Best fullscreen screen.

havoc · Post by **havoc** » Sat Apr 30, 2005 9:53 am

odd...

here's my entire setup just before running the demo without major problems (some screens still bug, but mainmenu etc works ok (and a small update, just noticed it crashes saint entirely upon selecting certain screens)):

http://img235.echo.cx/img235/6042/saintconfig9qu.jpg

Cyprian · Post by **Cyprian** » Sat Apr 30, 2005 10:13 am

I was test DRAGONLS.MSA and under Saint 1.99d its not working

havoc · Post by **havoc** » Sat Apr 30, 2005 10:17 am

check the screenshot... i am using saint1.99, not saint1.99d

have to say this is one of the pickiest demos ever indeed.. i guess i was lucky to get it to run at all ;)

unseenmenace · Post by **unseenmenace** » Sat Apr 30, 2005 10:23 am

Well that explains it then. The image you posted works but the other one that muguk provided a link to doesn't.

Cyprian · Post by **Cyprian** » Sat Apr 30, 2005 10:24 am

havoc wrote:check the screenshot... i am using saint1.99, not saint1.99d

have to say this is one of the pickiest demos ever indeed.. i guess i was lucky to get it to run at all

sorry, my english is very poor

You're using it udner Saint 1.99 with success but I'm testing it with Saint1.99d and it's not working

unseenmenace · Post by **unseenmenace** » Sat Apr 30, 2005 10:44 am

I got havocs image to work on Saint 1.99b but the image that muguk put a link to doesn't work at all.

havoc · Post by **havoc** » Sat Apr 30, 2005 11:19 am

Cyprian_K wrote:sorry, my english is very poor :)

no problem! my french is not so good either ;)

You're using it udner Saint 1.99 with success but I'm testing it with Saint1.99d and it's not working

that's useful info for sure.

so far, we know this:
- it works in 1.99 and 1.99b
- it doesn't work in 1.99d

anyone else who has success running the image on any emu, please post specs/version numbers/etc! :-)

prog99 · Post by **prog99** » Sat Apr 30, 2005 12:55 pm

havoc wrote: anyone else who has success running the image on any emu, please post specs/version numbers/etc!

Saint 1.99c works but at least one screen crashed the emu.

ljbk · Post by **ljbk** » Sat Apr 30, 2005 1:58 pm

Well thanks for this working disk image

!

I just tested the demo with my STF and SainT 1.99c, so i can tell you a bit about my opinion on the demo and about the screens emulation.

The demo is overall very good, but i don't think the Arkanoid/Rainbow Wall fullscreen is that impressive: the number and size of the simultaneous sprites is not that big and is probably handled at the time where the lower border fixed bitmap is displayed until next frame bitmap starting again to avoid problems. But the design of that fullscreen is very good.
The rest is color changing (probably the same HW register / logical color) and 2 bitplane background / tiles trick.
There is also a good mastering of the mouse with synchro code.
But that is only my modest opinion.

Now for the screens, one of them (Happy Islands) does not open the top border on my ST (it does with SainT) and flickers from time to time due to HW scrolling.
- 2 screens crash SainT: Blaupunkt and Pixel Land
- 4 fullscreens do not work correctly: Great Plasma (fake fullscreen), Rainbow Wall, Ace Ploter and Fractal Maker (ACIA read timings or pairing instructions ?) but Unlimited Bobs does work and so does the Reset demo
- there is a problem with Infinite Plan with the color at the bottom. But for this screen despite i can not see that color on my STF, it is like there is some kind of special sync occuring (screen at 60Hz to reduce color flicting perception ?)
- finally there is a problem with the menu as in the STF you get a continuous plasma (there is no repeated block): may be this problem has to do with the instruction being executed to be unique (movem.l) and the shifter emulation not reading the shadow color registers at the middle of an instruction execution ?

Leonard, it seems that you have some work for your spare time

PS: i justed tested it with SainT 1.99d and it WORKS. I did start with 1.99c, because it was said above that it did not work with 1.99d.
The behaviour is the same except that SainT does NOT crash with the 2 screens refered but the emulated ST crashes ...

alien · Post by **alien** » Sat Apr 30, 2005 4:33 pm

I like the creative use of Overscan in the Rainbow Wall demo. It's not as technically to the limit as some of the other overscan screens made so far, but... it's a nice idea, and it's prettier than Zarathoustra's Overscan Othello (the other overscan game that made it out). Although as Paulo said, I'm sure the real work is done in the lower border (not overscan): the overscan part is probably just changing the colours.

