Mortal Kombat II AGA Tech Test

[Old_Bob]

Hey, dudes.

I've uploaded a small test program for Mortal Kombat II AGA to The Zone. It's a teeny bit buggy but works well enough for testing. After a bit of fiddling, it *just* scrapes in at 50fps when running in WinUAE 4.4.0 at an emulated 14mhz with Fast RAM/quick launch option.

Testing in 5.3.0 though, it can't manage it unless the emulated clock speed is whacked up to 28mhz. Given WinUAE's inaccuracies in 68020+ emulation, it really needs testing on real hardware. I'd be somewhat surprised if a real A1200 at 14mhz could manage 50fps with it, but a 1220/4 accelerator board, or similar with a 28mhz CPU *should* hopefully?!

If you have a suitably specced machine and a few minutes to spare, it'd be nice to know how the program actually works and how it performs. If you could also maybe take a quick snap of the screen so I can see the timing bars, that would be useful, too.

This is the output in WinUAE, with the black border part being free CPU time where it's waiting on the next interrupt.





B

[alexh]

Thanks. Looking forward to trying it

[jotd]

very good!

Will you be able to use some of the original code or amiga code? The code for MK2 was available at some point but got withdrawn.


Is it using interleaved bitmaps? and how many colors?


I personally think that the first opus would be much more doable on amiga as AGA. Problem with those MK games is damn Shang Sung metamorph

[Old_Bob]

It's possibly worth pointing out that it's not in any way playable, the fighters just bounce around the screen. It's just a basic test of the graphics engine using a single frame for each fighter.



B

[Old_Bob]

Originally Posted by jotd

very good!

Will you be able to use some of the original code or amiga code? The code for MK2 was available at some point but got withdrawn.


Is it using interleaved bitmaps? and how many colors?


I personally think that the first opus would be much more doable on amiga as AGA. Problem with those MK games is damn Shang Sung metamorph

The demo runs in 8 planes/256 colours with interleaved bitplanes. The background is 128 colours with a separate 64 colour palette for each fighter. The CPU is rendering in a buffer in Fast RAM while the copper is executing a list to perform the erase step in Chip RAM. Then, it copies only the parts of the buffer overwritten in the draw step to Chip RAM when the last erase blit is done. With this, 50pfs is achievable if the CPU is fast enough. Where it would be simply impossible if relying on the blitter alone.

Annoyingly, I didn't find out about the MK2 source leak in time to get my hands on it before it was no longer available. But, yes, if a person was to take the task on, a conversion of the first game would be the sane option.

B

[AmigaBill]

I just tried it on my A1200 with TF1260. Hope this helps. It is just iPhone video https://www.dropbox.com/scl/fi/spj9i...b1ecgsz66&dl=0

[Old_Bob]

Originally Posted by AmigaBill

I just tried it on my A1200 with TF1260. Hope this helps. It is just iPhone video https://www.dropbox.com/scl/fi/spj9i...b1ecgsz66&dl=0

Thanks, dude.

Interestingly, your CPU isn't getting everything done in time to manage 50fps, which is why the border colour is flickering wildly. A bit surprising, given the spec of your machine and just the kind of thing I was concerned about. Especially considering the code for the entire program, which is less than 4kb, easily fits inside the instruction cache of your CPU?

B

[AmigaBill]

My pleasure Bob! Interesting. I don't know. I tried it with no startup sequence and also enabling and disabling the CPU Cache and the results are the same. Although I did run SetPatch each time. I don't know

[roondar]

Originally Posted by Old_Bob

Thanks, dude.

Interestingly, your CPU isn't getting everything done in time to manage 50fps, which is why the border colour is flickering wildly. A bit surprising, given the spec of your machine and just the kind of thing I was concerned about. Especially considering the code for the entire program, which is less than 4kb, easily fits inside the instruction cache of your CPU?

B

My guess would be that the TF1260 doesn't have good Chip RAM bandwidth for the CPU. Many 68060 accelerators and even some 68030 accelerators struggle with that. Since your code is copying from Fast RAM to Chip RAM, that might play a big role. To test this, I tried it on my A1200 with Blizzard 1230MK IV and it shows no flicker in the border, which seems to fit with this idea, as that card reaches the full 7MB/sec write speed for the CPU to Chip RAM. See video below:

[ Show youtube player ]

On a side note, it looks very nice to see the Sprites & background at such a high colour count
Also: sorry for the flickering video!

[alexh]

It was my understanding that the final firmware for the TF1260 can get 7MB/s chipRAM bandwidth.

Are you sure you're on the final Firmware (serialno 68090)?

