Betatesting SoftIEEE FPU emulator

[Thomas Richter]

Originally Posted by Marce

even Toni states u that u are not listened to him, when I replied that I was using winuae to test

I wasn't aware that the problem is only related to the 68060 emulation. But well, bugs happen, on Toni's side, and on mine. That's how life works. But unlike you, I'm not unfriendly, I reported a detailed report on how the problem could be reproduced, and Toni fixed it. That's how things work out in software development, you know?

Originally Posted by Marce

you tested simple mail yesterday ? the 1st program you have to be tested is simple mail

Good for you, but still, I'm not using it, I'm not reading mail on my Amiga. I'm happy to make it work, but I see nothing that stops it from working at this moment. But then, my setup is different from yours, so that is ok, but without knowing the differences, I need help, and that's pretty much what this thread is about, right?

Originally Posted by Marce

why ? because is the best Amiga mailer and is the only useful program which requires FPU

What's "useful" is pretty much in the eye of the beholder. Frankly, I would not consider the Amiga useful these days anymore, but I understand why people still play with it and I'm typically happy to help - until I run into trolls.

Originally Posted by Marce

so if you didn't know this is , because you are newbie and not ready to make this program

Just because I'm not testing with your favourite program first? Nice. Maybe you'd need to change your perspective a bit. The world isn't rotating around you and your preferences. Other people have other preferences.


Well, I'm surely a newbie. I'm only writing programs for Amiga since 1989. (-:


And now, MuForce output please, with "DISPC" option, and with SegTracker running. That would be helpful. It helped immediately with TVPaint. The more data you deliver, the less I have to guess.

[CodyJarrett]

Originally Posted by Marce

you are the one too lazy you were and the first one who responds disinterestedly
even Toni states u that u are not listened to him, when I replied that I was using winuae to test

you tested simple mail yesterday ? the 1st program you have to be tested is simple mail

why ? because is the best Amiga mailer and is the only useful program which requires FPU

so if you didn't know this is , because you are newbie and not ready to make this program

Marce - please stop trolling and insulting people here.

[Flash951]

Originally Posted by Marce

you are the one too lazy you were and the first one who responds disinterestedly
even Toni states u that u are not listened to him, when I replied that I was using winuae to test

you tested simple mail yesterday ? the 1st program you have to be tested is simple mail

why ? because is the best Amiga mailer and is the only useful program which requires FPU

so if you didn't know this is , because you are newbie and not ready to make this program

You're an 100% assh... I don't understand why he even waste his time answering your bullsh. Get real!

[stevelord]

Originally Posted by TCD

Marce has trolled other threads before. Seems like not enough people are reporting him (or it's now okay to troll here...).

I've reported him. Hopefully the mods will get round to it at some point.



Thomas, I don't need this tool but I know others will. Thankyou for writing it, and for everything you do. You make this hobby a better place.

[indigolemon]

Marce is, frankly, an idiot - and the sooner they get the boot from EAB the better. They seem to be utterly clueless about both Amiga hardware and software dev, and lash out at people that clearly have more knowledge than them.

I suggest simply ignoring them until the mods do something about it.

[Thomas Richter]

Thanks to Peter (again), here is another bugfix. This time, it is on an interpretation of a very bizarre instruction, namely "ftst.x fp2,fp0". ftst.x only tests the source register (here fp2) and ignores the testination. Typically, FPU instructions work such that if only one register is involved, source and destination are identical, and assemblers and compilers generate them this way. Apparently, this did not hold here, and unfortunately, my code grabbed the destination (meaningless) rather than the source.

So here is a fixed version.

[Thomas Richter]

...and here we have the first release candidate, the 40.2. The FPU emulation core did not change, though there was one change in the emulation core. This version replaces the algorithm for computing exp,log,2^x and log2 by a more efficient version that should approximately be twice as fast.

Again, please check and if there are no defects found, this will become the release.

[alenppc]

I was hoping to help you with the debug, as both of NovaCoder's Quake and Quake2 FPU ports crash with the SoftIEEE running, but I do not get any output from MuForce when the crash happens. I did set a log file and a console output but I get nothing. What am I doing wrong?

