void Renderer::threadLoop(int threadIndex)
{
while(!exitThreads)
{
taskLoop(threadIndex);
suspend[threadIndex]->signal();
resume[threadIndex]->wait();
}
}
int main(void)
{
init();
fdevopen(&serial_console_putc, NULL);
pinIsrSetup();
taskLoop(NULL);
return 0;
}
void Renderer::draw(VkPrimitiveTopology topology, VkIndexType indexType, unsigned int count, int baseVertex, bool update)
{
#ifndef NDEBUG
if(count < minPrimitives || count > maxPrimitives)
{
return;
}
#endif
context->topology = topology;
updateConfiguration();
int ms = context->sampleCount;
unsigned int oldMultiSampleMask = context->multiSampleMask;
context->multiSampleMask = context->sampleMask & ((unsigned)0xFFFFFFFF >> (32 - ms));
if(!context->multiSampleMask)
{
return;
}
context->occlusionEnabled = false;
for(auto query : queries)
{
if(query->type == VK_QUERY_TYPE_OCCLUSION)
{
context->occlusionEnabled = true;
break;
}
}
sync->lock(sw::PRIVATE);
if(update || oldMultiSampleMask != context->multiSampleMask)
{
vertexState = VertexProcessor::update(topology);
setupState = SetupProcessor::update();
pixelState = PixelProcessor::update();
vertexRoutine = VertexProcessor::routine(vertexState);
setupRoutine = SetupProcessor::routine(setupState);
pixelRoutine = PixelProcessor::routine(pixelState);
}
int batch = batchSize / ms;
int (Renderer::*setupPrimitives)(int batch, int count);
if(context->isDrawTriangle())
{
setupPrimitives = &Renderer::setupTriangles;
}
else if(context->isDrawLine())
{
setupPrimitives = &Renderer::setupLines;
}
else // Point draw
{
setupPrimitives = &Renderer::setupPoints;
}
DrawCall *draw = nullptr;
do
{
for(int i = 0; i < DRAW_COUNT; i++)
{
if(drawCall[i]->references == -1)
{
draw = drawCall[i];
drawList[nextDraw & DRAW_COUNT_BITS] = draw;
break;
}
}
if(!draw)
{
resumeApp->wait();
}
}
while(!draw);
DrawData *data = draw->data;
if(queries.size() != 0)
{
draw->queries = new std::list<vk::Query*>();
for(auto &query : queries)
{
++query->reference; // Atomic
draw->queries->push_back(query);
}
}
draw->topology = topology;
draw->indexType = indexType;
draw->batchSize = batch;
vertexRoutine->bind();
setupRoutine->bind();
pixelRoutine->bind();
draw->vertexRoutine = vertexRoutine;
draw->setupRoutine = setupRoutine;
draw->pixelRoutine = pixelRoutine;
draw->vertexPointer = (VertexProcessor::RoutinePointer)vertexRoutine->getEntry();
draw->setupPointer = (SetupProcessor::RoutinePointer)setupRoutine->getEntry();
draw->pixelPointer = (PixelProcessor::RoutinePointer)pixelRoutine->getEntry();
draw->setupPrimitives = setupPrimitives;
draw->setupState = setupState;
data->descriptorSets = context->descriptorSets;
data->descriptorDynamicOffsets = context->descriptorDynamicOffsets;
for(int i = 0; i < MAX_VERTEX_INPUTS; i++)
{
data->input[i] = context->input[i].buffer;
data->stride[i] = context->input[i].vertexStride;
}
data->indices = context->indexBuffer;
if(context->vertexShader->hasBuiltinInput(spv::BuiltInInstanceIndex))
{
data->instanceID = context->instanceID;
}
data->baseVertex = baseVertex;
if(pixelState.stencilActive)
{
data->stencil[0].set(context->frontStencil.reference, context->frontStencil.compareMask, context->frontStencil.writeMask);
data->stencil[1].set(context->backStencil.reference, context->backStencil.compareMask, context->backStencil.writeMask);
}
data->lineWidth = context->lineWidth;
data->factor = factor;
if(pixelState.transparencyAntialiasing == TRANSPARENCY_ALPHA_TO_COVERAGE)
{
float ref = context->alphaReference * (1.0f / 255.0f);