I was surprised but his happy islands demo's top border routine is flaky on my ST as well (and mine is no where as picky as Paulos).

Demos without music... wow!

Shame his fractal explorer isn't as optimized as Leonard's, because it displays things nicely.

Overall a nice demo and impressive to have written all that by himself. Thanks for posting it Havoc.

Now... why doesn't it even boot with STEEM?

Post by **ggn** » Sat Apr 30, 2005 4:53 pm

Alien,

Some years back I didn't have a mouse on my STE and I wanted to see the rest of the screens, so I thought I'd take a look under the hood... As far as I can remember the protection was weird, it dumped a lot of data starting from a low memory point ($8.w?) and clobbered all system registers. Perhaps STEEM doesn't emulate some of these registers?

George

ljbk · Post by **ljbk** » Sat Apr 30, 2005 4:54 pm

alien wrote:I like the creative use of Overscan in the Rainbow Wall demo. It's not as technically to the limit as some of the other overscan screens made so far, but... it's a nice idea

That is exactly what i meant, as i said, i think the design of the screen is very good.

alien wrote:Now... why doesn't it even boot with STEEM?

I would bet on prefetch protection problems ...

Post by **ijor** » Sat Apr 30, 2005 5:22 pm

ggn wrote:As far as I can remember the protection was weird, it dumped a lot of data starting from a low memory point ($8.w?) and clobbered all system registers. Perhaps STEEM doesn't emulate some of these registers?

Can't be. That sound a too simple protection from the point of view of an emulator.

ljbk wrote:I would bet on prefetch protection problems ...

Don't think so. Prefetch management in Steem is superb. Certainly much better than Saint's one, as Leonard admits.

I took a quick look. There is a protection/decryption based on syncing with the Video registers. So, either the demo doesn't run in STe at all (somebody tested?). Or it depends in perfect Video registers timing, something that, as we already know, is much better emulated in Saint.

Post by **ijor** » Sat Apr 30, 2005 8:52 pm

alien wrote:Now... why doesn't it even boot with STEEM?

Wow, look like the protection was designed on purpose against emulators! It’s amazing that Saint runs this correctly. It depends on (almost) perfect timing of the video address ($8209) and a DIVS instruction !!!

The basic technique is the same as the CopyLock packer. Each instruction is decrypted, traced and re-encrypted. But, unlike CopyLock, the decryption key depends on the current video address. Everything fine with Steem up to this point… until reaching a DIVS instruction. Seems like Steem timing here is wrong (no surprise) and then it gets the wrong key from the video register. This results in a bad decryption with a subsequent crash.

As said, no surprise that the DIVS timing in Steem is wrong. It’s one of the undocumented features of the 68000. Now, how Saint manages to make it right … Leonard???

leonard · Post by **leonard** » Sun May 01, 2005 10:21 pm

Now, how Saint manages to make it right … Leonard???

Well again I'm afraid it's just luck

SainT does not have any info about div exact timing (we have code to handle exact timing of MUL/MULS but no info about DIV).
For DIV, saint returns 136 (unsigned) or 150 (unsigned) cycles + number of cycle of adressing mode.

anyway I never seen that demo, but fullscreen of arkanoid-remake does not work on saint 1.99d ? did I miss something ? ( the first lines are working, but the other are not fullscreen). That's the same with another screen dispolaying a mandelbrot in fullscreen. I only see the beginning of the image in fullscreen.
Did someone get these screens work on any SainT version ?

leonard · Post by **leonard** » Sun May 01, 2005 10:34 pm

Leonard, it seems that you have some work for your spare time

How I think you're right Paulo

I just see the demo again, and I really like it ( maybe because of the technical look and feel ). Anyone know how old is that demo ? I never heard of it before !

Post by **ijor** » Sun May 01, 2005 11:28 pm

leonard wrote:Well again I'm afraid it's just luck SainT does not have any info about div exact timing (we have code to handle exact timing of MUL/MULS but no info about DIV).
For DIV, saint returns 136 (unsigned) or 150 (unsigned) cycles + number of cycle of adressing mode.

May be it is just luck, then. But may be it is also because of better timing emulation of the $ff8209 register.

May be the other version (the link muguk posted) works on a real ST, (somebody tried?) But fails with Saint because it uses a slight variation of the DIVS protection?

leonard · Post by **leonard** » Mon May 02, 2005 6:27 pm

May be it is just luck, then. But may be it is also because of better timing emulation of the $ff8209 register.

That's possible, even if ff8209 is very well handled in steem too (coz many demos are working).

