void main()
{
uint tileIndex = gl_WorkGroupID.x;
tileCounterBuffer.elements[tileIndex] = 0u;
}

View File

@ -1,6 +1,9 @@
#version 310 es
layout(local_size_x = 16, local_size_y = 16, local_size_z = 1) in;
precision mediump float;
precision mediump image2D;
layout(std430) buffer;
struct vertex {
@ -12,19 +15,19 @@ struct vertex {
int zIndex;
};
layout(binding = 0) buffer vertexBufferSSBO {
layout(binding = 0) restrict readonly buffer vertexBufferSSBO {
vertex elements[];
} vertexBuffer ;
layout(binding = 1) buffer indexBufferSSBO {
layout(binding = 1) restrict readonly buffer indexBufferSSBO {
uint elements[];
} indexBuffer ;
layout(binding = 2) coherent buffer tileCounterBufferSSBO {
layout(binding = 2) restrict readonly buffer tileCounterBufferSSBO {
uint elements[];
} tileCounterBuffer ;
layout(binding = 3) coherent buffer tileArrayBufferSSBO {
layout(binding = 3) restrict readonly buffer tileArrayBufferSSBO {
uint elements[];
} tileArrayBuffer ;
@ -32,9 +35,14 @@ layout(location = 0) uniform uint indexCount;
layout(location = 1) uniform uvec2 tileCount;
layout(location = 2) uniform uint tileSize;
layout(location = 3) uniform uint tileArraySize;
layout(rgba8, binding = 0) uniform restrict writeonly image2D outTexture;
layout(location = 0) out vec4 fragColor;
void main()
{}
#if 0
bool is_top_left(ivec2 a, ivec2 b)
{
@ -49,14 +57,21 @@ int orient2d(ivec2 a, ivec2 b, ivec2 p)
void main()
{
uvec2 tileCoord = uvec2(gl_FragCoord.xy)/tileSize;
ivec2 pixelCoord = ivec2(gl_WorkGroupID.xy*uvec2(16, 16) + gl_LocalInvocationID.xy);
imageStore(outTexture, pixelCoord, vec4(1, 0, 1, 1));
return;
uvec2 tileCoord = uvec2(pixelCoord) / tileSize;
uint tileIndex = tileCoord.y * tileCount.x + tileCoord.x;
uint tileCounter = tileCounterBuffer.elements[tileIndex];
const float subPixelFactor = 16.;
const int sampleCount = 8;
ivec2 centerPoint = ivec2(round((vec2(pixelCoord) + vec2(0.5, 0.5)) * subPixelFactor));
ivec2 centerPoint = ivec2(round(gl_FragCoord.xy * subPixelFactor));
/*
const int sampleCount = 8;
ivec2 samplePoints[sampleCount] = ivec2[sampleCount](centerPoint + ivec2(1, 3),
centerPoint + ivec2(-1, -3),
centerPoint + ivec2(5, -1),
@ -65,28 +80,38 @@ void main()
centerPoint + ivec2(-7, 1),
centerPoint + ivec2(3, -7),
centerPoint + ivec2(7, 7));
*/
const int sampleCount = 4;
ivec2 samplePoints[sampleCount] = ivec2[sampleCount](centerPoint + ivec2(-2, 6),
centerPoint + ivec2(6, 2),
centerPoint + ivec2(-6, -2),
centerPoint + ivec2(2, -6));
//DEBUG
//*
{
vec4 fragColor = vec4(0);
/*
if( int(gl_FragCoord.x - 0.5) % 16 == 0
||int(gl_FragCoord.y - 0.5) % 16 == 0)
{
fragColor = vec4(0, 0, 0, 1);
if( pixelCoord.x % 16 == 0
||pixelCoord.y % 16 == 0)
{
fragColor = vec4(0, 0, 0, 1);
}
else if(tileCounterBuffer.elements[tileIndex] == 0xffffu)
{
fragColor = vec4(1, 0, 1, 1);
}
else if(tileCounter != 0u)
{
fragColor = vec4(0, 1, 0, 1);
}
else
{
fragColor = vec4(1, 0, 0, 1);
}
*/
imageStore(outTexture, pixelCoord, fragColor);
return;
}
else if(tileCounterBuffer.elements[tileIndex] == 2u)
{
fragColor = vec4(1, 1, 0, 1);
}
else if(tileCounter != 0u)
{
fragColor = vec4(0, 1, 0, 1);
}
else
{
fragColor = vec4(1, 0, 0, 1);
}
return;
//*/
//----
@ -181,5 +206,8 @@ void main()
}
pixelColor += sampleColor[sampleIndex];
}
fragColor = pixelColor/float(sampleCount);
imageStore(outTexture, pixelCoord, pixelColor/float(sampleCount));
}
#endif

