FrontBench benchmark results

[Badwolf]

Good morning all,

Prompted by Christian I've put together a way to make my 'FrontBench' Frontier: Elite 2 intro frame counting hack from the shareware version of the full game.

Please see https://github.com/dh219/FrontBench for details and download.

The process requires extracting the frontier.bin (really a PRG) file from the shareware release and running my python script against it.

I've started tabulating some results in the readme of that repository. If anyone would like to contribute a figure for their hardware combination, please leave it below recording:-

1) Machine
2) PAL/NTSC/VGA/etc. (the VBL frequency makes a difference!)
3) Any pertinent hardware specifics
4) Your frames figure after the first pass has completed.

Cheers,

BW.

[Badwolf]

Some starting figures:-


Stock 8MHz PAL STE: ~850 frames.

16MHz PAL STE: ~940 frames.
32MHz PAL STE: ~1100 frames [exxos].
16MHz PAL STE + 7MB/s AltRAM: ~1320 frames.

Stock Falcon @60Hz VGA mode: ~1650 frames.
Stock Falcon @50Hz RGB mode: ~2070 frames.

50MHz PAL Mega ST+TF536: ~2440 frames [coonsgm].
50MHz PAL H5[STF]+TF536: ~2700 frames [davec].

50MHz DFB1 Falcon + 33MB/s AltRAM @50Hz RGB mode: ~4160 frames.

[czietz]

Thank you! (Too bad my machines (the interesting at least) are in their respective storage boxes, right now.)

[Badwolf]

Thank you! (Too bad my machines (the interesting at least) are in their respective storage boxes, right now.)

I've got some money on your 48MHz TT as having the winning figures in the 'classic' genre, although the CT60 might be the ultimate winner.

I think Frontier caps the frame rate at 25fps for the PAL version so we shouldn't see more than about 4900 (3m15s).

The NTSC version will be capped at 30fps, if I've understood the code correctly, but only takes 2m40s so the upper limit should be about the same.

BW

[Cyprian]

Good morning all,

Prompted by Christian I've put together a way to make my 'FrontBench' Frontier: Elite 2 intro frame counting hack from the shareware version of the full game.

great.

Can you please elaborate more about that path?
What was changed? Maybe assembly source code?

[Badwolf]

Maybe assembly source code?

I can't publish the disassembled source any more than I can publish the finished binary -- it's under the copyright of a still trading company. I would not have gone to all this trouble if I could just publish a listing!

What the patch does, roughly, is:

• Sets the AltRAM flags and corrects the program size;
• Repurposes some unused text to say 'frames' and 'lastcount';
• Adds a couple of words to store the frame count and last frame count;
• Adds a couple of subroutines to increment the counter and copy the counter to lastcount;
• Adds a subroutine to set up the text rendering routines and draw the counter and lastcount lines;
• Injects a JSR into the intro sequence when the elapsed intro time is between two existing points to increment the counter;
• Injects a JSR when that elapsed time passes that final point to copy counter to lastcount;
• Injects a JSR in the screen flipping routine to call the variable printing subroutine.

Most of the patch is correcting offsets that have changed by the injection of the JSRs.

BW

[viking272]

Good work BW.

Maybe we can ask David Braben if he will allow it to be modified and released as a PRG?!
Afterall he coded the original Atari version in 68000 I believe so may have some affinity with Atari users?

[Badwolf]

Maybe we can ask David Braben if he will allow it to be modified and released as a PRG?!
Afterall he coded the original Atari version in 68000 I believe so may have some affinity with Atari users?

I've never met him but the impression I have is that we might be better off keeping our heads down, to be honest.

If I ever bump into him in person, perhaps over a glass or two or port, I might try to put it in his mind that releasing the source code would be a great gesture.

But I think I know the answer.

BW

[christos]

I wonder if some 3d scenes from demos might be a good alternative. I am thinking of the illusion demo for starters. Maybe it's easier to get hold of the people involved.
Btw I find your videos very interesting. Looking forward to your next one.

[ThorstenOtto]

You might also consider supplying a binary diff using diffy/patchy. That might be easier to use than a python script, and also generates a checksum to ensure that the correct program version is patched.

[Badwolf]

You might also consider supplying a binary diff using diffy/patchy. That might be easier to use than a python script, and also generates a checksum to ensure that the correct program version is patched.

I wasn't familiar with it. Does it produce decent results?

Unfortunately all the binary diffing tools I looked at were pretty ordinary, or otherwise commercial so, since I knew where the insertions were, I thought I'd do it by hand. It wasn't much fun

BW

[czietz]

I've got some money on your 48MHz TT as having the winning figures in the 'classic' genre

Yes, I was indeed thinking about my TT - and I'll provide the number as soon as it's out of storage the next time.

Meanwhile, I have one other interesting machine to contribute: A PAL MegaST with Hypercache accelerator (68000 at 16 MHz + cache): 1505 framecount. Turning off the cache reduces the framecount to 953, which kinda shows why cache is vital for an accelerator.

[ThorstenOtto]

Diffy/patchy are freeware programs from Holger Weets. I currently can't find the original archives, but i can provide the binaries if you need them.

[susher]

I've got some money on your 48MHz TT as having the winning figures in the 'classic' genre

Yes, I was indeed thinking about my TT - and I'll provide the number as soon as it's out of storage the next time.

Meanwhile, I have one other interesting machine to contribute: A PAL MegaST with Hypercache accelerator (68000 at 16 MHz + cache): 1505 framecount. Turning off the cache reduces the framecount to 953, which kinda shows why cache is vital for an accelerator.

My TT gave 2876 frames. (Stock other than Thunder (64MB EDO TTRAM)).

[Badwolf]

Diffy/patchy are freeware programs from Holger Weets. I currently can't find the original archives, but i can provide the binaries if you need them.

Thanks, Thorsten. I'm happy to give it a try if you can provide them. Sounds like they'd have to be package with the patch too, then.

My TT gave 2876 frames. (Stock other than Thunder (64MB EDO TTRAM)).

Thanks, susher. 60Hz VGA, I presume?

Regards,

BW

[ThorstenOtto]

Here they are. I found an older archive, but i think the ttps are a bit newer. And yes, you should provide patchy.ttp along with your patch It's only ~6k

Edit: ah, finally found the newer archive too. It's available at https://www.maedicke.de/atari/download.htm

[czietz]

My TT (at 48 MHz and with EDO RAM) yields 3426 frames. Note that a TT always runs at 60 Hz in all color modes. Rescaling by 60/50 (to make it comparable to a "PAL" result) would give 4111, very close to your 50 MHz Falcon.

[viking272]

Maybe we can ask David Braben if he will allow it to be modified and released as a PRG?!
Afterall he coded the original Atari version in 68000 I believe so may have some affinity with Atari users?

I've never met him but the impression I have is that we might be better off keeping our heads down, to be honest.
BW

Just noticed Frontier Developments has nearly 700 staff and a £90m turnover...somehow I don't think he'll care for Atari's now!

Did anyone submit a CT6x frame count?

[Badwolf]

Hi all.

Further to Thorsten's contribution of Diffy, here's a binary diff that can be used to reconstruct FRNTBNCH.PRG from the shareware FRONTIER.BIN file. It's actually extremely effective and I'm very impressed.

This allows the file to be generated on the ST side without needing a python scripting environment.

Instructions contained within are repeated below.

Regards,

BW.




FrontBench generator via Diffy.

1) Download the shareware version of Frontier from, for example, https://www.frontierastro.co.uk/Files/files.html

2) From the .ST disc image extract FRONTIER.BIN and place it in the same directory as DIFFY.TTP and FRNTBNCH.DFY

3) In the ST (or emulator) drag FRNTBNCH.DFY to DIFFY.TTP.

4) If the patch completes without error, you can now rename FRONTIER.BIN to FRNTBNCH.PRG

[frank.lukas]

Mega ST4 with PAK68/3-030 50Mhz and Fastram : 3830 frames

[Badwolf]

My TT (at 48 MHz and with EDO RAM) yields 3426 frames. Note that a TT always runs at 60 Hz in all color modes. Rescaling by 60/50 (to make it comparable to a "PAL" result) would give 4111, very close to your 50 MHz Falcon.

Thanks, Christian. I'd not considered the 60Hz factor with the TT. I'll do a run on my Falcon in VGA mode and see how we compare.

Mega ST4 with PAK68/3-030 50Mhz and Fastram : 3830 frames

Thanks, Frank. Is that PAL or NTSC?

Regards,

BW

[frank.lukas]

... don't know what PAL or NTSC means. The Mega ST run ST Low at 50Hz and I start frontier.prg from Floppy ...

Mega ST4 with PAK68/3-020 40Mhz and Fastram : 3264 frames

[Badwolf]

... don't know what PAL or NTSC means. The Mega ST run ST Low at 50Hz and I start frontier.prg from Floppy ...

OK, thanks. Frank.

I can't help but noticing your figures are considerably off the normal sequence. Does the PAK do any STRAM acceleration? Do you have a graphics card or any other hardware in there? They're not in keeping with a 030 & 3.7MB/s ST-RAM combination.

Cheers,

BW.

[frank.lukas]

The PAK68/3 have L1 Cache into the CPU and 32kB L2 Cache to speed up the transfer with the ST Ram ...

PAK030 at 40Mhz because I have only 35ns Tag Ram. For 50Mhz I need 20ns ones. Basis is a 50Mhz PAK without L2 Cache. See the RAM ACCESS ...

L2CACHE.jpg

[DarkLord]

I just set this up here and ran it on 2 machines:

Atari Mega STe (stock) - 1227 frames

Atari STacy (Pak 68/3 030 @ 40mhz, no FASTRAM) - 3163 frames

So I'd say Franks' number are probably correct.

BTW, watching the difference in Frontier between the Mega STe
and the Pak 68/3 equipped STacy is like night and day...