A.D.A. Amiga Demoscene Archive

Hey guys. Still in Thailand ready to do some hardcore coding again. The c2p is finnished. The SMC txturemapper maps to scrambled buffer
1 instruction pr pixel + littlebitmore.. (Current version is mapping to a nibble chunky buffer and have to do byte scrambling in a secound pass.(To be optimized :D)
The c2p source is already published in another thread here at Ada.

The demo will be 2 disks + trackloaded. It Will work on 1 mb vanilla amigas. 512chip, 512 fast(fake)

Hopefully released at Breakpoint 2008

I will post my progress in this thread.
To be continued....

btw; If you are a former Mc680x0 coder with spare time I need some help. Skilled low poly 3d artist also welcome. Contact me at sp (a) kvarteret.uib.no

I've been waiting to see your first demo with your new blitter c2p can't wait to see it.

Is the whole demo going to use this effect or are you planning any other type of effects ? Would be nice to see an AGA 020 version for the A1200 :0)

I have outlined 3 other ideas/effects wich will push the hardware, but not sure If I have time. Playing around
with the copper,blitter and hardware registers take alot of time. But it's fun. :D
.
Basicly I like to spend my time optimizing and make fast routines. Demo coding, arranging the effects,
choosing colors, timing to music etc. I find borring. With only 1 meg of memory I guess its going
to be more tricky to pack it all together.. When I was active on the AGA chipset I normally had 2-16
megs of ram to play with. I need some help here :D

@sp it's great someone is pushing the classic standard hardware A500/A1200 those machine were the golden period after all.

Can you put in a 102 effect for me :0) Now thats fast bitmap manipulation.

Would be good to see someone implement an effect using that multi tap idea for blitter blending.

@sp: respect for making a trackmo! Did you write your own trackloader for the project?

But will there be 4X AA supersampled copper bars?

sp very nice to hear @demo... i'm working on one too... maybe you remember... almost 1 year delay till now... but anyway i'm good in doing delayed demos ;) but didn't do something on it since 6 monthes now... the vc20 demo was somehow more interesting..anyway

yup @byte swap i'm still doing it like this... dunno all i tried now was slower than a second pass... but well maybe i'll find a solution... but with second pass it's still ok...

jo @1mb is heavy... esp when you use div tables and stuff... i also decided to do a trackloader otherwise it will not fit...

good luck with your demo ;)

I work fulltime with my demo now.

Yesterday I rewrote my txturemap-renderer from scratch.

0 divs. using 64kb precalc. Poligon loop is less than 256 bytes, so it will fit in the 020/030 cache. Poligon size has a limit of 128x128 and precition is 8:8 (-) (Selfmodified code). Currently it writes to a non scrambled byte chunkybuffer. I Will work some more to make it optimal for 000 and my scrambled blitter chunky format. The byteswap pass can be exchanged with multiple smc loops. but I get a headache each time I try to code it.

This is the first txturemapper I have written in 10 years, but I am pleased with the result. It's ugly but faaaaaast. My 2008 eyes are used to perspectivecorrect 3dcard smooth mapping..
Linear mapping in 2x2 is not state of the art.. But on 7mhz it will do.

Hopefully Stingray will help me with some hardware code. I have memory and busproblems.

With two coders we hopefully manage to finish something for Breakpoint.

@sp:

Great to hear and good luck. The same goes for Ultra aswell. It would be great to see a new Ultra release on Amiga.

Not for being an irritation object, but the rules for the amiga demo compo:

http://breakpoint.untergrund.net/compos_amiga.php


A demo only working on A500 will crash on the compomachine, so there migth be a problem of getting in into the compo.

There have been several A500 demos on Breakpoint in recent years with similar compo rules, so apparently it is not a problem to get them to run it on an A500. It might be a good idea to contact them in advance and ask, of course. At least to make sure they have an A500 to use as compo machine.

Or you could just code the demo in a way that it also works on faster amigas ;)

hear hear :)

I am sure it won't be a problem. The guy who was responsible for the compo last year was really really helpful, he was also a true "amigan" knowing his way around both hardware and software. They are really keen on having as many releases as possible, and quite willingly go out of their way to get things released.

The problem with using self modified code is that it will crash on higher processors with Cache. Shouldn't be a problem to code a workaround though. Simply a cacheclear or using more memory. But Winuae with JIT enabled will cause unexpected results. Anyone know how to detect winuae?

