Http://www.intel.com/cd/software/products/asmo-na/eng/302910.htm
It is too powerful. Among them, vector/matrix mathematics, image processing, JPEG deCODE/encode, and data compression libraries are very constructive for improving the performance of GIS software. I remember that the Linux version seemed to be free at the time, and now I have to ask for money.
I did a comparative test last year. I forgot the specific test environment. Here I will only explain one problem.
Encode a 20 m BMP image as JPEG. The DOTNET framework comes with a processing capability of about 2000 Ms. It uses cximage, which is the open-source fuse Lib, which is about 1700 ms and uses ijl15 (Intel JPEG Lib, I only got this last year, only lib files and header files, no source code), only about Ms up and down.
The image encoding speed accounts for a significant proportion of the performance of WebGIS that uses raster images to publish maps, improving the image generation speed per unit of time.
If you can use the kml (kernel math lib core mathematical library, it seems like this vector/matrix mathematics) to increase the speed of geometric computing, it will be a good thing.