Tfx Aga?

[jotd]

If someone has a HDF with oxypatcher running the FPU version, I'd like it. I had a lot of crashes when trying to use oxypatcher on a 68040 and that game.

game starts but right before playing it reboots guru 8000000B

if I have a working oxypatcher setup + 040 or 060 I'll be able to get code diffs and statically patch oxypatcher patches into the fpu program so 040/060 users are able to use whdload and fpu version.

In the meantime I'm releasing the version as is. 68040/68060 users will have to use the non FPU version.

[Angus]

Originally Posted by jotd

My understanding was that only 68060 had unimplemented (emulated) FPU instructions. Angus seems to have successfully tested the 020+FPU version (probably 68881 FPU instructions only)

No, it was just the 68020 version - not the 68020+FPU version.

[jotd]

yeah, I just read your messages again. 68060 can't run FPU version for this game without oxypatcher or another fpu patcher. patch is about to be released on whdload page now.

[d4rk3lf]

Originally Posted by Angus

and the most advanced and ambitious sim released for the Amiga.

How about: The Shadow of the third moon?
I admit, it's not really a flight sim, and it's very arcadish, but graphics seems nicer to me then TFX... but I might be wrong, because haven't played any of these, so I can compare.

[ Show youtube player ]

[zipper]

TSottM was quite fun - got played about 3/4 but then I was stuck.

[Angus]

My recollection is the game was quite impressive visually but felt a bit shallow and underdeveloped, and as you say, wasn't a sim - more like a shoot-em-up as I recall. Having said all that, it was spectacular and reflected well on the skills of the coders.

[AMike]

Originally Posted by jotd

So Oxypatcher handles the trap only once, and the next time the program is executed at the same location, the instruction is replaced.

I use on my 060 machines MURedox - does this software do the same?
http://aminet.net/package/util/boot/MuRedox
Thanks for your hard work.

[jotd]

Looks very much like the same principle, yes, except that maybe that one isn't going to crash Will try it!

[Angus]

There's 3 pages of TFX player's Guide here, from page 43.

http://amr.abime.net/issue_657_pages

[skan]

Originally Posted by jotd

Looks very much like the same principle, yes, except that maybe that one isn't going to crash Will try it!

Dunno about TFX, but I experienced many issues (slow, buggy) with demoscene stuff back in the days with that. Oxy never failed to deliver on the other hand. My 2 cents.

[jotd]

the software seems new (2016)... anyway, if it works, I can use that as a model for the patches.

[DaveTaylor]

Originally Posted by Angus

There's 3 pages of TFX player's Guide here, from page 43.

http://amr.abime.net/issue_657_pages

Wow, thats so useful!
All keyboard functions explained.

And there is a comment about running TFX on 040 and 060.
Could be interesting to read for JOTD.

[Steve]

Originally Posted by Angus

There's 3 pages of TFX player's Guide here, from page 43.

http://amr.abime.net/issue_657_pages

Interesting. If you look on page 44 there is a section detailing the very same 68040+FPU issues you've been discussing here.

Bah you beat me to it Dave! lol.

[apex]

The TFX history is really confusing, is there a complete working package?

[Don_Adan]

Originally Posted by jotd

@DonAdan thanks for your input, precious as always!

yeah maybe I'm being unfair with this program. The fact that the devs used a 1988 program on their executables and didn't test the output like AT ALL is another big mistake from them.

I have one original TFX file (DamienD TFX.040 has a lot of sections and even debug symbols which is extremely handy).
You would have to figure out the broken logic and hope it's reversible. One thing I noticed is that on a non-processed exe, there are 3000+ relocs, and with hunkwizard there are 7000+ relocs. Clearly the tool is damn buggy.

Also the fact that it merges data hunks when some are "chip" and some not is clearly a bug. It's ALWAYS been a bug, even on old machines. Why the "chip" flag then?

Maybe disassemble prog with IRA, hunkwizard-trash it and disassemble again, compare to see if there's not a simple way of fixing it. We'll see.



yeah, that's another challenge. The 040 version with symbols gives a lot of info, there are math symbols like "atan" ... that use FPU. I disassembled atan with a python program using capstone and that worked pretty well.

Now I tested the program in 040 mode and all "healthy" executables worked.

My understanding was that only 68060 had unimplemented (emulated) FPU instructions. Angus seems to have successfully tested the 020+FPU version (probably 68881 FPU instructions only)

BTW program looks for AFB_68881 but as the includes read:



it seems that 68040 CPU doesn't set this flag. I forced the test in case of a 68040 else it didn't work. If the emulation code is loaded then the flags are set, but it's worthless in whdload since whdload cannot take math emulation code into account.

