[Précédent (date)] [Suivant (date)] [Précédent (sujet)] [Suivant (sujet)] [Index par date] [Index par sujet]

3Dfx Support Status

http://glide.xxedgexx.com/status.html

3dfx/xfree support is finally out.  Let's wait for the tnt2!



-- 
 | martin blais*
 |
 | *from discreet logic
 | E [email protected]   T (514) 954-7127
Title: 3Dfx Support Status







3Dfx Glide Porting


Current TODO List (in rough priority order):


Cleanup for glide (read /etc/conf.voodoo, use /dev/3dfx if possible as
root, use symlink for glide lib, support >4 boards)


Port Glide 3.0x (V1, VR, V2, VB, V3)


Release Glide 2.54 for V2.


Cleanup website.


Possible improvements for glide (dynamically load page flip for SVGA, Allow
Rush/Banshee to switch render windows)


Consider porting to other hardware and other OSes. (Alpha/PPC) Other OSes
would require legal modification. *sigh*


Consider adding support for these
3D glasses. They are nice little units and not very expensive. ELSA and
Metabyte are selling them as well. The big problem is finding someone who
can OK the release of source and specs.


May 16th, 1999
Well the time has arrived. There is now 3D support for the Voodoo Banshee
and Voodoo3 boards on the server page.
I will be out of town for the next week. Please do NOT mail me any
questions about this release until I get back. I will just delete them
when I return. In the mean time please use the 3dfx
newsgroup. I'm sorry about requiring that, but I don't want my mailbox
to explode.
I've got some fixes for the X server portion as well, but I'm not sure
I'm going to get them out before I leave.

May 14th, 1999
I've been pounding on the VB/V3 version of Glide. I basically got it working
last weekend. The problem is that there is still a show stopper bug. It
will hang the system after playing Q3A. So, this needs to be resolved before
I release it. The problem is that it takes a while for the crash to occur
and it doesn't happen with debugging turned on. So, it'll be nasty to find.
3dfx and I are working on it.
I spent a couple of days up at Quantum3D.
They do high end hardware for Visual Simulation and arcade systems. It
was interesting and useful to talk to them and some of their customers
about Linux.
I'm doing a presentation at Linux
Expo next week. This one is titled Implementing 3D hardware acceleration
for Linux. That's a bit of a misnomer, as the paper and presentation is
more like introduction to 3D hardware accleration. There's a couple other
papers on how it is done.
There will be a BOF at Linux Expo on OpenGL/3D on Friday 5/21 from
3:00-4:00 that I'll be co-hosting.

Apr 30th, 1999
X server code for 3.3 sent to XFree for upcoming release.
Released new Banshee/V3 X server
to fix a bug with small pixmaps.
Glide 2.54 for V2 is ready. It has 3DNow support and the code cleanup.
Needs to be packaged up and released.
Glide 2.46 for VG is ready. It has code cleanup added. Needs to be
packaged up and released.
Most of the Glide for Banshee/V3 is compiled and together except the
interaction with the X server.
More updates on my MASM->GAS translator to support
all these.
It's official. I'll be joining Precision
Insight to work on 3D support for Linux.

Apr 19th, 1999
I know updates have been a bit scarce lately. Real life has been getting
in the way.
Scott Bertin did a ton of new acceleration functions for the Banshee/V3
X server which I just released. Except for bugs and a few small improvements
this is done. Now it needs to roll into 4.0 for the next XFree release.
I've been working on Glide stuff.
A few people asked me about my MASM->GAS translator. It is an awful
hack. Just a perl script that knows almost nothing about MASM, but understands
enough to translate the text. It isn't anywhere near complete, but it mostly
works for my stuff. I just go through it by hand to fix other problems.
Anyone who wants it can grab a copy from here.

Apr 4th, 1999
Updated a few random things up on the website, including getting the 2.2
compatible Device3Dfx updated.
Got Glide 3.01 working. Mostly this was assembly fixes in my MASM->GAS
converter. A few key users need Glide 3.0.
My life has been a bit hectic lately, but this is changing. I'm leaving
Digital
Domain after more than four years. I haven't made up my mind as to
exactly where I want to go next, but I have some excellent choices. Not
to worry, I expect I'll be promoting 3D and Linux one way or another. It
also means I'll have a couple weeks of time off this month.

