Tornado AGA - Speed up code!

[paraj]

Originally Posted by Aardvark

Would it be possible to use NoMMU option without performance hit?

To be honest I don't know enough about WHDload to answer that, but I'd be interested to know. Maybe you can ask in project.WHDLoad?


However I've just noticed that VBlank interrupt routine calls ScrollVPort and in both of you're crashes that's happening concurrently. Surely that can't be a coincidence! Originally the game called WaitTOF just before LoadRGB32 which ensure they'd be very unlikely to overlap, but I patched that out.


If you enable the "Frame rate limit" option the WaitTOF call will not be patched, and if I'm right the crash should not happen. Game will be limited to 12.5fps, but better than not working. I'll try to confirm this theory and get it fixed.

[Aardvark]

Originally Posted by paraj

To be honest I don't know enough about WHDload to answer that, but I'd be interested to know. Maybe you can ask in project.WHDLoad?

Well, NoMMU option shouldn't have any performance hit, but in this case it seems to turn off cpu caches too.
I can't fix that with "NoMMU Cache" because weirdly enough Cache option also kills framerate, even when its a sole argument.

[paraj]

Originally Posted by Aardvark

Well, NoMMU option shouldn't have any performance hit, but in this case it seems to turn off cpu caches too.
I can't fix that with "NoMMU Cache" because weirdly enough Cache option also kills framerate, even when its a sole argument.

I don't know, sorry. But you can try attached test version instead. It disables "shake" but I hope should fix crash you're seeing (while I work better fix).

[Aardvark]

Originally Posted by paraj

I don't know, sorry. But you can try attached test version instead. It disables "shake" but I hope should fix crash you're seeing (while I work better fix).

No crashes so far with the new slave, but there is regression with keyboard handling. Many key presses counts as two, e.g. can't toggle views with Numpad7 or F2, and after takeoff I couldn't retract landing gear with G.

[paraj]

Originally Posted by Aardvark

No crashes so far with the new slave, but there is regression with keyboard handling. Many key presses counts as two, e.g. can't toggle views with Numpad7 or F2, and after takeoff I couldn't retract landing gear with G.

My fix for "1/3" numpad issue is wrong then (it's included in that build). Thanks for reporting. Back to the drawing board.

[paraj]

Update:
- Quit key changed to numpad '/'
- Think keyboard issues should be handled
- Added dirty hack that may fix crashes Aardvark were seeing

Known issues:
- Being downed by rocket (I think) doesn't show properly, so it looks like plane is just rotating for no reason and you die. (Missing blitter replacement function).

[Aardvark]

Seems good

[Angus]

The Glance Left/Right function is working great now.

[paraj]

RC1. I will do RC2 if there are any changes (or joystick calibration thing requires it)
Updates:
- Smoke should now be rendered.
- Changed slave mem requirements (from 2MB chip, 1 MB other to 1 MB chip 2 other). This should give a few more FPS.
- FPS display is optional (and off by default).

[vroom6sri]

A quick test on my A1200 with ACA1230/56 seems stable and a noticeable difference over the original.

Thank you very much for your efforts with making these improvements.

BTW what does RC1 and RC2 refer to? I read through the thread again but couldn't see it referenced.

[Angus]

Really Cool? ;-)

No, I think it's Release Candidate.

[Angus]

Guys, a short clip of the Paraj's latest patch.

[ Show youtube player ]

[vroom6sri]

Wow! And I was pleased with 6-7 FPS from my ACA1230/56. What are you running under the hood to get those speeds?

[sokolovic]

Maybe it could be nice to apply the same patch to the ECS version. The fewer colors in it could be useful to gain some more FPS, even on AGA machines.

[Karlos]

@paraj

Is the approach/workflow here the same as it was for TFX?

[Angus]

Originally Posted by vroom6sri

Wow! And I was pleased with 6-7 FPS from my ACA1230/56. What are you running under the hood to get those speeds?

It's a PiStorm32 with a Pi 4 attached. I love my Blizzard 060 but it's the fastest accelerator my 1200 has had inside it, and the cheapest.

[pixie]

Originally Posted by Angus

Guys, a short clip of the Paraj's latest patch.

[ Show youtube player ]

It's strange, it doesn't seem as smooth as the numbers imply

[pipper]

It’s possible that the math involved is terribly quantized and so we don’t get to see smooth rotations even though the framerate is high.

[Karlos]

Originally Posted by pipper

It’s possible that the math involved is terribly quantized and so we don’t get to see smooth rotations even though the framerate is high.

Could be. You tend to notice this more in rotation than translation in my experience.

[tomcat666]

Originally Posted by pipper

It’s possible that the math involved is terribly quantized and so we don’t get to see smooth rotations even though the framerate is high.

It is similar with TFX ... where you still get "jumps" when rotating around.