hmn
/
orca
5
Fork 2
Code Issues 21 Pull Requests 1 Packages Projects Releases Wiki Activity

[mtl canvas] Tiling per rectangles into per-shape tile queues (linked lists of triangle indices) to avoid sorting

This commit is contained in:
Martin Fouilleul 2023-03-20 16:46:12 +01:00
parent 11113f597c
commit a4ef58f2d8
6 changed files with 304 additions and 262 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
examples/canvas/main.c
View File

@ -162,6 +162,7 @@ int main()
// head // head
mg_set_color_rgba(1, 1, 0, 1); mg_set_color_rgba(1, 1, 0, 1);
mg_circle_fill(x, y, 200); mg_circle_fill(x, y, 200);
// smile // smile

45
examples/tiger/main.c
View File

@ -82,6 +82,11 @@ int main()
mp_window_bring_to_front(window); mp_window_bring_to_front(window);
mp_window_focus(window); mp_window_focus(window);
bool tracked = false;
vec2 trackPoint = {0};
f32 zoom = 1;
f32 startX = 0, startY = 0;
f64 frameTime = 0; f64 frameTime = 0;
while(!mp_should_quit()) while(!mp_should_quit())
@ -99,8 +104,35 @@ int main()
mp_request_quit(); mp_request_quit();
} break; } break;
case MP_EVENT_KEYBOARD_KEY: case MP_EVENT_MOUSE_BUTTON:
{ {
if(event.key.code == MP_MOUSE_LEFT)
{
if(event.key.action == MP_KEY_PRESS)
{
tracked = true;
vec2 mousePos = mp_mouse_position();
trackPoint.x = mousePos.x/zoom - startX;
trackPoint.y = mousePos.y/zoom - startY;
}
else
{
tracked = false;
}
}
} break;
case MP_EVENT_MOUSE_WHEEL:
{
vec2 mousePos = mp_mouse_position();
f32 trackX = mousePos.x/zoom - startX;
f32 trackY = mousePos.y/zoom - startY;
zoom *= 1 + event.move.deltaY * 0.01;
zoom = Clamp(zoom, 0.2, 10);
startX = mousePos.x/zoom - trackX;
startY = mousePos.y/zoom - trackY;
} break; } break;
default: default:
@ -108,13 +140,20 @@ int main()
} }
} }
if(tracked)
{
vec2 mousePos = mp_mouse_position();
startX = mousePos.x/zoom - trackPoint.x;
startY = mousePos.y/zoom - trackPoint.y;
}
mg_surface_prepare(surface); mg_surface_prepare(surface);
mg_set_color_rgba(1, 0, 1, 1); mg_set_color_rgba(1, 0, 1, 1);
mg_clear(); mg_clear();
mg_matrix_push((mg_mat2x3){1, 0, 300, mg_matrix_push((mg_mat2x3){zoom, 0, 300+startX*zoom,
0, 1, 200}); 0, zoom, 200+startY*zoom});
draw_tiger(); draw_tiger();

20
src/graphics.c
View File

@ -208,6 +208,7 @@ typedef struct mg_canvas_data
mp_rect srcRegion; mp_rect srcRegion;
vec4 shapeExtents; vec4 shapeExtents;
vec4 shapeScreenExtents;
u32 nextShapeIndex; u32 nextShapeIndex;
u32 vertexCount; u32 vertexCount;
u32 indexCount; u32 indexCount;
@ -789,6 +790,13 @@ void mg_finalize_shape(mg_canvas_data* canvas)
mg_vertex_layout* layout = &canvas->backend->vertexLayout; mg_vertex_layout* layout = &canvas->backend->vertexLayout;
*(mg_mat2x3*)(layout->uvTransformBuffer + index*layout->uvTransformStride) = uvTransform; *(mg_mat2x3*)(layout->uvTransformBuffer + index*layout->uvTransformStride) = uvTransform;
//TODO: transform extents before clipping
mp_rect clip = {maximum(canvas->clip.x, canvas->shapeScreenExtents.x),
maximum(canvas->clip.y, canvas->shapeScreenExtents.y),
minimum(canvas->clip.x + canvas->clip.w, canvas->shapeScreenExtents.z),
minimum(canvas->clip.y + canvas->clip.h, canvas->shapeScreenExtents.w)};
*(mp_rect*)(((char*)layout->clipBuffer) + index*layout->clipStride) = clip;
} }
} }
@ -799,17 +807,12 @@ u32 mg_next_shape(mg_canvas_data* canvas, mg_attributes* attributes)
canvas->transform = attributes->transform; canvas->transform = attributes->transform;
canvas->srcRegion = attributes->srcRegion; canvas->srcRegion = attributes->srcRegion;
canvas->shapeExtents = (vec4){FLT_MAX, FLT_MAX, -FLT_MAX, -FLT_MAX}; canvas->shapeExtents = (vec4){FLT_MAX, FLT_MAX, -FLT_MAX, -FLT_MAX};
canvas->shapeScreenExtents = (vec4){FLT_MAX, FLT_MAX, -FLT_MAX, -FLT_MAX};
mg_vertex_layout* layout = &canvas->backend->vertexLayout; mg_vertex_layout* layout = &canvas->backend->vertexLayout;
int index = canvas->nextShapeIndex; int index = canvas->nextShapeIndex;
canvas->nextShapeIndex++; canvas->nextShapeIndex++;
mp_rect clip = {canvas->clip.x,
canvas->clip.y,
canvas->clip.x + canvas->clip.w,
canvas->clip.y + canvas->clip.h};
*(mp_rect*)(((char*)layout->clipBuffer) + index*layout->clipStride) = clip;
*(mg_color*)(((char*)layout->colorBuffer) + index*layout->colorStride) = attributes->color; *(mg_color*)(((char*)layout->colorBuffer) + index*layout->colorStride) = attributes->color;
return(index); return(index);
