Making textures from memory

Hoodlum Smurf · Post by **Hoodlum Smurf** » Wed Jul 06, 2005 4:39 pm

I just recently ported my application from Ogre 0.14 to 1.0. I have a section of code that runs each frame where a texture is updated based on a new image which is in memory.

The code I have to do this is below:

Code: Select all

	HardwarePixelBufferSharedPtr pBuf = m_pTexture.get()->getBuffer();
	pBuf->blitFromMemory(PixelBox(m_pImage->GetPrevWidth(),m_pImage->GetPrevHeight(),nDepth,PF_A8R8G8B8,m_pImage->GetByteBuffer()));

The blitFromMemory call works correctly and there are no errors, but it is very slow. Noticeably slower than when I had it running on Ogre 0.14. Is there a more efficient way to perform this operation under Ogre 1.0?

Post by **DWORD** » Wed Jul 06, 2005 5:09 pm

See Creating dynamic textures in the wiki, and the DynTex demo.

Hoodlum Smurf · Post by **Hoodlum Smurf** » Wed Jul 06, 2005 8:46 pm

Thanks for the reply DWORD. I took a look at the link you sent and the DynTex demo. I am basically doing the same thing.

I did take a look at the source for blitFromMemory though and I think I know why I'm having performance issues. I think it's the mipmap generation. I don't want Ogre to automatically generate mipmaps even if the card is capable of it. From what I see the best way to prevent mipmaps from automatically being generated is to affect the value of TU_AUTOMIPMAP. Could anyone tell me how to set TU_AUTOMIPMAP to "false"? Or another way of preventing the automatic generation of mipmaps?

Post by **DWORD** » Wed Jul 06, 2005 9:37 pm

You pass the number of mipmaps to generate in the call to TextureManager::createManual(), so just set it to 0 there. There are more options, see the API reference for that function for further details.