Author |
Message |
skan
Member |
I just have to quote the whole Bonkers' post along with shitloads of +1s!!! ;)
|
SoLO
Member |
I was a bit surprised too about the Skarla last demo's comments. They are not that bad actually, most people enjoyed the demo, but they didn't like the music, especially the vocal. I still dont know why because i really like it, i like the vocal, it fits to the demo, but it seems it's a big deal for many people. I am sure if it was the same song but with a woman's voice, they will think it sounds great ;)
|
sp_
Member |
Jamies Comeback was a nice Surprise. Does it run smooth on the real hardware?
|
jamie2010
Member |
Actually it's not really smooth on my 68060 at 50 mhz, most effects run in 3vbl, with 2 more weeks all screen could be run at 25 fps, but you know for this we need time.
|
sp_
Member |
hehe. with 25fps on 50mhz it will be massive :D I iked that you used the old objects from the older demos and rendered them all together in the same scene.. That is what I call Optimizing.. :D
I have to disassemble to see which tricks you used this time :D
|
jamie2010
Member |
Parallelization and prefetch are the tricks)
|
sp_
Member |
Are you using the FPU for perspective correcting? While interpolating 8-16 pixels?
|
jamie2010
Member |
yep always, i use the winden division trick for parallelize all fdiv.
|
sp_
Member |
That is beatiful. :D The King of optimize is back!
|
kiero
Member |
oh come on, parallelizing divs is easy;) and in perspectivecorrect rasterizer basicly essential.
btw, nice work on the engine. i never realy liked spanbuffer (or derivate) based renderers tho. they are too limiting imo.
hope you will not make it one production comeback too.
|
sp_
Member |
hehe. Jamie was the king of optimize in 1998. You are the king of 2011. (Or maybe it is Blueberry?) To bad they don't produce Eurochart anymore.
anyway,
The old optimizer is giving you some challenges. :D
|
jamie2010
Member |
Parallelizing divs is easy, parallelizing all divs need more work and it will made a huge difference at the end. Btw stan's fluid or particle are easy too:)
Spanbuffer or coverage buffer in our case has some advantages and some inconvenients, our end scene must be faster without the occlusion part. If i choose this direction it's for the polygon material sorting before the renderer pass.
It was not a one production comeback, see you at Revision 2012.
|
sp_
Member |
I also liked the demo from Dekadence. The "shader like" GLOW filter made it look good. This is something you should try to implement Jamie. (Or just rip it from Britelite's EXE :)
It will give the 3d scenes 2011 look.
|
kiero
Member |
hehe, i never said fluids are hard. just lots of code uglification;) back to divisions i agree that they can be a pita. my texturemapper (except perspective correct one) has lame workaround by using lookups. never bothered to kill that. but i think i had integer divisions in transformations code till last year. paralelized fpu divisions gave both speedup and accuracy (mixed fpu and integer code). oh, and clipping code. maybe next year;)
about particles. there is one tricky/smart part in my particle rasterizer. it allows for big amounts of big and blended particles by utilizing simple trick (i wonder if everyone knew it before me). whole particle rasterizer is still in optimized C and it managed to render between 4 and 11(or was it 12)k particles in various effects (depending on how heavy rest of the effect is).
"It was not a one production comeback, see you at Revision 2012."
hopefully:)
|
jamie2010
Member |
You see my point, you're actually one or maybe the best coder on amiga scene and you don't use the amiga at full capacity. We use fpu-cpu in transformation for age now, add the prefetch and write all your vertex information in 16 bytes format and you will see a real boost. I tried to share edge computation but it was not a succes so i tried the cpu with a looked table(the very old way) and finally back to the fpu with the subpixel precision for avoid edge crack.
How you synchronized the particle effect on your music? Your synchro are really top.
|
britelite
Member |
sp_: or if someone asks nicely I can share the code for the glowfilter, when I've optimized it a bit more :)
|
jamie2010
Member |
Sp is ironic sometines:) I must optimize my slow radial blur too, process 2 pixel at once and it will be a lot faster.
|
kiero
Member |
@jamie: sure. i have lots of places to optimize my code. my problem is i don't want to focus on one thing for too long. sure it would be nice to optimize 3d engine (or any other effect) more and more but instead i prefer to add something new. this leaves your code slower but adds more variety to the demos and helps to keep you interested in coding. different schools maybe;)
as for particles, it really is just timer/envelope based. takes time but we have envelope reloading at runtime, variables which can be tweaked without recompilation (http://wozniak-m.w.interia.pl/guigrab.jpg) and timeline rewinding, so having sound editor on second screen it is not that consuming:)
|
skan
Member |
@Kiero Does it need 64mb this time again? Can't get it running here...
|
skan
Member |
@sp_
King of optimization? Blueberry, hands down. Or better: a weird Blueberry-StingRay hybrid!!! ;P
|
kiero
Member |
@skan
[blah, edited].
i just cut down 1-2 mb so it might run. will prepare special binary for testing
|
skan
Member |
@kiero Cool, thx! =)
|
kiero
Member |
ok, try this one: http://wozniak-m.w.interia.pl/eld-sotd_lowmem.exe it will be little bit slower but shouldn't require more than 32MB of memory. not tested on amiga so beware;)
|
skan
Member |
@kiero
Thanks man, gonna test RIGHT NOW! ;D
You see, my odd setup is 48MB, but coders tend to consider either 32 or 64 megs only, goddammit!!! ;P
EDIT:
No way, it gurus badly quite soon while loading... :(
|
kiero
Member |
ok, it might be something else then. just tried it under uae with 32MB and it runs together with final release. mail me at kiero@theelude.eu so we won't polute forum anymore.
|
z5_
Member |
Anybody got any idea when the final versions of the demos will be released? At least Shake off the dust and Human Traffic are confirmed to get a final release. Can't wait to watch them.
|
Charlie
Member |
I finally found some time to post here. Thank you for all the Amiga entries at Revision 2011! And even if the demos where the best again, lets not forget the 64k intros and the oldschool 4k intros as well, thank you for all of them. Because of the quality of your entries, I can only say, it's been a real pleasure to serve as your Amiga compo orga again.
I hope you all, and everyone who couldn't make this year will be back for Revision 2012, to shake off the dust from the old Amigas again, and let the world know by causing some massive human traffic, it's not the swansong of the Amiga scene yet! :)
|
d0DgE
Member |
z5: the final version to our humble intro has been posted on pouet today
Charlie: BESTEST Skoda compo master ev4r \o/
|
z5_
Member |
The first batch of releases is added. I'll try to do the others tomorrow. Anyone can help on the correct credits for Spring Onion?
Also, if anyone has Winuae settings for Massive, that would be appreciated.
|
jamie2010
Member |
The demo doesn't work with latest version of winuae, try the winuae 2.2.0. I can send you my config, but it's a really standard configuration. Give me your mail and i will send you the config.
|