Here's a little Mode X story for you:

Well, I should finally say goodbye to the "good old" days of programming
demos, and while it saddens me, I've just proved something to
myself--but first, some background:

On old ISA cards, good democoders spent a lot of time optimizing their
Mode X polygon fillers to take advantage of Mode X's ability to write
out four pixels with one write.  In addition to this, the nature of the
memory organization allowed for two or more video pages, which gave you
flicker-free animation (amongst other things).

Well, I tried an experiment tonight:  I merged Phred's 3D engine with my
Mode X high-speed semi-clipped use-the-latches polyfillers, and guess
what I found?  On any VLB or PCI video card with a large number of
polygons (over 200), it is no longer a speed advantage to use Mode X.
:-(  On my P120/PCI video card using my modex fillers and flipping video
pages without a vertical-retrace interrupt, I get about 35 fps.  If I
use the Mode 13h buffer-to-screen copy method, I get *over* 70 fps, but
I wait for retrace to get smooth animation and then I get 69-70 fps.

I'm sure many of you are saying, "No duh, that's common knowledge."
Well, look, I'm an old codgety demo guy, and I've been in denial for
quite some time, so be nice to me.  :-)

Try the enclosed programs:  trymo13h will be faster on a VLB or PCI
card, and trymodex will be faster on an ISA card.  Please let me know
what fps you get for both.

Actually, there is *one* advantage to the Mode X program: Without the
rendering buffer, you save 64K.  trymodex uses 177K RAM, and trymo13h uses
277K.  And, if you use a verticle-retrace interrupt and triple buffering, you
can get the same framerate.  Check out trymodx3 for an example.
Verticle-retrace interrupts aren't as feasible as they were back in the old
days, though, because of Win95 messing up the interrupts.
