• 11
• 9
• 10
• 9
• 10
• ### Similar Content

• By lxjk
Hi guys,
There are many ways to do light culling in tile-based shading. I've been playing with this idea for a while, and just want to throw it out there.
Because tile frustums are general small compared to light radius, I tried using cone test to reduce false positives introduced by commonly used sphere-frustum test.
On top of that, I use distance to camera rather than depth for near/far test (aka. sliced by spheres).
This method can be naturally extended to clustered light culling as well.
The following image shows the general ideas

Performance-wise I get around 15% improvement over sphere-frustum test. You can also see how a single light performs as the following: from left to right (1) standard rendering of a point light; then tiles passed the test of (2) sphere-frustum test; (3) cone test; (4) spherical-sliced cone test

I put the details in my blog post (https://lxjk.github.io/2018/03/25/Improve-Tile-based-Light-Culling-with-Spherical-sliced-Cone.html), GLSL source code included!

Eric

• Good evening everyone!

I was wondering if there is something equivalent of  GL_NV_blend_equation_advanced for AMD?
Basically I'm trying to find more compatible version of it.

Thank you!

• Hello guys,

How do I know? Why does wavefront not show for me?
I already checked I have non errors yet.

And my download (mega.nz) should it is original but I tried no success...
- Add blend source and png file here I have tried tried,.....

PS: Why is our community not active? I wait very longer. Stop to lie me!
Thanks !

• I wasn't sure if this would be the right place for a topic like this so sorry if it isn't.
I'm currently working on a project for Uni using FreeGLUT to make a simple solar system simulation. I've got to the point where I've implemented all the planets and have used a Scene Graph to link them all together. The issue I'm having with now though is basically the planets and moons orbit correctly at their own orbit speeds.
I'm not really experienced with using matrices for stuff like this so It's likely why I can't figure out how exactly to get it working. This is where I'm applying the transformation matrices, as well as pushing and popping them. This is within the Render function that every planet including the sun and moons will have and run.
if (tag != "Sun") { glRotatef(orbitAngle, orbitRotation.X, orbitRotation.Y, orbitRotation.Z); } glPushMatrix(); glTranslatef(position.X, position.Y, position.Z); glRotatef(rotationAngle, rotation.X, rotation.Y, rotation.Z); glScalef(scale.X, scale.Y, scale.Z); glDrawElements(GL_TRIANGLES, mesh->indiceCount, GL_UNSIGNED_SHORT, mesh->indices); if (tag != "Sun") { glPopMatrix(); } The "If(tag != "Sun")" parts are my attempts are getting the planets to orbit correctly though it likely isn't the way I'm meant to be doing it. So I was wondering if someone would be able to help me? As I really don't have an idea on what I would do to get it working. Using the if statement is truthfully the closest I've got to it working but there are still weird effects like the planets orbiting faster then they should depending on the number of planets actually be updated/rendered.

• Hello everyone,
I have problem with texture

# OpenGL Organizing rendering code

This topic is 998 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

## Recommended Posts

I'm writing a small engine, mainly for educational purposes. Where I want it to be able to use differente APIs (OpenGL, DirectX, etc).

So In my intend to separate API specific code from the engine's logic I thought on using pImpl Idiom, something like this:

template<class T>
class RenderManager
{
public:
// pImpl-> Create & Release resources
// pImpl-> Draw scene objects
private:
T pImpl;  // DirectX/OpenGL specific class
};


Everything was ok with that approach, until I needed to add rendering code to the RenderManager which is not API specific (like adding passes for Shadows, AlphaBlend, etc).

Adding that code to my pImpl class doesn't seem right

template<class T>
class RenderManager
{
public:
// pImpl-> Create & Release resources
void Update( )
{
myObjects = // retrive objects for shadow pass
pImpl->Pass( myObjects );

myObjects = // retrive objects for diffuse pass
pImpl->Pass( myObjects );

myObjects = // retrive objects for alpha blend pass
pImpl->Pass( myObjects );
}
private:
T pImpl;  // DirectX/OpenGL specific class
};


So I could use an abstract class from which API specific code inherits:
class RenderManagerBase
{
virtual void Pass( objectList ) = 0;
virtual void // Create & Release resources = 0;

void Update( )
{
myObjects = // retrive objects for shadow pass
Pass( myObjects );

myObjects = // retrive objects for diffuse pass
Pass( myObjects );

myObjects = // retrive objects for alpha blend pass
Pass( myObjects );
}
};

class RenderManager_dx11 : public RenderManagerBase
{
};

class RenderManager_ogl : public RenderManagerBase
{
};


But using virtual functions in rendering code could bring performance issues to something very time-critical such as the rendering.

So I thought about using some kind of reverse inheritance (the general class inheriting from the API specific class):


class RenderManager_dx11
{
};

class RenderManager_ogl
{
};

template<class T>
class RenderManagerBase : public T
{
void Update( )
{
myObjects = // retrive objects for shadow pass
Pass( myObjects );

myObjects = // retrive objects for diffuse pass
Pass( myObjects );

myObjects = // retrive objects for alpha blend pass
Pass( myObjects );
}
};


Not very "standard" way of doing it, but since I know there will only be two classes (one inherited from the other), this way the calls to the API specific methods wouldn't have the performance problems of using virtual functions.

So.... I'd like to know your oppinion...  how do you happen to handle your rendering code to make it API agnostic and writing rendering code not API-specific?

