Alain <kua### [at] videotronca> wrote:
> When using method 2, you NEED to provide intervals, variance, confidance
> and ratio, or rely on the default values. Also, you should use samples
> min_samples, max_samples.

That is pretty much what I have going on, save for variance, confidence and
ratio. I do not really know the details of cause and effect well enough to mess
with them, and I am extremely close to my goal of replicating MP Glows [types 1
and 2] now. I have a total of 11 Glow types at this point, and they are all
distinct enough where it isn't easy to confuse them.

One thing I tried was converting the C code from the Glow patch to SDL, as much
as was possible. I spent an hour or so doing that, but it was cumbersome because
it required parameters that I know of no way to access in SDL, such as ray
origin and various extents. I did this back in my first stab a few years ago;
perhaps I'll look at this again with an eye towards simply extracting the
falloff expressions.

One outlier is attached...I have no idea what is up with this odd cube
artifact... It almost looks like bounding boxes, but I wouldn't put money on
that being what it is. I replicated this accident once through overlapping media
containers and looks_like [quickly fixed through consolidation], but such is not
the case in this instance. This is the shabbiest example of what I have
accomplished toward my goal, by the way.

Once I am happy with the 11 types, I will consolidate the SDL to remove
duplicate code across types, remove anything that ends up being too close to
anything else, and to minimize light sources.

The FixedSimpleVector exceptions I was getting are gone now through prior
consolidation efforts, but I have not yet established a comparison of max glows
between MP and the media variant. Obviously the media implementation is slower
to varying degrees according to type.

Thanks for the tips man! :)

Ian

Post a reply to this message

Attachments:
Download 'mediaglow 2014_0_2_.png' (258 KB)

Preview of image 'mediaglow 2014_0_2_.png'