DWORD __stdcall GameThreadMain(void* args)
{
HWND hWnd = (HWND) args;
Render::Init();
SwapChain swapChain;
swapChain.Create(hWnd, OUTPUT_SIZE_X, OUTPUT_SIZE_Y);
Core::Load();
Core::SetSwapChain(&swapChain);
for (;;)
{
Core::Update();
swapChain.Present();
}
return 0;
}
void Queue::present(
Semaphore& signalSemaphore,
SwapChain& swapChain,
unsigned int chainIndex)
{
auto transitionCmd = m_device.createCommandList();
transitionCmd.transition(swapChain.renderTarget(chainIndex), ResourceState::Present);
submit(transitionCmd);
m_impl->present(signalSemaphore, swapChain, chainIndex);
}
// ---------------------------------------------------------------------------------------------------------
LVEDRENDERINGENGINE_API void __stdcall LvEd_End()
{
RenderContext* rc = RenderContext::Inst();
RenderSurface* surface = s_engineData->pRenderSurface;
s_engineData->basicRenderer->End();
LineRenderer::Inst()->RenderAll(rc);
ErrorHandler::ClearError();
LvEdFonts::FontRenderer::Inst()->FlushPrintRequests( rc );
if(surface->GetType() == RenderSurface::kSwapChain)
{
RenderWorldAxis();
SwapChain* swapchain = static_cast<SwapChain*>(surface);
HRESULT hr = swapchain->GetDXGISwapChain()->Present(0,0);
Logger::IsFailureLog(hr, L"presenting swapchain");
}
s_engineData->renderableSorter.ClearLists();
s_engineData->pRenderSurface = NULL;
RenderContext::Inst()->LightEnvDirty = false;
}
void mainLoop(double deltaTime) override
{
auto cmd = device.graphicsCommandList();
auto backbuffer = swapChain.backbuffer();
cmd.imguiBeginFrame(swapChain, deltaTime);
ImGui::ShowTestWindow();
cmd.clearRTV(backbuffer, float4(0, 0, .25f, 1));
cmd.setRenderTargets({backbuffer});
cmd.bind(hello);
cube.setForRendering(cmd);
float objectPhase = frac(time.secondsF() / ObjectPeriod) * 2 * Pi;
float cameraPhase = frac(time.secondsF() / CameraPeriod) * 2 * Pi;
Hello::Constants c;
float3 cameraPos;
cameraPos.x = cos(cameraPhase) * CameraDistance;
cameraPos.z = sin(cameraPhase) * CameraDistance;
cameraPos.y = 2;
Matrix view = Matrix::lookAt(cameraPos, 0);
Matrix proj = Matrix::projectionPerspective(size());
c.viewProj = proj * view;
c.model = Matrix::axisAngle({1, 0, 0}, Angle(objectPhase));
cmd.setConstants(c);
cmd.setShaderView(Hello::tex, lena);
cmd.drawIndexed(cube.numIndices());
cmd.setRenderTargets();
cmd.imguiEndFrame(swapChain);
device.execute(cmd);
device.present(swapChain);
}
static void runTestsForDevice(GpuDevice& dev, GraphicsQueue& queue, Window& window, SwapChain& sc, ViewPort& port, GfxCommandList& gfx, faze::Logger& log)
{
faze::TestWorks t("advtests");
t.setAfterTest([&]()
{
// clean up
auto fence = dev.createFence();
queue.insertFence(fence);
fence.wait();
if (!gfx.isClosed())
{
gfx.closeList();
}
gfx.resetList();
});
t.addTest("Lots of drawcalls bench, (single thread), rough baseline", [&]()
{
using namespace faze;
struct buf
{
float pos[4];
};
auto triangleCount = 1000000;
auto currentTriangleCount = triangleCount;
auto srcdata = dev.createBuffer(ResourceDescriptor()
.Width(triangleCount)
.Format<buf>()
.Usage(ResourceUsage::UploadHeap));
auto dstdata = dev.createBuffer(ResourceDescriptor()
.Width(triangleCount)
.Format<buf>());
auto dstdataSrv = dev.createBufferSRV(dstdata);
std::random_device rd;
std::mt19937 gen(rd());
std::uniform_real_distribution<> dis(-0.8f, 0.8f);
std::uniform_real_distribution<> dis2(0.f, 1.f);
{
auto tmp = srcdata.Map<buf>();
