[PD] array-abs

[Jonathan Wilkes]

Kind of off topic, but...If we're just talking about iterating through an array, how could hand-optimized assemblerpossibly beat a modern compiler?
-Jonathan 


     On Monday, October 5, 2015 1:40 PM, IOhannes m zmölnig <zmoelnig at iem.at> wrote:
   

 On 10/05/2015 05:39 PM, Christof Ressi wrote:
> You're right that it's easy to implement as an abstraction but I was rather thinking about execution speed. 
> I guess a fictional object like [array drip] that would just iterate through the array using a C loop would be the fastest possible method. 

yes.
well no: the fastest possible way would probably be hand-optimized
assembler.

> I don't see how you could even get close to that with abstractions, especially when using only vanilla objects.

i don't think that in practice this would matter though.
an [array-drip]¹ implementation in Pd-vanilla should be O(n).
an [array drip] implementation in C should be O(n).

sure, the abstraction implementation will perform worse by a (constant)
factor but the complexity stays the same which is the important part.

i assume that the cases where you do need that extra speed boost are
rather seldom, and do not warrant an extra built-in object.


gfamrds
IOhannes

¹ it always causes me pain to see the name "drip" used so widely.

_______________________________________________
Pd-list at lists.iem.at mailing list
UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list


  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.puredata.info/pipermail/pd-list/attachments/20151005/a1a2480b/attachment.html>