Initial Benchmark Results for Trunk, PI5, PI5 ILS & PI5 ILS/Cycle

[Mike Katz]

Here are the initial Benchmark Results:



Trunk =                 fossil open ~/museum/pidp8i.fossil
PI 5 =                    fossil open ~/museum/pidp8i.fossil pi5
Pi 5 ILS =              fossil open ~/museum/pidp8i.fossil pi5-ils2-bookworm
Pi 5 Cycle/ILS =   fossil open ~/museum/pidp8i.fossil pi5-ils2-bworm-cyclerealistic

All of the scripts and results can be found here:

https://drive.google.com/drive/folders/1gqcaIQveI0iRxoHGpogkj6muWeZLja1s?usp=drive_link

The Benchmark was running the Pascal Prime program modified to calculate the first 1000 prime numbers and not print them out.

As you can see from the results Cycle Accurate W/Incandescent Lamp Simulation is approximately 50% slower than the non-cycle accurate.
Incandescent Lamp Simulation is only slightly slower than without.

I don't understand why trunk is faster on the 4B and 4B and slower on the rest.  Unless there were other boot processes that were still running when I ran the benchmark.

Please let me know if you have any suggestions, questions or comments.

[William Cattey]

I suspect that the reason why the number of instructions per second on the Pi4 and Pi5 is so much higher with the trunk code line is that it is not even bothering to try griot work.

-Bill