for (int i = 0;i < triangleCount; ++i)
{
auto& it = tmp[i].pos;
it[0] = static_cast<float>(dis(gen));
it[1] = static_cast<float>(dis(gen));
it[2] = static_cast<float>(dis2(gen));
}
}
gfx.CopyResource(dstdata, srcdata);
GpuFence fence = dev.createFence();
gfx.closeList();
queue.submit(gfx);
queue.insertFence(fence);
auto pipeline = dev.createGraphicsPipeline(GraphicsPipelineDescriptor()
.PixelShader("tests/stress/pixel")
.VertexShader("tests/stress/vertex_triangle")
.setRenderTargetCount(1)
.RTVFormat(0, FormatType::R8G8B8A8_UNORM_SRGB)
.DepthStencil(DepthStencilDescriptor().DepthEnable(false)));
auto vec = faze::vec4({ 0.2f, 0.2f, 0.2f, 1.0f });
fence.wait();
gfx.resetList();
WTime t;
t.firstTick();
float frameTime = 30.f;
float cpuTime = 30.f;
Bentsumaakaa b;
while (cpuTime > 14.f)
//while(true)
{
if (window.simpleReadMessages())
break;
// Rendertarget
b.start(false);
gfx.setViewPort(port);
auto backBufferIndex = sc->GetCurrentBackBufferIndex();
gfx.ClearRenderTargetView(sc[backBufferIndex], vec);
gfx.setRenderTarget(sc[backBufferIndex]);
// graphics begin
{
auto bind = gfx.bind(pipeline);
bind.SRV(0, dstdataSrv);
gfx.drawInstanced(bind, 3, 1, 0, 0);
for (int i = 1; i < currentTriangleCount; ++i)
{
gfx.drawInstancedRaw(3, 1, 0, i); // this is the cheat.
}
}
// submit all
gfx.closeList();
queue.submit(gfx);
cpuTime = b.stop(false) / 1000000.f;
// present
sc->Present(1, 0);
queue.insertFence(fence);
fence.wait();
gfx.resetList();
t.tick();
frameTime = static_cast<float>(t.getCurrentNano())*0.000001f;
if (cpuTime > 16.f)
{
currentTriangleCount -= currentTriangleCount/100;
}
}
F_LOG("frametime: %.3fms CpuTime: %.3fms TriangleCount: %d\n", frameTime, cpuTime, currentTriangleCount);
log.update();
//fence.wait();
return true;
});
t.addTest("Lots of drawcalls bench, (single thread), myapi", [&]()
{
using namespace faze;
struct buf
{
float pos[4];
};
auto triangleCount = 200000;
auto currentTriangleCount = triangleCount;
auto srcdata = dev.createBuffer(ResourceDescriptor()
.Width(triangleCount)
.Format<buf>()
.Usage(ResourceUsage::UploadHeap));
auto dstdata = dev.createBuffer(ResourceDescriptor()
.Width(triangleCount)
.Format<buf>());
auto dstdataSrv = dev.createBufferSRV(dstdata);
std::random_device rd;
std::mt19937 gen(rd());
std::uniform_real_distribution<> dis(-0.8f, 0.8f);
std::uniform_real_distribution<> dis2(0.f, 1.f);
{
auto tmp = srcdata.Map<buf>();
for (int i = 0;i < triangleCount; ++i)
{
auto& it = tmp[i].pos;
it[0] = static_cast<float>(dis(gen));
it[1] = static_cast<float>(dis(gen));
it[2] = static_cast<float>(dis2(gen));
}
}
gfx.CopyResource(dstdata, srcdata);
GpuFence fence = dev.createFence();
gfx.closeList();
queue.submit(gfx);
queue.insertFence(fence);
auto pipeline = dev.createGraphicsPipeline(GraphicsPipelineDescriptor()
.PixelShader("tests/stress/pixel")
.VertexShader("tests/stress/vertex_triangle")
.setRenderTargetCount(1)
.RTVFormat(0, FormatType::R8G8B8A8_UNORM_SRGB)
.DepthStencil(DepthStencilDescriptor().DepthEnable(false)));
auto vec = faze::vec4({ 0.2f, 0.2f, 0.2f, 1.0f });
fence.wait();
gfx.resetList();
WTime t;
t.firstTick();
float frameTime = 30.f;
float cpuTime = 30.f;
Bentsumaakaa b;
LBS lbs;
while (cpuTime > 14.f)
//while (true)
{
if (window.simpleReadMessages())
break;
// Rendertarget
b.start(false);
gfx.setViewPort(port);
auto backBufferIndex = sc->GetCurrentBackBufferIndex();
gfx.ClearRenderTargetView(sc[backBufferIndex], vec);
