Re: do you like this idea?

From: Johan Klockars <>
Date: Fri, 8 May 1998 11:34:19 +0200 (MET DST)

> > I discussed this with Geir Řyvind last weekend, and we came up with a
> > (possible) way of fooling NVDI to use videoRAM for it's buffers:
> >
> > 1. Run a program that allocates all available fast-RAM.
> > 2. Use Maddalt() to register some of the videoRAM as alternate RAM
> > (e.g. the upper 1 Mb on a 2 or 4 Mb card).
> > 3. Run NVDI, run in ST-RAM, Malloc from alternate RAM.
> > 4. Release fast-RAM.
> So have you tried it? :-)

Even if you got something like that to work, it would only be worse than
what usually happens. There's no way for the NOVA VDI to know that the RAM
it uses in this case is on the graphics card, so it will use its normal
RAM<->screen blitting routine. This will be even slower than usual because
of the slower accesses to the screen memory.

> It would be nice if fVDI could use video RAM in a clever way.

That would of course be no problem, but for things like fonts it would not
necessarily be faster. It depends on how much setup is needed to blit a
single character compared to the time it takes so send a larger monochrome
bitmap from normal RAM.
For things like background saves for drop down menus and such, using screen
memory would of course be a big win. That's something that the AES and/or
application has to take care of, however. I'll definitely add support for
allocating 'off-screen but on-card' buffers for such things.

In case anyone is interested, fVDI is coming along rather nicely. My latest
beta version is now (in monochrome mode) just about up to NVDIs speed on the
normal text test in GEMBench and up to _five_ times faster (on an AB040) on
the small text test.
The following numbers are from fVDI in 1008x608x1bit with shadow buffer and
copyback cache turned on, compared to a normal Falcon (NVDI was loaded before,
but I don't _think_ anything should fall through to that):

GEM Dialog Box: 0.720 486%
VDI Text: 0.180 1888%
VDI Text Effects: 0.315 2377% (no effects are done)
VDI Small Text: 0.085 4411%
VDI Graphics: 0.545 1601% (no ellipse or XOR effects)
GEM Window: 0.555 252%
Blitting: 0.625 179%
VDI Scroll: 1.230 214%
Justified Text: 0.565 576% (nothing whatsoever is done)
VDI Enquire: 0.065 2430%
New Dialogs: 0.905 491%

With NVDI the numbers are (copyback cache on here too, but NVDI actually does
slightly better without it on two or three of the tests):

GEM Dialog Box: 0.625 560%
VDI Text: 0.235 1446%
VDI Text Effects: 0.375 1997%
VDI Small Text: 0.410 914%
VDI Graphics: 0.860 1015%
GEM Window: 0.490 285%
Blitting: 0.865 129%
VDI Scroll: 1.230 214%
Justified Text: 0.755 431%
VDI Enquire: 0.080 1975%
New Dialogs: 0.785 566%

I'd really appreciate some help with this.
Sure, I can do it all myself, but it's going to take a while...

  Chalmers University   | Why are these |  e-mail:
     of Technology      |  .signatures  |  
                        | so hard to do |  WWW/ftp:
   Gothenburg, Sweden   |     well?     |            (MGIFv5, QLem, BAD MOOD)
Received on fr. mai 08 1998 - 13:14:00 CEST

This archive was generated by hypermail 2.3.0 : ti. nov. 03 2015 - 20:07:54 CET