[NovaCoder]

Originally Posted by alenppc

I was hoping to help you with the debug, as both of NovaCoder's Quake and Quake2 FPU ports crash with the SoftIEEE running, but I do not get any output from MuForce when the crash happens. I did set a log file and a console output but I get nothing. What am I doing wrong?

Are you actually using an 060 (sans FPU)? My 060 ports are actually compiled for an 060 so you need an 060.

I see lots of people setup WinUAE to use an 020 + 68882 to try and run 060 ports, no idea why they don't just configure WinUAE and AmigaOS properly

[alenppc]

Originally Posted by NovaCoder

Are you actually using an 060 (sans FPU)? My 060 ports are actually compiled for an 060 so you need an 060.

I see lots of people setup WinUAE to use an 020 + 68882 to try and run 060 ports, no idea why they don't just configure WinUAE and AmigaOS properly

Yes, LC060 is all the rage these days since we can't get full 060s anymore. If you remember you also did a -nofpu compile of Quake, and it actually runs really well at 100 Mhz (it's mostly playable), but that's not the point here. I just figured that with SoftIeee the FPU version should run in a similar fashion to the -nofpu compile but at the moment it crashes, so I wanted to figure out why, but I can't figure out how to set up MuForce to report the crash details.

[Thomas Richter]

Originally Posted by alenppc

I was hoping to help you with the debug, as both of NovaCoder's Quake and Quake2 FPU ports crash with the SoftIEEE running, but I do not get any output from MuForce when the crash happens. I did set a log file and a console output but I get nothing. What am I doing wrong?

Your best option is of course a null-modem cable on the serial port and another system to read the output. Failing that, you could redirect the output to a RAD: and try to read it there after a reboot. Edit: If you test under WinUAE, please do not use its 68LC060 emulation as this has currently a known defect where upon entry of the line F exception vector, one register is trashed. You may try the 68LC040.

[Thomas Richter]

Originally Posted by NovaCoder

Are you actually using an 060 (sans FPU)? My 060 ports are actually compiled for an 060 so you need an 060.

It is here really only providing a test case for the SoftIEEE FPU emulation, and not so much as a realistic use case. It's all understood that the result will not be playable because the FPU emulation will be much to slow. Yet, it should not crash by any means.

[alexh]

Some of the XC68LC060 branded CPUs I've had are actually full 060's with FPUs that did not pass factory testing.

AFAIK there is no way to tell them apart from full 060's (except the FPU doesn't work).

Can you use this software with these CPUs?

[Thomas Richter]

The 68060 PCR register tells if there is an FPU on board or not, and if it is not, the 68060.library does not attempt to enable the FPU, or load the FPU support package. If that identification does not work, the MMU-Configuration file (for the MMULib based libraries) allows you to disable the FPU ("FPU off" as command).

SoftIEEE can be used whenever there is no FPU detected by the operating system, thus depending on the line F exception vector. Luckily, the 68LC060 (and 68060) reacts as if there is no FPU present if the FPU is disabled in the PCR register, and thus then can depend on emulation.

So, long story short: Yes.

[Reynolds]

Okay, now a theoretic question.
Could that happen, if SoftIEEE will cover all FPU functions, to redirect it using a PowerPC FPU in a BlizzardPPC-kind of card for example or in Ragnarok? I know it is something different, "rewiring" instead of emulating, but...

Do I ask something stupid?

[Thomas Richter]

Originally Posted by Reynolds

Okay, now a theoretic question.
Could that happen, if SoftIEEE will cover all FPU functions, to redirect it using a PowerPC FPU in a BlizzardPPC-kind of card for example or in Ragnarok? I know it is something different, "rewiring" instead of emulating, but...

Do I ask something stupid?

No, that's not stupid. Please remember that SoftIEEE (the executable) redirects all but trivial (e.g. fmove, fbcc) instructions to the softieee.library to perform the mathematical operations. Thus, if you want to to that, you would need to replace the softieee.library by a library that forwards the requested computations to an external FPU.


As I'm a fan of open documentation and open specification (but not open source), you'll find all the information needed to implement an alternative softieee.library in the above archive. Essentially, you need to implement all the functions of the library (more or less one per FPU instruction) and provide a sufficiently large library base as the SoftIEEE binary will want to place of its internal state information there.


Hence, that's all well possible, the architecture was designed to allow that, just somebody needs to pick it up.

