Have a Macintosh-68k? Please run this for me!

[Vol]

I have written my first assembly program for the old good Mac. It is a number pi calculator that can give us any number of digits of the pi. It also outputs the amount of time spent on each calculation.
I am gathering timings for 100, 1000, and 3000 digits for the 68k system (up to the 68040). I have found out that emulators are rather inaccurate with timings and they especially inaccurate for the 68020+ systems. So I need data from real hardware. I need just 3 timings for a system, optional screenshots are welcome too. For the 68020+ systems, would you like also to run the 68020 optimized program (pi20)?
I have tested programs under MacOS 1.1-7.5.5 - they work fine on emulators. It seems I need to send a copy of my request to the Compact Macs thread too. Maybe later, because my primarily interest is the Mac II family.
Thank you.

 

[bakkus]

Had the IIfx running, so I loaded this.
I'm having a bit of an issue with the program filling _more_ than the available screen estate horizontally., so it makes things a bit hard to make out.y
Here are the results for 100, 1000 and 3000 - if you can make them out:

 

[Vol]

Had the IIfx running, so I loaded this.
I'm having a bit of an issue with the program filling _more_ than the available screen estate horizontally., so it makes things a bit hard to make out.y
Here are the results for 100, 1000 and 3000 - if you can make them out:

Thank you very much. It is unexpected for me that the first left digits are slightly cut on your photos. Emulators show the correct picture. It seems real iron has some peculiarities which I missed in my code. I have attached a screenshot from the MAME/MESS emulator of the Mac II - there is no any issue there.
I use DrawString to print digits, so I doubt that it is easy to make any out-file. I use this function because it is the most compatible way to print some text.
BTW What is your MacOS version?

 

[bakkus]

BTW What is your MacOS version?

This was on 7.1.1.

 

[Vol]

The timing is the last number printed. So the results for Pi20 on the IIfx are
100 - 0.5
1000 - 1.62
3000 - 11.80
You can check these numbers with a stopwatch too. It is also interesting to get results for the Pi program that is optimized for the 68000. I am curious how faster is the program for the 68020.

 

[Vol]

This was on 7.1.1.

Thanks. Do you use CPU overclocking and faster RAM?

 

[retr01]

Hey @Vol, does your Pi calculator program take advantage of the FPU? Or is this to see the timings of the CPU?

 

[speakers]

These times for Pi. Pi20 is only very slightly faster.

SE, 7.1 020 accelerator @16MHz:

100 0.12​

1000 4.05​

3000 30.57​

IIci, 7.1, 030 + cache @25Mhz:

100 0.08​

1000 2.53​

3000 19.5​

Q650, 7.6, 040 @33MHz:

100 0.03​

1000 0.93​

3000 7.33​

 

[Vol]

Hey @Vol, does your Pi calculator program take advantage of the FPU? Or is this to see the timings of the CPU?

My code uses only integers.

 

[Vol]

These times for Pi. Pi20 is only very slightly faster.

SE, 7.1 020 accelerator @16MHz:

100 0.12​

1000 4.05​

3000 30.57​

IIci, 7.1, 030 + cache @25Mhz:

100 0.08​

1000 2.53​

3000 19.5​

Q650, 7.6, 040 @33MHz:

100 0.03​

1000 0.93​

3000 7.33​

Thank you very much. Is your accelerator board for the SE the Radius Accelerator SE? The results for Pi20 will also be valuable to me...
Did you have issues with the output to the screen?

 

[speakers]

Thank you very much. Is your accelerator board for the SE the Radius Accelerator SE? The results for Pi20 will also be valuable to me...
Did you have issues with the output to the screen?

The accelerator is a General Computer HyperCharger 020.

For the ci and Quadra, I see as much variability in timings between individual runs than between Pi and Pi20. Also, runtimes vary with different OS versions.

For the SE, 7.1 020 accelerator @16MHz:

​

	Pi	Pi20
100	0.12	0.12
1000	4.00	3.90
3000	31.62	30.37

On the SE and ci under MacOS 7.1, I see "(up to z764)" instead of "(up to 9228)".

