Fast Red Zone and other fastmem-relocated projects

[paraj]

Originally Posted by Retroplay

Pressing F2 ingame in Killing Cloud v2.0 (fast relocated slave) makes it quit back to Workbench.

If I do the same with old v1.0 or v2.0 NoReloc slave, it switches to rear view.

Good catch, there's an issue with the keyboard hook, though it's F4 that accidentally quits for me.

[jotd]

ok found the issue, released a new 2.1 with the fix. I'm really too confident with minor changes nowadays

[Reynolds]

Just thinking... Powermonger isn't a clearly vector gfx based game, but has stencil vectors, and is definitely faster on a Vampire for example, can it get some boost as well with Fastram relocation? Same with Populous 2, some speed fix would be awesome, as on high spec configs the standard screen is quite fast, but when I change to the fullscreen view it drops down the speed. Can be there any solution to get balanced somehow?

[PascalDe73]

@JOTD,
@RetroPlay,

Hi,
For the game Killing Cloud, I need some informations about the version supported :
Is it US version from SPS 3019 ?
Because I tried Your latest slave 2.1 with SPS 3031 Version (RetroPlay Pack v2.1).
With that version I need to use Noreloc.slave and use CUSTOM3=1
because without that the game crash on the menu when I select City Info.

It was tested on my A500 ACA500Plus & my A1200

Best regards,

Pascal

[Retroplay]

US version (SPS 3019) MUST have CUSTOM3=1 set even with v2.1 NoReloc slave, if not it crashes with an error.
I set this already in icon to avoid this.

PAL version (SPS 3031) support fast relocation, noreloc file is added to data dir next to disk images.

Btw, it's not my slave, it's JOTD's.

[PascalDe73]

Originally Posted by Retroplay

US version (SPS 3019) MUST have CUSTOM3=1 set even with v2.1 NoReloc slave, if not it crashes with an error.
I set this already in icon to avoid this.

PAL version (SPS 3031) support fast relocation, noreloc file is added to data dir next to disk images.

Btw, it's not my slave, it's JOTD's.

@Retroplay
I know the slave is made by JOTD, it was just that caught your eye.

When I use the PAL version (SPS 3031)

On my ACA500PLUS:

If I use the slave with the reloc with or without custom3=1
when I click on the City info menu, I get the error :

Exception "Address Erroe" ($F00C) PC = $A37D48 (Expmem $FD48) Read from $A377AF (ExpMem $F7AF)

With the alternative slave (NOReloc) AND CUSTOM3=1 the game works with that menu.

On my A1200 :

If I use the slave with the reloc with or without custom3=1
when I click on the City info menu, I get the error :
Exception Illegal instruction ($10)
at $5AEC4 occured.

With the alternative slave (NOReloc) AND CUSTOM3=1 the game works with that menu.

That's my observation

[jotd]

bug confirmed. not cool...

[jotd]

aaaand fixed. Thanks git and frequent commits. It was working on an earlier version and I could track down the change! That was close

[Reynolds]

Regarding the Killing cloud... The cheat of turning in/out the cloud is available only in the first mission. Can it be extended to all others too?

[jotd]

That would be awesome, but in the end I think that the detailed data just isn't there, else they would have left it in.

My theory is that they created level 1 with a lot of detail, saw that it was a slideshow on A500 and created the other levels with less detail, adding the "cloud off" option for level 1 so they could show their detailed 3D model (instead of removing it)

[Reynolds]

Originally Posted by jotd

That would be awesome, but in the end I think that the detailed data just isn't there, else they would have left it in.

My theory is that they created level 1 with a lot of detail, saw that it was a slideshow on A500 and created the other levels with less detail, adding the "cloud off" option for level 1 so they could show their detailed 3D model (instead of removing it)

I think as you can fly the same map, it should be kinda similar, but you can have right, with remarking that this could be a possible "cheat" or similar which wasn't removed before releasing the final version...?

For me viewing distance was always too short to travel with adequate speed, but can happen that with the actual speedup I can learn to control the vehicle better, so it won't be a problem from then...

[jotd]

If I can enable "cloud off" in other missions believe me I will.

[Reynolds]

Originally Posted by jotd

If I can enable "cloud off" in other missions believe me I will.

Yay!

[Reynolds]

Maybe a bit lame question, but regarding the A500, if it is a 68000-driven config but with real fastram, can that give any boost? I can't recall any difference when I use my A500 with an A590 for example, but there should be some...?
Other thing is, with the fastram relocation may that happen, if a game was asking for 1.5MB chipram and X MB Fastram, then it will run with less chip ram as well? My A1000 is asking it...

[jotd]

Originally Posted by Reynolds

Maybe a bit lame question, but regarding the A500, if it is a 68000-driven config but with real fastram, can that give any boost?

I think so, since code now runs in fast. But on a 7MHz I'm not sure that replacing the blitter by the CPU helps a lot.

Other thing is, with the fastram relocation may that happen, if a game was asking for 1.5MB chipram and X MB Fastram, then it will run with less chip ram as well? My A1000 is asking it...

1.5MB chip? which one? Generally it's because the slave is old and doesn't bother about handling fast mem. Report those.

[alexh]

Originally Posted by Reynolds

Maybe a bit lame question, but regarding the A500, if it is a 68000-driven config but with real fastram, can that give any boost? I can't recall any difference when I use my A500 with an A590 for example, but there should be some...?

The amount of work a 7MHz 68000 with no instruction cache can do during the time it would have been stalled had it been running from chip-ram is probably not that much.

Add in a CPU instruction cache and increase clock speed and you're loosing a lot of processing power running from chip-ram when stalled waiting for chip-ram access.

[Reynolds]

Originally Posted by jotd

I think so, since code now runs in fast. But on a 7MHz I'm not sure that replacing the blitter by the CPU helps a lot.


1.5MB chip? which one? Generally it's because the slave is old and doesn't bother about handling fast mem. Report those.

Okay, that makes sense with speed.
I wrote from memory that there were games which required more chip ram than normally would.

Good to know that this has changed. Now just have to work out how to upgrade my whd stuff in it's easiest way, instead of one by one. :P

[alexh]

Originally Posted by jotd

I think so, since code now runs in fast. But on a 7MHz I'm not sure that replacing the blitter by the CPU helps a lot.

I didn't even consider that. It is surprising that it is not slower than the non-reloc version that uses the blitter.

[paraj]

Originally Posted by alexh

I didn't even consider that. It is surprising that it is not slower than the non-reloc version that uses the blitter.

The new slave is probably not going to get good performance on a plain 68000. The fastram stuff probably gives the ~25% you would expect from not having to wait on chip ram, but you lose (I think) quite a bit from it replacing the SMC. The CPU replacement code is optimized for 020+ (really 060) so I expect it to be much slower than the original code, though trying it in WinUAE it doesn't seem noticeable (probably due to the SMC thing).

[Angus]

Originally Posted by jotd

If I can enable "cloud off" in other missions believe me I will.

As Reynolds said, I think you're always flying in the same city (San Francisco) and it's just that you fly to different parts of it - so maybe there is a fog off/on switch somewhere. Showing my deep understanding of coding once again. ;-)