Thanks

PD.- Also... If I have a pure virtual function, which gets implemented in its only derived class, I'm guessing it probably wouldn't have such a performance impact since there won't be more than one function to look up to...  (but still would have better performance calling the function directly)

Edited by Menny Rivers

##### Share on other sites
This is an easy implementation that doesn't require any virtuals or inheritance/templating

//put in some header, lets say "RenderingAbstraction.h"

#if defined(USE_D3D11)

typedef CD3D11API CRenderingAPI;

#elif defined(USE_OGL)

typedef COGLAPI CRenderingAPI;

#endif

#include "RenderingAbstraction.h"

class CRenderManager
{

void Update()
{
API->Pass(....);

}

CRenderingAPI * API;
};



##### Share on other sites

Thanks AThompson,

That's pretty much what the pImpl Idiom does...  it works, but using a pattern not in its 'pure' form can be a philosofical dilemma to some...

Anyway... I've just read about Static Polymorfism, which is a Curiously Recurring Template Pattern, and I think that pretty much solves the problem to what I'm trying to achieve...

I'll give it a try.

Thanks!

##### Share on other sites

I would be weary of using CRTP. I've done a fair amount of tests with that. All my cases showed that it was up to two times slower than even a virtual call. I haven't done extensive tests, but I ran a hundred or so test runs in the past week or so while refactoring my rendering code. I'd like to see if anyone else had the same issue.

##### Share on other sites

I would be weary of using CRTP. I've done a fair amount of tests with that. All my cases showed that it was up to two times slower than even a virtual call. I haven't done extensive tests, but I ran a hundred or so test runs in the past week or so while refactoring my rendering code. I'd like to see if anyone else had the same issue.

That doesn't sound right... CRTP code should get optimized out and leave a regular function call. A regular function call should not be slower than a virtual function call...

##### Share on other sites
I thought the same, possibly I have a compiler setting that isn't optimizing fully. I don't have the source with me, but it was the generic crtp method found on wiki or google. I did a simple count to 10000 I believe, and standard/virtual calls were at around .3 milliseconds, while crtp was at 0.64 . I was on win32 using high performance counters.

##### Share on other sites

I would be weary of using CRTP. I've done a fair amount of tests with that. All my cases showed that it was up to two times slower than even a virtual call. I haven't done extensive tests, but I ran a hundred or so test runs in the past week or so while refactoring my rendering code. I'd like to see if anyone else had the same issue.

I don't belive it eigther. Most likely, first of all you are not an optimized build, because 10000 iterations for a simple count should not take 0.3 miliseconds for eigther crpt or virtual calls, or do you have an extraordinarily weak CPU?

Aside from that, even if not running a fully optimized build, you must make sure that your compiler isn't optimizing things out that he would normally not be able to - like devirtualization, etc...

I've run my own test case, using 5 million iterations, and the results where

CRTP: 0.0040035
Virtual: 0.0160113

This is the code I've been using:

const int NUM_ITERATIONS = 5000000;

template<typename T>
class CRTPTest
{
public:

float Testing(int i) const
{
return m_test.Testing(i);
}

private:

T m_test;
};

class CRTPTestImpl
{
public:

float Testing(int i) const
{
return (float)i;
}
};

class InheritanceTest
{
public:

virtual float Testing(int i) = 0;
};

class InheritanceTestImpl :
public InheritanceTest
{
float Testing(int i) override final
{
return (float)i;
}
};

class InheritanceTestImpl2 :
public InheritanceTest
{
float Testing(int i) override final
{
return (float)(i + 5);
}
};

int WINAPI WinMain(HINSTANCE hInstance, HINSTANCE hPrevInstance, LPSTR lpCmdLine, int nCmdShow)
{
{
CRTPTest<CRTPTestImpl> CRTP;

float result = 0.0f;

core::Timer timer;
for(int i = 0; i < NUM_ITERATIONS; i++)
{
result += CRTP.Testing(i);
}

sys::log->Out(sys::LogModule::CUSTOM, sys::LogType::INFO, "CRPT:", result, timer.Duration());
}

{
InheritanceTest* pInheritance = new InheritanceTestImpl;
InheritanceTest* pDummy = new InheritanceTestImpl2; // to hopefully keep the compiler from performing devirtualization

float result = 0.0f;

core::Timer timer;
for(int i = 0; i < NUM_ITERATIONS; i++)
{
result += pInheritance->Testing(i);
}

sys::log->Out(sys::LogModule::CUSTOM, sys::LogType::INFO, "Virtual:", result, timer.Duration());
sys::log->Out(sys::LogModule::CUSTOM, sys::LogType::INFO, pDummy->Testing(std::rand()));
}

}


You can run it yourself, you just need to replace the log calls and the timer (interally uses std::chrono, but you might use QueryPeformanceTimer as well).
Notice I've taking great care to ensure that the compiler is not able to optimize things he should not be able to in a real world use case of CRTP vs. inheritance.

Edited by Juliean

##### Share on other sites

You appear to be way too focused on patterns. Don't do that - don't just apply a pattern because its a pattern. Use the solution that fits the case best - most patterns are just names for common programming solutions with potentially different variations, but if you insist on doing things exactly like some pattern just because someone showed it that way, you are in for a world of hurt ;)

Thanks Juliean.... yeah, I guess you're right... maybe I'm just overthinking it too much.

I would be weary of using CRTP. I've done a fair amount of tests with that. All my cases showed that it was up to two times slower than even a virtual call. I haven't done extensive tests, but I ran a hundred or so test runs in the past week or so while refactoring my rendering code. I'd like to see if anyone else had the same issue.

I agree with everyone else here... I'm guessing maybe your tests didn't had inlining enabled?

check out this artcle about the cost of dynamic virtual calls vs static crtp dispatch in C++:

http://eli.thegreenplace.net/2013/12/05/the-cost-of-dynamic-virtual-calls-vs-static-crtp-dispatch-in-c

Thanks!