 

[Vol]

The accelerator is a General Computer HyperCharger 020.

For the ci and Quadra, I see as much variability in timings between individual runs than between Pi and Pi20. Also, runtimes vary with different OS versions.

For the SE, 7.1 020 accelerator @16MHz:

​

	Pi	Pi20
100	0.12	0.12
1000	4.00	3.90
3000	31.62	30.37

On the SE and ci under MacOS 7.1, I see "(up to z764)" instead of "(up to 9228)".

Thank you very much for your information. It has confirmed other results that the 68020 optimization has no effect for the 68030. This is odd for me, theoretically this optimization should be useful for the 68030 too.
Thank you very much for your bug report. I successfully tested my code with 1.1, 2.0, 3.2, 6.0.3, 7.5.5. It seems a system call to _NewPtr corrupts D4 in 7.1 - so it is rather a bug in this version of MacOS. Do you know where can I take MacOS 7.1 disk image? I need it to fix the issue.

 

[speakers]

... it is rather a bug in this version of MacOS. Do you know where can I take MacOS 7.1 disk image? I need it to fix the issue.

Macintosh Garden is always the first place to go: https://macintoshgarden.org/apps/mac-os-71

 

[Vol]

Macintosh Garden is always the first place to go: https://macintoshgarden.org/apps/mac-os-71

Thank you very much. However it was not an OS issue, it was my miss. Sorry. I only tested the previous version with 7.5.5 and 7.0.1 and this did not allow me to detect my bug. Now the bug is fixed. This bug just corrupted the number printed and didn't affect computations.

 

[Vol]

@speakers The results for the Quadra 650 are fantastic for me. The 68040 shows unbelievable speed that surpasses the 80486!
It would be great if you can get results for your SE without the accelerator card. But thank you anyway.
EDIT. It would also be great to get a screenshot.
BTW I couldn't publish results from @bakkus 's Mac because I didn't know what is the CPU freq.

 

[bakkus]

@speakers The results for the Quadra 650 are fantastic for me. The 68040 shows unbelievable speed that surpasses the 80486!
It would be great if you can get results for your SE without the accelerator card. But thank you anyway.
EDIT. It would also be great to get a screenshot.
BTW I couldn't publish results from @bakkus 's Mac because I didn't know what is the CPU freq.

Sorry, missed the question.
No overclocking, and no special RAM - the IIfx runs a stock 68030 at 40MHz: https://en.wikipedia.org/wiki/Macintosh_IIfx

 

[speakers]

@speakers The results for the Quadra 650 are fantastic for me. The 68040 shows unbelievable speed that surpasses the 80486!
It would be great if you can get results for your SE without the accelerator card. But thank you anyway.

There's no way to software-disable the accelerator so I'd need to pull the board .. which I don't want to do right now. Sorry.

EDIT. It would also be great to get a screenshot.

What screenshot exactly: Q650? 3000?

 

[Vol]

What screenshot exactly: Q650? 3000?

I will be happy with any but as a maximum it's the best to get a shot from every piece of your hardware (SE+020, IIsi, QQ650). For 1000 or 3000. Thank you in advance.

 

[speakers]

I will be happy with any but as a maximum it's the best to get a shot from every piece of your hardware (SE+020, IIsi, QQ650). For 1000 or 3000. Thank you in advance.

Unfortunately, your code doesn't allow the system to get a (Shift+Cmd+3) screenshot at completion. I'll have to find another way.

 

[JDW]

My code uses only integers.

If ever you write a little app (or mod of the existing one) that can take advantage of the 68881 and 68882 FPUs, that would be really amazing! The Levco SuperMac SpeedCard I have now is having 68881 FPU issues. To be honest, I'm not sure if it is working. Neither Speedometer nor Norton System Info recognize its FPU chip, and yet a friend of mine told me his Radius accelerator (also used in a Mac SE) shows his FPU just fine in both apps. So having a nifty little app to test the FPU would probably be a help to any 68K owner who has an accelerator of some sort, or who has a 68K Mac with a built-in FPU.