Posts: 469 from 2003/4/11
many thanks for explanation and nice plans.
I like to contribute to the bounty if it will be.
And I have some questions. I am not developer or expert on computer graphics, but want to ask, what part of graphics cards hardware this future MorphOS OpenGL implementation utilize?
R200 - R500 have in HW spec (techpowerup.com):
TMUs (texture mapping units)
ROPs (render output units)
R600, Evergreen, Northern Island have:
So what is most important for TinyGL/OpenGL speed in MorphOS? Can MorphOS utilize all Shaders or cores?
Please, in short is enough. I make often a benchmarking and it helps a lot.
I'll try to make it as short as possible :)
Prior to R600 (so up to and including R500), vertex shaders and pixel shaders are handled completely differently by completely different parts of the hardware. Starting with R600, Radeons implemented what is known as unified shaders
, where it's the same hardware that executes every kind of shader. That's why you see this difference.
When it comes to performance, it's pretty hard to generalise just based on specs alone. For current
MorphOS OpenGL applications, the limiting factor in OpenGL performance is almost always memory bandwidth. For future ones that will use complex shaders, having more and higher clocked shader cores will become more important than memory bandwidth.
As for actually using all the cores, that is pretty much transparently handled on the hardware itself. The driver doesn't program individual cores, it just submits a job to the GPU, and the GPU internally handles splitting up this job into smaller parts and handing those parts to individual cores. So short answer: Yes, MorphOS will be able to handle any shader core count.
I rarely log in to MorphZone which means that I often miss private messages sent on here. If you wish to contact me, please email me at [username]@asgaard.morphos-team.net, where [username] is my username here on MorphZone.