SysInfo 4.3 ready, finally

[Paperboy]

Hi,
I'm trying to get Sysinfo 4.3 working with a LC CPU but after several hours I can't get any further.

The LC CPU is supposed to run with Thor MMU Library but I didn't manage it.

As soon as I click on Speed Guru

[Paperboy]

Hello
Did anyone make this work with a LC CPU and speed test ?

[Toni Wilen]

It can't work. It tries to execute non-existing FPU instructions when CPU is 68LC060. (=68060 without FPU)

btw, it executes code like this:

...
0804E306 f200 0328                FSUB.X FP0,FP6
0804E30A f280 0000                FNOP
0804E30E f200 0080                FMOVE.X FP0,FP1
0804E312 f280 0000                FNOP
0804E316 f200 0080                FMOVE.X FP0,FP1
0804E31A f280 0000                FNOP
0804E31E f200 0080                FMOVE.X FP0,FP1
0804E322 f280 0000                FNOP
0804E326 f200 0080                FMOVE.X FP0,FP1
0804E32A f280 0000                FNOP
0804E32E f200 0080                FMOVE.X FP0,FP1
0804E332 f280 0000                FNOP
0804E336 f200 0080                FMOVE.X FP0,FP1
...

No one writes that kind of totally pointless and time wasting code in real world. Even 68882 should be able to partially execute 2 FPU instructions simultaneously but FNOP prevents it. Sysinfo CPU/FPU speed tests never made any sense.

[CmdrAmiga]

Originally Posted by Toni Wilen

Sysinfo CPU/FPU speed tests never made any sense.

I couldn't agree more! Toni, I know it's probably the last thing you'd like to touch but could you please help making this piece of software more useful? Too many people use it as a reference. It just begs for mercy . It's speed test module needs a serious update. And finally a fix to this nonsense where an LC CPU, whereas detected as a non FPU unit, is being tested as such(!)

[quantum8]

Originally Posted by Paperboy

Hi,
I'm trying to get Sysinfo 4.3 working with a LC CPU but after several hours I can't get any further.

The LC CPU is supposed to run with Thor MMU Library but I didn't manage it.

As soon as I click on Speed Guru

This might only be a workaround, but have you tried the LC libraries + an FPU emulator?


https://eab.abime.net/showthread.php?t=101115


Not sure where to find an FPU emulator though, the search result escapes me.

[SpeedGeek]

Originally Posted by quantum8

This might only be a workaround, but have you tried the LC libraries + an FPU emulator?


https://eab.abime.net/showthread.php?t=101115


Not sure where to find an FPU emulator though, the search result escapes me.

Sorry, but this library won't work either. This is a bug in SysInfo and is better fixed in the source code. Also, you might want to read the FPU emulation comments on the above thread.

[Paperboy]

Hello,
The problem is that more and more Amiga users are buying a LC CPU because nobody wants to buy a 68060 at the super prices that are currently being offered.

It is very difficult to change the code of this software.
It who also cool if the software also recognizes the type of CPU in this case " LC " EC " Full so 68060 " but I think all this is too much for such an old software.
Thanks for the work anyway.

[Paperboy]

But you also don't know if an Amiga operating system with a not 68060 runs clean.

I have already done some tests with the LC library linked above and it doesn't work with my operating system, it is an OS 3.1.4 .
With OS 3.1 there are only crashes .

With OS 3.9 it runs but there are also crashes .

If a USER has a clean running LC CPU who is nice when he writes Which OS and which libraries he uses .

[SpeedGeek]

Originally Posted by Paperboy

But you also don't know if an Amiga operating system with a not 68060 runs clean.

I have already done some tests with the LC library linked above and it doesn't work with my operating system, it is an OS 3.1.4 .
With OS 3.1 there are only crashes .

With OS 3.9 it runs but there are also crashes .

If a USER has a clean running LC CPU who is nice when he writes Which OS and which libraries he uses .

Please post your results on the thread linked above, and they will be addressed as a possible bug report. This library was NOT intended for use with OS 3.1.4. However, I will attempt to resolve any bugs with OS 3.1 or OS 3.9.

Also, please specify which version of the library you were using.

If your system crashes with a Software Failure Error: 8000 000B it's likely that you still have some FPU dependent libraries or you attempted to use FPU dependent Apps...

[abbub]

I notice that on my Amiga 2000, it's identifying my 1 MB ECS Agnus (NTSC 8372A) as a 2 MB Agnus. I wish it was.

[Geijer]

Please try SysInfo 4.4b4 to see if it works better on your LC/EC CPUs, thanks.

[Geijer]

Originally Posted by abbub

I notice that on my Amiga 2000, it's identifying my 1 MB ECS Agnus (NTSC 8372A) as a 2 MB Agnus. I wish it was.

Strange, unfortunately I don't have the hardware to reproduce the error

I've also tried several combinations in both fs-uae and WinUAE but still no error.

