Re: [Dri-users] Radeon 7000 benchmarks on bug #4197

[Declan M. <jun...@io...>]

Recently, Somebody Somewhere wrote these words
> 
> > What interests me is this: The hardened system should have a
> > speed penalty, and a video penalty, as it uses a
> > (theoretically) slower libGL.so.1.2 to link everything
> > against.
> 
> Since you're trying to make about about a change in libGL you
> might want to consider some very popular benchmarks, even though
> you're not a gamer: Quake 3, 4, Doom 3, UT. These games have a
> benchmark ("timedemo") mode. I do not know if the demo versions
> are enough to try that. glbench is a very synthetical benchmark
> which might help to find out which OpenGL commands are affected
> by changes.
> 
The thought of compiling these large gaming packages  on a
hardened system makes me shudder as they are built, I am sure,
without any thought for system security. I'll look at glbench, and
you gave me a url for the spec, which I also will try.

Any Assembler routines have to go. All code has to be relocatable,
i.e. position independent. If quake, doom, or UT (whassat?) boasts
of being pic code or producien Position Independent Executables  
(and libraries), then let me know. My experience is gamers cut all
the corners available.

If you look at the bug discussion

http://bugs.freedesktop.org/show_bug.cgi?id=4197

and comment 8 in particular, you get the programmers description
of what is affected in libGL.so by the patch. That's the peak of the 
technical discussion, as it gets a tad personal after that.

-- 

	With best Regards,


	Declan Moriarty.