A.D.A. Amiga Demoscene Archive

  Welcome guest! Please register a new account or log in

  

  

  

log in with SceneID

  

Demos Amiga Demoscene Archive Forum / Demos / Swansong

 

Author Message
Corial
Member
#1 - Posted: 29 Apr 2011 08:45
Reply Quote
Britelite, what is actually going on regarding your 3d? It a really nice looking glow you have made, and how did you achieve that nice effect?

What format is the music? It sounds so freaking great!
d0DgE
Member
#2 - Posted: 29 Apr 2011 08:58 - Edited
Reply Quote
...and where the heck can one retrieve that soundtrack !!!one1!


EDIT

found your link on pouet
z5_
Member
#3 - Posted: 29 Apr 2011 12:23 - Edited
Reply Quote
When watching the demo, i knew there was something different. It was as if every scene was "washed in light" (for lack of a better description) but i didn't realise it was a glow effect around the objects. Works really well with this kind of sad autumn atmosphere.

Also, any official word on wether Britelite will continue coding on Amiga? I truly hope so as it wouldn't be the same without him.
StingRay
Member
#4 - Posted: 29 Apr 2011 13:49
Reply Quote
z5_:
Also, any official word on wether Britelite will continue coding on Amiga? I truly hope so as it wouldn't be the same without him.

As far as I understand it, "only" Dekadence is dead. Doesn't necessarily mean he completely quits Amiga coding, I'm sure we'll see some Supergroup or RNO release from him sooner or later. :)
britelite
Member
#5 - Posted: 29 Apr 2011 15:11
Reply Quote
Corial: I've applied a glowfilter on top of the 3D (and also making sure it only reacts to the colors in the 3D-objects themselves, and not the background). And the texts also have some inverted glow applied and are drawn with alpha-blending to get that extra smoothness.

The glowfilter itself is a version using two passes, one for blurring the stuff to a separate buffer horizontally, and then blurring that buffer vertically and at the same time applying it to the original screenbuffer through a shadetable. And the glowbuffer is in 2x2 precision to get a more decent framerate.
britelite
Member
#6 - Posted: 29 Apr 2011 15:14
Reply Quote
And about my coding-activities, I'm still a bit unsure about me continuing at all as I've really been running out of inspiration and ideas the past few years. But then again, it's probably not like would be able to stop completely anyway ;)
sp_
Member
#7 - Posted: 29 Apr 2011 20:20
Reply Quote
I really liked the glow filter. I have seen other groups use similar tricks, but not as beatiful as this. Rendering in 2x2 was a good idea. It's difficult to see the difference. I wonder what the flow will look like with an env mapped object. Maybe the object can be rendered plainshaded in 2x2 just to calculate the glow. and then apply it to a envmapped object screen afterwards..

You need to render the object 2 times though. Maybe even 4x4 for the plainshade part is enough..

Just to give you some more inspiration. Can't quit now! :D
sp_
Member
#8 - Posted: 29 Apr 2011 20:29 - Edited
Reply Quote
BTW; loaderror also used to use two buffers. one for lighthning and one for data. and then combine the buffer with a shadetable access per pixel. To speed up I rember I helped him with adding a Merge8 C2P pass with longword reads. (But this was a looong time ago. 2000?) When processing 8 pixels per loop. you also remove the ALU stall from the shadetable accesses..

Much faster than this:

move.w (a0),d0
move.b (a1)+,d0
addq.l #1,a0
move.b (a2,d0.w),(a3)+

edit:

but when doing 2x2 shadetable accesses there is probobly bether loops. To increase cachehits all 4 pixels should be processed close to eachother in the loop. and you reduce memory reads from the lightningbuffer by 1/4
britelite
Member
#9 - Posted: 29 Apr 2011 21:10
Reply Quote
sp_: yeah, I currently do the merge of the buffers while applying the verticalblur, so not too much time wasted there.
kiero
Member
#10 - Posted: 30 Apr 2011 00:04 - Edited
Reply Quote
didn't look at the screenshots really so maybe you do it already, but to improve image you can do simple average to calculate middle pixel getting 1x2 precision almost for free (add+shift+holding previous value). i do that in lots of cases when merging passes rendered in different resolutions. if you have some spare cache available (assuming no shadetable) you can even go further and store previous line to get 1x1 back;)
z5_
Member
#11 - Posted: 30 Apr 2011 13:01 - Edited
Reply Quote
britelite:
And about my coding-activities, I'm still a bit unsure about me continuing at all as I've really been running out of inspiration and ideas the past few years.

