[Pc_Support] Avalon/WGF 1.0 in Longhorn release will _lack_ punch of Apple's QuartzExtreme ...

[Bryan J. Smith]

ExtremeTech tried to put a good face on a sprawling clusterfsck in NT6.0
Longhorn known as Avalon, which is now a "WinFX technology".  Avalon is
the new replacement for the 14 year-old Graphics Display Interface
(GDI), the thorn in every Win32 program's side from NT3's original
design.  Of course, it's really just another layer atop of it for now,
based on DirectX 9.

Avalon is supposed to use the GPU's processing and framebuffer to off-
load from the normal CPU.  Your GPU sits there, unused, for most
operations in the GDI (or, in Linux, X11**), while your CPU
inefficiently copies blocks of system memory back and forth, back and
forth, and then makes memory mapped I/O calls to the video card.  It's
far more efficient to just put the whole window management and
framebuffer in the GPU, where it's directly rendered.

[ **NOTE:  The Linux world has several implementations.  There are a few
window managers/environments that do the same thing, only with OpenGL
and it works well, via the "Cairo" (not to be confused with the NT 4.0
Release Codename) subsystem for GLX (OpenGL on X11).  Then there is
Sun's greater Looking Glass project which actually replaces X11 (while
offering legacy X11, GLX, etc... compatibility). ]

Apple has done this for years now via QuartzExtreme.  QuartzExtreme
merely leverages OpenGL inherent to modern ATI Radeon or nVidia GeForce
GPUs -- easily anything produced in the last 5 years.  Even most chipset
embedded GPUs could probably take it now, and it drastically improves
performance of the window management, while adding eye candy at the same
time.  Win-win.

<pun>Longhorn story short</pun>, they didn't get done what they needed
to do, so they are pushing back all the goodies to the so-called Windows
Graphics Foundations (WGF) version 2.0.  In other words, in unless you
have a serious $400+ graphics card, NT6.0 Longhorn with WGF 1.0 is going
to trade eyecandy for performance.  This should _not_ be the case, as a
_good_ API design would result in what Apple gets out of QuartzExtreme
-- "textures for nothin' and your vids for free."

The "non-happy-world, non-we-always-love Microsoft" low-down:

DirectX 9 still _lacks_ major geometry setup and capabilities that
OpenGL has inherently.  WGF 1.0 is based on existing DX9.  The overhead
will be massive, requiring top-of-the-line graphics cards, or the
disabling of features.  DirectX 9 is 4-5 years old, and Microsoft was
hoping that its more complete replacement would be available for
Longhorn.  It isn't, so WGF suffers as a result.  Oh joy!

WGF 2.0 will, of course, add all the goodies need "very soon."
Heard that before!  ;->

Reality:  
Once Microsoft introduces all the limitations of WGF 1.0, then they will
have put all that added layering and overhead in and the result will be
continued inefficiency and issues.  I'm not here to demonize Microsoft,
but to point out that their half-ass efforts have just polluted them
time and time again.  Instead of introducing a good design, they will
rush a bad one to market -- which will make the vendors happy because
it'll sell more top-quality hardware.  Sigh, circular reference.

Now more than ever, they should go back to OpenGL, the standard they
_did_, in fact, adopt because it solved the GDI issues in the NT3 days.
But because they couldn't get it to work well with DOS7
"Chicago" (Win9x/Me), they have hacked Direct DOS Memory Map aka, today,
DirectX with geometry, T&L and other things added for gaming, and not
professional usage. 

In fact, imagine if Microsoft wouldn't have introduced "Chicago" and
pressed forth with pure Win32 and GDI/OpenGL in NT.  Imagine where we'd
be right now in terms of stability, reduced viruses and eye candy
without overhead.  Microsoft could have focused on producing something
like QuartzExtreme, instead of messing with DirectX for almost a decade
to just bring it to the level of professional/reliable use as OpenGL.

Sigh, I just have to shake my head.

Happy-Feel-Good Version:  
http://www.extremetech.com/article2/0,1558,1791702,00.asp


-- 
Bryan J. Smith                                  b.j.smith at ieee.org 
------------------------------------------------------------------ 
Community software is all about choice, choice of technology.
Unfortunately, too many Linux advocates port over the so-called
"choice" from the commercial software world, brand name marketing.
The result is false assumptions, failure to focus on the real
technical similarities and loyalty to blind vendor alignments.