WValidator::Result WTimeValidator::validate(const WT_USTRING& input) const
{
if (input.empty())
return WValidator::validate(input);
for (unsigned i = 0; i < formats_.size(); ++i) {
try {
WTime d = WTime::fromString(input, formats_[i]);
if (d.isValid()) {
if (!bottom_.isNull())
if (d < bottom_)
return Result(Invalid, invalidTooEarlyText());
if (!top_.isNull())
if (d > top_)
return Result(Invalid, invalidTooLateText());
return Result(Valid);
}
} catch (std::exception& e) {
LOG_WARN("validate(): " << e.what());
}
}
return Result(Invalid, invalidNotATimeText());
}
bool WTime::operator<= (const WTime& other) const
{
if (!isValid() && !other.isValid())
return true;
if (!isValid() || !other.isValid())
throw InvalidTimeException();
return time_ <= other.time_;
}
int WTime::msecsTo(const WTime& t) const
{
if (!isValid() || !t.isValid())
throw InvalidTimeException();
return t.time_ - time_;
}
void WTimePicker::setTime(const WTime& time)
{
if (!time.isValid()) {
LOG_ERROR("Time is invalid!");
return;
}
int hours = 0;
if (formatAp()) {
hours = time.pmhour();
if (time.hour() < 12)
cbAP_->setCurrentIndex(0);
else
cbAP_->setCurrentIndex(1);
} else
hours = time.hour();
int minutes = time.minute();
int seconds = time.second();
int millisecond = time.msec();
sbhour_->setValue(hours);
sbminute_->setValue(minutes);
sbsecond_->setValue(seconds);
if (formatMs()) {
sbmillisecond_->setValue(millisecond);
}
}
void WTimePicker::setTime(const WTime& time)
{
if(!time.isValid()) {
LOG_ERROR("Time is invalid!");
return;
}
std::string hoursStr("0"), minutesStr("00"), secondsStr("00");
try {
hoursStr = time.toString("hh").toUTF8();
minutesStr = time.toString("mm").toUTF8();
secondsStr = time.toString("ss").toUTF8();
} catch(const boost::bad_lexical_cast& ex) {
LOG_ERROR("boost::bad_lexical_cast caught in WTimePicker::time()");
}
hourText_->setText(hoursStr);
minuteText_->setText(minutesStr);
secondText_->setText(secondsStr);
}
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();
}