Looking at your release history, there are 28 prods on ada that you coded, starting from 2004. Add two revisions entries and a couple of productions that aren't on ada (yet), and i think we're looking at 35 to 40 releases in about 7 years. That really is an insane amount of activity so it seems only normal that you're running out of inspiration and ideas. Maybe it would help if you adopted the "one or two releases each year" strategy that most groups have now.

Also, it has to be said that the lack of inspiration wasn't visible in your demos and the quality of your demos is still awesome.

In short, i really hope you find inspiration and continue releasing these great demos, even if it means only one demo each year.
britelite
Member
#12 - Posted: 30 Apr 2011 13:31
Reply Quote
z5: truth is, since 96 I've made over 150 demos/intros if you combine all platforms, groups and even fakegroups :)
kiero
Member
#13 - Posted: 30 Apr 2011 20:30
Reply Quote
In my opinion people get washed out because they try too hard. Try doing it like i do. Prepare some framework in which you can prototype in high level language. Like C/processing. try multiple effects which run really slow. Under emulation it really doesn't matter. then check if any of them is a candidate for a demo. This is the pass when you thin about smart optimizations. Not cycle counting. SMART. sometimes the slowest effect turns out to be the easiest. if you think about it DIFFERENTLY. This is the most important part. Think. THINK. Think how you can fake it. in my opinion this is the spirit of the amiga demoscene. Make Amiga scene move forward.

PS. And Ask . Asking never hurt nobody
sp_
Member
#14 - Posted: 30 Apr 2011 22:24
Reply Quote
Britelite: That must be a world record!

Kiero:You are right. Youtube videos are realtime too.
kiero
Member
#15 - Posted: 2 May 2011 12:44
Reply Quote
sp: no idea what youtube has to do with it really. my point is that you can make slow code to run fast with clever thinking instead of cycle counting which can be tiresone and time consuming for more complex effects.
d0DgE
Member
#16 - Posted: 2 May 2011 12:54
Reply Quote
My opinion on the idea-issue is - as I'm probably the weakest kid on the block regarding the Kung-Fu - that a coherent motto/theme and its subsequent presentation outranks any load of effect-mayhem in a production ... unless one tries to deliver a pure "show-off" production.
Crumb
Member
#17 - Posted: 4 May 2011 00:32
Reply Quote
@britelite

please don't stop coding, imho the quality of your releases is impressive, perhaps you could relax a little and release "just" a pair of demos each year ;-)

Just relax a little and new ideas will flow to your mind... you are one of the few who still dare to code for plain A500, I guess coding for it is a great challenge and you got your 030-like effects running in 1 or 2 frames (without fastram). I bet than some weeks after you take your well deserved holidays you'll be doing something completely different and will think suddenly "omg! what would it look if I did this and that..." ideas will come to your mind, if you have made so many releases it's pretty obvious that you have plenty of imagination and ideas so don't worry, you'll be back here soon with new ideas for releases!

I was going to praise metropolice but swansong was also beautiful, as supergroup releases, as rno oldskool demos, as dekadence ones as other releases you made... you even composed some SID music! and your code always runs smooth and fast...
Crumb
Member
#18 - Posted: 4 May 2011 00:52
Reply Quote
@sp

Why not making some release together with Britelite for A500? I bet you could make Paula&Denise sweat :-)

My interpretation of Kiero wise words (these are the usual good coding practices used worldwide):
1. Finding good algorithms is prioritary over optimization, specially at the beginning of the development phase
2. Optimization should be done only when the algorithm is good enough and just in the parts that require it, other parts can use C code happily
3. There are various ways to get similar effects that should be investigated (even if some parts are fake or inexact from a mathemathical point of view)
4. this doesn't mean demo is composed of faked animations

@Kiero
Releasing 150 productions would be tiring for anybody even with the best IDE in the universe, most coders have not released a quarter of that in all their life, but I see your point :-)

 

  Please register a new account or log in to comment

  

  

  

 

A.D.A. Amiga Demoscene Archive, Version 3.0