DOOM - Unofficial CD32 Release - FastRAM Required!

[NovaCoder]

Originally Posted by SKOLMAN_MWS

c2p test

A4000
68040/25+fast - 18.6 ms

CD32 with SX32Pro
68030/50+fast - 22 ms
68030/50+fast+akiko - 13.3 ms

Strange, maybe that 030 c2p wasn't the fastest implementation?

[wairnair]

Higher ms values are better? (is ms=millisecs?)
Did you turn off cpu caches for the 68030/50+fast+akiko test?

[NovaCoder]

Originally Posted by wairnair

Higher ms values are better? (is ms=millisecs?)
Did you turn off cpu caches for the 68030/50+fast+akiko test?

No higher would be slower.

Read all about it here -> http://eab.abime.net/showthread.php?t=74008

[wairnair]

Originally Posted by NovaCoder

No higher would be slower.

Read all about it here -> http://eab.abime.net/showthread.php?t=74008

Thanks. Then akiko does have a ~40% advantage over 030+fast ram, that's very noticeable during gameplay I think.

@SKOLMAN_MWS can we please have info on how you managed to run Doom with Akiko and 030?

[emiespo]

Hey, I managed to run DoomAttack (CD32) on a CD32 with the TF330 (and firmware v1).

Edit: forgot to add that either you disable the DATA CACHE - it's a *feature* of the 68030 - or you install the 68030 + MMU library (on Aminet) that prevents the data cache from being active while talking to Akiko.

Here are my findings (also posted in the exxoshost forum):

single buffer
akiko2 = 8.161fps
020_optimized = 7.461fps
blitter - crashes

double buffer
akiko2 - 8.06
020_opt - 7.43
blitter - 8.02

Interesting to see how close the Blitter assisted C2P is to Akiko.

WingCommander seems notably faster with the optimized C2P than with Akiko. And this one is still much faster than with the V2 firmware and the fallback (original) C2P.

Unfortunately due to the "simple" design (write chunky - read planar - write planar to the screen buffer), Akiko doesn't really free the CPU, it only prevents doing many logic operations in between reads and writes. This is a minor improvement over one read - many rols/ands/ors - one write.

Anyone with a TF330 and the V2 firmware? It should get one extra FPS with the 020 routine, it would be interesting to see what happens with Akiko, but on my CD32, Akiko disappears with the TF330 v2 firmware.

[NovaCoder]

Originally Posted by emiespo

single buffer
akiko2 = 8.161fps
020_optimized = 7.461fps
blitter - crashes

double buffer
akiko2 - 8.06
020_opt - 7.43
blitter - 8.02

Interesting to see how close the Blitter assisted C2P is to Akiko.

Very interesting, like I said there is not much in it (you would hardly notice the difference).

Can someone see if they can get Wolfenstein 3D running with a TF330 in AGA mode, would be interested to know how playable it is.

Wolfenstein 3D

[wairnair]

Originally Posted by emiespo

install the 68030 + MMU library (on Aminet)

Can you please provide a link?
Is it this: http://aminet.net/package/util/libs/MMULib

Thanks for the test.

[emiespo]

It's this one: http://aminet.net/package/util/sys/Mu680x0Libs

It seems not to support OS3.9 though, different SetPatch command, but apparently the docs have an alternative solution.

[emiespo]

Originally Posted by NovaCoder

Very interesting, like I said there is not much in it (you would hardly notice the difference).

Well in single buffer the display flickers a lot with the 020 C2P. The difference between that and the double buffered version is almost 1fps and that is quite noticeable (not that it makes a huge difference, but...).

I'd be curious to see what happens with the TF330 V2 firmware, as it gives an extra 25% speed boost (according to Sysinfo). Unfortunately Akiko stops working for some (weird) reason.

Can someone see if they can get Wolfenstein 3D running with a TF330 in AGA mode, would be interested to know how playable it is.

Wolfenstein 3D

Tested this one as well. I'd say it's poorly playable at full screen, ~6/7FPS (this is my feeling, is there a way to activate a FPS counter?). Reducing the screen size helps a lot.

Is it using the OS routines or has its own C2P?

[SKOLMAN_MWS]

Originally Posted by emiespo

Interesting to see how close the Blitter assisted C2P is to Akiko.

I would like notice, that the Blitter assisted c2p is of lower quality, the graphics are simplified.

[NovaCoder]

Originally Posted by emiespo

Well in single buffer the display flickers a lot with the 020 C2P. The difference between that and the double buffered version is almost 1fps and that is quite noticeable (not that it makes a huge difference, but...).

I'd be curious to see what happens with the TF330 V2 firmware, as it gives an extra 25% speed boost (according to Sysinfo). Unfortunately Akiko stops working for some (weird) reason.



Tested this one as well. I'd say it's poorly playable at full screen, ~6/7FPS (this is my feeling, is there a way to activate a FPS counter?). Reducing the screen size helps a lot.

Is it using the OS routines or has its own C2P?

Thanks for testing amiwolf, yes it has it's own c2p for fastram 030.

[Wachi]

So ... The conclusion I draw is that the best c2p option for DoomCD32 on a CD32 with TF330 is c2p_020_optimized ....

[wairnair]

Originally Posted by Wachi

So ... The conclusion I draw is that the best c2p option for DoomCD32 on a CD32 with TF330 is c2p_020_optimized ....

akiko still has a slight advantage, doesn't it?

[Wachi]

Yes, but it does not look good just looks good with c2p_020_blitter and c2p_020_optimized?

[Anachronist]

This is awesome! And as luck has it I just won a CD32 assembled Terriblefire today >

[Wachi]

I'm sure you already know him, but if anyone is interested ...

DoomAttack-AIO

http://rhz1.com/miscstuff/doomattack-aio.htm

[emiespo]

Originally Posted by Wachi

Yes, but it does not look good just looks good with c2p_020_blitter and c2p_020_optimized?

Who said that?

Easy recap: Akiko gives ~.5 extra fps *when the CPU is busy enough* so around 7/8fps.

Best results are in single buffer mode, where Akiko works OK, Blitter crashes, and 020 flickers.

All of this with the TF330 V1 firmware. I tried the V2 which gives a nice 25% speed boost, but Akiko disappears from the system (even the OS doesn't detect it anymore).

I'd like to understand if it happens to everyone or it's just me. And also see if the Akiko C2P of DoomAttack is still slightly above the 020_optimized in single buffer mode.

Extra note: if a game is not as CPU intensive as Doom (ie Wing Commander), Akiko still does its job, but on a 030@50Mhz the CPU patched C2P is faster. On a TF328 the opposite is experienced though, Akiko gives some extra fps*.

So I'd say that Akiko is still useful in the range of 0 < fps < 15 :-)

* Edit: I'd also say that maybe the WC C2P takes advantage of the data caches on a 030?

[Wachi]

Yes, it is clear that the TF328 goes better with c2p_akiko, but I was referring to TF330 in what only looks good with c2p_020_blitter and c2p_020_optimized .....

[emiespo]

Originally Posted by Wachi

Yes, it is clear that the TF328 goes better with c2p_akiko, but I was referring to TF330 in what only looks good with c2p_020_blitter and c2p_020_optimized .....

I never said that, though! All my findings were on the TF330. It looks exactly the same with 020* or Blitter or Akiko, and with Akiko DoomAttack is still ~.5fps faster


* actually the 020 C2P flickers terribly in single-buffer mode.

[turrican3]

Hi guys,
could it be possible to make some parts of the 3d calculate by the main cpu and others parts by akiko ???
I'm not a coder don't kick me too hard !