topic Re: Working in VC7 in Intel® oneAPI Math Kernel Library

Working in VC7

limchiwa — Wed, 19 Jul 2006 10:43:58 GMT

Sorry about another post. I just need an answer regarding using Intel Kernel Math Library in a VC7 environment.

Is there any setting/optimization setting needed? Or i just link the library and use the library function in order to see some performance gain?

I trying using just the Dot Product function in the kernel math library, but it seems to be much slower than the C++ version.

Can someone pls provide some light on this matter for me? Been looking around but just seems to any clues on this.

thanks a million

Calvin

Re: Working in VC7

TimP — Wed, 19 Jul 2006 13:15:53 GMT

The options used for your compiler affect only the code generated by the compiler, not the code in the MKL library. Even with a non-vectorizing compiler, you may be able to execute dot products of vectors up to length about 50 faster than the MKL function does. The advantage of MKL comes into play where aggressive use of parallel instructions and cache optimizations are useful. Dot products of short vectors depend on in-line expansion for efficiency.

Re: Working in VC7

limchiwa — Wed, 19 Jul 2006 13:23:59 GMT

Hmm.. so am i to assume that if i'm only going to employ simple 3D vector operations, like dot product, magnitude, crossproduct, etc.. and simple trigonometry functions like sin, cos, tan..... then i'm not going to see much use for the MKL library?

Calvin