[Reynolds]

Originally Posted by Thomas Richter

No, that's not stupid. Please remember that SoftIEEE (the executable) redirects all but trivial (e.g. fmove, fbcc) instructions to the softieee.library to perform the mathematical operations. Thus, if you want to to that, you would need to replace the softieee.library by a library that forwards the requested computations to an external FPU.


As I'm a fan of open documentation and open specification (but not open source), you'll find all the information needed to implement an alternative softieee.library in the above archive. Essentially, you need to implement all the functions of the library (more or less one per FPU instruction) and provide a sufficiently large library base as the SoftIEEE binary will want to place of its internal state information there.


Hence, that's all well possible, the architecture was designed to allow that, just somebody needs to pick it up.

It won't be me sadly who will bring this solution, as I know nothing about coding, so all what fellow Amigans like you and others do is truly unbelievable and I fully respect all coded lines

Hope someone will see the potential to prove if it can be done in reality. Just have been thinking on it hipothetically. Glad to know that my theory can work.

[alenppc]

Originally Posted by Thomas Richter

Your best option is of course a null-modem cable on the serial port and another system to read the output. Failing that, you could redirect the output to a RAD: and try to read it there after a reboot. Edit: If you test under WinUAE, please do not use its 68LC060 emulation as this has currently a known defect where upon entry of the line F exception vector, one register is trashed. You may try the 68LC040.

Thank you for the tip! I don't use WinUAE, if I did, I wouldn't care about this project.

This is the output I gathered, is this helpful?

Exception !!   0000000B     TCB: 08380B88     CTX: 080477D8     SSP: 08002310
USP : 083E2510 SR: 0004  (U0)(-)(-)  TCB: 08380B88
  PC: 082E56B4
Data: 08448C9C 00000073 00000000 3F800000 00000001 00000001 082E9C0C 082E9BF8
Addr: 082DBCEC 08448C9C 0829DA34 0829DABC 0829C10C 083E259C 0828B528 08002310
Stck: 00000580 00000000 0829DA34 0829DABC 0829C10C 40050000 8A21BCAE A747D000
Stck: 083F12CC 083E257C 0000000B 083F11E4 083F11E4 083E257C 00000580 00000000
082e5694 :  4e55 ff94                  link.w a5,#-$6c
082e5698 :  f227 e004                  fmovem.x fp2,-(a7) ;extended opcode
082e569c :  48e7 3038                  movem.l d2-d3/a2-a4,-(a7)
082e56a0 :  4282                       clr.l d2
082e56a2 :  263c 3f80 0000             move.l #$3f800000,d3
082e56a8 :  2b43 ffdc                  move.l d3,-$24(a5)
082e56ac :  2b42 ffe0                  move.l d2,-$20(a5)
082e56b0 :  2b42 ffe4                  move.l d2,-$1c(a5)
082e56b4 : *f202 455a                  fsneg.s d2,fp2 ;extended opcode
082e56b8 :  f22d 6500 ffe8             fmove.s fp2,-$18(a5) ;extended opcode
082e56be :  2b43 ffec                  move.l d3,-$14(a5)
082e56c2 :  2b42 fff0                  move.l d2,-$10(a5)
082e56c6 :  2b42 fff4                  move.l d2,-$c(a5)
082e56ca :  2b42 fff8                  move.l d2,-$8(a5)
082e56ce :  2b43 fffc                  move.l d3,-$4(a5)
082e56d2 :  2b43 ffb8                  move.l d3,-$48(a5)
Name: "AmiQuake.FPU"  Hunk 0000 Offset 00050B34