[AmigaBill]

Originally Posted by alexh

Are you sure you're on the final Firmware (serialno 68090)?

No, I am not sure. I don't think I am. I was updated once, but I think there may be a newer one. I have to check.

[Old_Bob]

Originally Posted by roondar

My guess would be that the TF1260 doesn't have good Chip RAM bandwidth for the CPU. Many 68060 accelerators and even some 68030 accelerators struggle with that. Since your code is copying from Fast RAM to Chip RAM, that might play a big role. To test this, I tried it on my A1200 with Blizzard 1230MK IV and it shows no flicker in the border, which seems to fit with this idea, as that card reaches the full 7MB/sec write speed for the CPU to Chip RAM.

Thanks.

The copy step is the green bar, which is more or less the same as the screen shot in my first post. If your CPU is already maxxing out the Chip RAM write speed then are surely no further gains to had there with any other chip. The free raster time remaining also looks to be similar.

Since this is part where all the logic code would need to execute, as well as some additional graphical stuff like the shadows and fireballs along with all the blood and guts flying around the place, it seems in this case, even a very speedy 50mhz 68030 wouldn't be enough to maintain 50fps in gameplay.

It would be interesting to see if one of the newer ACA or Terrible Fire 030 trapdoor expansions performed much differently to the classic old Blizzard board.

B

[Superman]

Here is a quick test on my A1200 with TF1230 50mhz 128mb. It’s very impressive

[ Show youtube player ]

[Old_Bob]

Originally Posted by Superman

Here is a quick test on my A1200 with TF1230 50mhz 128mb. It’s very impressive

This one is surprising in that it isn't getting through the frame quite as fast as Roondar's Blizzard MKIV. Looking at the timing bars, it seems the writes to Chip RAM aren't quite as quick, although, it is still just managing to update at 50fps.

An interesting result. Thanks.

B

[BigD]

Originally Posted by Old_Bob

Hey, dudes.

I've uploaded a small test program for Mortal Kombat II AGA to The Zone. It's a teeny bit buggy but works well enough for testing. After a bit of fiddling, it *just* scrapes in at 50fps when running in WinUAE 4.4.0 at an emulated 14mhz with Fast RAM/quick launch option.

Testing in 5.3.0 though, it can't manage it unless the emulated clock speed is whacked up to 28mhz. Given WinUAE's inaccuracies in 68020+ emulation, it really needs testing on real hardware. I'd be somewhat surprised if a real A1200 at 14mhz could manage 50fps with it, but a 1220/4 accelerator board, or similar with a 28mhz CPU *should* hopefully?!

If you have a suitably specced machine and a few minutes to spare, it'd be nice to know how the program actually works and how it performs. If you could also maybe take a quick snap of the screen so I can see the timing bars, that would be useful, too.

This is the output in WinUAE, with the black border part being free CPU time where it's waiting on the next interrupt.





B

How do you use this? What sort of file is it? It just hangs Workbench when I execute it under OS3.9. Get a green and black chequered pattern under OS3.1...and yes it works! Very pretty!

[Old_Bob]

Originally Posted by BigD

How do you use this? What sort of file is it? It just hangs Workbench when I execute it under OS3.9. Get a green and black chequered pattern under OS3.1...and yes it works! Very pretty!

Yes, you just need to click the left mouse button when you see the grid at startup. You can also scroll the screen around with the stick in port 2. The OS gets completely killed off when it takes the machine over, so you'll need to restart once you're done.

I haven't got any OS version later than 3.1 to try it on, so I've no idea why it might fail on anything later than that.

B

[BigD]

Even with OS3.1 it still doesn't do anything beyond display the statue and background of the sewer level though! I can't seem to scroll with the joystick/pad! Setup: 060/AGA with loads of Fast Ram. Real Amiga.

[tomcat666]

This is A1200 with Pistorm32 with 2.2Ghz rpi4:

[Old_Bob]

Originally Posted by BigD

Even with OS3.1 it still doesn't do anything beyond display the statue and background of the sewer level though! I can't seem to scroll with the joystick/pad! Setup: 060/AGA with loads of Fast Ram. Real Amiga.

Not too sure what's going on here. Testing with the quicklaunch 68060 option in WinUAE works fine, along with all the others. If you're seeing one of the statues at the edge of backdrop image then it's not setting the scroll register and bitplane pointers correctly, for some reason?!

Is it displaying any hex value in the top left of the screen when the startup grid is displayed?

B

[Old_Bob]

Originally Posted by tomcat666

This is A1200 with Pistorm32 with 2.2Ghz rpi4:

Good to know that it works with this setup. It also seems that 50fps wouldn't present too much of a challenge for it, either...

B