@ -842,6 +845,11 @@ void mg_push_vertex_cubic(mg_canvas_data* canvas, vec2 pos, vec4 cubic)
vec2 screenPos = mg_mat2x3_mul(canvas->transform, pos); vec2 screenPos = mg_mat2x3_mul(canvas->transform, pos);
canvas->shapeScreenExtents.x = minimum(canvas->shapeScreenExtents.x, screenPos.x);
canvas->shapeScreenExtents.y = minimum(canvas->shapeScreenExtents.y, screenPos.y);
canvas->shapeScreenExtents.z = maximum(canvas->shapeScreenExtents.z, screenPos.x);
canvas->shapeScreenExtents.w = maximum(canvas->shapeScreenExtents.w, screenPos.y);
mg_vertex_layout* layout = &canvas->backend->vertexLayout; mg_vertex_layout* layout = &canvas->backend->vertexLayout;
ASSERT(canvas->vertexCount < layout->maxVertexCount); ASSERT(canvas->vertexCount < layout->maxVertexCount);
ASSERT(canvas->nextShapeIndex > 0); ASSERT(canvas->nextShapeIndex > 0);

143
src/mtl_canvas.m
View File

@ -34,9 +34,8 @@ typedef struct mg_mtl_canvas_backend
mg_color clearColor; mg_color clearColor;
// permanent metal resources // permanent metal resources
id<MTLComputePipelineState> shapePipeline;
id<MTLComputePipelineState> trianglePipeline; id<MTLComputePipelineState> trianglePipeline;
id<MTLComputePipelineState> tilingPipeline;
id<MTLComputePipelineState> sortingPipeline;
id<MTLComputePipelineState> computePipeline; id<MTLComputePipelineState> computePipeline;
id<MTLRenderPipelineState> renderPipeline; id<MTLRenderPipelineState> renderPipeline;
@ -53,9 +52,10 @@ typedef struct mg_mtl_canvas_backend
id<MTLBuffer> shapeBuffer[MG_MTL_MAX_BUFFER_AVAILABLE]; id<MTLBuffer> shapeBuffer[MG_MTL_MAX_BUFFER_AVAILABLE];
id<MTLBuffer> vertexBuffer[MG_MTL_MAX_BUFFER_AVAILABLE]; id<MTLBuffer> vertexBuffer[MG_MTL_MAX_BUFFER_AVAILABLE];
id<MTLBuffer> indexBuffer[MG_MTL_MAX_BUFFER_AVAILABLE]; id<MTLBuffer> indexBuffer[MG_MTL_MAX_BUFFER_AVAILABLE];
id<MTLBuffer> tileCounters; id<MTLBuffer> shapeQueueBuffer;
id<MTLBuffer> tileArrayBuffer;
id<MTLBuffer> triangleArray; id<MTLBuffer> triangleArray;
id<MTLBuffer> arenaBuffer;
id<MTLBuffer> arenaOffset;
} mg_mtl_canvas_backend; } mg_mtl_canvas_backend;
@ -212,15 +212,33 @@ void mg_mtl_canvas_draw_batch(mg_canvas_backend* interface, mg_image_data* image
int triangleCount = indexCount/3; int triangleCount = indexCount/3;
//----------------------------------------------------------- //-----------------------------------------------------------
//NOTE(martin): encode the clear counter //NOTE(martin): encode the clear arena offset
//----------------------------------------------------------- //-----------------------------------------------------------
id<MTLBlitCommandEncoder> blitEncoder = [surface->commandBuffer blitCommandEncoder]; id<MTLBlitCommandEncoder> blitEncoder = [surface->commandBuffer blitCommandEncoder];
blitEncoder.label = @"clear counters"; blitEncoder.label = @"clear arena";
[blitEncoder fillBuffer: backend->tileCounters range: NSMakeRange(0, RENDERER_MAX_TILES*sizeof(uint)) value: 0]; [blitEncoder fillBuffer: backend->arenaOffset range: NSMakeRange(0, sizeof(int)) value: 0];
[blitEncoder endEncoding]; [blitEncoder endEncoding];
//----------------------------------------------------------- //-----------------------------------------------------------
//NOTE(martin): encode the triangle prepass //NOTE(martin): encode the shape setup pass
//-----------------------------------------------------------
id<MTLComputeCommandEncoder> shapeEncoder = [surface->commandBuffer computeCommandEncoder];
shapeEncoder.label = @"shape pass";
[shapeEncoder setComputePipelineState: backend->shapePipeline];
[shapeEncoder setBuffer: backend->shapeBuffer[backend->bufferIndex] offset:backend->shapeBufferOffset atIndex: 0];
[shapeEncoder setBuffer: backend->shapeQueueBuffer offset:0 atIndex: 1];
[shapeEncoder setBuffer: backend->arenaBuffer offset:0 atIndex: 2];
[shapeEncoder setBuffer: backend->arenaOffset offset:0 atIndex: 3];
[shapeEncoder setBytes: &scale length: sizeof(float) atIndex: 4];
MTLSize shapeGroupSize = MTLSizeMake(backend->shapePipeline.maxTotalThreadsPerThreadgroup, 1, 1);
MTLSize shapeGridSize = MTLSizeMake(shapeCount, 1, 1);
[shapeEncoder dispatchThreads: shapeGridSize threadsPerThreadgroup: shapeGroupSize];
[shapeEncoder endEncoding];
//-----------------------------------------------------------
//NOTE(martin): encode the triangle setup and binning
//----------------------------------------------------------- //-----------------------------------------------------------
