sieve benchmark

Sieve Benchmark and Windows CE Handhelds

  • by

What started as a simple program to compare performance between PC and Amiga accelerators is now a project with single source code that can be compiled on many computer platforms: DOS (Borland 16bit, Metaware 32bit), Linux/gcc (x86+asm, x86-64+asm, ARM+asm), IRIX (MIPS), Mac/gcc/codewarrior (68k, PPC classic, PPC carbon, PPC/x86 OS X native), Amiga (68k up to 68040; StormC/gcc) and now even Windows CE (SH3, MIPS, ARM).

For me, this is an entertaining way to understand several OS platforms and their popular development environments. This is mostly about C/C++ compilers but we’ve made one BASIC (MS QuickBasic 4.5) and few assembly implementations to see efficiency of compiled code.

… back to Windows CE handhelds: I used several during my studies (Jornada 680, 690, 720, 728) for taking notes (as a replacement for Psion 5MX). Those with Hitachi SH-3 were unbelievably slow and you could see how GUI was drawn item by item. Later, I switched to ARM-based ones and everything was super-fast.  Now I see – low-power Intel StrongARM on 206MHz is more efficient (integer performance) than Pentium MMX with the same clock if data can fit in its 8K cache. For larger data sets it is still as fast as my 133-MHz Pentium MMX laptop. Nice results for something that small.

Web page with results (70+) from all machines is under construction.