Jaguar XJ220 WHDLoad runs too fast/too smooth

[rsn8887]

I made an issue here, maybe someone can double check:
http://mantis.whdload.de/view.php?id=4738

Check game speed in first race of Jaguar Xj220 floppy version running on 1MB A500 versus WHDLoad version on A1200. WHDLoad is way too fast. Makes 1st place on first race almost impossible to reach for me.

[daxb]

Have you tried ExpChip and/or NoCache tooltype as mentioned in the Readme?

[rsn8887]

That works I guess I should have read the manual. Also interesting that these things are fixed just recently, 2019 after all those years of being faulty. Thanks!

[rsn8887]

Never mind, it is still a bit too fast. Problem with this game seems to be that it becomes more difficult even if it slightly runs too fast.

[jotd]

ExpChip and/or NoCache tooltype just make the game not use fastmem / caches. But it's not proper speed regulation.

Also interesting that these things are fixed just recently, 2019 after all those years of being faulty.

Well, I think i can explain why.

First, only a few people are still active in whdload team. Stingray, Psygore and myself (sorry if I forget someone). Bert (the author of whdload) is focusing on whdload itself.
So if the whdload slave source isn't available, we have to ask the author/disassemble/resource it so we can work on it.

Then I can only speak for myself but I didn't really know how games worked on the amiga until I code my own game. The fact that winuae is also more powerful, has MMU so we can code without testing too much on a real machine, and cross assemblers + lots of scripts on the windows side help a lot.

I also changed my method of fixing games. I first get clean dissassemblies of each part & each version of the game, so I can get a global picture. In the old days, I used to write down addresses/offsets on papers and that's it...

To fix a game which is too fast, you have to identify the main game loop. Then figure out the speed on a A500/A1200 (WinUAE is VERY useful, with a real amiga you just can't do that) and try to match it.

I suppose that the most common issue is for games running at 25Hz. Game HAS to sync on raster beam else there's flicker on the screen. It's often done properly (except for Test Drive II...) but the game doesn't check that TWO beam cycles have passed, so if the machine is fast enough, only ONE beam cycle can have passed. Forcing wait for 2 beam cycles (at the start of the mainloop) is the way to regulate the speed evenly. Using delay loops (even calibrated) doesn't work properly. I dropped that method. I applied the new method on recent fixes (Led storm, Populous 2, Magic Pockets to some extent, more I can't remember right now)

[rsn8887]

Thank you jotd for the detailed explanation. I understand the reason now why those cache and memory tooltypes will not give the exact same speed as A500 what the game was designed for.

[kot_behemot53]

I know that this thread is 3 years old, but I just want to chime in that the game also sometimes does it on stock A500, so it doesn't have to be a WHDLoad issue. I recently had this when I fiddled with the radio before a race and clicked on the "seek" buttons that go through those faux radio frequencies. I'm not sure if it's a glitch or intended behaviour.

[Predseda]

No, the game includes this hidden speed mode (activated with the radio) from the beginning.

https://www.lemonamiga.com/games/cheats.php?id=1199

[DanyPPC]

Anyway it needs a better whdload fix to have a game that runs as on a stock A500.
Like the Buggy Boy recent slave..

[derSammler]

I'm currently trying to get that game running at all. It has some very odd behaviour, which seems to be related to the cheat mode that was added some updates ago (I immediately win ever race without doing anything, but no cheats are active).

Need to test older slaves...

//edit: issue found, posting in a new thread.

[tomcat666]

It is a problem on most very fast accelerators, even with EXPCHIP and NOCACHE options it cannot be slowed down enough to make it run in proper 25 fps

[derSammler]

For Emu68, use these settings to have the game run at normal speed.

This is for the 3A+ only. May not work for other Pi models.

And yes, ICNT=1 works fine with this slave.

ps: Make sure your Cleanup script reverts these options, or you end up on a very slow Workbench.

[jotd]

it could be a known 68040 bug where vblank interrupt needs to be acknowledged twice.

This bug also affects PiStorm apparently.

[derSammler]

If you refer to my post, using an older slave fixed my issue.