I think using total time gives too much weight to day 14. If one compares the average ratio, thenSince the Pico's arrival I've tried running C++ Advent of Code solutions on it. This year it gets more interesting with the addition of the Pico 2. These solutions are single threaded pure CPU and memory system (including XIP) stressors using integer arithmetic, without IO. I wanted some idea of CM3 vs. CM0+ performance. Again this is pure CPU and does not take advantage of the CM3's DSP and floating point extensions. All solutions were compiled with GCC 14.2 Rel 1 provided by Arm here and using the -std=c++23 -O3 compiler options.
pico.png
Adjusting for different system clocks, the CM3 core is 1.98 * 133 / 150 = 1.75 CM0+ cores.
1.80*133/150=1.6
implies the CM3 core is 1.6 CM0+ cores.
The Pico 2 also has a RISC-V mode. I wonder if that's faster or slower on these puzzles.
Statistics: Posted by ejolson — Mon Dec 30, 2024 12:49 am
