Re: GLX in bad state

I would like to clarify that my objective is not to publicize a comparison 
between WineX and DirectX.  I am trying to isolate performance problems in 
WineX.  To this end I believe that benchmark statistics will be beneficial.

When I use the terms WineX and DirectX I mean the gaming platform and all 
the application levels used on these platforms.  You will notice that I 
discuss OpenGL performance, which is not a part of the WineX code itself. 

I have added some comments below.

On Mon, 12 Apr 2004, "LaPeer, Jerry K." wrote:

> I may be jumping into a pool that is considerably over my head, but...
> 
> Being the curious person that I am...
>
> I found this to be an interesting concept
>                                                                            
> "information which will allow a performance comparison between WineX and
> DirectX"
>                                                                                
> How would one design a set of test, to measure performance between WineX
> and DirectX when one does not have the ability to factor out
> (presumably) the Linux vs MSWindows (and the various versions).

First establish that there is a performance difference.  Then try to find 
the reasons and try to fix them.  OProfile is one tool which tells you 
quite accurately what your CPU has been busy doing while you were playing 
a game.  It will give you the amount of time spent in the Linux kernel and 
even break it down into functions.
                                                
> As an example, I have asserted that EverQuest plays better for me on
> Mandrake 9.2, WineX than it does on Windows XP Pro or Home.  When we
> discussed this (on an Everquest forum), I was unsure how to quantify why
> this is true.  I would like to believe it is all WineX, but quite
> honestly I can't bring myself to accept that as true.  I assume, there
> would need to be a NTFS vs JFS comparison.  I assume there would need to
> be a Linux vs MSWindows comparison.  I assume there would need to be an
> accounting for (in Everquest context) of memory foot print models (OS
> context) and the effects that would have on paging.

Firsly, quantify what you mean by "plays better".

> I have both Windows Prof with all the MS approved patches, etc.  You
> could say I have an exemplary Out of the Box MS Windows installation.  I
> also have a Mandrake 9.2 Linux with WineX.  They are both on the same
> machine.  I would be more than willing/happy to run what ever
> performance test you can design that would isolate WineX vs DirectX.  I
> would find this most illuminating.  I am not sure I would be up for
> testing the MS Windows Package vs the Linux Package though as that
> sounds like a bit more work.

Why don't you run the Comanche 4 Benchmark?  With your dual-boot machine 
we will be able to quantify the performance difference.  Even better if 
you have a 128 MB GPU.  Remember to change glx_private.h to have a cache 
size of 128, and don't forget to disable hardware shaders and sound.

This is not a Linux vs. Windows show-down.  If TG wishes they can move 
this discussion to the Beta Testers forum, and perhaps allow me access to 
the discussion even though I am not a Beta Tester.  Maybe create a new 
forum for performance testers.

Nadav Frum
----------