[Insight-developers] Implementing optimized image copy method

[Bradley Lowekamp]

Hello,

There is a topic/task I have had on my plate for quite a while now. The latest work in progress is here:

http://review.source.kitware.com/#change,3196

Unfortunately, I have been dropping this topic for the past couple of months, and now I will be unable to work on if for the next ~3weeks, which is really when it needed to be finished if not sooner. I am very excited about this topic as it can provide as much as 30x-100x performance improvement.

Would some one care to pick up this issue?

The interface has been defined in the toolkit for months, and it's been used by many filters for many months too. Additionally, the WIP in the above patch should have a robust algorithm to copy block from one region to another.

What is really left is to determine when this memcpy optimized method can be used. I currently think that when the image is a itk::Image or itk::VectorImage  and the pixel type is (stored as ) POD that the optimized method could be used. The is pod part is the trickier one it should be easies to add an additional numeric trait try or try_compile for tr1::is_pod ( and pity those with out ). Additionally, robust testing is always desired, and double checking there are no wacky pixel buffer structures or arrangements may be needed too.

If some one is interested in picky this up we could schedule a mini-tcon on monday or tuesday to estrange information.

Thanks,
Brad

========================================================
Bradley Lowekamp  
Medical Science and Computing for
Office of High Performance Computing and Communications
National Library of Medicine 
blowekamp at mail.nih.gov



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.itk.org/mailman/private/insight-developers/attachments/20111123/02538dcb/attachment.htm>