This is the performance based implementation of the of the two part pair, for which Multiversioning for range check elimination was the first part. This part implements masked vectorization for fully enabled EVEX x86/x64 targets.
Tested as follows:
SpecJvm2008 windows: HSW(64/32 bit): SKX(64/32 bit) : KNL (64/32 bit)
jtreg: windows: HSW(64/32 bit) , skx(64/32-bit)