Some errors and misinformation has been pointed out to me about my previous
posting about the GX accelerator.
Firstly my incompatibility problems between the xnews server and X11R4
clients is probably due to sun's implementation of the Xt library (which is
still at R3 in parts). (Although this doesn't explain bad colour errors to
me).
Openwindows 2.0 does use the GX accelerator (as does SunGKS etc), however
the GX is not that fast a piece of hardware, to quote from
richard@unipalm.co.uk
----- Begin Included Message -----
1. Line drawing is very fast in software. The GX only produces a 2.85 times
improvement over MIT. Filled rectangles are only 14.5 times quicker.
THE GX IS NOT A VERY FAST PIECE OF HARDWARE. .....
2. Size of server is important relative to memory in your machine. If your
application causes any significant swapping, any speed improvements by using
the GX card are completely wiped out by the swapping overhead. ......
----- End Included Message -----
Here is a list of what operations the server performs from mayer@wrc.xerox.com
----- Begin Included Message -----
First, here is (roughly) what the GX board does:
(1) It has a hardware matrix multiplier that provides direct support
for 2D AND 3D transformations.
(2) It supports solid (but not vertex shaded) area fills. Vector
drawing is
supported as a special case of area filling.
(3) It supports area moves.
(4) It supports a 1 -> 8 bit mode for fast text support.
(5) It has a hardware cursor.
(6) It has some hardware support for anti-aliased graphics, including a
"high-resolution" monochrome mode.
(7) It supports direct access to the 8-bit frame buffer.
Second, according to MIT, it is actually slightly slower than Sun's
"dumb" 8 bit board when the accelerator is not being used.
Finally, OpenWindows 2.0 definitely DOES use the GX accelerator.
I belive that you will find the MIT R4 release to be significantly
faster than OpenWindows 2.0 for window creation, somewhat (but not
dramatically) faster for graphics operations that are very small (short
lines, tiny areas) or that do not invoke the GX (curves), and
dramatically slower than the OpenWindows 2.0 server for operations that
involve moving large numbers of bits around. For fun, try dragging a
window around while running "olwm" with the "OpenWindows.DragWindow"
resource set to "true". Also, I think that you will notice that text
scrolling works much better with the GX.
Finally, I don't want to knock the MIT server; it is quite fast, and
will both be getting faster and picking up scalable fonts (which I
consider a necessity) in R5. I used it quite happily for a long time,
and if I had to run on an 8meg machine again I would probably switch
right back. On the OpenWindows side, I hope that the window creation
in the OpenWindows server will get faster in the 3.0 release. Also,
you may want to consider whether or not the Postscript and scalable
font capability available in OpenWindows 2.0 is important to you.
----- End Included Message -----
So in conclusion, for big things xnews will be faster, but you probably want
an extra 4 or 8MB to make it run well, there is a patch for the X11R4 server
to let it use some GX features in a minimal way. Who knows how fast the R5
stuff will be ?
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Bill Hart Internet : hart@ocean.ml.csiro.au
CSIRO Division of Oceanography Phone : +61 02 206 446
Hobart, Tas., 7000 Australia Fax : +61 02 240 530
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:06:12 CDT