Driver resources management
Author
Hello,
My question is simple : what about this : D3DCREATE_DISABLE_DRIVER_MANAGEMENT ?
What is generaly the behavior of driver-driven managed resources ? What's the main diference between RUN-TIME and Driver driven resource ?
Thank You
Quote:Original post by Woodchuck
Hello,
My question is simple : what about this : D3DCREATE_DISABLE_DRIVER_MANAGEMENT ?
What is generaly the behavior of driver-driven managed resources ? What's the main diference between RUN-TIME and Driver driven resource ?
To be honest, nobody knows the exact specifics of driver internals (except for the IVH driver programmers [wink]). However, they move resources around as they see fit, depending on the current hardware. For example, ATI drivers may swap a texture in and out of video memory when event XXX occurs. However, Nvidia drivers may swap it when some other event YYY occurs. It all depends on what the IHVs have determined that their cards are best at doing. There are bazillions of tricks they employ just to squeeze as much as they can out.
Sorry I can't be of more substance...perhaps some Nvidia and ATI guy will spot this though.
Author
Thank You. I wanted to know if some things are well known about this. Seems that no. =)
But maybe it is what I wanted to know. We don't know what the driver do, then it is maybe a good things to avoid his management.
But maybe it is what I wanted to know. We don't know what the driver do, then it is maybe a good things to avoid his management.
Quote:Original post by Woodchuck
But maybe it is what I wanted to know. We don't know what the driver do, then it is maybe a good things to avoid his management.
I think you should put some faith in the hard working guys writing drivers for ATI and NVidia. I'm quite confident they know what they're doing. After all, it's what they do.
Author
Quote:Original post by sirobQuote:Original post by Woodchuck
But maybe it is what I wanted to know. We don't know what the driver do, then it is maybe a good things to avoid his management.
I think you should put some faith in the hard working guys writing drivers for ATI and NVidia. I'm quite confident they know what they're doing. After all, it's what they do.
Sorry, this is not what I wanted to say. I just wanted to say that for a specific 3D engine you can maybe write your resources managment in a way that you test with run-time management and have bad results in other machines cause of the driver management. I don't really know if this is founded...
Quote:Original post by Woodchuck
Sorry, this is not what I wanted to say. I just wanted to say that for a specific 3D engine you can maybe write your resources managment in a way that you test with run-time management and have bad results in other machines cause of the driver management. I don't really know if this is founded...
Your engine is not going to decide driver-level details like that though. For example, the driver may determine where in video memory it wants to place a texture and where it may want buffers for the fastest access. This is just information that you don't have, and even if you did, you couldn't act on it. As an end-user of D3D, you do not have the access to do stuff like that.
The bottom line is, the drivers know best. Even if they do slightly naughty things every once in a while [wink]
Author
What you said is true, but what i'm saying is that when you set (resources priorities in our case) your application for a type of software\hardware architecture, and when a part of this architecture is hiden, you do a lots of bench and try to make appear some rules, some bench rules...
If you take this application and you put it in an other software\hardware architecture, your bench rulez become wrong.
Then if your 3D engine run fastly with a run-time resource managment on every machine, I think this is not a good idea to enable driver management. In fact, it depend on how the driver 'jungle' is unforeseeable (don't sure of the english word sorry:p) or if your setting make the application run faster with some main driver.
In other term, for a specific application this is not obligatorily a good idea to let the driver put his hand in it concerning resource management.
If you take this application and you put it in an other software\hardware architecture, your bench rulez become wrong.
Then if your 3D engine run fastly with a run-time resource managment on every machine, I think this is not a good idea to enable driver management. In fact, it depend on how the driver 'jungle' is unforeseeable (don't sure of the english word sorry:p) or if your setting make the application run faster with some main driver.
In other term, for a specific application this is not obligatorily a good idea to let the driver put his hand in it concerning resource management.
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement
