Quantcast
Channel: Raspberry Pi Forums
Viewing all articles
Browse latest Browse all 5351

General discussion • Re: A Pi Pie Chart

$
0
0
It's astonishing but Xcode on MacOS is still missing OpenMP support. Here are the results for the single-processor version of pichart on a Mac Mini M4.

Code:

% ./pichart-serialpichart -- Raspberry Pi Performance Serial version 40Prime Sieve          P=14630843 Workers=2 Sec=0.573956 Mops=1627.88Merge Sort           N=16777216 Workers=1 Sec=0.732116 Mops=549.985Fourier Transform    N=4194304 Workers=2 Sec=0.111412 Mflops=4141.16Lorenz 96            N=32768 K=16384 Workers=1 Sec=0.0733727 Mflops=43902.3My Computer has Raspberry Pi ratio=100.297Making pie charts...done.
For reference the exact system reported

Code:

olson@Mac pichart-40 % system_profiler SPHardwareDataTypeHardware:    Hardware Overview:      Model Name: Mac mini      Model Identifier: Mac16,10      Model Number: MU9D3LL/A      Chip: Apple M4      Total Number of Cores: 10 (4 performance and 6 efficiency)      Memory: 16 GB      System Firmware Version: 11881.41.5      OS Loader Version: 11881.41.5
Presumably the scheduler is smart enough to run a compute bound task on one of the performance cores. I don't know how to lock a process to an efficiency core as Linux tools like taskset and numactl are missing on MacOS.
There are not many single core benchmarks on this page so I ran pichart on one core of a first-generation AMD Threadripper 1920X 12-core processor.

Code:

taskset -c 0,12 ./pichart-openmp pichart -- Raspberry Pi Performance OPENMP version 40Prime Sieve          P=14630843 Workers=4 Sec=0.543909 Mops=1717.8Merge Sort           N=16777216 Workers=4 Sec=0.927773 Mops=434Fourier Transform    N=4194304 Workers=2 Sec=0.451203 Mflops=1022.54Lorenz 96            N=32768 K=16384 Workers=1 Sec=0.190552 Mflops=16904.7My Computer has Raspberry Pi ratio=53.2023Making pie charts...done.
and a one core of Power 9 LC922 system.

Code:

$ numactl -i 0 -C 0,1,2,3 ./pichart-openmp -tLC922pichart -- Raspberry Pi Performance OPENMP version 25Prime Sieve          P=14630843 Workers=8 Sec=0.933369 Mops=1001.03Merge Sort           N=16777216 Threads=8 Sec=1.08134 Mops=372.366Fourier Transform    N=4194304 Workers=8 Sec=0.309852 Mflops=1489.01Lorenz 96            N=32768 K=16384 Workers=4 Sec=0.444361 Mflops=7249.11The LC922 has Raspberry Pi ratio=39.7688 Making pie charts...done.
Although the Threadripper has 2 threads per core and the Power9 has 4 threads per core, that was not enough to make up for the progress since 2017 and respectively 2015 when those older processors were released.

Statistics: Posted by ejolson — Tue Dec 31, 2024 8:38 pm



Viewing all articles
Browse latest Browse all 5351