[PATCH][Aarch64] Add vectorized mersenne twister

[Ulrich Drepper]

On Tue, Jun 6, 2017 at 12:07 PM, James Greenhalgh
<james.greenhalgh@arm.com> wrote:
> We're a good number of years late to do that without causing some pain.

Well, it's pain for those who deserve it.  Who thought it to be a
smart idea to pollute the global namespace?

It's a one-time deal.


> So we have a few solutions to choose from, each of which invokes a trade-off:
>
>   1 Use the current names and pollute the namespace.

IMO unacceptable.


>   2 Use the GCC internal __builtin_aarch64* names and tie libstdc++ to GCC
>     internals.

Maybe.


>   3 Define a new set of namespace-clean names and complicate the Neon
>     intrinsic interface while we migrate old users to new names.

See Jonathan's proposal.  I never suggested that those who don't care
about namespace pollution would have to change their code.  Add
appropriate aliases.

There is perhaps number 4:

- use the x86-64 intrinsics which you map to aarch64 intrinsics.
Isn't this compatibility layer planned anyway?  I don't know whether
everything maps 1-to-1 and you don't lose performance but you could
this way use the arch-specific code I wrote a long time ago for
x86-64.