id<MTLComputeCommandEncoder> triangleEncoder = [surface->commandBuffer computeCommandEncoder]; id<MTLComputeCommandEncoder> triangleEncoder = [surface->commandBuffer computeCommandEncoder];
triangleEncoder.label = @"triangle pass"; triangleEncoder.label = @"triangle pass";
@ -229,8 +247,11 @@ void mg_mtl_canvas_draw_batch(mg_canvas_backend* interface, mg_image_data* image
[triangleEncoder setBuffer: backend->indexBuffer[backend->bufferIndex] offset:backend->indexBufferOffset atIndex: 1]; [triangleEncoder setBuffer: backend->indexBuffer[backend->bufferIndex] offset:backend->indexBufferOffset atIndex: 1];
[triangleEncoder setBuffer: backend->shapeBuffer[backend->bufferIndex] offset:backend->shapeBufferOffset atIndex: 2]; [triangleEncoder setBuffer: backend->shapeBuffer[backend->bufferIndex] offset:backend->shapeBufferOffset atIndex: 2];
[triangleEncoder setBuffer: backend->triangleArray offset:0 atIndex: 3]; [triangleEncoder setBuffer: backend->triangleArray offset:0 atIndex: 3];
[triangleEncoder setBuffer: backend->shapeQueueBuffer offset:0 atIndex: 4];
[triangleEncoder setBuffer: backend->arenaBuffer offset:0 atIndex: 5];
[triangleEncoder setBuffer: backend->arenaOffset offset:0 atIndex: 6];
[triangleEncoder setBytes: &scale length: sizeof(float) atIndex: 4]; [triangleEncoder setBytes: &scale length: sizeof(float) atIndex: 7];
MTLSize triangleGroupSize = MTLSizeMake(backend->trianglePipeline.maxTotalThreadsPerThreadgroup, 1, 1); MTLSize triangleGroupSize = MTLSizeMake(backend->trianglePipeline.maxTotalThreadsPerThreadgroup, 1, 1);
MTLSize triangleGridSize = MTLSizeMake(triangleCount, 1, 1); MTLSize triangleGridSize = MTLSizeMake(triangleCount, 1, 1);
@ -238,56 +259,15 @@ void mg_mtl_canvas_draw_batch(mg_canvas_backend* interface, mg_image_data* image
[triangleEncoder dispatchThreads: triangleGridSize threadsPerThreadgroup: triangleGroupSize]; [triangleEncoder dispatchThreads: triangleGridSize threadsPerThreadgroup: triangleGroupSize];
[triangleEncoder endEncoding]; [triangleEncoder endEncoding];
//-----------------------------------------------------------
//NOTE(martin): encode the tiling pass
//-----------------------------------------------------------
id<MTLComputeCommandEncoder> tileEncoder = [surface->commandBuffer computeCommandEncoder];
tileEncoder.label = @"tiling pass";
[tileEncoder setComputePipelineState: backend->tilingPipeline];
[tileEncoder setBuffer: backend->triangleArray offset:0 atIndex: 0];
[tileEncoder setBuffer: backend->tileCounters offset:0 atIndex: 1];
[tileEncoder setBuffer: backend->tileArrayBuffer offset:0 atIndex: 2];
[tileEncoder setBytes: &triangleCount length:sizeof(int) atIndex: 3];
[tileEncoder setBytes: &viewportSize length: sizeof(vector_uint2) atIndex: 4];
[tileEncoder setBytes: &scale length: sizeof(float) atIndex: 5];
MTLSize tileGroupSize = MTLSizeMake(1, 1, 16);
MTLSize tileGridSize = MTLSizeMake(nTilesX, nTilesY, 16);
[tileEncoder dispatchThreads: tileGridSize threadsPerThreadgroup: tileGroupSize];
[tileEncoder endEncoding];
//-----------------------------------------------------------
//NOTE(martin): encode the sorting pass
//-----------------------------------------------------------
/*
id<MTLComputeCommandEncoder> sortEncoder = [surface->commandBuffer computeCommandEncoder];
sortEncoder.label = @"sorting pass";
[sortEncoder setComputePipelineState: backend->sortingPipeline];
[sortEncoder setBuffer: backend->triangleArray offset:0 atIndex: 0];
[sortEncoder setBuffer: backend->tileCounters offset:0 atIndex: 1];
[sortEncoder setBuffer: backend->tileArrayBuffer offset:0 atIndex: 2];
u32 nTilesX = (viewportSize.x + RENDERER_TILE_SIZE - 1)/RENDERER_TILE_SIZE;
u32 nTilesY = (viewportSize.y + RENDERER_TILE_SIZE - 1)/RENDERER_TILE_SIZE;
MTLSize sortGroupSize = MTLSizeMake(backend->sortingPipeline.maxTotalThreadsPerThreadgroup, 1, 1);
MTLSize sortGridSize = MTLSizeMake(nTilesX*nTilesY, 1, 1);
[sortEncoder dispatchThreads: sortGridSize threadsPerThreadgroup: sortGroupSize];
[sortEncoder endEncoding];
*/
//----------------------------------------------------------- //-----------------------------------------------------------
//NOTE(martin): encode drawing pass //NOTE(martin): encode drawing pass
//----------------------------------------------------------- //-----------------------------------------------------------
id<MTLComputeCommandEncoder> drawEncoder = [surface->commandBuffer computeCommandEncoder]; id<MTLComputeCommandEncoder> drawEncoder = [surface->commandBuffer computeCommandEncoder];
drawEncoder.label = @"drawing pass"; drawEncoder.label = @"drawing pass";
[drawEncoder setComputePipelineState:backend->computePipeline]; [drawEncoder setComputePipelineState:backend->computePipeline];
[drawEncoder setBuffer: backend->tileCounters offset:0 atIndex: 0]; [drawEncoder setBuffer: backend->shapeQueueBuffer offset:0 atIndex: 0];
[drawEncoder setBuffer: backend->tileArrayBuffer offset:0 atIndex: 1]; [drawEncoder setBuffer: backend->triangleArray offset:0 atIndex: 1];
[drawEncoder setBuffer: backend->triangleArray offset:0 atIndex: 2]; [drawEncoder setBuffer: backend->arenaBuffer offset:0 atIndex: 2];
[drawEncoder setTexture: backend->outTexture atIndex: 0]; [drawEncoder setTexture: backend->outTexture atIndex: 0];
int useTexture = 0; int useTexture = 0;
@ -298,8 +278,9 @@ void mg_mtl_canvas_draw_batch(mg_canvas_backend* interface, mg_image_data* image
useTexture = 1; useTexture = 1;
} }
[drawEncoder setBytes: &useTexture length:sizeof(int) atIndex: 3]; [drawEncoder setBytes: &shapeCount length:sizeof(int) atIndex: 3];
[drawEncoder setBytes: &scale length: sizeof(float) atIndex: 4]; [drawEncoder setBytes: &useTexture length:sizeof(int) atIndex: 4];
[drawEncoder setBytes: &scale length: sizeof(float) atIndex: 5];
//TODO: check that we don't exceed maxTotalThreadsPerThreadgroup //TODO: check that we don't exceed maxTotalThreadsPerThreadgroup
DEBUG_ASSERT(RENDERER_TILE_SIZE*RENDERER_TILE_SIZE <= backend->computePipeline.maxTotalThreadsPerThreadgroup); DEBUG_ASSERT(RENDERER_TILE_SIZE*RENDERER_TILE_SIZE <= backend->computePipeline.maxTotalThreadsPerThreadgroup);
@ -391,8 +372,13 @@ void mg_mtl_canvas_destroy(mg_canvas_backend* interface)
} }
//NOTE: semaphore does not have a destructor? //NOTE: semaphore does not have a destructor?
[backend->tileArrayBuffer release]; [backend->shapeQueueBuffer release];
[backend->triangleArray release]; [backend->triangleArray release];
[backend->arenaBuffer release];
[backend->arenaOffset release];
//////////////////////////////////////////
//TODO release all pipelines
[backend->computePipeline release]; [backend->computePipeline release];
} }
} }
@ -524,18 +510,18 @@ mg_canvas_backend* mg_mtl_canvas_create(mg_surface surface)
options: bufferOptions]; options: bufferOptions];
} }
backend->tileArrayBuffer = [metalSurface->device newBufferWithLength: RENDERER_TILE_BUFFER_SIZE*sizeof(int)*RENDERER_MAX_TILES
options: MTLResourceStorageModePrivate];
backend->triangleArray = [metalSurface->device newBufferWithLength: MG_MTL_CANVAS_DEFAULT_BUFFER_LENGTH*sizeof(mg_triangle_data) backend->triangleArray = [metalSurface->device newBufferWithLength: MG_MTL_CANVAS_DEFAULT_BUFFER_LENGTH*sizeof(mg_triangle_data)
options: MTLResourceStorageModePrivate]; options: MTLResourceStorageModePrivate];
//TODO(martin): retain ? backend->shapeQueueBuffer = [metalSurface->device newBufferWithLength: MG_MTL_CANVAS_DEFAULT_BUFFER_LENGTH*sizeof(mg_shape_queue)
//----------------------------------------------------------- options: MTLResourceStorageModePrivate];
//NOTE(martin): create and initialize tile counters
//----------------------------------------------------------- backend->arenaBuffer = [metalSurface->device newBufferWithLength: MG_MTL_CANVAS_DEFAULT_BUFFER_LENGTH*sizeof(mg_queue_elt)
backend->tileCounters = [metalSurface->device newBufferWithLength: RENDERER_MAX_TILES*sizeof(uint) options: MTLResourceStorageModePrivate];
options: MTLResourceStorageModePrivate];
backend->arenaOffset = [metalSurface->device newBufferWithLength: sizeof(int)
options: MTLResourceStorageModePrivate];
//----------------------------------------------------------- //-----------------------------------------------------------
//NOTE(martin): load the library //NOTE(martin): load the library
//----------------------------------------------------------- //-----------------------------------------------------------
@ -551,9 +537,8 @@ mg_canvas_backend* mg_mtl_canvas_create(mg_surface surface)
LOG_ERROR("error : %s\n", errStr); LOG_ERROR("error : %s\n", errStr);
return(0); return(0);
} }
id<MTLFunction> shapeFunction = [library newFunctionWithName:@"ShapeSetup"];
id<MTLFunction> triangleFunction = [library newFunctionWithName:@"TriangleKernel"]; id<MTLFunction> triangleFunction = [library newFunctionWithName:@"TriangleKernel"];
id<MTLFunction> tilingFunction = [library newFunctionWithName:@"TileKernel"];
id<MTLFunction> sortingFunction = [library newFunctionWithName:@"SortKernel"];
id<MTLFunction> computeFunction = [library newFunctionWithName:@"RenderKernel"]; id<MTLFunction> computeFunction = [library newFunctionWithName:@"RenderKernel"];
id<MTLFunction> vertexFunction = [library newFunctionWithName:@"VertexShader"]; id<MTLFunction> vertexFunction = [library newFunctionWithName:@"VertexShader"];
id<MTLFunction> fragmentFunction = [library newFunctionWithName:@"FragmentShader"]; id<MTLFunction> fragmentFunction = [library newFunctionWithName:@"FragmentShader"];
@ -566,6 +551,14 @@ mg_canvas_backend* mg_mtl_canvas_create(mg_surface surface)
error:&error]; error:&error];
ASSERT(backend->computePipeline); ASSERT(backend->computePipeline);
MTLComputePipelineDescriptor* shapePipelineDesc = [[MTLComputePipelineDescriptor alloc] init];
shapePipelineDesc.computeFunction = shapeFunction;
backend->shapePipeline = [metalSurface->device newComputePipelineStateWithDescriptor: shapePipelineDesc
options: MTLPipelineOptionNone
reflection: nil
error: &error];
MTLComputePipelineDescriptor* trianglePipelineDesc = [[MTLComputePipelineDescriptor alloc] init]; MTLComputePipelineDescriptor* trianglePipelineDesc = [[MTLComputePipelineDescriptor alloc] init];
trianglePipelineDesc.computeFunction = triangleFunction; trianglePipelineDesc.computeFunction = triangleFunction;
@ -574,22 +567,6 @@ mg_canvas_backend* mg_mtl_canvas_create(mg_surface surface)
reflection: nil reflection: nil
error: &error]; error: &error];
MTLComputePipelineDescriptor* tilingPipelineDesc = [[MTLComputePipelineDescriptor alloc] init];
tilingPipelineDesc.computeFunction = tilingFunction;
backend->tilingPipeline = [metalSurface->device newComputePipelineStateWithDescriptor: tilingPipelineDesc
options: MTLPipelineOptionNone
reflection: nil
error: &error];
MTLComputePipelineDescriptor* sortingPipelineDesc = [[MTLComputePipelineDescriptor alloc] init];
sortingPipelineDesc.computeFunction = sortingFunction;
backend->sortingPipeline = [metalSurface->device newComputePipelineStateWithDescriptor: sortingPipelineDesc
options: MTLPipelineOptionNone
reflection: nil
error: &error];
//----------------------------------------------------------- //-----------------------------------------------------------
//NOTE(martin): setup our render pipeline state //NOTE(martin): setup our render pipeline state
//----------------------------------------------------------- //-----------------------------------------------------------

28
src/mtl_shader.h
View File

@ -59,4 +59,32 @@ typedef struct mg_triangle_data
} mg_triangle_data; } mg_triangle_data;
#ifndef __METAL_VERSION__
#define device
#else
using namespace metal;
#endif
typedef struct mg_tile_elt
{
int triangleIndex;
int next;
} mg_queue_elt;
typedef struct mg_tile_queue
{
atomic_int first;
} mg_tile_queue;
typedef struct mg_shape_queue
{
vector_int4 area;
device mg_tile_queue* tileQueues;
} mg_shape_queue;
#ifndef __METAL_VERSION__
#undef device
#endif
#endif //__MTL_RENDERER_H_ #endif //__MTL_RENDERER_H_

329
src/mtl_shader.metal
View File

@ -46,13 +46,60 @@ int orient2d(int2 a, int2 b, int2 c)
return((b.x-a.x)*(c.y-a.y) - (b.y-a.y)*(c.x-a.x)); return((b.x-a.x)*(c.y-a.y) - (b.y-a.y)*(c.x-a.x));
} }
device uchar* arena_allocate(device uchar* arenaBuffer,
device volatile atomic_uint* arenaOffset,
uint size)
{
uint index = atomic_fetch_add_explicit(arenaOffset, size, memory_order_relaxed);
return(&arenaBuffer[index]);
}
//NOTE: shape setup allocates tile queues for each shape
kernel void ShapeSetup(constant mg_shape* shapeBuffer [[buffer(0)]],
device mg_shape_queue* shapeQueueBuffer [[buffer(1)]],
device uchar* arenaBuffer [[buffer(2)]],
device volatile atomic_uint* arenaOffset [[buffer(3)]],
constant float* scaling [[buffer(4)]],
uint gid [[thread_position_in_grid]])
{
float4 box = shapeBuffer[gid].clip * scaling[0];
int2 firstTile = int2(box.xy)/RENDERER_TILE_SIZE;
//WARN: the following can result in a 1x1 tile allocated even for empty boxes. But if we didn't allocate
// any tile queue, the tileQueues pointer for that shape would alias the tileQueues pointer of another
// shape, and we would have to detect that in the tiling and drawing kernels. Instead, just accept some
// waste and keep the other kernels more uniforms for now...
int nTilesX = int(box.z)/RENDERER_TILE_SIZE - firstTile.x + 1;
int nTilesY = int(box.w)/RENDERER_TILE_SIZE - firstTile.y + 1;
int tileCount = nTilesX * nTilesY;
int tileArraySize = tileCount * sizeof(mg_tile_queue);
shapeQueueBuffer[gid].area = int4(firstTile.x, firstTile.y, nTilesX, nTilesY);
shapeQueueBuffer[gid].tileQueues = (device mg_tile_queue*)arena_allocate(arenaBuffer, arenaOffset, tileArraySize);
for(int i=0; i<tileCount; i++)
{
atomic_store_explicit(&shapeQueueBuffer[gid].tileQueues[i].first, -1, memory_order_relaxed);
}
}
//NOTE: setup triangle data and bucket triangle into tile queues
kernel void TriangleKernel(constant mg_vertex* vertexBuffer [[buffer(0)]], kernel void TriangleKernel(constant mg_vertex* vertexBuffer [[buffer(0)]],
constant uint* indexBuffer [[buffer(1)]], constant uint* indexBuffer [[buffer(1)]],
constant mg_shape* shapeBuffer [[buffer(2)]], constant mg_shape* shapeBuffer [[buffer(2)]],
device mg_triangle_data* triangleArray [[buffer(3)]], device mg_triangle_data* triangleArray [[buffer(3)]],
constant float* scaling [[buffer(4)]], device mg_shape_queue* shapeQueueBuffer [[buffer(4)]],
device uchar* arenaBuffer [[buffer(5)]],
device volatile atomic_uint* arenaOffset [[buffer(6)]],
constant float* scaling [[buffer(7)]],
uint gid [[thread_position_in_grid]]) uint gid [[thread_position_in_grid]])
{ {
//NOTE: triangle setup
uint triangleIndex = gid * 3; uint triangleIndex = gid * 3;
uint i0 = indexBuffer[triangleIndex]; uint i0 = indexBuffer[triangleIndex];
@ -95,7 +142,6 @@ kernel void TriangleKernel(constant mg_vertex* vertexBuffer [[buffer(0)]],
triangleArray[gid].p1 = ip1; triangleArray[gid].p1 = ip1;
triangleArray[gid].p2 = ip2; triangleArray[gid].p2 = ip2;
//NOTE(martin): compute triangle orientation and bias for each edge
int cw = orient2d(ip0, ip1, ip2) > 0 ? 1 : -1; int cw = orient2d(ip0, ip1, ip2) > 0 ? 1 : -1;
triangleArray[gid].cw = cw; triangleArray[gid].cw = cw;
@ -103,92 +149,45 @@ kernel void TriangleKernel(constant mg_vertex* vertexBuffer [[buffer(0)]],
triangleArray[gid].bias1 = is_top_left(p2, p0) ? -(1-cw)/2 : -(1+cw)/2; triangleArray[gid].bias1 = is_top_left(p2, p0) ? -(1-cw)/2 : -(1+cw)/2;
triangleArray[gid].bias2 = is_top_left(p0, p1) ? -(1-cw)/2 : -(1+cw)/2; triangleArray[gid].bias2 = is_top_left(p0, p1) ? -(1-cw)/2 : -(1+cw)/2;
triangleArray[gid].tileBox = int4(fbox)/RENDERER_TILE_SIZE; int4 tileBox = int4(fbox)/RENDERER_TILE_SIZE;
} triangleArray[gid].tileBox = tileBox;
kernel void TileKernel(const device mg_triangle_data* triangleArray [[buffer(0)]],
device uint* tileCounters [[buffer(1)]],
device uint* tileArrayBuffer [[buffer(2)]],
constant int* triangleCount [[buffer(3)]],
constant uint2* viewport [[buffer(4)]],
constant float* scaling [[buffer(5)]],
uint3 gid [[thread_position_in_grid]])
{
uint2 tilesMatrixDim = (*viewport - 1) / RENDERER_TILE_SIZE + 1;
int nTilesX = tilesMatrixDim.x;
int tileX = gid.x; //NOTE: bucket triangle into tiles
int tileY = gid.y; device mg_shape_queue* shapeQueue = &shapeQueueBuffer[shapeIndex];
int tileIndex = tileY * nTilesX + tileX;
int groupIndex = gid.z;
const int groupSize = 16; int xMin = max(0, tileBox.x - shapeQueue->area.x);
int count = 0; int yMin = max(0, tileBox.y - shapeQueue->area.y);
int mask = 0xffff>>(16-groupIndex); int xMax = min(tileBox.z - shapeQueue->area.x, shapeQueue->area.z-1);
int yMax = min(tileBox.w - shapeQueue->area.y, shapeQueue->area.w-1);
for(int triangleBatchIndex=0; triangleBatchIndex<triangleCount[0]; triangleBatchIndex += groupSize) //NOTE(martin): it's important to do the computation with signed int, so that we can have negative xMax/yMax
// otherwise all triangles on the left or below the x/y axis are attributed to tiles on row/column 0.
