| From: | Jesper Svennevid |
| Date: | 12 Jul 2000 at 14:24:09 |
| Subject: | FW: Re: gfx card speed issues |
> From: Paul Hill [mailto:paul@lagernet.clara.co.uk]
>
> Daithi O'Cuinn wrote:
> >
> >
> > This is correct. The reason for drawing into the buffer
> > instead of direct to the screen is to prevent flicker.
> > Bitmaps can be held in fast-ram, but there will be a very
> > noticeable slowdown.
>
> But surely fast ram is better because the CPU can access this
> a *lot* faster?
Yes, but when using the gfxboards native memory, the on-board
blitter can directly access the buffer, and blitting is almost
instantaneous. This is not the case when rendering in fast-ram.
The problem with the CyberGraphX software of today is that it
doesn't allow for bitmaps in video memory that doesn't need to
be locked in place, and when locking them, you need to be fast
on updates. Using the MMU (through mmu.library preferably),
the video-memory could be made completely virtual for the user.
If the bitmap is swapped into fast-ram, this is made transparent
to the user, and thanks to access-flags in the mmu-tables,
refreshes would be quite easy to process.
Only problem with this approach is that the computer needs an
MMU, but we can't have an "everybody wins" solution.
(I have never used Picasso96, so I can't say anything about it)
------------------------------------------------------------------------
Wish you had something rad to add to your email?
We do at www.supersig.com.
http://click.egroups.com/1/6810/1/_/451227/_/963408320/
------------------------------------------------------------------------