Farewell Thorsten! (A.K.A I take your word on this)
Posted: Thu Sep 22, 2016 4:11 pm
Farewell Thorsten! (A.K.A I take your word on this!)
Thorsten Renk on the Devel.List wrote:I'm sorry, I've reached the point where I no longer enjoy looking into the
forum or joining mailing list discussions. I'm going to just *not*
participate in discussions, answer questions, fix bugs, or do
documentation etc. for a while.
Gene - it's not your fault in particular, it's just... one discussion of a
particular kind too many.
I've actually spent a few days of my life trying to figure out whether we
can have clouds generated by numerical integration over a true 3d
distribution. The answer is no. and I know of a few good math reasons why
this scales how it does. Last time I checked, FSX had pretty much our
cloud technology (rotated stacks of billboards). Last time I checked, REX
had pretty impressive 2D cloud textures - but the 3d clouds seemed to work
like ours. I know some people have used Perlin 3d noise - but that's not
going to result in Cu clouds, it only works for some cloud types.
I might not up to date with everything (it's been two years at least since
I dealt with clouds), but I'm not going to spend hours verifying what you
say and how other renderers actually generate clouds.
If you (or anyone else) knows how to code fast 3d clouds which don't rely
on billboards, you're welcome to do it, and I'll happily port the shader
to ALS and insert the clouds into AW. Otherwise it's just not helpful.
It costs 5 minutes to write a quick claim how we *should* be able to do
something and how FG is really on a bad way if we don't do something
(plenty of forum posts along these lines) See - what we usually see in
stunning videos of XY is binary assets - texturing work, high-quality
meshes - we don't really see *technical* aspects of how it's done, and we
most definitely don't see where it fails, and if we'd use the same
techniques on FG terrain mesh and textures, it'd still largely look like
FG terrain mesh and textures and not like super-fancy video X.
It usually takes a few hours for me to verify what the rendering
background of a feature shown in screenshot XY actually is and how it is
done, lots of words to explain it once I understood it, usually nobody
likes the information why we can't do something easily because we. unlike
the other renderer, are committed to some procedure or format, and
actually coding some super-fancy graphic feature may be a months of my
life. And I'm at the point where I am simply tired.
So, Gene - you win, my assertion is demonstrably wrong - but I can't do
it anyway. Happy?
Stuart has tried impostors before (the code is to my knowlegde still in),
I've tried a couple of other techniques over the times which also don't
work - in my humble opinion, the main deficiency of our clouds is
texturing work (which is limited by the fact that I did most of it, and
I'm not good at it). If someone actually knows how to make it better and
not just that we should be able to - please go ahead.
In a similar vein, there's hundreds of absolutely cool effects renderer XY
has - which we all don't, unless someone actually figures out a way to
make them work in our framework with enough performance left. Everyone
happy now?
I'm going to focus for a few weeks on things I actually like, finish the
work on the Shuttle avionics and perhaps tinker again with my expolanet
simulation code. If there's rendering or weather related bugs, they'll
just have to wait. There's the search for the next FG location in the
forum - anyone feel free to take over and make a poll, I won't bother any
more. I think if no one does anything we default to Prague (LKPR).
See you in a few weeks,
* Thorsten
