shared graphics memory
category: general [glöplog]
I'm trying to do a little something for low end crappy cheap-o-puters onboard GPUs and I came up with this dilema.
Since they all have shared memory when I copy the textures and framebuffers to the "video memory" I get a copy of the same thing, maybe in a different format, on the same place. And since it's being read from the same place and all does it make any performance difference using vertexbuffers and framebuffers and stuff?
Since they all have shared memory when I copy the textures and framebuffers to the "video memory" I get a copy of the same thing, maybe in a different format, on the same place. And since it's being read from the same place and all does it make any performance difference using vertexbuffers and framebuffers and stuff?
Yes it makes a difference.
The separate "upload pass" allows for optimizations, pre-processing and format conversions which makes the data more suitable for GPU processing.
And seriously; in the case of textures, what else would you do anyway? Create a new API where the GPU is supposed to read directly from a memory buffer allocated by the application?
Also; what do you mean by copying the framebuffer to video memory? I prefer the GPU to produce the framebuffer for me..
The separate "upload pass" allows for optimizations, pre-processing and format conversions which makes the data more suitable for GPU processing.
And seriously; in the case of textures, what else would you do anyway? Create a new API where the GPU is supposed to read directly from a memory buffer allocated by the application?
Also; what do you mean by copying the framebuffer to video memory? I prefer the GPU to produce the framebuffer for me..
Ever heard of Amiga's chipram ?
:)
:)
no
That's more or less what I had in mind.
I meant render to texture.
So, I don't need to make any special optimization, just a regular "optimization"?
I meant render to texture.
So, I don't need to make any special optimization, just a regular "optimization"?