Mar 30th, 1999
The new server seems to be functioning correctly, but there's been complaints
of the performance being slower and/or jerky. I suspect this is the Rush
extension. It changes the way the server handles pixmap caching. It caches
a lot more pixmaps which I don't care about particularly.  It also
caches arbitrary sized pixmaps. That I do care about, because that's how
I do 3D in a window. So, I'm going to have to rethink how I handle that
problem.
Scott Bertin has been cranking on acceleration features. He sent me
a patch that added a bunch. I need to check them out and roll them in.

Mar 28th, 1999
New packages have been released for my X
server.. It should fix server problems including the Rush and Diamond
problems.

Mar 26th, 1999
Scott Bertin tracked down what looks like the problem with the Diamond
Monster Fusion cards and sent me the details.
I found that I wasn't saving clipping information which caused problems
for people running two servers at once.
It looks like the Rush patch is causing some pixmap problems. I need
that for Glide, but I'll pull it or fix it in my next release.

Mar 23rd, 1999
Got permission last Saturday to release the source for the Banshee X server.
Built RPMS and an SRPM of the material. I'll try to get them posted
tonight.
A new binary (RPM) of the server that includes VMWare support and the
Rush patch is now on my Banshee X server
page.
The Banshee X server page also
includes the source (43MB) to the X server.
I barely got all this in by the end of the day. :-)

Mar 18th, 1999
A few people have asked why I haven't updated lately.  I spent a week
at Linux Expo, and then came home and had a nasty cold for a week. So,
I've been out of touch for a while.
I went to the 3Dfx developers conference. 3Dfx said good things about
Linux support. SGI said good things about Linux support. 3Dfx has some
very excitiing stuff planned. (The conference is all under NDA so don't
ask for details)

Mar 6th, 1999
Installed my V3 and it worked fine with the 2/23 X server.
I'm waiting on 3Dfx to OK my X source for release. Since I only use
the functions in the spec it should be fine.

Mar 1st, 1999
The big news today is that 3Dfx just released the specs for the 2D portion
of the Banshee. When I get back from LWCE
I will great full releases (binary and source) of the server which can
go on the 3Dfx site. I'll also be contributing 4.0 source to XFree.
If you haven't noticed 3Dfx is now housing copies of all my RPMs on
their site.
I'll be giving session 42 on Wednesday from 2:30-3:45 on the state
of 3D for Linux at LWCE.
Wednesday evening from 5:30 to 7:30 in room A4 will be the OpenGL/Linux
BOF hosted by Jon Leech and myself.
I'll be at LWCE March 2nd through 4th.

Feb 21st, 1999
The "Full House" (nines over twos) release of my Banshee
X server. Please let this one be it. I found a case that casued crashes
on startup when you ran 8bpp mode at <1024 across. It took me a little
longer to track this bug down.
I'm now running an ELSA Victory AGP card with no ill effects. (Thanks
ELSA!)

Feb 19th, 1999
I fixed up a few more glitches from reports from people. I'm forcing more
of the registers (that I don't use directly) to known states. I fixed >135Mhz
(again, since I broke it in 2/15). I adjusted the video timing to hopefully
be more stable for everyone.
>From my web logs I noticed that this page is getting a ton of hits.
Don't forget the news page for general news about
3D for Linux, and the rest of the site.
I'll release this tomorrow, since I'm just wiped out tonight.

Feb 15th, 1999
The Bullwinkle release of my Banshee X
server. As in his famous saying "This time for sure!" I found a bad
reference which would cause a SEGV during startup leaving you screen black.
I also save a couple more registers so that you can run multiple copies
of the server at different depths on different virtual terminals. (No this
isn't a common thing to want to do, but it was a bug that should have been
handled.) It also correctly requires that modelines have widths that are
multiples of 16 pixels when running at >135Mhz pixel clock.

Feb 14th, 1999
The Valentines Day release of my Banshee
X server.  I missed a few bugs in the last one. This fixes a bug
with clipping the lower end of the display when using 8bpp mode. It also
fixes a problem with large vertical resolutions. My monitor supports 1600x1200@60Hz,
and that now works. I think this really is it. Let me know if you see any
problems. Thanks to everyone who reported problems up till now!

Feb 13th, 1999
A new Banshee server release is available.
This seems to clear my buglist. It now supports modes requiring pixel clocks
>135Mhz. It works when you use the 2.2 framebuffer code at startup. There's
a workaround for the "slashdot" bug.  I also simplified and cleaned
up a bunch of the initialization and shutdown code. It now makes more sense
and hopefully avoids the problems with some card manufacturers.
The "slashdot" bug is a problem when you are
using netscape and loading pages from slashdot while scrolling the display
you can hang the system. If I turn on deubgging output in that section
the problem goes away, so I suspect it is timing related. I introduced
a small delay by syncing the server before these operations which seems
to remove the problem.
This version also includes the first support
for the Voodoo3 boards, but that's completely untested at this point.
So, I really want people to try this version.
I think it is ready to roll. If it goes well I'll package this up and get
on with the rest of my work.

Feb 11th, 1999
Made some good progress on how the Banshee shuts down. In now works (most
of the time) in framebuffer mode. It never hangs for me, but it sometimes
leaves the screen a little strange during startup or shutdown. I think
I'm getting close to nailing this.
I heard from the Linux PPC folks. They've had delays, but they do intend
to get me a box.
I got graphics cards and a server from Xi Graphics to do some testing.
I got a server from Metro Link.

Feb 9th, 1999
I released another copy of the server on 2/7. This did the minimum amount
necessary to reset the video to VGA mode to hopefully avoid the lockups
that occurred on Diamond cards. It did. but it also failed to reset to
a clean state every time.
So, now I've got it bracketed. 2/6 resets things nicely, but crashed
on Diamonds. 2/7 didn't crash, but didn't reset. So, I've done a little
more in the reset code to hopefully get both parts right. That's the 199902090
server.

Feb 6th, 1999
Big news on the server front.
I got hardware acceleration working. I had to tell the server to sync
the board during operations. Most cards seem to need this.
I spent a bunch of time trying to work out the Diamond/ELSA BIOS, and
I seem to have hit on an order of operations that works!

Feb 4th, 1999
Added BitBlit acceleration and rectangles support any raster operation,not
just copy. There are still synchronization bugs that I'm working on.
These additions make my server feature complete for the first Alpha
release. Now i just need to work out all the bugs: crashes, synchronization,
high resolution modes.
ELSA is kindly providing me with some boards to do testing for my presentation
at Linux World Conference and Expo.
Looks like Jon Leech and I will be hosting a BOF on 3D for Linux as
well. If you're just going to the expo try to catch the BOF.

Feb 2nd, 1999
It looks like the ELSA Victory II boards have the same BIOS issues as the
Diamond. I'm not saying their BIOS is broken, it just does work with my
code. That could be my code or the BIOS that is at fault.
I added my first accelerated function: Filled rectangles in GXCOPY
mode. Next we do blits, that'll be a big improvement.

Jan 31th, 1999
It appears from reports that it is only the Diamond boards that hang on
exit. If you're board hangs on exit and is not a Diamond, please let me
know. I can reproduce the problem on my machine when I flashed Diamond's
BIOS on my reference board.
Did some other cleanup and preparation for building a bunch of Glide
releases.
Did some work with various 3D benchmarks in preparation for my presentation
in March.

Jan 30th, 1999
Definetly BIOS related. I just flashed my board and sure enough I crash
on exit. It works fine with 3Dfx's BIOS. I'm going to work with them to
get this figured out.

Jan 29th, 1999
Talking more with the folks at 3Dfx we're starting to think the crash problems
people have with exiting the Banshee X server may be some interaction with
the BIOS versions people are using. I'm going to try flashing my board
with some different versions to see if I can reproduce it.
I added hardware cursor support to the server.

Jan 27th, 1999
OK, this version (199901271)
really does idle your card. People ran the version from the 26th and instantly
died. This one should work.
Now you might ask, "How did he release one that instantly died? Doesn't
he test this stuff?" The answer is yes I do, but I tested with debugging
turned on, and when I release I turn some of the debugging off. A print
statement made all the difference. Which leads to the lesson of the day.
Volatile is an important keyword in the C language, particularly when you're
dealing with memory mapped hardware! Without the print statement and the
volatile keyword, the compiler optimized away an important part of my idle
code.

Jan 26th, 1999
Went through a couple more Banshee
X servers without substantial changes. Now, at the recommendation of
3Dfx, I'm explicitly idling the board before doing the mode switches. Please
let me know if this makes allows you to switch modes without a crash.

Jan 23th, 1999
Some folks are having trouble with the server crashing when you switch
virtual terminals or exit the server.
There have been two versions today. Please pick today's
latest version and report back the debugging information.

Jan 20th, 1999
Getting mostly success reports from the X
server.
Fixed a problem with Glide for people with >4 boards (3 SLI config
for a simulation is 6 boards from Glide's point of view)
Did a little work on hardware cursor support.
Folks if you try a version of the server and it messes up, power down
your machine. Don't just reboot. That's won't reset everything.

Jan 18th, 1999
X server has made some big improvements. You can get it from here.
    Detects chipset correctly.
    Works at 8bpp, 16bpp, 24bpp, and 32bpp.
    Works for any resolution up to 1280x1024@72hz. (135Mhz
clock)
    Exits back to console mode correctly (at least for
me and most users).
    (This is getting close to what I want to call an
Alpha. I've got a few more fixes and then I need to package it up.)
Downloaded a bunch of benchmarks for 3D. I'll use these for my Linux
World Conference & Expo talk.
Lined up some hardware to benchmark.
Started trying to track down a PPC machine to port with. I had a couple
offers a while back.

Jan 16th, 1999
A pre Alpha release of my Banshee server is now available from this page.
Please read the caveats carefully.
More updates will come to the server shortly. If you're not comfortable
with the caveats, please wait for a better release.

Jan 15th, 1999
The 3.3.3 version is running, but once it switches to X, it makes the screen
black when it returns to VGA mode (for virtual terminals or by exiting
the server) This is a pretty bad gotcha. It also has software cursor instead
of hardware which is minor. I may release this one for those people that
are desparate and then work some more over the long weekend.
Got DSL at my house. More bandwidth is always good!

Jan 10th, 1999
Rough draft of the 3.3.3 version of the Banshee X server is written and
compiles. Now testing.

Jan 2nd, 1999
Fixed the problem with the mode changes. This seems to be a working minimal
server using XFree 4.0.
Minimal means: Works at 640x480, 1024x768, and 1280x1024 at 16bpp and
24bpp. Anything higher or lower will probably break badly.
Hardware cursor, but no other hardware acceleration, so it is slow
dragging around opaque windows. No 3D support.
Let the backporting begin. I need to roll this into a XFree 3.3.3 server
so I can release it.

Jan 1st, 1999
Happy new year! May Linux have more success in the coming year.

Dec 28th, 1998
Hardware cursor works.
It looks like I'm going to have to back port my code from XFree 4.0
to 3.3.3 if I want to release it any time soon.

Dec 27th, 1998
Fixed the root weave. I now have a really minimal working server.
The screen gets somewhat corrupted when I change resolutions. The top
lines change from a nice root weave to colors.

Dec 26th, 1998
My first results where the screen isn't just black! A really messed up
root weave appeared!

Dec 25th, 1998
Happy Holidays to everyone! I hope everyone has a safe and fun holiday
and a rewarding new year.

Dec 21st, 1998
The server launches and switches video modes. At this point it seems to
be doing what I tell it, but I'm missing something.
If I do an xwd I see a root window. The problem is that nothing is
visible on the screen.
You can get some cool bugs by doing stuff with the X server. Somehow
my system clock was running about 10x realtime.
Please don't email me asking to be a tester. When it is working I'll
put it up on the website for everyone.
I got a 4 port video/keyboard/monitor switcher. It makes working on
the card much easier. I keep X up on my Alpha and switch back to my Intel
to run tests.

Dec 17th, 1998
The board is installed. It runs real well in text mode. :-) Attempted to
boot the X server. It loads my module, locates the board, and gets some
of the configuration parameters right. Then it decides it only allows displays
that are 0 pixels tall and exits.

Dec 16th, 1998
Lots of code cleanup for the Banshee driver. It all compiles and I've built
an X server. I'll test when I can afford to break my machine.

Dec 13th, 1998
I've got a working draft of all the required code for a non-accelerated
Banshee. At the moment, it's not compiling.
3Dfx should be sending me a big batch of Glide 3 sources soon since
they've all finalized.

Dec 6th, 1998
Got most of the VGA register and timing code written. There's now a bunch
of support routines to write and I should have the first version which
could possibly boot.

Nov 11th, 1998
The 3.01 release didn't make it before my vacation. (Now) 3Dfx is still
changing these rapidly, so it isn't worth a lot of effort anyway. They
should have final versions by the time I come back.
I've had an offer to provide me with a PPC machine. No more excuses
for not supporting it. :-)
I've swiped a bunch of hardware setup code from the 3Dfx D3D driver,
so it's moving along nicely now.
I'm gone for a little over two weeks. Budgets are DD are done. I should
be able to crank in December.

Oct 29th, 1998
Got the 3.01 sources for V2 and seem to be working. I don't want to spend
too much time on these early 3.x versions, since they are still prereleases
from 3Dfx and I'll just need to replace them with the official versions
when they come out in a couple weeks. I'll package this one up and release
it for those hardcore V2 programmers to start working with.
Started seriously into Banshee X server support.

Oct 18th, 1998
Spent most of this weekends work time writing a perl script that translates
MASM->Gas. It's something like a 90% solution, but as 3Dfx has added more
assembly to their libraries it makes my life easier. I should have done
this a while ago.
Working on Glide 3.01.
I was sent a version of /dev/3dfx code that works for 2.1, so I'll
probably stick that out as "really unsupported" on my site after I review
it.

Oct 15th, 1998
Glide 3.0 for V2 went really smoothly. I'm tracking down a bug in the strips
and fans code, but that shouldn't be too bad.
I got 3.01 sources, so I'll move to those before I release.

Oct 11th, 1998
The same link problem existed for libc5 versions. Also, I didn't include
the libc5 libraries in the glibc package. Both packages have been updated
to be Glide_VR-2.46-3. Hopefully, this should work for all Rush users.
Pick'em up from the drivers
page.
I got Glide3.0 for V2 running with C triangle setup. This is a good
start. Now I need to port the assembly language.

Oct 5th, 1998
It looks like the DGA problem was really some incorrect link information
in my assembly module. Rush users should get Glide_VR-2.46-2 from my driver
page for the fix. Hopefully this will do it.

Sept 27th, 1998
Glide version 2.53 is now available in libc5 and glibc versions on the
software page. This should be a minor performance improvement for V2 users
and should be more reliable about locating the boards. If you had a problem
with your V2 board not being detected try this version.
Device3Dfx RPM was upgraded. This turns off debugging, which was leaving
messages in your syslog and compiles with module versions, if appropriate,
to get rid of the unknown symbols message when the module is loaded.
Still tracking the Rush DGA problem. Not making much progress yet.

Sept 26th, 1998
Fixed up the MTRR page to
no longer talk about proform now that the device driver has it built it.

Sept 25th, 1998
Still working a problem with the Rush. Some users hang as soon as I switch
to DGA mode.
I've made some progress with H3D. They now have a developer program
for hobbiest. I'm thinking of making an H3D release.

Sept 18th, 1998
It appears that the people having trouble with the Rush are actually having
problem with DGA mode. That's how I implement full screen page flipping.
Somehow this is locking the machine. If you run in a window it runs fine.
The 3dfx module always complained about unknown symbols, but it worked.
This is cause by the kernel versioning system. Kernel symbols are checksummed
and their names mangled. I need to do that in my driver.
I got the O'Reilly book Linux
Device Drivers after reading the table the table of contents on
slashdot.
Looks like a useful text.

Sept 16th, 1998
Today is a big software update. The following packages were updated on
the 3Dfx Software page:
    Glide runtime for Voodoo Rush version 2.46 (libc5
and glibc)
    Glide runtime for Voodoo Graphics version 2.46 (libc5
and glibc)
    Device3Dfx version 1.1
    X server and Rush libraries (glibc only at this
time)
This version of the Rush libraries allows me to play Quake2 version
3.19a in a window or full screen with no corruption and no special hacks
on my RH 5.1 system. Hopefully this will work for you as well.
The known fixes include:
Several libc5 versus glibc cleanups.
Fixes to pixmap handling in the server.
Better memory management for Rush.
MTRR support.
Better installation and removal of Device3Dfx.
Glide 2.46
I still need to compile the X stuff for libc5 (that's easy but time consuming)
and do the Glide 2.53 for V2 port and builds.
My Rush hang on my PII was actually a stupid bug. I was pushing non-snapped
triangles to the board. (Instant hang) So, I am currently unable to hang
this collection of the packages.