gfx.setRenderTarget(sc[backBufferIndex]);
// graphics begin
{
for (int i = 0; i < currentTriangleCount; ++i)
{
auto bind = gfx.bind(pipeline);
bind.SRV(0, dstdataSrv);
gfx.drawInstanced(bind, 3, 1, 0, i);
}
}
// submit all
gfx.closeList();
queue.submit(gfx);
cpuTime = b.stop(false) / 1000000.f;
// present
sc->Present(1, 0);
queue.insertFence(fence);
fence.wait();
gfx.resetList();
t.tick();
frameTime = static_cast<float>(t.getCurrentNano())*0.000001f;
frameTime = t.analyzeFrames().x();
if (cpuTime > 16.f)
{
currentTriangleCount -= currentTriangleCount / 100;
}
}
//fence.wait();
F_LOG("frametime: %.3fms CpuTime: %.3fms TriangleCount: %d\n", frameTime, cpuTime, currentTriangleCount);
log.update();
return true;
});
t.addTest("Lots of drawcalls bench, (Multithread), baseline", [&]()
{
using namespace faze;
struct buf
{
float pos[4];
};
auto triangleCount = 4000000;
auto currentTriangleCount = triangleCount;
auto srcdata = dev.createBuffer(ResourceDescriptor()
.Width(triangleCount)
.Format<buf>()
.Usage(ResourceUsage::UploadHeap));
auto dstdata = dev.createBuffer(ResourceDescriptor()
.Width(triangleCount)
.Format<buf>());
auto dstdataSrv = dev.createBufferSRV(dstdata);
std::random_device rd;
std::mt19937 gen(rd());
std::uniform_real_distribution<> dis(-0.8f, 0.8f);
std::uniform_real_distribution<> dis2(0.f, 1.f);
{
auto tmp = srcdata.Map<buf>();
for (int i = 0;i < triangleCount; ++i)
{
auto& it = tmp[i].pos;
it[0] = static_cast<float>(dis(gen));
it[1] = static_cast<float>(dis(gen));
it[2] = static_cast<float>(dis2(gen));
}
}
gfx.CopyResource(dstdata, srcdata);
GpuFence fence = dev.createFence();
gfx.closeList();
queue.submit(gfx);
queue.insertFence(fence);
auto pipeline = dev.createGraphicsPipeline(GraphicsPipelineDescriptor()
.PixelShader("tests/stress/pixel")
.VertexShader("tests/stress/vertex_triangle")
.setRenderTargetCount(1)
.RTVFormat(0, FormatType::R8G8B8A8_UNORM_SRGB)
.DepthStencil(DepthStencilDescriptor().DepthEnable(false)));
auto vec = faze::vec4({ 0.2f, 0.2f, 0.2f, 1.0f });
fence.wait();
gfx.resetList();
WTime t;
t.firstTick();
float frameTime = 30.f;
float cpuTime = 30.f;
Bentsumaakaa b;
LBS lbs;
std::vector<GfxCommandList> m_cmds;
for (size_t i = 0; i < lbs.threadCount(); ++i)
{
m_cmds.push_back(dev.createUniversalCommandList());
}
//while (true)
while (cpuTime > 14.f)
{
if (window.simpleReadMessages())
break;
// Rendertarget
b.start(false);
m_cmds[0].setViewPort(port);
auto backBufferIndex = sc->GetCurrentBackBufferIndex();
m_cmds[0].ClearRenderTargetView(sc[backBufferIndex], vec);
m_cmds[0].setRenderTarget(sc[backBufferIndex]);
for (size_t i = 1; i < lbs.threadCount(); ++i)
{
m_cmds[i].setViewPort(port);
//m_cmds[i].ClearRenderTargetView(sc[backBufferIndex], vec);
m_cmds[i].setRenderTarget(sc[backBufferIndex]);
}
// graphics begin
lbs.addParallelFor<1>("fillCommands", {}, {}, 0, 100, [&](size_t id, size_t threadIndex)
{
auto& gfx2 = m_cmds[threadIndex];
unsigned workAmount = currentTriangleCount / 100;
unsigned startIndex = static_cast<unsigned>(workAmount * id);
auto bind = gfx2.bind(pipeline);
bind.SRV(0, dstdataSrv);
gfx2.drawInstanced(bind, 3, 1, 0, startIndex);
for (unsigned i = startIndex+1; i < startIndex + workAmount; ++i)
{
gfx2.drawInstancedRaw(3, 1, 0, i);
}
});
lbs.sleepTillKeywords({ "fillCommands" });
for (size_t i = 0; i < lbs.threadCount(); ++i)
{
m_cmds[i].closeList();
queue.submit(m_cmds[i]);
}
cpuTime = b.stop(false) / 1000000.f;
// present
sc->Present(1, 0);
queue.insertFence(fence);
fence.wait();
for (size_t i = 0; i < lbs.threadCount(); ++i)
{
m_cmds[i].resetList();
}
t.tick();
frameTime = static_cast<float>(t.getCurrentNano())*0.000001f;
frameTime = t.analyzeFrames().x();
if (cpuTime > 16.f)
{
currentTriangleCount -= currentTriangleCount / 100;
}
}
F_LOG("frametime: %.3fms CpuTime: %.3fms TriangleCount: %d\n", frameTime, cpuTime, currentTriangleCount);
log.update();
return true;
});
t.addTest("Lots of drawcalls bench, (Multithread), myapi", [&]()
{
using namespace faze;
struct buf
{
float pos[4];
};
auto triangleCount = 2000000;
auto currentTriangleCount = triangleCount;
auto srcdata = dev.createBuffer(ResourceDescriptor()
.Width(triangleCount)
.Format<buf>()
.Usage(ResourceUsage::UploadHeap));
auto dstdata = dev.createBuffer(ResourceDescriptor()
.Width(triangleCount)
.Format<buf>());
auto dstdataSrv = dev.createBufferSRV(dstdata);
std::random_device rd;
std::mt19937 gen(rd());
std::uniform_real_distribution<> dis(-0.8f, 0.8f);
std::uniform_real_distribution<> dis2(0.f, 1.f);
{
auto tmp = srcdata.Map<buf>();
for (int i = 0;i < triangleCount; ++i)
{
auto& it = tmp[i].pos;
it[0] = static_cast<float>(dis(gen));
it[1] = static_cast<float>(dis(gen));
it[2] = static_cast<float>(dis2(gen));
}
}
gfx.CopyResource(dstdata, srcdata);
GpuFence fence = dev.createFence();
gfx.closeList();
queue.submit(gfx);
queue.insertFence(fence);
auto pipeline = dev.createGraphicsPipeline(GraphicsPipelineDescriptor()
.PixelShader("tests/stress/pixel")
.VertexShader("tests/stress/vertex_triangle")
.setRenderTargetCount(1)
.RTVFormat(0, FormatType::R8G8B8A8_UNORM_SRGB)
.DepthStencil(DepthStencilDescriptor().DepthEnable(false)));
auto vec = faze::vec4({ 0.2f, 0.2f, 0.2f, 1.0f });
fence.wait();
gfx.resetList();
WTime t;
t.firstTick();
float frameTime = 30.f;
float cpuTime = 30.f;
Bentsumaakaa b;
LBS lbs;
std::vector<GfxCommandList> m_cmds;
for (size_t i = 0; i < lbs.threadCount(); ++i)
{
m_cmds.push_back(dev.createUniversalCommandList());
}
//while (true)
while (cpuTime > 14.f)
{
if (window.simpleReadMessages())
break;
// Rendertarget
b.start(false);
m_cmds[0].setViewPort(port);
auto backBufferIndex = sc->GetCurrentBackBufferIndex();
m_cmds[0].ClearRenderTargetView(sc[backBufferIndex], vec);
m_cmds[0].setRenderTarget(sc[backBufferIndex]);
for (size_t i = 1; i < lbs.threadCount(); ++i)
{
m_cmds[i].setViewPort(port);
//m_cmds[i].ClearRenderTargetView(sc[backBufferIndex], vec);
m_cmds[i].setRenderTarget(sc[backBufferIndex]);
}
// graphics begin
lbs.addParallelFor<1>("fillCommands", {}, {}, 0, 100, [&](size_t id, size_t threadIndex)
{
auto& gfx2 = m_cmds[threadIndex];
unsigned workAmount = currentTriangleCount / 100;
unsigned startIndex = static_cast<unsigned>(workAmount * id);
for (unsigned i = startIndex; i < startIndex + workAmount; ++i)
{
auto bind = gfx2.bind(pipeline);
bind.SRV(0, dstdataSrv);
gfx2.drawInstanced(bind, 3, 1, 0, i);
}
});
lbs.sleepTillKeywords({ "fillCommands" });
for (size_t i = 0; i < lbs.threadCount(); ++i)
{
m_cmds[i].closeList();
queue.submit(m_cmds[i]);
}
cpuTime = b.stop(false) / 1000000.f;
// present
sc->Present(1, 0);
queue.insertFence(fence);
fence.wait();
for (size_t i = 0; i < lbs.threadCount(); ++i)
{
m_cmds[i].resetList();
}
t.tick();
frameTime = static_cast<float>(t.getCurrentNano())*0.000001f;
frameTime = t.analyzeFrames().x();
if (cpuTime > 16.f)
{
currentTriangleCount -= currentTriangleCount / 40;
}
}
F_LOG("frametime: %.3fms CpuTime: %.3fms TriangleCount: %d\n", frameTime, cpuTime, currentTriangleCount);
log.update();
return true;
});
t.addTest("Lots of drawcalls bench, (single thread), rough baseline, frametime 20ms", [&]()
{
using namespace faze;
struct buf
{
float pos[4];
};
auto triangleCount = 800000;
auto currentTriangleCount = triangleCount;
auto srcdata = dev.createBuffer(ResourceDescriptor()
.Width(triangleCount)
.Format<buf>()
.Usage(ResourceUsage::UploadHeap));
auto dstdata = dev.createBuffer(ResourceDescriptor()
.Width(triangleCount)
.Format<buf>());
auto dstdataSrv = dev.createBufferSRV(dstdata);
std::random_device rd;
std::mt19937 gen(rd());
std::uniform_real_distribution<> dis(-0.8f, 0.8f);
std::uniform_real_distribution<> dis2(0.f, 1.f);
{
auto tmp = srcdata.Map<buf>();
for (int i = 0;i < triangleCount; ++i)
{
auto& it = tmp[i].pos;
it[0] = static_cast<float>(dis(gen));
it[1] = static_cast<float>(dis(gen));
it[2] = static_cast<float>(dis2(gen));
}
}
gfx.CopyResource(dstdata, srcdata);
GpuFence fence = dev.createFence();
gfx.closeList();
queue.submit(gfx);
queue.insertFence(fence);
fence.wait();
gfx.resetList();
auto pipeline = dev.createGraphicsPipeline(GraphicsPipelineDescriptor()
.PixelShader("tests/stress/pixel")
.VertexShader("tests/stress/vertex_triangle")
.setRenderTargetCount(1)
.RTVFormat(0, FormatType::R8G8B8A8_UNORM_SRGB)
.DepthStencil(DepthStencilDescriptor().DepthEnable(false)));
auto vec = faze::vec4({ 0.2f, 0.2f, 0.2f, 1.0f });
WTime t;
t.firstTick();
float frameTime = 30.f;
float cpuTime = 30.f;
Bentsumaakaa b;
int beenUnderLimit = 0;
while (beenUnderLimit < 10)
{
if (frameTime < 20.f)
beenUnderLimit++;
if (window.simpleReadMessages())
break;
// Rendertarget
b.start(false);
gfx.setViewPort(port);
auto backBufferIndex = sc->GetCurrentBackBufferIndex();
gfx.ClearRenderTargetView(sc[backBufferIndex], vec);
gfx.setRenderTarget(sc[backBufferIndex]);
// graphics begin
{
auto bind = gfx.bind(pipeline);
bind.SRV(0, dstdataSrv);
gfx.drawInstanced(bind, 3, 1, 0, 0);
for (int i = 1; i < currentTriangleCount; ++i)
{
gfx.drawInstancedRaw(3, 1, 0, i); // this is the cheat.
}
}
// submit all
gfx.closeList();
queue.submit(gfx);
cpuTime = b.stop(false) / 1000000.f;
// present
sc->Present(1, 0);
queue.insertFence(fence);
fence.wait();
gfx.resetList();
t.tick();
frameTime = static_cast<float>(t.getCurrentNano())*0.000001f;
if (frameTime > 20.f)
{
currentTriangleCount -= currentTriangleCount / 100;
}
}
F_LOG("frametime: %.3fms CpuTime: %.3fms TriangleCount: %d\n", frameTime, cpuTime, currentTriangleCount);
log.update();
//fence.wait();
return true;
});
t.addTest("Lots of drawcalls bench, (single thread), myapi, frametime 20ms", [&]()
{
using namespace faze;
struct buf
{
float pos[4];
};
auto triangleCount = 100000;
auto currentTriangleCount = triangleCount;
auto srcdata = dev.createBuffer(ResourceDescriptor()
.Width(triangleCount)
.Format<buf>()
.Usage(ResourceUsage::UploadHeap));
auto dstdata = dev.createBuffer(ResourceDescriptor()
.Width(triangleCount)
.Format<buf>());
auto dstdataSrv = dev.createBufferSRV(dstdata);
std::random_device rd;
std::mt19937 gen(rd());
std::uniform_real_distribution<> dis(-0.8f, 0.8f);
std::uniform_real_distribution<> dis2(0.f, 1.f);
{
auto tmp = srcdata.Map<buf>();
for (int i = 0;i < triangleCount; ++i)
