Support for 128-bit ints

[bradcray]

As time goes on, interest in int(128) seems to be growing. This issue is a placeholder requesting to add such support for 128 bit integers, at least on platforms and back-end compilers for which it's reasonably supported.

My assumption is that the work here would involve the following major components:

• extending the compiler to support 128-bit ints in addition to 8, 16, 32, and 64
• adding overloads to common 128-bit math routines
• ensuring that resolution and related compiler logic handled this new width appropriately
• plumbing the type to LLVM and C back-ends as appropriate

A lot of this seems pretty reasonable and mechanical as long as the LLVM/C support is there. The part I fear the most is the degree to which we've (intentionally or un-) baked in assumptions about the default int width also being the max.

[damianmoz]

Wise move.

There was discussion about real(128) and int(128) in the email discussion (predating discourse) from about December 2017 which noted both would need to happen eventually. It involved Michael, Jeff Hammond and myself and it came from the 128-bit floating point perspective. Not sure that we said anything startling or even momentous but rather noted it for the future. Jeff mentioned

https://en.wikipedia.org/wiki/Intel_ADX

To your list, add fundamental operations like transmute (eventually).

[bradcray]

I'm giving this a bump due to additional user interest since posting it.