Profiled the terrain demo, including rendering lots of cubes

[Crivens]

I just profiled the terrain demo, plus rendering about 100 small cubes. The function that used the most time?

irr::video::SColor::toOpenGLColor() const

Odd!

This is just the first part of the results:

Code: Select all

Flat profile:

Each sample counts as 0.01 seconds.
  %   cumulative   self              self     total           
 time   seconds   seconds    calls  ms/call  ms/call  name    
 14.26     15.73    15.73                             irr::video::SColor::toOpenGLColor() const
  7.60     24.12     8.39                             irr::core::matrix4::operator*(irr::core::matrix4 const&) const
  6.24     31.01     6.89                             irr::core::matrix4::operator*=(irr::core::matrix4 const&)
  5.67     37.27     6.26                             irr::core::matrix4::makeIdentity()
  4.12     41.81     4.55                             irr::scene::CTerrainSceneNode::getIndex(int const&, int const&, int const&, unsigned int, unsigned int)
  4.11     46.34     4.53 44648444     0.00     0.00  irr::core::vector3d<float>::operator=(irr::core::vector3d<float> const&)
  4.01     50.76     4.42                             irr::core::matrix4::operator=(irr::core::matrix4 const&)
  3.50     54.62     3.86                             irr::core::vector2d<float>::operator=(irr::core::vector2d<float> const&)
  3.20     58.16     3.54                             irr::video::COpenGLDriver::drawIndexedTriangleList(irr::video::S3DVertex2TCoords const*, int, unsigned short const*, int)
  2.61     61.03     2.88                             irr::scene::ISceneNode::OnPostRender(unsigned int)
  2.51     63.81     2.78                             irr::video::S3DVertex2TCoords::operator=(irr::video::S3DVertex2TCoords const&)
  2.15     66.18     2.37                             irr::video::COpenGLDriver::createGLMatrix(float*, irr::core::matrix4 const&)
  2.12     68.52     2.34                             irr::scene::CTerrainSceneNode::OnPreRender()
  1.85     70.56     2.04                             irr::core::matrix4::transformVect(irr::core::vector3d<float>&) const
  1.60     72.32     1.76                             irr::core::matrix4::setRotationRadians(irr::core::vector3d<float> const&)
  1.42     73.89     1.57                             irr::video::COpenGLDriver::drawIndexedTriangleList(irr::video::S3DVertex const*, int, unsigned short const*, int)
  1.33     75.36     1.47                             irr::core::array<irr::video::S3DVertex2TCoords>::operator[](unsigned int)
  1.23     76.72     1.36                             irr::video::CImage::getPixel(int, int)
  1.19     78.03     1.31                             irr::scene::CSceneManager::registerNodeForRendering(irr::scene::ISceneNode*, irr::scene::E_SCENE_NODE_RENDER_PASS)
  1.04     79.17     1.15                             irr::core::array<int>::operator[](unsigned int)

[disanti]

Funny, that function looks VERY optimized. I would assume that it takes that much time because it is called a LOT in COpenGLDriver.cpp.

In SColor.h:

Code: Select all

		//! Converts color to open gl color format.
		//! \return Returns the 32 bit openGL color value.
		inline s32 toOpenGLColor() const 
		{
			return (((color>>24) & 0xff)<<24) |
					(((color)& 0xff)<<16) |
					((color>>8 & 0xff)<<8) |
					((color>>16) & 0xff);
		};

________
Teen vids

[niko]

It's true, this function is called once for every vertex which is renderered in OpenGL mode. But I think I could optimize that away too