The official way is to search for uaescsi or uaehf device.
But you may also try to fool the cpu emulation with some prefetch tricks :-)

The problem with using self modified code is that it will crash on higher processors with Cache. Shouldn't be a problem to code a workaround though. Simply a cacheclear or using more memory. But Winuae with JIT enabled will cause unexpected results. Anyone know how to detect winuae?


You might want to check this thread. :)

@sp_,

What I meant was that, I am pretty sure that if a a500 demo is released they will dig up a computer from somewhere to make sure that it's shown properly. You could also warn them before by contacting the Amiga Compo Organiser, I am sure he will be happy to help.

Isn't Charlie compo organiser? I think he was last year.

Most probably he is compo organiser this year again. If so there shouldn't be any problems as he's very easy to deal with (last year I fixed a bug HOURS after the official deadline in the organiser's room while he was already recording the entries ;D).

The problem with using self modified code is that it will crash on higher processors with Cache. Shouldn't be a problem to code a workaround though. Simply a cacheclear or using more memory. But Winuae with JIT enabled will cause unexpected results. Anyone know how to detect winuae?

How about just disabling caches? :)

If the demo is supposed to run on A500, it will surely run fast enough on any Amiga even with caches disabled?

If it is most likely some A500 prods will be finished until BP I can bring my 500 with fake fast to the party. Did that the last two years for gaming sessions

I managed to fix the SMC mapper to plot in my scrambled chunkyformat. Only one smc loop is required and no byteswapping pass(!), The routine still have some possible speed imprvements, so I will use the day tomorrow to optimize further. Turning off the cache is a good Idea, and thats probobly the solution I will go for. JIT enabled in WINUAE will cause problems.. But who uses a a500 with JIT??

My wife is a little upset now, because I use to much time in my own coder-world. :D

The JIT is the 'instruction cache' of the emulated Amiga. I am quite sure that if the emulated Amiga disables the instruction cache, the JIT gets disabled.

And don't forget that you also make some people happy by spending time in your own coder-world. ;D

My triangle linear txturemapper is getting faster..

The innerloop is down to 16 cycles pr txturemapped pixel. including the dbf. (Loop can be unrolled for more speed) This Pretty good, considering that one move like this is 16 cycles alone.
move.b $0000(a0),(a1)+

An optimal unrolled loop will require only 14 cycles pr txturemapped pixel, But then the Selfmodified code routine will have to calculate more offsets.
I have to do some more testing to see what's fastest. I think the blitterc2p can use the free buscycles while I am interpolating. Add and addx is 4 cycles each and no extra memory fetching required.

For fullscreen Table effects less cycles pr pixel can be achieved.

The "Faster than copyspeed" is possible because each pixel is 4 bits and not 8 bits.

Hi folks,

Great to hear that someone is working on a kick-ass A500 prod for BP. ;)

Yes, i am the BP compo organizer this year again. Please make sure that the demo works on a 060 too. We cannot make exceptions, it only creates unnecessary flames and discussions after the party, because then everyone want to have his own tiny little exception then... (For example Karate users, who always mean trouble with OS3.1 installed on the compomachine, or TBL with their UAE-o-mania... ;) If the demo doesn't work on A1200/060, we will have no other choice but to disqualify it, or to record it from an A500, and put it into the Console/Wild competition. If you want, i can disable the 060 caches for you, but really, that's all, i'm afraid, sorry.

The A500 entries from last years worked on A1200/060 too, as far as i know.

If you want, i can test the code for you, on the compomachine anytime before the party too. (Not in the 2-3 days before the party tho', because i'll leave to Bingen on Wednesday.)

So, to sum up, i want to do my best, to help fixing the code for the compomachine, but i don't really want to make exceptions, and let any prod use a differently specced compomachine.

Ok, let the flamefest start. ;)

SP_: I missed your demo at Breakpoint 2008! :-(

Still working on it. Not enough time to release on Breakpoint this year.

Does that mean we have to wait another year ? :0(

How about a preview of the textue mapper running on www.youtube.com :0)

Yes, another year or maybe for the oldschool compo at assembly? I will put some stuff on youtube soon.

2x2 4 bpl look ugly, so I am considering running in 1x1 if AGA is detected. I simply switch to hires mode and change the fetchmode to gain DMA. The 3/4 extra bitplanes can be used to give me free hardware motionblur.