Sept 12th, 1998
I'm still trying to track down the Rush hang. Now I can reproduce it on
my PII system quite easily. I couldn't on my Pentium.
I've got all the Glide 3.0 sources.
I've got Banshee hardware and software, but I'm just starting to look
at it. I really want to finish earlier hardware work first.

Sept 2nd, 1998
I think I found a bug with the pixmap handling in the Rush X server. It
should remove a few glitches.
Memory management has been reworked. It may get you some free 3D space
for Rush.
Fixed a bug in Emil's MTRR code. We shouldn't set MTRRs for Rush boards.
I updated my 2.43 code base to 2.46.
I got a nice PII-400 system. It speeds up my compiling by about 4x-5x!
My P120 was slow! Also allowed me to test MTRR code.
I've got a Banshee and the appropriate documentation. I'm starting
to work on that.
I've got an engineer grad student from 3Dfx who is going to be involved
with Banshee work.
I appologize for not updating some of this sooner. I've just been swamped.
I need to get out a new release of just about everything, but that
probably won't happen until after Labor Day.

Aug 16th, 1998
Improved memory management can't help with 800x600 in a window with a 1024x768
root, but it can help fullscreen mode.
New versions of the glibc packages for VG and VR are now on the website.
These cleanup a bunch of problems with the libc5 build and add better error
reporting for the VR users. This should tide people over until I get everything
worked out for the 2.46 and 2.53 release.

Aug 14th, 1998
Found a compiler bug with gcc 2.7.2.3 so that forced the move to egcs.
I believe I got all the compiling libc5 on RH 5.1 issues straightened
out.  YUCK!
Got 2.46 and 2.53 from 3Dfx building against those.
Better error reporting for Rush users who can't grSstWinOpen.
Corrected the documentation on the non-RPM
page in a few places.

July 22nd, 1998
The SIG at SIGGRAPH went really well. You can read
all about it.

July 21st, 1998
The Digital Domain party at SIGGRAPH was incredible as usual. Wilson Pickett
is definetly an act to see.

July 12th, 1998
A complete set of glibc libraries are now available.
Broke the Rush library into its own RPM instead of installing all of
the X libs.

July 5th, 1998
Got my system running RH 5.1 and building both libc5 and glibc libraries.
Released glibc versions of Glide_VG and Glide_V2 RPMS. See the release
page.

June 29th, 1998
Well I kind of took it easy last week. Mostly I let people play with the
big release and report problems. There's still a few unexplained issues,
but mostly it is working so that makes me fairly happy.
I had planned to work on glibc support this weekend. Then one of my
disks died. *sigh* I'm pretty good about doing backups, which means I only
lost a weekend instead of any real work. Do your backups boys and girls!
:-)
So, about all I can report is that I now have RH 5.1 installed on my
system.

June 24th, 1998
Fixed the link on the 3Dfx RPMS page to the X server.
Fixed the device driver so that it creates /dev/3dfx for you, and so
it doesn't try to be relocatable. (You can't relocated devices anyway)

June 22nd, 1998
The SDK was missing the headers from /usr/include/glide now fixed. I also
called the package version 2.1, so that it matches 3Dfx naming.
Fixed the SDK's makefiles and linking of test applications.
Fixed the test application in the runtime packages.
Fixed the device driver so it works with older kernels that are missing
some of the #defines.
Improved the instructions for the device driver. Doing --rebuild is
not enough, you must also install the result.

June 21st, 1998
I had some bad links on my new RPMS page. If you had trouble downloading
the packages, please try again.

June 20th, 1998
An anniversary of my birth. My present to all of you is a new release.
Release day! New RPMS for everyone. Get them here.
Wow! It takes 4 hours to build the X RPMS from source on my P120.