 
Exception !!   0000000B     TCB: 08380B88     CTX: 080477D8     SSP: 08002308
USP : 083E2510 SR: 0004  (U0)(-)(-)  TCB: 08380B88
  PC: 082E56B4
Data: 08448C9C 00000073 00000000 3F800000 00000001 00000001 082E9C0C 082E9BF8
Addr: 082DBCEC 08448C9C 0829DA34 0829DABC 083E2500 083E259C 0828B528 08002308
Stck: 00000580 00000000 0829DA34 0829DABC 0829C10C 40050000 8A21BCAE A747D000
Stck: 083F12CC 083E257C 0000000B 083F11E4 083F11E4 083E257C 00000580 00000000
082e5694 :  4e55 ff94                  link.w a5,#-$6c
082e5698 :  f227 e004                  fmovem.x fp2,-(a7) ;extended opcode
082e569c :  48e7 3038                  movem.l d2-d3/a2-a4,-(a7)
082e56a0 :  4282                       clr.l d2
082e56a2 :  263c 3f80 0000             move.l #$3f800000,d3
082e56a8 :  2b43 ffdc                  move.l d3,-$24(a5)
082e56ac :  2b42 ffe0                  move.l d2,-$20(a5)
082e56b0 :  2b42 ffe4                  move.l d2,-$1c(a5)
082e56b4 : *f202 455a                  fsneg.s d2,fp2 ;extended opcode
082e56b8 :  f22d 6500 ffe8             fmove.s fp2,-$18(a5) ;extended opcode
082e56be :  2b43 ffec                  move.l d3,-$14(a5)
082e56c2 :  2b42 fff0                  move.l d2,-$10(a5)
082e56c6 :  2b42 fff4                  move.l d2,-$c(a5)
082e56ca :  2b42 fff8                  move.l d2,-$8(a5)
082e56ce :  2b43 fffc                  move.l d3,-$4(a5)
082e56d2 :  2b43 ffb8                  move.l d3,-$48(a5)
Name: "AmiQuake.FPU"  Hunk 0000 Offset 00050B34

 
LONG READ from 3FB99999                        PC: 0948FD48
USP : 0AD0E99C SR: 0010  (U0)(-)(-)  TCB: 0ACC5638
Data: 3FEFFFFC C0000000 0000009F 00000053 00000053 000002FD 00000010 00000000
Addr: 00F8781C 0A4D5882 09402374 0941DED6 094A2CF8 0AD0E9C4 0ACC41A0 08002308
Stck: 09402374 0941DED6 094A2CF8 0ACC41A0 00000000 00000000 00000000 7FFF0000
Stck: FFFFFFFF FFFFFFFF 0AD0EA00 0949746E FFFFFFFF FFFFFFFF FFFFFFBB FFFFFF89
0948fd28 :  0440 f21a                  subi.w #-$de6,d0
0948fd2c :  6400 f23c                  bcc $948ef6a
0948fd30 :  55e6                       scs.b -(a6)
0948fd32 :  3fb9 9999 a000 0000        move.w -$66666000,$0(a7,d0.w)
0948fd3a :  b5fc 0a4f c16c             cmpa.l #$a4fc16c,a2
0948fd40 :  6fa2                       ble.s $948fce4
0948fd42 :  4ced 5c00 ffd8             movem.l -$28(a5),a2-a4/a6
0948fd48 : *f22d d030 ffe8             fmovem.x -$18(a5),fp2-fp3 ;extended opcode
0948fd4e :  4e5d                       unlk a5
0948fd50 :  4e75                       rts
0948fd52 :  4e55 ff94                  link.w a5,#-$6c
0948fd56 :  f227 e004                  fmovem.x fp2,-(a7) ;extended opcode
0948fd5a :  48e7 3038                  movem.l d2-d3/a2-a4,-(a7)
0948fd5e :  4282                       clr.l d2
0948fd60 :  263c 3f80 0000             move.l #$3f800000,d3
0948fd66 :  2b43 ffdc                  move.l d3,-$24(a5)
Name: "AmiQuake2"  Hunk 0000 Offset 0008D700

