Reading/storing speed for multiple operations

[bibi88]

Hello,

I'm using IPP for image processing since a short time and I was wondering if it increases the speed when I'm doing MANY DIFFERENT operations on pixels.
For example, I would like to apply a lot of arithmetic operations on an image. As far I've understand the library, it enables to apply one arithmetic operation on the image and store the result in an ippiMalloc memory. Imagine now that I've 100 different arithmetic operations to do sequentially on this picture. Because of the multiple reading/storing access times (reading the result of the previous arithmetic operation, process it, and store it in a new ippiMalloc) of each function, I'm wondering if it's not more suitable to read a pixel of the first image only once, do the operation, and go to another pixel, and so on, until storing the global result?
I hope I've been clear enough in my question and that it will avoid me to have to compare each time the difference of speed with/without IPP :)...

Best regards,

Bibi

[Vladimir_Dudnik]

Hello,

Intel Deffered Mode Image Processing (DMIP) library which is built on top of IPP was designed to enable threaded computational pipeline of consequtive image operations. Intel DMIP library available in IPP Samples download page,you can check it by the link below
http://software.intel.com/en-us/articles/intel-integrated-performance-primitives-code-samples/

Regards,
Vladimir

[bibi88]

Thanks for the so fast response. Is there any way to use it (or to do otherwise) for a Linux user?

Best regards,

Bibi

[Vladimir_Dudnik]

Unfortunately no. At the current version DMIP is only available for Windows platforms. We do not have immediate plans to provide it on Linux

Vladimir