Intel® Integrated Performance Primitives

Size of temporary Buffer for ippmInvert


I was having a lot of weird problems in my app and solved them increasing the size of the tempbuffer variable in the ippmInvert function.

I was using ippsMalloc_64f(numRows * numCols). And the problems were solved when I increased the length to (numRows * numCols * 2).

But it was a guess and I'm worried it might not work depending on the size of the original matrix.

what would be a better choice of buffer size?

Thanks very much.


Problem creating .lib that uses IPP functions


I have been using IPP 4.1 for some time now and it all works fine. Recently I have tried to create a static lib of my class files which use the IPP. Unfortunately it doesn't work. The code all compiles fine, but then it says "Creating Library..", the CPU jumps to 100% and it stays there for what seems to be eternity. Eventually the computer becomes unusable. I am using Visual Studio .NET, and like I said the same class with the same project properties/settings, works fine. This problem only seems to appears when tryingto make a static library.

Any ideas?

Image resizing question


I'm resizing large images (>10000x10000, rgb or cmyk) using the ippiResize functions.

Am I correct in thinking that the functions require a whole image to be loaded into memory? (not neseccarily a problem as memory is cheap)

Also, are there some sizes & scaling which work best? eg I know there are optimal ways of scaling to 50%, 33%, does the ippiResize also benefit at these scalings?

I've been musing about loading either a few scanlines at a time, or breaking the images into tiles - would a 10x10 set of 1024x1024 image tiles scale faster?

S’abonner à Intel® Integrated Performance Primitives