you're not supposed to get that yet. Basically, during sse2 optimization, i re-wrote the pointer math in gabor.c. so that all operations are performed "forward" in memory. which requires flipping the data from the "natural" direction.
My next patch is "ready", and gets the rest of the 2x speed improvement in, but i'm not quite happy with it, as i do four discrete things at once in the patch. However, the patch broken in four will be 3X the size of gabor.c. I'll see if i can push this/these patch/patches out the door today, for your perusal. i'm very excited to get this next patchset in, as it'll let me start working on moving the feature extractor to 512x512 images. :D Julia Longtin <[EMAIL PROTECTED]> On Thu, Jul 20, 2006 at 05:57:17PM +0200, Wolfgang Müller wrote: > Hi, > > > My next set will come with changelog entries. I'll use the ones you wrote > > as an example template, when writing new ones. > > In fact, I am not happy with what I wrote for your patches, but I think it's > a > start. I think it would be good to have a one-liner which is barely longer > than the file name. I found the file name a good approximation for what I > would have liked to write. What is the double-reverse stuff for? I did not > quite get that. > > Thanks again, > > Cheers, > Wolfgang > > -- > Dr. Wolfgang Müller > LS Medieninformatik > Universität Bamberg > Check out the SIG MM web site http://www.sigmm.org _______________________________________________ help-GIFT mailing list [email protected] http://lists.gnu.org/mailman/listinfo/help-gift