May be the other version (the link muguk posted) works on a real ST, (somebody tried?)

Don't know if it can help but the first version I get from that thread was a ".DIM" image format, showing "0 sides" in saint. So maybe the DIM image is corrupted, or maybe SainT does not handle DIM properly. I'll check this.

leonard · Post by **leonard** » Mon May 02, 2005 6:45 pm

I confirm there was a mistake in the DIM format support in saint. Did anyone has an official DIM header description ? Actually I modify the code to get the "lastSide" at offset 6 (and you can get the value "1" for a 2 sides disk) but I'm not sure of that ( no official doc)

ljbk · Post by **ljbk** » Mon May 02, 2005 9:13 pm

Hi !

I just tested the DIM image file, by removing the 32 bytes header.
It works just as the MSA disk image with SainT 1.99d .
The files are NOT equal.
But there is only 1 byte different on the part that matters: inside the 2 sides and 80 tracks with 10 sectors each (if you believe the intro screen) at offset 0x863F4: one file has 0xDA and the other 0xD8.
May be 1 bit got corrupted. If it is inside a sample, it has no real consequences.

Paulo.

Post by **ijor** » Tue May 03, 2005 6:52 pm

leonard wrote:
May be it is just luck, then. But may be it is also because of better timing emulation of the $ff8209 register.
That's possible, even if ff8209 is very well handled in steem too (coz many demos are working).

I didn't make a detailed analysis. But as I remember, it seems that Steem reaches the relevant point a little bit too early. Just one NOP too early, I think. And it seems Steems uses a slighly smaller timing for DIVS. I guess you both estimated an average, but Steem authors choose a slightly smaller one.

I said this to Russ once, and I'll take the opportunity to say this to you and to everybody else. We should all make a joint effort to find out the exact DIVU/DIVS timing. There are a lot of people interested. Mame users, Atari, Amiga, Mac, and even more.

alien · Post by **alien** » Wed May 04, 2005 5:11 am

This is a hard question, and I don't think I can come up with a single demo.

- Level 16 in the Union was a jaw dropper to me, and made me search for left/right Overscan & I liked the way left/right borders were used in the main menu of the Union Demo
- TCB's super scroller in the Cuddly Demos was pretty damn cool, and just by looking at it I knew I had to think outside the box. My first hardscroller was running a few hours after seeing that demo. I liked the way TCB did their grey TV screen in overscan IIRC.
- Gunstick's mega disting scroller in the Dark Side of the Spoon was very cool
- I like MJJ proj's Full shade demo (colourful!) although it's obviously similar to Gunstick's demo.
- The ACF main menu's also nice.
- Paulo Simoes' Fullscreen Spectrum 512 is impressive technically.
- Finally I like my demo in the Ventura (nice KRX font) and in the Punish The Machine (4 bit hardscroller!)... and then there's my unreleased pac-man that had beautiful KRX gfx.
- Zarathoustra made a technically cool Othelo game, but I found it ugly
- Equinox' Shadow of the Beast was promising, although incomplete.
- Mcoder & Ziggy did the 3d overscan screen which was the only timer based left/right overscan demo I know of.
- Checkpoint's rotozoomer in Posh is also very cool.
- Next's main menu of the Phaleon is also really nice
- Although it's not overscan, Enchanted land deserves a mention because it was technically amazing (synchronizing the sprite drawing to the electron gun position)

Overall I'm a design freak so I'd have to go with one of Krazy Rex's fonts. For instance if Gunstick had KRX's gfx, and I hadn't, I'd probably prefer his demo to mine.

Unfortunately most overscan demos were... rather boring, which is understandable because it's hard to code overscan. But still it's a shame.

karlm · Post by **karlm** » Wed May 04, 2005 11:03 am

unreleased overscan pacman!? c'mon now alien, release it or bear the wrath of untold emails from atari-forum members

karlm.

alien · Post by **alien** » Wed May 04, 2005 7:51 pm

karlm wrote:unreleased overscan pacman!? c'mon now alien, release it or bear the wrath of untold emails from atari-forum members

karlm.

It's in Germany, somewhere in my parent's in law's house. I'm in the US... Don't even know if the harddrive still works.

Heaven/Taquart · Post by **Heaven/Taquart** » Fri Sep 02, 2005 2:50 pm

my one are ULM's Dark Side of the Spoon intro with the bubbles... and the wiggle scroller in DSOS.

Atari-Forum

Best fullscreen screen.

rainbox wall

Re: rainbox wall

Re: rainbox wall

Best Full Screen Ever