On 07/17/2012 09:20 AM, Gael Varoquaux wrote:
> On Tue, Jul 17, 2012 at 09:01:28AM -0700, Jake Vanderplas wrote:
>> I put this together really quickly this morning - let me know if you see
>> anything I missed that makes the comparisons unfair.
> You're at scipy, aren't you? Maybe it would be useful to grab some of the
> cython folks to have their input on this.
>
Will do.
I did a bit more experimentation, and it looks like memoryviews with
boundscheck and wraparound turned off are largely comparable to pointers
for *pure indexing*. Slicing is what kills things.
With Ball Tree in mind, passing around slices (i.e. points represented
by rows of the data matrix) is what we need to do, so I think for speed
it would be better to stick with pointers for now.
Jake
------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and
threat landscape has changed and how IT managers can respond. Discussions
will include endpoint security, mobile security and the latest in malware
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
