Are Math Libraries optimized to use the Vector Unit ?

Are Math Libraries optimized to use the Vector Unit ?

I was wondering if Math Libraries (for Linux) are optimized to use the Vector unit when it come to heavy floating point operations ..
or they use the FPU ??

2 posts / 0 new
Last post
For more complete information about compiler optimizations, see our Optimization Notice.

If you're considering architectures supported by MKL, there is no specialized vector unit. MKL for IA32 and EM64T architectures, evidently, makes heavy use of vectorized code, which employs the parallel SSE, SSE2, or SSE3 instructions. So, MKL takes advantage of the ability of the FPU to produce 2 or 4 results in parallel. So do the IPP and svml libraries.

Leave a Comment

Please sign in to add a comment. Not a member? Join today