Ok with my project I am drawing an image, but the problem is that i need to perform an operation on every pixel. Currently its running at about 5-10fps and it needs to be real time. I have tried most simplistic speed measures. I am using draw pixels, as I thought that the constant loading of textures to memory would make texturing a quad slower? The image can get to about 500^2. Currently using gl bytes and GL_RGB, any other pointers or things I can turn on/off that will make an impact? My client is suggesting that I pre create all the images… uuuggghh, hardly ideal, especially due to the amount of code that would become redundant!
From the Red Book:
glDrawPixels() - Writes a rectangular array of pixels from data kept in processor memory into the framebuffer at the current raster position specified by glRasterPos*().
Try using textures - a 512 x 512 RGBA image is 1 MB, so you can keep a fair number of them resident even on a card with low memory (say, 32 MB, by todays standards), plus they can be DMA’d to the card, which is probably the real advantage.
“I am drawing an image”
What type of application is that?
“the problem is that i need to perform an operation on every pixel”
What operation exactly?
Could you do that using fragment shaders on the GPU?
“I am using draw pixels”
Generally not a fast solution.
Make sure everything affecting a simple 2D copy is off, like depth test, lighting, texturing, vertex and fragment programs, blending, etc.
“Currently using gl bytes and GL_RGB”
Try GL_UNSIGNED_BYTE in GL_BGR or GL_BGRA order as input. Some hardware uses that as native format and wouldn’t need to swap R and B of your input data.
“My client is suggesting that I pre create all the images”
Do you get them from a changing input or are they all based on one single input with the “operation you need to perform on each pixel”?