{
auto& it = tmp[i].pos;
it[0] = static_cast<float>(dis(gen));
it[1] = static_cast<float>(dis(gen));
it[2] = static_cast<float>(dis2(gen));
}
}
gfx.CopyResource(dstdata, srcdata);
GpuFence fence = dev.createFence();
gfx.closeList();
queue.submit(gfx);
queue.insertFence(fence);
auto pipeline = dev.createGraphicsPipeline(GraphicsPipelineDescriptor()
.PixelShader("tests/stress/pixel")
.VertexShader("tests/stress/vertex_triangle")
.setRenderTargetCount(1)
.RTVFormat(0, FormatType::R8G8B8A8_UNORM_SRGB)
.DepthStencil(DepthStencilDescriptor().DepthEnable(false)));
auto vec = faze::vec4({ 0.2f, 0.2f, 0.2f, 1.0f });
fence.wait();
gfx.resetList();
WTime t;
t.firstTick();
float frameTime = 30.f;
float cpuTime = 30.f;
Bentsumaakaa b;
LBS lbs;
int beenUnderLimit = 0;
while (beenUnderLimit < 10)
{
if (frameTime < 20.f)
beenUnderLimit++;
if (window.simpleReadMessages())
break;
// Rendertarget
b.start(false);
gfx.setViewPort(port);
auto backBufferIndex = sc->GetCurrentBackBufferIndex();
gfx.ClearRenderTargetView(sc[backBufferIndex], vec);
gfx.setRenderTarget(sc[backBufferIndex]);
// graphics begin
{
for (int i = 0; i < currentTriangleCount; ++i)
{
auto bind = gfx.bind(pipeline);
bind.SRV(0, dstdataSrv);
gfx.drawInstanced(bind, 3, 1, 0, i);
}
}
// submit all
gfx.closeList();
queue.submit(gfx);
cpuTime = b.stop(false) / 1000000.f;
// present
sc->Present(1, 0);
queue.insertFence(fence);
fence.wait();
gfx.resetList();
t.tick();
frameTime = static_cast<float>(t.getCurrentNano())*0.000001f;
//frameTime = t.analyzeFrames().x();
if (frameTime > 20.f)
{
currentTriangleCount -= currentTriangleCount / 100;
}
}
//fence.wait();
F_LOG("frametime: %.3fms CpuTime: %.3fms TriangleCount: %d\n", frameTime, cpuTime, currentTriangleCount);
log.update();
return true;
});
t.addTest("Lots of drawcalls bench, (Multithread), baseline, frametime 20ms", [&]()
{
using namespace faze;
struct buf
{
float pos[4];
};
auto triangleCount = 1000000;
auto currentTriangleCount = triangleCount;
auto srcdata = dev.createBuffer(ResourceDescriptor()
.Width(triangleCount)
.Format<buf>()
.Usage(ResourceUsage::UploadHeap));
auto dstdata = dev.createBuffer(ResourceDescriptor()
.Width(triangleCount)
.Format<buf>());
auto dstdataSrv = dev.createBufferSRV(dstdata);
std::random_device rd;
std::mt19937 gen(rd());
std::uniform_real_distribution<> dis(-0.8f, 0.8f);
std::uniform_real_distribution<> dis2(0.f, 1.f);
{
auto tmp = srcdata.Map<buf>();
for (int i = 0;i < triangleCount; ++i)
{
auto& it = tmp[i].pos;
it[0] = static_cast<float>(dis(gen));
it[1] = static_cast<float>(dis(gen));
it[2] = static_cast<float>(dis2(gen));
}
}
gfx.CopyResource(dstdata, srcdata);
GpuFence fence = dev.createFence();
gfx.closeList();
queue.submit(gfx);
queue.insertFence(fence);
auto pipeline = dev.createGraphicsPipeline(GraphicsPipelineDescriptor()
.PixelShader("tests/stress/pixel")
.VertexShader("tests/stress/vertex_triangle")
.setRenderTargetCount(1)
.RTVFormat(0, FormatType::R8G8B8A8_UNORM_SRGB)
.DepthStencil(DepthStencilDescriptor().DepthEnable(false)));
auto vec = faze::vec4({ 0.2f, 0.2f, 0.2f, 1.0f });
fence.wait();
gfx.resetList();
WTime t;
t.firstTick();
float frameTime = 30.f;
float cpuTime = 30.f;
Bentsumaakaa b;
LBS lbs;
std::vector<GfxCommandList> m_cmds;
for (size_t i = 0; i < lbs.threadCount(); ++i)
{
m_cmds.push_back(dev.createUniversalCommandList());
}
//while (true)
int beenUnderLimit = 0;
while (beenUnderLimit < 10)
{
if (frameTime < 20.f)
beenUnderLimit++;
if (window.simpleReadMessages())
break;
// Rendertarget
b.start(false);
m_cmds[0].setViewPort(port);
auto backBufferIndex = sc->GetCurrentBackBufferIndex();
m_cmds[0].ClearRenderTargetView(sc[backBufferIndex], vec);
m_cmds[0].setRenderTarget(sc[backBufferIndex]);
for (size_t i = 1; i < lbs.threadCount(); ++i)
{
m_cmds[i].setViewPort(port);
//m_cmds[i].ClearRenderTargetView(sc[backBufferIndex], vec);
m_cmds[i].setRenderTarget(sc[backBufferIndex]);
}
// graphics begin
lbs.addParallelFor<1>("fillCommands", {}, {}, 0, 100, [&](size_t id, size_t threadIndex)
{
auto& gfx2 = m_cmds[threadIndex];
unsigned workAmount = static_cast<unsigned>(currentTriangleCount / 100);
unsigned startIndex = static_cast<unsigned>(workAmount * id);
auto bind = gfx2.bind(pipeline);
bind.SRV(0, dstdataSrv);
gfx2.drawInstanced(bind, 3, 1, 0, startIndex);
for (unsigned i = startIndex + 1; i < startIndex + workAmount; ++i)
{
gfx2.drawInstancedRaw(3, 1, 0, i);
}
});
lbs.sleepTillKeywords({ "fillCommands" });
for (size_t i = 0; i < lbs.threadCount(); ++i)
{
m_cmds[i].closeList();
queue.submit(m_cmds[i]);
}
cpuTime = b.stop(false) / 1000000.f;
// present
sc->Present(1, 0);
queue.insertFence(fence);
fence.wait();
for (size_t i = 0; i < lbs.threadCount(); ++i)
{
m_cmds[i].resetList();
}
t.tick();
frameTime = static_cast<float>(t.getCurrentNano())*0.000001f;
//frameTime = t.analyzeFrames().x();
if (frameTime > 20.f)
{
currentTriangleCount -= currentTriangleCount / 120;
}
}
F_LOG("frametime: %.3fms CpuTime: %.3fms TriangleCount: %d\n", frameTime, cpuTime, currentTriangleCount);
log.update();
return true;
});
t.addTest("Lots of drawcalls bench, (Multithread), myapi, frametime 20ms", [&]()
{
using namespace faze;
struct buf
{
float pos[4];
};
auto triangleCount = 400000;
auto currentTriangleCount = triangleCount;
auto srcdata = dev.createBuffer(ResourceDescriptor()
.Width(triangleCount)
.Format<buf>()
.Usage(ResourceUsage::UploadHeap));
auto dstdata = dev.createBuffer(ResourceDescriptor()
.Width(triangleCount)
.Format<buf>());
auto dstdataSrv = dev.createBufferSRV(dstdata);
std::random_device rd;
std::mt19937 gen(rd());
std::uniform_real_distribution<> dis(-0.8f, 0.8f);
std::uniform_real_distribution<> dis2(0.f, 1.f);
{
auto tmp = srcdata.Map<buf>();
for (int i = 0;i < triangleCount; ++i)
{
auto& it = tmp[i].pos;
it[0] = static_cast<float>(dis(gen));
it[1] = static_cast<float>(dis(gen));
it[2] = static_cast<float>(dis2(gen));
}
}
gfx.CopyResource(dstdata, srcdata);
GpuFence fence = dev.createFence();
gfx.closeList();
queue.submit(gfx);
queue.insertFence(fence);
auto pipeline = dev.createGraphicsPipeline(GraphicsPipelineDescriptor()
.PixelShader("tests/stress/pixel")
.VertexShader("tests/stress/vertex_triangle")
.setRenderTargetCount(1)
.RTVFormat(0, FormatType::R8G8B8A8_UNORM_SRGB)
.DepthStencil(DepthStencilDescriptor().DepthEnable(false)));
auto vec = faze::vec4({ 0.2f, 0.2f, 0.2f, 1.0f });
fence.wait();
gfx.resetList();
WTime t;
t.firstTick();
float frameTime = 30.f;
float cpuTime = 30.f;
Bentsumaakaa b;
LBS lbs;
std::vector<GfxCommandList> m_cmds;
for (size_t i = 0; i < lbs.threadCount(); ++i)
{
m_cmds.push_back(dev.createUniversalCommandList());
}
//while (true)
int beenUnderLimit = 0;
while (beenUnderLimit < 10)
{
if (frameTime < 20.f)
beenUnderLimit++;
if (window.simpleReadMessages())
break;
// Rendertarget
