x86 NCG SIMD: Implement pack/insert/broadcast/unpack for integer vectors
Motivation
Support of integer vectors in NCG would be nice. As a start, I'd like to implement non-arithmetic basic operations (pack/insert/broadcast/unpack) on integer vectors.
Proposal
Implement pack/insert/broadcast/unpack for Int8X16#
, Word8X16#
, Int16X8#
, Word16X8#
, Int32X4#
, Word32X4#
, Int64X2#
, Word64X2#
in x86 NCG.
Some of them (64X2) are already implemented in !12857 (closed)/!12860 (closed), but they could be enhanced to use SSE4.1 instructions if enabled.
See also #25030.