float margin = sw::min(ref, 1.0f - ref);
if(ms == 4)
{
data->a2c0 = replicate(ref - margin * 0.6f);
data->a2c1 = replicate(ref - margin * 0.2f);
data->a2c2 = replicate(ref + margin * 0.2f);
data->a2c3 = replicate(ref + margin * 0.6f);
}
else if(ms == 2)
{
data->a2c0 = replicate(ref - margin * 0.3f);
data->a2c1 = replicate(ref + margin * 0.3f);
}
else ASSERT(false);
}
if(pixelState.occlusionEnabled)
{
for(int cluster = 0; cluster < clusterCount; cluster++)
{
data->occlusion[cluster] = 0;
}
}
#if PERF_PROFILE
for(int cluster = 0; cluster < clusterCount; cluster++)
{
for(int i = 0; i < PERF_TIMERS; i++)
{
data->cycles[i][cluster] = 0;
}
}
#endif
// Viewport
{
float W = 0.5f * viewport.width;
float H = 0.5f * viewport.height;
float X0 = viewport.x + W;
float Y0 = viewport.y + H;
float N = viewport.minDepth;
float F = viewport.maxDepth;
float Z = F - N;
if(context->isDrawTriangle())
{
N += context->depthBias;
}
data->Wx16 = replicate(W * 16);
data->Hx16 = replicate(H * 16);
data->X0x16 = replicate(X0 * 16 - 8);
data->Y0x16 = replicate(Y0 * 16 - 8);
data->halfPixelX = replicate(0.5f / W);
data->halfPixelY = replicate(0.5f / H);
data->viewportHeight = abs(viewport.height);
data->slopeDepthBias = context->slopeDepthBias;
data->depthRange = Z;
data->depthNear = N;
}
// Target
{
for(int index = 0; index < RENDERTARGETS; index++)
{
draw->renderTarget[index] = context->renderTarget[index];
if(draw->renderTarget[index])
{
VkOffset3D offset = { 0, 0, static_cast<int32_t>(context->renderTargetLayer[index]) };
data->colorBuffer[index] = (unsigned int*)context->renderTarget[index]->getOffsetPointer(offset, VK_IMAGE_ASPECT_COLOR_BIT);
data->colorPitchB[index] = context->renderTarget[index]->rowPitchBytes(VK_IMAGE_ASPECT_COLOR_BIT, 0);
data->colorSliceB[index] = context->renderTarget[index]->slicePitchBytes(VK_IMAGE_ASPECT_COLOR_BIT, 0);
}
}
draw->depthBuffer = context->depthBuffer;
draw->stencilBuffer = context->stencilBuffer;
if(draw->depthBuffer)
{
VkOffset3D offset = { 0, 0, static_cast<int32_t>(context->depthBufferLayer) };
data->depthBuffer = (float*)context->depthBuffer->getOffsetPointer(offset, VK_IMAGE_ASPECT_DEPTH_BIT);
data->depthPitchB = context->depthBuffer->rowPitchBytes(VK_IMAGE_ASPECT_DEPTH_BIT, 0);
data->depthSliceB = context->depthBuffer->slicePitchBytes(VK_IMAGE_ASPECT_DEPTH_BIT, 0);
}
if(draw->stencilBuffer)
{
VkOffset3D offset = { 0, 0, static_cast<int32_t>(context->stencilBufferLayer) };
data->stencilBuffer = (unsigned char*)context->stencilBuffer->getOffsetPointer(offset, VK_IMAGE_ASPECT_STENCIL_BIT);
data->stencilPitchB = context->stencilBuffer->rowPitchBytes(VK_IMAGE_ASPECT_STENCIL_BIT, 0);
data->stencilSliceB = context->stencilBuffer->slicePitchBytes(VK_IMAGE_ASPECT_STENCIL_BIT, 0);
}
}
// Scissor
{
data->scissorX0 = scissor.offset.x;
data->scissorX1 = scissor.offset.x + scissor.extent.width;
data->scissorY0 = scissor.offset.y;
data->scissorY1 = scissor.offset.y + scissor.extent.height;
}
// Push constants
{
data->pushConstants = context->pushConstants;
}
draw->primitive = 0;
draw->count = count;
draw->references = (count + batch - 1) / batch;
schedulerMutex.lock();
++nextDraw; // Atomic
schedulerMutex.unlock();
#ifndef NDEBUG
if(threadCount == 1) // Use main thread for draw execution
{
threadsAwake = 1;
task[0].type = Task::RESUME;
taskLoop(0);
}
else
#endif
{
if(!threadsAwake)
{
suspend[0]->wait();
threadsAwake = 1;
task[0].type = Task::RESUME;
resume[0]->signal();
}
}
}
