Closed gadfly16 closed 10 years ago
The Attrib Transfer qL SOP should probably become obsolete (as sesi claims drastic improvements to the original Attrib Transfer SOP). We should do some quick speed comparisons, and dump the qL version if it's not faster any more.
qL! We should agree on a speed factor that makes it feasible to maintain a qL version of things. I suggest 10% or 20%..
On Fri, Apr 19, 2013 at 9:25 PM, Imre Tuske notifications@github.comwrote:
The Attrib Transfer qL SOP should probably become obsolete (as sesi claims drastic improvements to the original Attrib Transfer SOP). We should do some quick speed comparisons, and dump the qL version if it's not faster any more.
— Reply to this email directly or view it on GitHubhttps://github.com/qLab/qLib/issues/21#issuecomment-16678325 .
Right now the original attrib xfer seems ~2 times faster than the qL one (for 2 attribs), and slightly faster for a single attrib. I suggest graveyard. :))
You know I'm happy to bury. By the way I expect this to happen to more and more performance inspired qL operators. After the geometry engine rewrite, SESI told that operator specific improvements will be rolling out slowly as they reimplement the ops one by one in a multithreaded fashion.
On Wed, Aug 28, 2013 at 4:32 AM, Imre Tuske notifications@github.comwrote:
Right now the original attrib xfer seems ~2 times faster than the qL one. I suggest graveyard. :))
— Reply to this email directly or view it on GitHubhttps://github.com/qLab/qLib/issues/21#issuecomment-23385927 .
Man, you should be able to limit the effect of this node too.