I just tested the 020+FPU code without the "unimplemented" feature of WinUAE and it crashes because it uses unimplemented 040 FPU instructions... So I understand now how hard it is to make this program work on whdload or even on a standard config (and why Oxypatcher can help). BUT I think we can plug the 040 versions of the fpu routines over the 020+68881 ones on 020 exes.

And now got to test the 040 version that on a real 060 (that I have), and later the whd-patched 020+FPU versions

Sorry for the long post. This is interesting as this is the first game that I'm trying to fix that uses a FPU (other than the FRESTORE call of lattice/manx startup)

@Others I usually don't like flight sims, but that one seems pretty good & technically optimized. Nothing to do with the countless Microprose ports of F[0-9]*. Tornado AGA is probably a good challenger though.
Play with settings for more colors if you can afford the CPU. The fact that the arcade game starts in-flight means that you don't have to take off, and can enjoy fight right away. Reminds me loosely of F/18. The hardest part is to learn all the keys and all (but I added joypad controls now)

7000+ relocs? I think that Hunk Wizard dont handle correctly debug/symbols hunk, simple you must remove all debug/symbols relocs from final/fixed exe. For merging data hunks, i thinked that Hunk Wizard merging these hunks to chip ram only, for fast memory this is not good, but again no problem in 1988 year

[skan]

Originally Posted by jotd

the software seems new (2016)... anyway, if it works, I can use that as a model for the patches.

It's been updated then, good! Maybe now is useful at least.

[jotd]

7000+ relocs? I think that Hunk Wizard dont handle correctly debug/symbols hunk, simple you must remove all debug/symbols relocs from final/fixed exe. For merging data hunks, i thinked that Hunk Wizard merging these hunks to chip ram only, for fast memory this is not good, but again no problem in 1988 year

too late for that

It's been updated then, good! Maybe now is useful at least.

I'm pretty sure the author is an eab member too always telling everyone to "use the OS and don't bang the hardware".

Even if I disagree with the last advice in my context, I wouldn't be able to code the stuff he coded. Seems like a very knowledgeable expert. I'd trust that tool (and if it doesn't work, I can still yell here )

Interesting. If you look on page 44 there is a section detailing the very same 68040+FPU issues you've been discussing here.

I came to the same conclusion. There is no real 040 version with FPU unless you're using oxypatcher or similar (in which case you have trouble running the game because it's picky).
Personally I get guru with oxypatcher... didn't try without. whdload it works but I need to adapt the patches. That'll be the next move.

[Don_Adan]

Originally Posted by jotd

too late for that



I'm pretty sure the author is an eab member too always telling everyone to "use the OS and don't bang the hardware".

Even if I disagree with the last advice in my context, I wouldn't be able to code the stuff he coded. Seems like a very knowledgeable expert. I'd trust that tool (and if it doesn't work, I can still yell here )



I came to the same conclusion. There is no real 040 version with FPU unless you're using oxypatcher or similar (in which case you have trouble running the game because it's picky).
Personally I get guru with oxypatcher... didn't try without. whdload it works but I need to adapt the patches. That'll be the next move.

If OxyPatcher direct replaced FPU code with own routines called by jmp or jsr then must alloc a part of memory from first 32KB chip area for handling 4 bytes long FPU instructions. Here is some info (initialization program):


"Achim Koyen's (New Generation Software) Oxyron Patcher, more commonly known as OxyPatcher, provides similar functionality to that of CyberPatcher and adds various optimization and configuration options, the ability to monitor its activity, and support for most accelerators. In addition to the main program going in WBStartUp, it is also necessary to run an initialization program from your startup-sequence. When booting, this startup program will install itself in memory and then reset your system. On the test system this added about twenty seconds to the booting process. Like CyberPatcher, illegal instructions are trapped and library emulation routines are replaced with much faster functions. OxyPatcher is commercially available for about US$30."

[Angus]

I have Cyberpatcher installed on my Blizz 1260 system. I have tried replacing it with Oxypatcher but could not get it to work without crashing - which might just be down to me.

Does anybody know for definite if, as Jean-Francois suggested, Cyberpatcher has to "repatch" every time an unsupported instruction occurs, or if it only needs to do this on the first occurence and then "remembers" the adjustment like Oxypatcher?

If Oxypatcher does have this advantage, I will go back and try getting it to work again - I have no more hair left to lose. :-)

[skan]

Originally Posted by jotd

I'm pretty sure the author is an eab member too always telling everyone to "use the OS and don't bang the hardware".

Even if I disagree with the last advice in my context, I wouldn't be able to code the stuff he coded. Seems like a very knowledgeable expert. I'd trust that tool (and if it doesn't work, I can still yell here )

...Yet he thinks it's an outdated toy OS not worth the effort... and yet he keeps on working on it...
Yeah, he has some radical views I often don't agree with, but ok.