pouët.net

I found out I can precalc the last two divses pr poligon, but it will cost memory... my mapper only has 8:8 presition and txtures jump, due to the self modified code. But hey its really smooth on the Mc68000!!
.
Since Each effect in my intro will require wast amount of memory, maybe I go for a demo instead. :D
.
And one more thing.. Atari STE coders:
The amiga has the copper, multitasking blitter and the 5th bitplane. It will be difficult to beat the speed of my production.

He-He.

Yeah! New stuff for old A500 is coming. :)
Well, I was looking for old contraz's demos and I've found a lot of dead links on pouet.

all this hardware-banging-talk is making me excited inside my underpants.

Looking forward to the final prod SP! ;-)

Thats nice sp^ctz, but Atari coders don't use div at all, even without resorting to precalculations. =P
(use a log table and subtract instead, then use an exp table to get the real value again)

Kudos for the hard work.

Sorry to highjack a bit this thread but could guys please enlight me about the copper: what does it control ? the palette ? the address of the screen ? ... since the bitplanes are not interleaved on Amiga, can it manipulate them separetely ?

p01:

The Amiga Hardcore Hardware Banging Coders Forum:

http://ada.untergrund.net/forum/index.php?action=vtopic&forum=4

it's a place for all your hardcore Amiga coding related queries...especially those about its hardware! :)

Ok guys, time for an update.
Work is going forward. The txture-mapper is smaller and smoother, and I found some tricks to optimize further. Now i finally remembered why I never made a descent demo when I was active. TO MUCH WORK!! I miss the youth spirit, but I have plenty of spare time. I have very little of my old sourcecode left so I have to rewrite alot. It is time consuming and borring. With the txturemapper up and running I can make some new effects never seen before on the 500. One of my favorite intros on amiga. is Chaosland from Virtual dreams (the party 1993 vinner) I am writing a similar blitter/chunkyversion in 4 bpl. This morning I watched a couple of nice demos from Sanity. Arte and Roots 2.0. in winuae with disaassembler ofcourse. :D
.
I have contacted my friend Dran/ctz wich will help me put together a production and design but he is pretty busy with work now maybe we delay the release to Assembly's oldschool compo in August.No "Coders colors" this time. :-D Isaac has made a couple of THX tunes we plan to use. I need all the chipmem for the effects, and c64 chip music is beatiful for my coder ears. :D

.
I really miss women and booze now. My thai girlfriend say I'm married to my computer. I tell her I'm just banging Alise and Denise. :D

sp - Thailand

btw. be careful with fastram on winaue... fastram cheats you ;) for example

>The SMC table renderer now goes 25fps on 320*220 2x2 ... is for sure not correctly... try with chipmem only... ;)
i did some tests with winaue and my a600 and the max you can do with it is 25fps with 320*140 (or a bit more) in 2*2 with this technique... even with a movem write to the chunky

Finally a real interesting thread in pouet BBS! :)

This is hardcore coding - cool stuff!
After seeing the nice stuff done with chunky effects on the C64 in the later half of the nineties, I always wondered why noone put some effort into A500 chunky.
Looking forward to the result!

ah ok... you used a 16*16 texture in your test... i did so too now... then i'm able to do 320*265 (or so) in 2*2 25fps... my test before was a 128*128 texture...

Sdw there are some chunky things ... like rno, da jormas or my 18 years thing... ok this technique is faster on amiga... so there will be more to come for sure ;)

sp: OCS showdown at asm then! (Yes, I know I've said this to others in 04, 05 and 06 aswell...)

No I am using a 256x256 txture. I did some more testing now. You are right about chipmem only slows down. In the newest version of the sourcecode the 2 blitter passes now run in a blitter interrupt. In the source I applied for testing only one blitter pass multitask with the cpu.
In the source I released you can changing the raster position of the vblwait and notice that the effect might use less rasterlines. this is because when the blitter is working outside screen DMA the effect goes faster. In the new version the blitterc2p will start on the end of the display wich makes it optimal. Turning of spriteDMA and vblwait interrupt speeds up a little bit too :D
.
In my newest version I also use a 2x2 copper stretch with the modulo I am not sure if this is faster than displaying a 2x1 screen.

btw; stingrays font plotter has divses in it. so it takes 25 rasterlines to plot the text :D

The final version will be able to render more pixels. Probobly i end using word txtures to optimize the loop as you suggested. in the next version..

ic @chipmem... no i don't use your source... i recoded it...

yep that's logical @end of the screen... well i read your tut about c2p on amycoders... that was something i wanted to try next ;)


i guess the result of stings test is unusable because the source to the timing was not correct... i think it's twice as much it shows... because of

a) VBLWAIT 200
and
b); bsr WAITBLITTER in the second call...

so the second call is not added to the time... but anyway

yep word textures speed up much... i write 8 words to the chunky at once...via movem

Still it only shows the real time your effect needs. :) And I couldn't be arsed to optimize that textwriter (which was just a quick hack anyway). :) Divisions are used for the Hex-Dec conversion. :) Feel free to add a divtable. :D

Hex-Ascii conversion rather. ;)

The textplotter works for me. I like it, and use it to test on my winuae a500 setup with wb1.3 dos disk. :D Seing the bluescreen with red pointer bring back memories.. :D
.
You are right its difficult to see the excact speed of the c2p using the source I supplied. But then again the final demo will be much more interesting to watch too.. Cant give away all the tricks. he-he

Hmm, also i don't use signed divisions for that. ;) So no divs but divu instead. :D

slummy: Should be enough time for you to make something for assembly compo or what? :D You too ultra, lets make the amiga 500 into a playstation :D

oehm anyway... the 25 rasterlines are not added to the result... because timer done is called before the textout ... ;)

Yep, only times the actual effect. :)

The vbl wait can be removed by using tripple buffer, and wait for blitter fished instead. if the effect goes slower than 25fps....
Funny, all my optimalizations require more memory. I end up having a intro with only one effect. he-he
.
My goal is 25fps or 50. Watching the old sanity demos today gave me a good felleing when I saw one fram effects again. You have teoritical speed, and you have cheat spead... Cheating involves precalculation. A good demo should have effekts where ppl say. How in the hell did he manage to code that. Only a few coders in the amiga scene has achieved this.

In search for more ECS demos I like this one: Coppermaster by Angels.. 1990 impressive!
http://ada.untergrund.net/showdemo.php?demoid=295

Yes, Coppermaster rocks. Corsair's masterpiece. :)

I have posted an optimal movem version of the SMC renderer on the ada site. I still use byte pr pixel txtures, but limited to 256*128 in size. maybe I update my speed test source later today.

i tried out yestereday to call the blitter after the effect was displayed on screen... unfortunately the speedgain is not very impressive... just a few rasterlines... ok i only timed the first call to the blitter... after i've seen the result i skipped it :/

tbh i don't unterstand the result... hm or winaue is cheating me... i mean after the display there are 100 mem cycles more to give to the blitter... i tried with a 70 lines efx... so 243 lines with 100 mem cycles more... hm hm... odd