View File

@ -0,0 +1,83 @@
#version 310 es
layout(local_size_x = 1, local_size_y = 1, local_size_z = 1) in;
precision mediump float;
layout(std430) buffer;
struct vertex {
vec2 pos;
vec4 cubic;
vec2 uv;
vec4 color;
vec4 clip;
int zIndex;
};
layout(binding = 0) restrict readonly buffer vertexBufferSSBO {
vertex elements[];
} vertexBuffer ;
layout(binding = 1) restrict readonly buffer indexBufferSSBO {
uint elements[];
} indexBuffer ;
layout(binding = 2) coherent readonly restrict buffer tileCounterBufferSSBO {
uint elements[];
} tileCounterBuffer ;
layout(binding = 3) coherent restrict buffer tileArrayBufferSSBO {
uint elements[];
} tileArrayBuffer ;
layout(location = 0) uniform uint indexCount;
layout(location = 1) uniform uvec2 tileCount;
layout(location = 2) uniform uint tileSize;
layout(location = 3) uniform uint tileArraySize;
int get_zindex(uint tileArrayOffset, uint tileArrayIndex)
{
uint triangleIndex = tileArrayBuffer.elements[tileArrayOffset + tileArrayIndex];
uint i0 = indexBuffer.elements[triangleIndex];
int zIndex = vertexBuffer.elements[i0].zIndex;
return(zIndex);
}
void main()
{
uint tileIndex = gl_WorkGroupID.x;
uint tileArrayOffset = tileArraySize * tileIndex;
uint tileArrayCount = tileCounterBuffer.elements[tileIndex];
for(uint tileArrayIndex=1u; tileArrayIndex < tileArrayCount; tileArrayIndex++)
{
for(uint sortIndex = tileArrayIndex; sortIndex > 0u; sortIndex--)
{
int zIndex = get_zindex(tileArrayOffset, sortIndex);
int prevZIndex = get_zindex(tileArrayOffset, sortIndex-1u);
if(zIndex >= prevZIndex)
{
break;
}
uint tmp = tileArrayBuffer.elements[tileArrayOffset + sortIndex];
tileArrayBuffer.elements[tileArrayOffset + sortIndex] = tileArrayBuffer.elements[tileArrayOffset + sortIndex - 1u];
tileArrayBuffer.elements[tileArrayOffset + sortIndex - 1u] = tmp;
}
}
//DEBUG
/*
int prevZIndex = -1;
for(uint tileArrayIndex=1u; tileArrayIndex < tileArrayCount; tileArrayIndex++)
{
int zIndex = get_zindex(tileArrayOffset, tileArrayIndex);
if(zIndex < prevZIndex)
{
tileCounterBuffer.elements[tileIndex] = 0xffffu;
break;
}
prevZIndex = zIndex;
}
//*/
}

