James Pond GFX glitch

[Anubis]

Just tried to run James Pond's first game, and there are graphics glitches at loading screen:






I tried both A500, A500+ setups (OCS and Full ECS), using IPF and ADF versions of the game, to the same results. (neither of games loads completely)


Am I missing something here?

[amilo3438]

Just tested James Pond - Underwater Agent (1990)(Millennium)[cr PDX] in last official and last winuae.7z = no problem! (QS A500 ECS 512+512 config)

[TCD]

Using most common A500 quickstart and the '3015' IPF it works fine for me in WinUAE 4.9.1.

[Anubis]

Must be my video card. Will test it at home.

Thank you both for checking it.

[Zarnal]

IPF 3015 is also Ok. Quickstart A500.

[turrican3]

which version of winuae do you use ?

[Zarnal]

4.9.1

[Anubis]

I used both IPFs - 3015 and 3016 and had the same results.

The same WinUAE version, 4.9.1, 64 bit exe.

[Zarnal]

Ok also with 4.9.1 64 bits and 3015.

[jotd]

those are blitter errors. use "immediate blitter" in WinUAE, or whdload version

[Toni Wilen]

It is 100% not display driver related. Do you really use proper A500 config? (=Quickstart A500 with max compatibility)

(You don't use immediate blitter when emulating A500 accurately..)

[Anubis]

Just used unmodified quick config for A500 and got the same screen.

Tried with FS-UAE and looks OK.


@jotd - that fixed it. 'Immediate blitter' instead of 'Wait for blitter' fixed the issue.

Not sure why 'Wait for blitter' is default.

Thank you!!


ps. Toni, can you add FIXED or SOLVED to title?

[Galahad/FLT]

Originally Posted by Anubis

Just used unmodified quick config for A500 and got the same screen.

Tried with FS-UAE and looks OK.


@jotd - that fixed it. 'Immediate blitter' instead of 'Wait for blitter' fixed the issue.

Not sure why 'Wait for blitter' is default.

Thank you!!


ps. Toni, can you add FIXED or SOLVED to title?

Wait for blitter is default because there is no immediate blitter option on a real Amiga, wait for blitter is what a real Amiga expects.

Some code however is only expecting 68000, and on that processor, the programmer has written the code to sometimes not have a blitter wait because they know that by the time the blitter is setup for another blit operation, it has finished the previous blit.

[jotd]

I've noticed that WinUAE wait for blitter option doesn't work or fix things for me when blitter waits are missing in the game. Immediate blitter does the trick, always.

Best way to code is to wait for blitter just before modifying anything for the next blit.

Sometimes blitter waits have been inserted in a lazy way (I plead guilty) to wait right after BLTSIZE because it's easier, but it's also slower.

[Anubis]

Thank you for explaining.

[Zarnal]

Originally Posted by Galahad/FLT

Wait for blitter is default because there is no immediate blitter option on a real Amiga, wait for blitter is what a real Amiga expects.

I don't quite understand anymore. In principle "Wait for Blitter" is a compatibility hack if CPU is too fast. I thought it was only useful at 020+. What interest on a Quickstart A500 (in the case which interests us with James Pond)? I am wrong ?

[ross]

Maybe it's better to clarify a bit..
Perfect and better emulation requires neither "Immediate Blitter" nor "Wait for blitter".

"Immediate Blitter" does what it says: the blitter operation is performed in 'no time' during emulation, so there is no possibility that an incorrect blitter wait or annoying setting of the registers will be performed during an 'active' operation.
But obviously in this way there is no emulation of what really happens because there are no waiting times for the result and the timed operations risk failing.
I advise against activating it when you want to test your code because it does not provide reliable results for the corresponding real machines.

"Wait for Blitter" is a little more complicated to understand and often misunderstood.
Basically in this case the Blitter works like usual, unless the CPU code (or Copper) attempts to write to blitter registers while previous blit is still active.
If that happens, it wait for the blitter to finish (just like a coded blitterwait would).
That's why it can be found preset in some configurations with a fast processor: because in these cases a failed blitterwait (code expected to be slower and blitter already finished its job) causes video glitches (and that's what is done programmatically in patches or WHDLoad slaves).
For a perfect emulation also in this case the check should be removed.


EDIT:

Originally Posted by Zarnal

... I thought it was only useful at 020+. What interest on a Quickstart A500 (in the case which interests us with James Pond)? I am wrong ?

There could also be a case related to a normal A500 (but I have no idea if it concerns the game in question because I have not checked the code).
If the registers of the blitter are written before the previous blit is finished (therefore without a blitterwait), the result is 'undefined' (or rather, with a perfect internal emulation of Agnus it could become 'defined').
In that case the emulation could be incorrect and the result unexpected; a "Wait for Blitter" might fix it.
Theoretically, for a normal A500 this should never help

[Mclane]

I presume this applies to the A500 plus.. As in neither are really needed to be set for correct emulation.

[ross]

Originally Posted by Mclane

I presume this applies to the A500 plus.. As in neither are really needed to be set for correct emulation.

Yep, it applies to any 7Mz 68k machine.

[Mclane]

Thanks Ross, appreciated..