Masking for AVX512 and complete mask boxing

[Lupusoru, Razvan A]

Hi everyone,

I am submitting a patch with the following features:

*         AVX512 mask support by converting mask to vector registers (and vice-versa)

*         Working comparisons and blend for float in AVX512

*         Improved architecture feature checking which ensures intrinsification will not happen if unsupported instruction is being used

o   I made sure that Vladimir's generalized intrinsics also use this improved checking mechanism

*         Complete mask boxing implementation for types and sizes

*         Bug fix for mulAll in generalized intrinsics implementation

Please let me know if you have any comments. The webrev is available here:
http://cr.openjdk.java.net/~rlupusoru/panama/webrev_maskboxingavx512_02/index.html

Thanks,
Razvan