[Geijer]

SysInfo 4.4b5 now includes both EC/LC fixes and better speed comparison number formatting (removed decimals if to big numbers).

[Geijer]

At last a reasonable bug free version so I created a Release Candidate for you to test:
SysInfo 4.4rc1

Fixed in this version:

• Changed handling of speed numbers, if big, don't print decimals
• Updated "Chip Speed vs A600" to handle much faster Chip Speed machines.
• Added support for AC68080 frequenc support
• Update will no longer try to open 68040/68060.library when there is no such CPU
• Bugfix: 68040/68060 non FPU guru fixed, again!
• Lots of updates/corrections in the SysInfo.guide documentation.
• Lots of code cleanup

Please test and report bugs, thanks.

[SnkBitten]

Originally Posted by Geijer

Changelog:

4.3:

• Bootable disk image (adf) distribution
• Identifies APOLLO-Core 68080 CPU/FPU
• Better hanfling of 68060 CPU/FPU identification
• Bugfix: 68040 non FPU guru fixed
• Experimental adf release http://sysinfo.d0.se
• Documentation in AmigaGuide format and experimental html

4.2

• Intermediate release with better utilization of pipelining in 68040 and newer CPUs in "Dhrystone" calculation implementation.

SysInfo_4.3.lha or test the new adf version SysInfo_4.3.adf

Looking forward for feedback.

Hi,

Is there a way to implement Amithlon detection instead of just saying "Virtual Amiga" in the comments? I believe this was done for the Vampire.

Maybe if amithlon device is present or amithlon resource or amithlon_x86cb task as these should only be present on Amithlon systems. These were listed uner the SYSTEM SOFTWARE INSTALLED (Devices, Resources and Tasks).

Not a big deal was just curious if possible without too much trouble.

[Geijer]

Originally Posted by SnkBitten

Hi,

Is there a way to implement Amithlon detection instead of just saying "Virtual Amiga" in the comments? I believe this was done for the Vampire.

Maybe if amithlon device is present or amithlon resource or amithlon_x86cb task as these should only be present on Amithlon systems. These were listed uner the SYSTEM SOFTWARE INSTALLED (Devices, Resources and Tasks).

Not a big deal was just curious if possible without too much trouble.

Attachment 69645

It's possible but I think that there is some other stuff to fix first.

Do the Amithlon emulation use JIT?

I've fixed the Chip Speed vs A600 calculation in RC2, can you please test to se if you get relevant value?

Thanks.

[CmdrAmiga]

Results from 4.4rc1:

Warp 1260 100 MHz (why Virtual Amiga? comment?)



------------------------------
Warp 1260 105 MHz



------------------------------
Apollo 1260 66 MHz with MC68LC060 CPU and 68060.library ver. 46.4 (30.05.2018)



------------------------------
Apollo 1260 66 MHz with MC68LC060 CPU and 680LC60.library ver. 40.1



------------------------------
Apollo 1260 66 MHz with MC68LC060 CPU and 680LC60.library ver. 40.2


------------------------------
WhichAmiga Apollo 1260



------------------------------
I'm disappointed with the dedicated LC library. As can be seen the MIPS result is significantly lower than on a regular library. But maybe there is something that can fix this(?).

Edit: Updated with the newer version of 68LC060.library and the result is the same like with a regular 68060.library.

Other than that, if I click EXPAND the red bar still shows sign + instead of showing actually the full scale (so we see in both states sign + which doesn't make much sense combined with EXPAND option).

[Geijer]

Originally Posted by CmdrAmiga

Results from 4.4rc1:

Warp 1260 100 MHz (why Virtual Amiga? comment?)

...

------------------------------
I'm disappointed with the dedicated LC library. As can be seen the MIPS result is significantly lower than on a regular library. But maybe there is something that can fix this(?).

Other than that, if I click EXPAND the red bar still shows sign + instead of showing actually the full scale (so we see in both states sign + which doesn't make much sense combined with EXPAND option).

The short answere is that I have never got any feedback from anyone with a 060 faster than 80MHz. So thank you very much! There is no, and (in my opinion) shouldn't be a universal way to see if your program is running on real hardware or in an emulator. The emulator should be that good

Shrink/Expand toggles between two predefined values, the ruler above the graphs is a picture with fixed values. It is to much work to change that in 4.4 but I will put it on the growing todo list.

If it's ok with you I would appreciate if you can help me with tests on you hardware in the future?

Thanks

[SnkBitten]

Originally Posted by Geijer

It's possible but I think that there is some other stuff to fix first.

Do the Amithlon emulation use JIT?

I've fixed the Chip Speed vs A600 calculation in RC2, can you please test to se if you get relevant value?

Thanks.

Yes, Amithlon uses JIT in it's emulation.

[Geijer]

Originally Posted by SnkBitten

Yes, Amithlon uses JIT in it's emulation.
Attachment 69655

Ok, then the Chip speed value is totally irrelevant, hmm, need to work something out...