🚀 go-pugleaf

adt@netcom.com (Anthony D. Tribelli) writes:

>karl unterkofler wrote:
>> A simple test with Mathematica:
>> 
>> Timing[N[Sin[1/2],2500]][[1]]
>> 
>>      PowerMac 8500/120, 16M RAM:             1.43 Second
>>      Dell Dimension XPS P120c, 16Mb, 512K:   4.27 Second
>> 
>> Timing[10001!][[1]]
>> 
>>      PowerMac 8500/120, 16M RAM:             2.32 Second
>>      Dell Dimension XPS P120c, 16Mb, 512K:   4.31 Second

>Which could be telling us more about Mathmatica's implementation and the
>compilers used to build Mathematica than the hardware Mathematica is being
>run on. Just something to consider. Using a commercial application to
>evaluate hardware is tricky. If you pick the right application you can
>"prove" nearly anything. In this specific case I think the general idea
>that the PPC system is faster for purely computational code is correct. 

>Tony

It is a bit tricky and neither side will ever give in.  "It wasn't 604 
optimized but it was P6 optimized..." "They didn't use the xyz libraries" 
"it was 16bit code..."  are just some of the excuses.  Bottom line IMO:
These are the apps that you use if one is implemented poorly on a platform 
it says nothing about machine speed or processor speed but you have to use the
app anyway and that platform may be the wrong choice for what you do.

The only alternative is to use suites of benchmarks that mean nothing.  Does
anyone care how fast a pentium can do IDEA encryption or start the seive of
Eratostophenes?  I'll give up a few KB/sec on IDEA speed if it means that my
compilers and editors will run 10% faster.