 
Exception !!   0000000B     TCB: 0ACC5638     CTX: 080477D8     SSP: 08002310
USP : 0AD0E938 SR: 0014  (U0)(-)(-)  TCB: 0ACC5638
  PC: 0948FD72
Data: 3FEFFFFC C0000000 00000000 3F800000 00000053 000002FD 00000010 00000000
Addr: 00F8781C 0A4D5882 09402374 0941DED6 094A2CF8 0AD0E9C4 0ACC41A0 08002310
Stck: 0000009F 00000053 09402374 0941DED6 094A2CF8 00000000 00000000 00000000
Stck: 00000053 00000053 000002FD 00000010 00000000 00F8781C 0A4D5882 0A4FB040
0948fd52 :  4e55 ff94                  link.w a5,#-$6c
0948fd56 :  f227 e004                  fmovem.x fp2,-(a7) ;extended opcode
0948fd5a :  48e7 3038                  movem.l d2-d3/a2-a4,-(a7)
0948fd5e :  4282                       clr.l d2
0948fd60 :  263c 3f80 0000             move.l #$3f800000,d3
0948fd66 :  2b43 ffdc                  move.l d3,-$24(a5)
0948fd6a :  2b42 ffe0                  move.l d2,-$20(a5)
0948fd6e :  2b42 ffe4                  move.l d2,-$1c(a5)
0948fd72 : *f202 455a                  fsneg.s d2,fp2 ;extended opcode
0948fd76 :  f22d 6500 ffe8             fmove.s fp2,-$18(a5) ;extended opcode
0948fd7c :  2b43 ffec                  move.l d3,-$14(a5)
0948fd80 :  2b42 fff0                  move.l d2,-$10(a5)
0948fd84 :  2b42 fff4                  move.l d2,-$c(a5)
0948fd88 :  2b42 fff8                  move.l d2,-$8(a5)
0948fd8c :  2b43 fffc                  move.l d3,-$4(a5)
0948fd90 :  2b43 ffb8                  move.l d3,-$48(a5)
Name: "AmiQuake2"  Hunk 0000 Offset 0008D72A

 
Exception !!   0000000B     TCB: 0ACC5638     CTX: 080477D8     SSP: 08002308
USP : 0AD0E938 SR: 0014  (U0)(-)(-)  TCB: 0ACC5638
  PC: 0948FD72
Data: 3FEFFFFC C0000000 00000000 3F800000 00000053 000002FD 00000010 00000000
Addr: 00F8781C 0A4D5882 09402374 0941DED6 0AD0E928 0AD0E9C4 0ACC41A0 08002308
Stck: 0000009F 00000053 09402374 0941DED6 094A2CF8 00000000 00000000 00000000
Stck: 00000053 00000053 000002FD 00000010 00000000 00F8781C 0A4D5882 0A4FB040
0948fd52 :  4e55 ff94                  link.w a5,#-$6c
0948fd56 :  f227 e004                  fmovem.x fp2,-(a7) ;extended opcode
0948fd5a :  48e7 3038                  movem.l d2-d3/a2-a4,-(a7)
0948fd5e :  4282                       clr.l d2
0948fd60 :  263c 3f80 0000             move.l #$3f800000,d3
0948fd66 :  2b43 ffdc                  move.l d3,-$24(a5)
0948fd6a :  2b42 ffe0                  move.l d2,-$20(a5)
0948fd6e :  2b42 ffe4                  move.l d2,-$1c(a5)
0948fd72 : *f202 455a                  fsneg.s d2,fp2 ;extended opcode
0948fd76 :  f22d 6500 ffe8             fmove.s fp2,-$18(a5) ;extended opcode
0948fd7c :  2b43 ffec                  move.l d3,-$14(a5)
0948fd80 :  2b42 fff0                  move.l d2,-$10(a5)
0948fd84 :  2b42 fff4                  move.l d2,-$c(a5)
0948fd88 :  2b42 fff8                  move.l d2,-$8(a5)
0948fd8c :  2b43 fffc                  move.l d3,-$4(a5)
0948fd90 :  2b43 ffb8                  move.l d3,-$48(a5)
Name: "AmiQuake2"  Hunk 0000 Offset 0008D72A

[Thomas Richter]

Yes, excellent! This is very helpful, and tells - stupid me - that I forgot another instruction.

[alenppc]

Originally Posted by alexh

Some of the XC68LC060 branded CPUs I've had are actually full 060's with FPUs that did not pass factory testing.

AFAIK there is no way to tell them apart from full 060's (except the FPU doesn't work).

This is a bit off topic, but I just wanted to add that I have seen a LOT of 68060 cpus with various prints and designations and everything points out to this "factory testing" to be nothing but a myth. I am yet to find a rev1 EC marked CPU that does not run everything perfectly.



The "factory testing" seems to have been an excuse for Motorola to say "we don't have an FPUless die yet, so don't go buying these cheapo ones because they won't work, we promise, go buy the expensive ones instead". Looks like even Phase5 knew it back in the 90s since all of my 060 accelerators came with EC marked rev1 060s, yet I never had an issue.


In fact I would be _very_ curious to find an 060 with a present but "defective" fpu.