b.start(false);
m_cmds[0].setViewPort(port);
auto backBufferIndex = sc->GetCurrentBackBufferIndex();
m_cmds[0].ClearRenderTargetView(sc[backBufferIndex], vec);
m_cmds[0].setRenderTarget(sc[backBufferIndex]);
for (size_t i = 1; i < lbs.threadCount(); ++i)
{
m_cmds[i].setViewPort(port);
//m_cmds[i].ClearRenderTargetView(sc[backBufferIndex], vec);
m_cmds[i].setRenderTarget(sc[backBufferIndex]);
}
// graphics begin
lbs.addParallelFor<1>("fillCommands", {}, {}, 0, 100, [&](size_t id, size_t threadIndex)
{
auto& gfx2 = m_cmds[threadIndex];
size_t workAmount = currentTriangleCount / 100;
size_t startIndex = workAmount * id;
for (unsigned i = static_cast<unsigned>(startIndex); i < static_cast<unsigned>(startIndex + workAmount); ++i)
{
auto bind = gfx2.bind(pipeline);
bind.SRV(0, dstdataSrv);
gfx2.drawInstanced(bind, 3, 1, 0, i);
}
});
lbs.sleepTillKeywords({ "fillCommands" });
for (size_t i = 0; i < lbs.threadCount(); ++i)
{
m_cmds[i].closeList();
queue.submit(m_cmds[i]);
}
cpuTime = b.stop(false) / 1000000.f;
// present
sc->Present(1, 0);
queue.insertFence(fence);
fence.wait();
for (size_t i = 0; i < lbs.threadCount(); ++i)
{
m_cmds[i].resetList();
}
t.tick();
frameTime = static_cast<float>(t.getCurrentNano())*0.000001f;
//frameTime = t.analyzeFrames().x();
if (frameTime > 20.f)
{
currentTriangleCount -= currentTriangleCount / 120;
}
}
F_LOG("frametime: %.3fms CpuTime: %.3fms TriangleCount: %d\n", frameTime, cpuTime, currentTriangleCount);
log.update();
return true;
});
t.runTests();
}
std::unique_ptr<SwapChain> SwapChain_<>::Create( const Config &config ) {
SwapChain* retval = nullptr;
#if defined(LUDI_PLATFORM_XENON)
retval = nullptr;
#elif defined(LUDI_PLATFORM_PS3)
retval = nullptr;
#elif defined(LUDI_PLATFORM_WII)
retval = nullptr;
#elif LUDI_PLATFORM_ID == LUDI_PLATFORM_WINDOWS
// Attempt to get the DXGI Factory.
Device* device = Device::Get();
Driver* driver = device->driver();
// Attempt to create the swap chain.
IDXGISwapChain* chain = nullptr;
// Create the swap chain
DXGI_SWAP_CHAIN_DESC desc;
desc = ConversionDx11::To(config);
HRESULT hresult = driver->factory()->CreateSwapChain(device->device(), &desc, &chain);
if (FAILED(hresult)) {
//Log::Get().Write( L"Failed to create swap chain!" );
ASSERT(false);
return nullptr;
}
// Retrieve the texture buffer (render-texture)
ID3D11Texture2D* texture = nullptr;
hresult = chain->GetBuffer( 0, __uuidof( ID3D11Texture2D ), reinterpret_cast< void** >( &texture ) );
if (FAILED(hresult)) {
//Log::Get().Write( L"Failed to get swap chain texture resource!" );
ASSERT(false);
return nullptr;
}
// Create the swap chain render target view
ID3D11RenderTargetView* view;
hresult = (*device)->CreateRenderTargetView(texture, nullptr, &view);
if (FAILED(hresult)) {
//Log::Get().Write( L"Failed to create render target view for swap chain!" );
ASSERT(false);
return nullptr;
}
// Create the swap chain, set the render texture and render target view
retval = new SwapChain_<LUDI_PLATFORM_WINDOWS>(chain);
retval->set_render_texture(new Texture(texture));
retval->set_render_target_view(new RenderTargetView(view));
#elif LUDI_PLATFORM_ID == LUDI_PLATFORM_OSX
retval = new SwapChain_<LUDI_PLATFORM_OSX>(nullptr);
retval->set_render_texture(new Texture());
retval->set_render_target_view(new RenderTargetView());
#else // NO VALID PLATFORM
retval = nullptr;
#endif
return std::unique_ptr<SwapChain>(retval);
}
