[SciPy-user] Benchmark data

[Travis Oliphant]

>
>
>> 
>>    
>>
>I understand that gcc uses inline assembler for simple math
>functions, so it is certainly something to look into.
>  
>
We should find out which functions and make sure they have their own 
inner loop instead of using the generic one.  

It definitely looks like gcc is able to optimize numarray code better on 
Intel processors.  It would be nice to have more people on Intel chips 
confirm this as well --- I have access to one so I can look into it. 

I'll play with some differences in the way the loops are written to see 
what effect it has.  As you can see, I've not noticed any effect on an 
AMD chip, but on INTEL chips gcc may be able to do more.

>I agree that more data is necessary and I remind everybody that my data is
>for an Intel CPU and that all other data (David Cooke, Arnd Backer and you)
>is for AMD CPU's.
>  
>
Yes, I think it's been very valuable to show us these cases.

>
>Gerard
>
>PS: the additional benchmark data:
>
>[snip]
>

Thanks for those additional data.  I think they show that it's more the 
INTEL architecture then the optimizations.

-Travis