for(int y = yMin; y <= yMax; y++)
{ {
int triangleIndex = triangleBatchIndex + groupIndex; for(int x = xMin ; x <= xMax; x++)
bool active = false;
// if(triangleIndex + groupIndex < triangleCount[0])
{ {
int4 box = triangleArray[triangleIndex].tileBox; int tileIndex = y*shapeQueue->area.z + x;
/*
if( tileX >= box.x && tileX <= box.z
&& tileY >= box.y && tileY <= box.w)
{
active = true;
}
*/
}
int vote = uint64_t(simd_ballot(active)); device mg_tile_queue* tileQueue = &shapeQueue->tileQueues[tileIndex];
if(active) device mg_queue_elt* elt = (device mg_queue_elt*)arena_allocate(arenaBuffer, arenaOffset, sizeof(mg_queue_elt));
{ int eltIndex = (device uchar*)elt - arenaBuffer;
int batchOffset = popcount(vote & mask);
tileArrayBuffer[tileIndex*RENDERER_TILE_BUFFER_SIZE + count + batchOffset] = triangleIndex;
}
count += popcount(vote);
}
if(groupIndex == 0)
{
tileCounters[tileIndex] = count;
}
}
kernel void SortKernel(constant mg_triangle_data* triangleArray [[buffer(0)]], elt->next = atomic_exchange_explicit(&tileQueue->first, eltIndex, memory_order_relaxed);
const device uint* tileCounters [[buffer(1)]],
device uint* tileArrayBuffer [[buffer(2)]],
uint gid [[thread_position_in_grid]])
{
uint tileIndex = gid;
uint tileArrayOffset = tileIndex * RENDERER_TILE_BUFFER_SIZE;
uint tileArrayCount = min(tileCounters[tileIndex], (uint)RENDERER_TILE_BUFFER_SIZE);
for(uint tileArrayIndex=1; tileArrayIndex < tileArrayCount; tileArrayIndex++) elt->triangleIndex = gid;
{
for(uint sortIndex = tileArrayIndex; sortIndex > 0; sortIndex--)
{
int shapeIndex = triangleArray[tileArrayBuffer[tileArrayOffset + sortIndex]].shapeIndex;
int prevShapeIndex = triangleArray[tileArrayBuffer[tileArrayOffset + sortIndex - 1]].shapeIndex;
if(shapeIndex >= prevShapeIndex)
{
break;
}
uint tmp = tileArrayBuffer[tileArrayOffset + sortIndex];
tileArrayBuffer[tileArrayOffset + sortIndex] = tileArrayBuffer[tileArrayOffset + sortIndex - 1];
tileArrayBuffer[tileArrayOffset + sortIndex - 1] = tmp;
} }
} }
} }
kernel void RenderKernel(const device uint* tileCounters [[buffer(0)]], kernel void RenderKernel(const device mg_shape_queue* shapeQueueBuffer [[buffer(0)]],
const device uint* tileArrayBuffer [[buffer(1)]], const device mg_triangle_data* triangleArray [[buffer(1)]],
const device mg_triangle_data* triangleArray [[buffer(2)]], const device uchar* arenaBuffer [[buffer(2)]],
constant int* useTexture [[buffer(3)]], constant int* shapeCount [[buffer(3)]],
constant float* scaling [[buffer(4)]], constant int* useTexture [[buffer(4)]],
constant float* scaling [[buffer(5)]],
texture2d<float, access::write> outTexture [[texture(0)]], texture2d<float, access::write> outTexture [[texture(0)]],
texture2d<float> texAtlas [[texture(1)]], texture2d<float> texAtlas [[texture(1)]],
@ -200,37 +199,7 @@ kernel void RenderKernel(const device uint* tileCounters [[buffer(0)]],
{ {
//TODO: guard against thread group size not equal to tile size? //TODO: guard against thread group size not equal to tile size?
const int2 pixelCoord = int2(gid); const int2 pixelCoord = int2(gid);
const uint2 tileCoord = uint2(pixelCoord)/ RENDERER_TILE_SIZE; const int2 tileCoord = pixelCoord/ RENDERER_TILE_SIZE;
const uint2 tilesMatrixDim = (gridSize - 1) / RENDERER_TILE_SIZE + 1;
const uint tileIndex = tileCoord.y * tilesMatrixDim.x + tileCoord.x;
const uint tileCounter = min(tileCounters[tileIndex], (uint)RENDERER_TILE_BUFFER_SIZE);
#ifdef RENDERER_DEBUG_TILES
//NOTE(martin): color code debug values and show the tile grid
{
float4 fragColor = float4(0);
if( pixelCoord.x % 16 == 0
||pixelCoord.y % 16 == 0)
{
fragColor = float4(0, 0, 0, 1);
}
else if(tileCounters[tileIndex] == 0xffffu)
{
fragColor = float4(1, 0, 1, 1);
}
else if(tileCounter != 0u)
{
fragColor = float4(0, 1, 0, 1);
}
else
{
fragColor = float4(1, 0, 0, 1);
}
outTexture.write(fragColor, gid);
return;
}
#endif
const int subPixelFactor = 16; const int subPixelFactor = 16;
const int2 centerPoint = int2((float2(pixelCoord) + float2(0.5, 0.5)) * subPixelFactor); const int2 centerPoint = int2((float2(pixelCoord) + float2(0.5, 0.5)) * subPixelFactor);
@ -258,89 +227,108 @@ kernel void RenderKernel(const device uint* tileCounters [[buffer(0)]],
currentColor[i] = float4(0, 0, 0, 0); currentColor[i] = float4(0, 0, 0, 0);
} }
for(uint tileArrayIndex=0; tileArrayIndex < tileCounter; tileArrayIndex++) for(int shapeIndex = 0; shapeIndex < shapeCount[0]; shapeIndex++)
{ {
int triangleIndex = tileArrayBuffer[RENDERER_TILE_BUFFER_SIZE * tileIndex + tileArrayIndex]; const device mg_shape_queue* shapeQueue = &shapeQueueBuffer[shapeIndex];
const device mg_triangle_data* triangle = &triangleArray[triangleIndex];
int2 p0 = triangle->p0; // get the tile queue that corresponds to our tile in the shape area
int2 p1 = triangle->p1; int2 tileQueueCoord = tileCoord - shapeQueue->area.xy;
int2 p2 = triangle->p2; if( tileQueueCoord.x >= 0
&& tileQueueCoord.y >= 0
int cw = triangle->cw; && tileQueueCoord.x < shapeQueue->area.z
&& tileQueueCoord.y < shapeQueue->area.w)
int bias0 = triangle->bias0;
int bias1 = triangle->bias1;
int bias2 = triangle->bias2;
float4 cubic0 = triangle->cubic0;
float4 cubic1 = triangle->cubic1;
float4 cubic2 = triangle->cubic2;
int shapeIndex = triangle->shapeIndex;
float4 color = triangle->color;
color.rgb *= color.a;
int4 clip = triangle->box;
matrix_float3x3 uvTransform = triangle->uvTransform;
for(int sampleIndex = 0; sampleIndex < sampleCount; sampleIndex++)
{ {
int2 samplePoint = samplePoints[sampleIndex]; int tileQueueIndex = tileQueueCoord.y * shapeQueue->area.z + tileQueueCoord.x;
device mg_tile_queue* tileQueue = &shapeQueue->tileQueues[tileQueueIndex];
if( samplePoint.x < clip.x int firstEltIndex = atomic_load_explicit(&tileQueue->first, memory_order_relaxed);
|| samplePoint.x > clip.z device mg_queue_elt* elt = 0;
|| samplePoint.y < clip.y
|| samplePoint.y > clip.w) for(int eltIndex = firstEltIndex; eltIndex >= 0; eltIndex = elt->next)
{ {
continue; elt = (device mg_queue_elt*)(arenaBuffer + eltIndex);
} const device mg_triangle_data* triangle = &triangleArray[elt->triangleIndex];
int w0 = cw*orient2d(p1, p2, samplePoint); int2 p0 = triangle->p0;
int w1 = cw*orient2d(p2, p0, samplePoint); int2 p1 = triangle->p1;
int w2 = cw*orient2d(p0, p1, samplePoint); int2 p2 = triangle->p2;
if((w0+bias0) >= 0 && (w1+bias1) >= 0 && (w2+bias2) >= 0) int cw = triangle->cw;
{
float4 cubic = (cubic0*w0 + cubic1*w1 + cubic2*w2)/(w0+w1+w2);
float eps = 0.0001; int bias0 = triangle->bias0;
if(cubic.w*(cubic.x*cubic.x*cubic.x - cubic.y*cubic.z) <= eps) int bias1 = triangle->bias1;
int bias2 = triangle->bias2;
float4 cubic0 = triangle->cubic0;
float4 cubic1 = triangle->cubic1;
float4 cubic2 = triangle->cubic2;
int shapeIndex = triangle->shapeIndex;
float4 color = triangle->color;
color.rgb *= color.a;
int4 clip = triangle->box;
matrix_float3x3 uvTransform = triangle->uvTransform;
for(int sampleIndex = 0; sampleIndex < sampleCount; sampleIndex++)
{ {
if(shapeIndex == currentShapeIndex[sampleIndex]) int2 samplePoint = samplePoints[sampleIndex];
if( samplePoint.x < clip.x
|| samplePoint.x > clip.z
|| samplePoint.y < clip.y
|| samplePoint.y > clip.w)
{ {
flipCount[sampleIndex]++; continue;
} }
else
int w0 = cw*orient2d(p1, p2, samplePoint);
int w1 = cw*orient2d(p2, p0, samplePoint);
int w2 = cw*orient2d(p0, p1, samplePoint);
if((w0+bias0) >= 0 && (w1+bias1) >= 0 && (w2+bias2) >= 0)
{ {
if(flipCount[sampleIndex] & 0x01) float4 cubic = (cubic0*w0 + cubic1*w1 + cubic2*w2)/(w0+w1+w2);
float eps = 0.0001;
if(cubic.w*(cubic.x*cubic.x*cubic.x - cubic.y*cubic.z) <= eps)
{ {
sampleColor[sampleIndex] = currentColor[sampleIndex]; if(shapeIndex == currentShapeIndex[sampleIndex])
{
flipCount[sampleIndex]++;
}
else
{
if(flipCount[sampleIndex] & 0x01)
{
sampleColor[sampleIndex] = currentColor[sampleIndex];
}
float4 nextColor = color;
if(useTexture[0])
{
float3 sampleFP = float3(float2(samplePoint).xy/(subPixelFactor*2.), 1);
float2 uv = (uvTransform * sampleFP).xy;
constexpr sampler smp(mip_filter::nearest, mag_filter::linear, min_filter::linear);
float4 texColor = texAtlas.sample(smp, uv);
texColor.rgb *= texColor.a;
nextColor *= texColor;
}
currentColor[sampleIndex] = sampleColor[sampleIndex]*(1.-nextColor.a) + nextColor;
currentShapeIndex[sampleIndex] = shapeIndex;
flipCount[sampleIndex] = 1;
}
} }
float4 nextColor = color;
if(useTexture[0])
{
float3 sampleFP = float3(float2(samplePoint).xy/(subPixelFactor*2.), 1);
float2 uv = (uvTransform * sampleFP).xy;
constexpr sampler smp(mip_filter::nearest, mag_filter::linear, min_filter::linear);
float4 texColor = texAtlas.sample(smp, uv);
texColor.rgb *= texColor.a;
nextColor *= texColor;
}
currentColor[sampleIndex] = sampleColor[sampleIndex]*(1.-nextColor.a) + nextColor;
currentShapeIndex[sampleIndex] = shapeIndex;
flipCount[sampleIndex] = 1;
} }
} }
} }
} }
} }
float4 pixelColor = float4(0); float4 pixelColor = float4(0);
for(int sampleIndex = 0; sampleIndex < sampleCount; sampleIndex++) for(int sampleIndex = 0; sampleIndex < sampleCount; sampleIndex++)
@ -353,4 +341,5 @@ kernel void RenderKernel(const device uint* tileCounters [[buffer(0)]],
} }
outTexture.write(pixelColor/float(sampleCount), gid); outTexture.write(pixelColor/float(sampleCount), gid);
} }