Have a Power Macintosh, please run it for me!

Vol

New Tinkerer
Jul 28, 2022
35
4
8
I am gathering results for the speed of the pi-spigot algorithm for different platforms. Now it is time for the Macintosh computers based on the PowerPC and running MacOS. It would be great to get results from the PPC601 and PPC603 but other models are welcome too. Three programs are attached:
pi-ppc
pi-68k
pi-pwr

The first is the main one, it uses code for the PowerPC. The second one uses the 68k code so it will use the built-in emulator. The last one uses code for the IBM Power. It is only interesting to run it on the PPC601 because the other PPC don't support the POWER instructions used in the code. I gather results for timings of calculations of 100, 1000, and 3000 digits, those timings are printed. It is also an option to get results for 9000 digits just for fun.
I only need results for machines below 400 MHz. Many thanks in advance.
Sources are on - https://github.com/litwr2/rosetta-pi-spigot/tree/master/mac
 

Attachments

  • pi-powermac.zip
    5.8 KB · Views: 131
Nov 4, 2021
126
98
28
Tucson, AZ
i finally got an ADB Wombat so I can type on my 7100/66 running System 8.5. PPC 601 @ 66MHz

pi-86k:
100: 0.07
1000: 1.85
3000: 13.12

pi-ppc:
100: 0.03
1000: 0.53
3000: 4.10

pi-pwr:
100: 0.02
1000: 0.48
3000: 3.80
 
  • Like
Reactions: Vol

Vol

New Tinkerer
Jul 28, 2022
35
4
8
i finally got an ADB Wombat so I can type on my 7100/66 running System 8.5. PPC 601 @ 66MHz

pi-86k:
100: 0.07
1000: 1.85
3000: 13.12

pi-ppc:
100: 0.03
1000: 0.53
3000: 4.10

pi-pwr:
100: 0.02
1000: 0.48
3000: 3.80
Thank you very much! However we have got a mystery. Results from another system show much lower performance.
How might this be possible? I have three ideas about this:
1) your system uses the L2 cache while other system doesn't use it;
2) OS 8.5 has a much faster routine to output characters than OS 7.6.1;
3) the other system is actually 60 MHz, not 66 (it is rather unlikely).
IMHO the L2 cache may be used to cache the system routine to output characters and this makes the difference. So the presence of the L2 cache on your system is crucial to solve the mystery. Does your system use the L2 cache? Do you have other options about the performance difference?
 
Nov 4, 2021
126
98
28
Tucson, AZ
I'll check for a cache simm tomorrow, but if draw speed is a factor it could be the video card. I'm using the DB15 on a card not the on-board video because I don't have the adapter cable for it. System Profiler lists it as an integrated vram card, where I presume the on board would be sharing the system ram bandwidth.
 
  • Like
Reactions: Vol

Vol

New Tinkerer
Jul 28, 2022
35
4
8
I'll check for a cache simm tomorrow, but if draw speed is a factor it could be the video card. I'm using the DB15 on a card not the on-board video because I don't have the adapter cable for it. System Profiler lists it as an integrated vram card, where I presume the on board would be sharing the system ram bandwidth.
One more reason for the difference may be the RAM speed. Systems may use 60-80 ns RAM chips.
 
Nov 4, 2021
126
98
28
Tucson, AZ
Same system, 640x480 256 colors, but booted from the OpenRetroSCSI 7.5.3 image. There does appear to be a bit of performance improvement on 8.5. The other guy was running at higher resolution which probably slowed the scrolling down slightly too.

pi-86k:
100: 0.07
1000: 1.88
3000: 13.17

pi-ppc:
100: 0.03
1000: 0.73
3000: 4.70

pi-pwr:
100: 0.03
1000: 0.70
3000: 4.38
 
  • Like
Reactions: Vol

Vol

New Tinkerer
Jul 28, 2022
35
4
8
Same system, 640x480 256 colors, but booted from the OpenRetroSCSI 7.5.3 image. There does appear to be a bit of performance improvement on 8.5. The other guy was running at higher resolution which probably slowed the scrolling down slightly too.

pi-86k:
100: 0.07
1000: 1.88
3000: 13.17

pi-ppc:
100: 0.03
1000: 0.73
3000: 4.70

pi-pwr:
100: 0.03
1000: 0.70
3000: 4.38
Excuse me, but what is OpenRetroSCSI ? I haven't been able to google this. :mad: Actually your latest results are much closer to the other guy's results than the previous ones for OS 8.5. I can also assume that his RAM is just running slower. And thanks a lot for your support.
EDIT. Your latest results confirm that the char output routine is much faster for 8.5 than for 7.5.3.
 
Last edited: