Very Computer

Hi There,

I know I asked this on algorithms a few back, but This is a little more
specific - and I appologise that this is not necessarily OpenGL/DirectX
centric, but I know people here have done this sort of thing and well, the
API's to a point can govern what is possible.

Anyhow,

Basically I am looking for fast shadow algorthims/methods, the faster the
better, but ones that can take advantage of common hardware functions.
Basically, I can be sure that the people using the program will have
hardware acceleration, but I don't know if they have a stencil buffer, nor
do I know if it is a quick implementation.  So please, any help here would
be greatly appreciated.

Kind Regards,

Ian Hawley

remove antespam. for direct replies

On the win32 platform, you can't actually check for a hardware
implementation, you can only check whether a given pixelformat is supported
via an icd. Commonly this means the pixelformat is accelerated. However, low
end hardware may provide a slow path for things such as the stencil buffer,
one of my dev cards being a voodoo 1 is one such beast. All you can do is
run some metrics to see how slow things go if you use the stencil buffer.
The pixelformat returns whether a stencil buffer exists or not, for the
given mode. See also glGet* with GL_STENCIL_BITS. If you're not to confident
about the stencil buffer, use a depth buffer or alternative method.

--
Anti-spam: Replace nospam with blueyonder to reply

Quote:> Basically I am looking for fast shadow algorthims/methods, the faster the
> better, but ones that can take advantage of common hardware functions.

Without knowing what requirements your shadow must meet, this will be
difficult to answer. For example, if it will lie only on a simple plane,
then a projected shadow is your fastest approach.
--

    Hewlett Packard Technical Solutions Lab
    OpenGL FAQ: http://www.opengl.org/About/FAQ/Technical.html
                http://www.frii.com/~martz/oglfaq

Hi there

Quote:> Charles Wrote:
> On the win32 platform, you can't actually check for a hardware
> implementation, you can only check whether a given pixelformat is
supported
> via an icd. Commonly this means the pixelformat is accelerated. However,
low
> end hardware may provide a slow path for things such as the stencil
buffer,
> one of my dev cards being a voodoo 1 is one such beast

Aye, this is what we have found; even if the stencil buffer is supported in
hardware, there can be a huge disparity from one card to the next.  Hence my
wish to look at all methods possible.  I think your right about the metrics
as well, it seems the only way to go there.

Quote:> Paul Wrote:
> Without knowing what requirements your shadow must meet, this will be
> difficult to answer. For example, if it will lie only on a simple plane,
> then a projected shadow is your fastest approach.

Well at least one of our methods will need to handle fairly complex meshes
being cast on other complex meshes and possibly self-shadowing as well.  I
believe we will probably need special case shadowing methods, for spheres
for instance, we would like to be able to cast at least one shadow and
possibly as may as four on as many as 25 spheres.  Largely the surface will
be a plane, but that cannot always be guarenteed.

I think we might take a number of algorithms, optimise and tune them and
create special case methods as well.  There may be an option to choose the
shadow method, number of shadows cast and possibly some metrics to determine
what settings - if any - should be used.

Thanks to you both for the help, anything else you can add would be greatly
appreciated.

Kind Regards,

Ian

remove antespam. for direct replies

hi,

Thanks for the help guys. What about SOFT shadows?
I tried using an accumulation buffer(jitting the lightsource), and it the
results look
really good..but the program grind to a halt :(
It seems as if the accumulation buffer (at least in my case, using opengl)
is done in software.
Don't cards support it?  Has anyone else had this problem, and know of an
alternative way to get soft shadows?

Thanks in advance
-Andrew Griffiths

Precompute.

--
Anti-spam: Replace nospam with blueyonder to reply

Hi, There

Quote:> Precompute.

Any chance you could expand on that, what is it that can be precomputed?

Also, can the stencil buffer be made to anti-alias the scene or geometry in
some way ?

Thanks

Ian

Precompute,... lightmaps.

--
Anti-spam: Replace nospam with blueyonder to reply