View File

@ -13,19 +13,19 @@ struct vertex {
int zIndex;
};
layout(binding = 0) buffer vertexBufferSSBO {
layout(binding = 0) restrict readonly buffer vertexBufferSSBO {
vertex elements[];
} vertexBuffer ;
layout(binding = 1) buffer indexBufferSSBO {
layout(binding = 1) restrict readonly buffer indexBufferSSBO {
uint elements[];
} indexBuffer ;
layout(binding = 2) coherent buffer tileCounterBufferSSBO {
layout(binding = 2) coherent restrict buffer tileCounterBufferSSBO {
uint elements[];
} tileCounterBuffer ;
layout(binding = 3) coherent buffer tileArrayBufferSSBO {
layout(binding = 3) coherent restrict writeonly buffer tileArrayBufferSSBO {
uint elements[];
} tileArrayBuffer ;
@ -36,41 +36,35 @@ layout(location = 3) uniform uint tileArraySize;
void main()
{
uint tileIndex = gl_WorkGroupID.y * tileCount.x + gl_WorkGroupID.x;
uint tileArrayOffset = tileArraySize * tileIndex;
uint triangleIndex = gl_WorkGroupID.x * 3u;
vec4 tileBox = vec4(gl_WorkGroupID.x * tileSize,
gl_WorkGroupID.y * tileSize,
gl_WorkGroupID.x * tileSize + tileSize,
gl_WorkGroupID.y * tileSize + tileSize);
uint i0 = indexBuffer.elements[triangleIndex];
uint i1 = indexBuffer.elements[triangleIndex+1u];
uint i2 = indexBuffer.elements[triangleIndex+2u];
uint tileCounter = 0u;
vec2 p0 = vertexBuffer.elements[i0].pos;
vec2 p1 = vertexBuffer.elements[i1].pos;
vec2 p2 = vertexBuffer.elements[i2].pos;
for(uint triangleIndex = 0u; triangleIndex < indexCount; triangleIndex += 3u)
vec4 fbox = vec4(min(min(p0.x, p1.x), p2.x),
min(min(p0.y, p1.y), p2.y),
max(max(p0.x, p1.x), p2.x),
max(max(p0.y, p1.y), p2.y));
uvec4 box = uvec4(floor(fbox))/tileSize;
uint xMin = max(0u, box.x);
uint yMin = max(0u, box.y);
uint xMax = min(box.z, tileCount.x - 1u);
uint yMax = min(box.w, tileCount.y - 1u);
for(uint y = yMin; y <= yMax; y++)
{
uint i0 = indexBuffer.elements[triangleIndex];
uint i1 = indexBuffer.elements[triangleIndex+1u];
uint i2 = indexBuffer.elements[triangleIndex+2u];
vec2 p0 = vertexBuffer.elements[i0].pos;
vec2 p1 = vertexBuffer.elements[i1].pos;
vec2 p2 = vertexBuffer.elements[i2].pos;
vec4 bbox = vec4(min(min(p0.x, p1.x), p2.x),
min(min(p0.y, p1.y), p2.y),
max(max(p0.x, p1.x), p2.x),
max(max(p0.y, p1.y), p2.y));
if(!( bbox.x > tileBox.z
|| bbox.z < tileBox.x
|| bbox.y > tileBox.w
|| bbox.w < tileBox.y))
for(uint x = xMin ; x <= xMax; x++)
{
tileArrayBuffer.elements[tileArrayOffset + tileCounter] = triangleIndex;
tileCounter++;
uint tileIndex = y*tileCount.x + x;
uint tileCounter = atomicAdd(tileCounterBuffer.elements[tileIndex], 1u);
tileArrayBuffer.elements[tileArraySize*tileIndex + tileCounter] = triangleIndex;
}
}
tileCounterBuffer.elements[tileIndex] = tileCounter;
// tileCounterBuffer.elements[tileIndex] = 1u;
}

View File

@ -125,6 +125,7 @@ mg_surface mg_gles_surface_create_for_window(mp_window window)
//TODO: reactivate this when finished testing!
// eglSwapInterval(surface->eglDisplay, 1);
eglSwapInterval(surface->eglDisplay, 0);
res = mg_surface_alloc_handle((mg_surface_data*)surface);
}