June 16th, 1998
Just bought Maximum
RPM so I'm brushing up on the packaging details. I'd like to do
them right.
No real luck with a good 320x240 mode yet. Has anyone else got one
they like? Email me.

June 15th, 1998
(Early update) Flashing during page flipping is fixed. Henrik pointed me
at the appropriate code to add to the APM driver.  We're still discussing
turning off the hardware cursor. It can be done in my extension, but doing
it in his driver actually makes more sense.
Henrik provided a quick fix to hide the cursor when I'm page flipping.
That seems to work.
Turned off the virtual root feature when page flipping.

June 14th, 1998
Found that a problem that I thought was my X server code was really caused
by the fact that I had a really old BIOS on my Rush board. *sigh* That
wasted a lot of time.
Assembly language for Rush is now working.
The page flipping currently flashes a little. I think that's because
the X server is page flipping immediately instead of waiting for the next
verticle refresh. I've asked Henrik Harmsen (author of the AT3D driver)
about it.
Removed the annoying screen black when shutting down the window.
Use VidMode to select the desired screen resolution for page flipped
mode.

May 26th, 1998
DGA is working fairly well for the page flipping. It does require some
careful configuration of your X server config file.
I'm in the middle of debugging the assembly language for the Rush.
(Translating between MASM and GAS is a pain)

May 22nd, 1998
Happy day! My Voodoo2 stuff is now available. Check here
for the details.
Started implementing DGA based page flipping for Rush. Looks like it'll
go smoothly.

May 18th, 1998
Took my first shot at make Glide2.5 into an RPM. It went pretty smoothly.
Now I need to generalize it and make it do all the packages I want. I started
looking at the package list. I think I need seven to do the release!
Here's the list


Glide v2.43 for Voodoo


Glide v2.43 for Voodoo Rush


Glide v2.51 for Voodoo2


Glide 2x development (I think I can make one for all of the above, instead
of 3)


XServer for VRush


XServer source


/dev/3dfx source



May 17th, 1998
Voodoo2 code updated to Glide 2.51.
Started researching RPM. It looks
like it should be reasonable to setup.
The SLI card does fail several tests. The failures appear to be releated.
I'm checking with 3Dfx to see if these are known problems.

May 15th, 1998
/dev/3dfx now has its own official registered Linux device number. Character
device with major number 107!
Sent my device driver source off to 3Dfx for release approval.

May 13th, 1998
VRush now works with /dev/3dfx.

May 12th, 1998
VRush assembly is ported and works. Started on supporting VRush in /dev/3dfx.

May 11th, 1998
Backed the V1/VRush code to 2.43 from 2.45, and started on the VRush assembly.

May 10th, 1998
MTRRs now really do work if you set them correctly. Here's my instructions.

May 9th, 1998
I spent the day up at 3Dfx working with one of their engineers. We got
a ton of stuff done. We appear to have fixed my illusive Rush bug! There
were some alignment issues I didn't know about. He spotted them almost
right away, since they had the same problem at one point. We also think
we found the MTRR problem.  More on both of these after I do some
more work and testing. I also got the updates to the code that I needed.

May 3rd, 1998
I had a real good time at Immersion'98. They covered some good information
on the Banshee, and it was great to meet a lot more of the people I've
been emailing.

May 2nd, 1998
I think /dev/3dfx is done.  I've closed the holes. You can still crash
the machine by sending bad data to the board, but my device prevents you
from sending commands to anything other than the board. I think that's
the best it can do.

April 11, 1998
All my Rush extensions are in the XFree3.3.2 code base.  That allows
them to be distributed.

April 1, 1998
It seems that I got Voodoo2 working. (No really, it seems to work :-) I'm
letting it out to a few special people for testing.

March 27, 1998
Source code for all the current Voodoo products showed up.

March 19, 1998
Got the legal paperwork for getting new Glide code done.

March 13th, 1998
Voodoo2 board arrived. No source, so all I can do is play Windows games
with it.

Feb 16th, 1998
Buffer copying was easy. The problem appears to be memory timing

January 9th, 1998
I've made some good progress on the Rush port. The setup seems to be correct
in X, and the card seems to be performing all of its functions. The only
problem is that the graphics aren't getting copied from the cards buffer
to the screen buffer. This shouldn't be too hard to track down.