From: Martin Erhardsen <mee@daimi.au.dk>
  To  : ggi-develop@eskimo.com
  Date: Sat, 29 May 1999 12:27:04 +0200

Re: Transparent Blit

Jon M. Taylor wrote:
> 
> On Fri, 28 May 1999, Thad Phetteplace wrote:
> 
> > >
> > >     Why do you want to concern yourself with the hardware accelerated
> > > part?  If you are writing to the LibGGI2D API, your only concern should be
> > > whether the API supports transparents blits The target code determines
> > > whether the transparent blit is accelerated or not.  It looks like
> > > ggi2dBlitOp and ggi2dStretchBlitOp() are what you are looking for, but they
> > > appear to be unimplemented as of yet.
> >
> > I only mentioned hardware acceleration to avoid being sent a software
> > only workaround (too slow for what I am doing).  If the groundwork is
> > in place in LibGGI2D to do direct manipulation of video memory, I might
> > take a shot at implementing ggi2dBlitOp.
> > A direct video mem copy with
> > with some bit masking would probably be fast enough for my purposes,
> > and it will give us some fallback code for when full accel is not
> > available.
> >
> > Is this type of blit a standard part of SVGA?
> 
>         No.  Most newer cards support the Win32 ROP256 system, but some
> older cards do ROP16 and older systems like the Amiga only do ROP8.
> 

The amiga has 256 ROPs, with up to three inputs and one output.
I don't know if you can use the same bitmap as both input and output,
so I'll have to check this, if we are to support amigas.

> > Or will we have to
> > implement it on a card-by-card basis.
> 
>         Yes.  I can help with a hardware implementation and some API work
> if you want to do the software implementation.  I need to do a LibGGI2D
> target for Savage4-KGI anyway at some point.
> 

I have a Voodoo3 2000, so I might write a driver if I have enough spare
time.