lkml.org 
[lkml]   [1998]   [Mar]   [29]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: GGI and cli/sti in X
From
Date
>>>>> "LT" == Linus Torvalds <torvalds@transmeta.com> writes:

LT> Oh. Even the GGI people agree that for performance we _have_ to do the IO
LT> in user space. Taking the overhead of a system call is simply not even an
LT> option.

No, now you are showing your ignorance.

The GGI design allows several ways to access a video card.
The last resort is the ioctl interface but that exists.
So syscall overhead is possible, just not desireable.
IT IS AN OPTION!

LT> What's the difference between mapping the frame buffer into user memory
LT> and doing any other IO access?

Well. When I correct that to say.
What's the difference between performing reads/writes on a memory
mapped frame buffer in user memory and a doing any other IO access.
There is no difference, in performing IO.

The important difference lies in the permissions checks of the system
calls that allow you to map that i/o region. With a well written
kernel system only safe ports are allowed. So you can't lock up the
kernel. And this are abitrated so if there are resouce conflicts
those will be handled.

An important message to get to Linus:
At least as of a couple of months ago, the GGI developers
were appreciative that you didn't let code that wasn't ready to go in,
go in. And they appreciated his fairness.

Now there is a problem in that there seems to be no dialog between
Linus and the GGI developers, so they can work out design glitches,
and explain reasons for design choices. That appears to be the
current problem.

From what I have seen reading the ggi development list the core group
in the design are a lot more flexible and reasonable than many of the
``GGI'' advocates, and if linus were to talk with them he probably get
progress made.

All of this GGI is good GGI is bad talk, simply leaves with predijuces
that can't help anything.

Eric




-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu

\
 
 \ /
  Last update: 2005-03-22 13:41    [W:0.454 / U:0.048 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site