Determine/document support for Intel compiler on macOS

[Paul Hargrove]

Intel is now providing free "Intel C++ Compiler Classic" for macOS with Intel CPUs. There is no DPC++ product for macOS at this time.

Testing on my laptop with UPCXX_INSTALL_NOCHECK=1 encountered only one issue in dev-check (smp-conduit).

This is currently a low priority, IMO. I can imagine HPC developers using Intel's compilers, rather than Apple's which lack OpenMP and Fortran (Intel's Fortran compiler is free too). However, if there was a DPC++ product for macOS, then I would take that possibility far more seriously.

I will see about getting automated nightly testing setup when I have some spare cycles (see Bug 4253).

[Paul Hargrove]

The "one issue in dev-check" mentioned above seems to have been a potentially genuine type error. Whether it was a bad example or a compiler bug is still uncertain (at least to me). However, it was fixed in ea21e24

[Paul Hargrove]

A EX-big_sur-smp-icc-pshm configuration now runs regression tests every Wednesday, and include the UPC++ tests. The first run, on July 14, was free of new errors.

[Paul Hargrove]

This was resolved in pull request #366, in which we documented the Intel oneAPI compilers as supported

[Paul Hargrove]

Oops!

While we have added a statement of support for the "oneAPI" compilers on Linux, this issue is for the "Classic" compilers on macOS.

I am considering Intel Classic compiler support for the next release (milestone update shortly) and thus am moving this issue to the "open" status (rather than "new").

[Paul Hargrove]

Though I am perceiving no urgency to claim support for Intel Classic C++ on macOS, I also don't see any reason not to (based on weeks of testing each Wed). So, unless I hear objections, I am planning to create a PR for (at least) the configure checks portion of this issue for consideration by the Sep 15 "feature freeze".

[Paul Hargrove]

Pull request 375 now contains the necessary configure logic updates to permit Intel Classic C++ on macOS as "has not been validated", but not as fully supported.

A 1-line change would update this status to supported, but would require corresponding documentation updates. That is the next logical step toward resolving this issue.

[Dan Bonachea]

Mass roll-over of open issues to next release milestone

[Dan Bonachea]

Mass roll-over of open issues to next release milestone

[Dan Bonachea]

This is now blocked by issue #559: Compile error on test/view from Intel compiler on macOS with optional.hpp

[Paul Hargrove]

In light of issue #559, and a perceived lack of end-user interest, I have no objection to resolving this "wontfix" if others feel that is appropriate.

[Dan Bonachea]

Until/unless a stakeholder expresses interest in the macOS/IntelC++ combination, we are not going to expend effort supporting it.