How to define contant look up tables (LUT) ??

I want to build hundreds of 32 bit input and 16 bit input look up tables inside the fpga. But they must work very fast and spend minimum logic cells.

there is also LPM_MUX plugin in the wizard.

- Is this a simple LUT generator? Are there wasted logic cells on the design output?

- is it the way for the fastest LUT generation? Or the primitives method output is faster?

Thanks..