initial Aarch64 support

[WalterBright]

Only works for a trivial function: void foo() { }

This is here mainly to ensure it stays in sync with the master branch as both evolve, and to not break the master.

Added undocumented -arm command line switch to enable it.

[dlang-bot]

Thanks for your pull request, @WalterBright!

Bugzilla references

Your PR doesn't reference any Bugzilla issue.

If your PR contains non-trivial changes, please reference a Bugzilla issue or create a manual changelog.

Testing this PR locally

If you don't have a local development environment setup, you can use Digger to test this PR:

dub run digger -- build "master + dmd#16554"

[ibuclaw]

I can't see you getting far mixing different architectures like this.

[ibuclaw]

In case this comes across as an unactionable comment. To clarify, changes like this make it quite clear that a lot more work needs to go into separation of IR and target modules in the back-end before the introduction of any new architecture can be done (sans this kind of spaghetti imports).

That alone would make introducing any new foreign architecture into dmd more palatable.

[WalterBright]

A good chunk of the code between the two will be identical. It's a balance between having them being totally separate and sharing common code. I don't know just where that balance will be, it will become apparent after I make more progress.

[Herringway]

Having every architecture depend on every other architecture is not sustainable. Isn't contributing a new architecture hard enough without technical debt like this?

[WalterBright]

It won't be clear where the separation should be until I make much more progress.

[WalterBright]

How does disasm86.d, with its last commit 2 weeks ago, suddenly start failing the LDC Ubuntu run of its unittests?

a746142

core.exception.AssertError@src/dmd/backend/x86/disasm86.d(3760): Assertion failure

[WalterBright]

The LDC Ubuntu test is now working again, though I still have made no changes in that direction.

[WalterBright]

The LDC Ubuntu is back to failing again.

[tgehr]

In target.d:

426 | isX86 = !isX86_64;

Looks wrong.

[WalterBright]

@tgehr the process of setting those variables is an ugly tangle that I had only partially straightened out. I added an assert that it was set up correctly.

[WalterBright]

P.S. I fixed the heisenbug in the disasm86 code.

[tgehr]

@tgehr the process of setting those variables is an ugly tangle that I had only partially straightened out. I added an assert that it was set up correctly.

isX86 = !isX86_64;
assert(isX86 + isX86_64 + isAArch64 == 1); // there can be only one

This assertion is equivalent to assert(!isAArch64).