new proposal to vector intrinsics

[Sa Liu]

Tobias Burnus <burnus@net-b.de> wrote on 15.01.2008 11:50:33:

> As written, I would prefer something like the following (ignoring "kind"
> for the moment). And I also do not see how xlf plans to handle the
> reverse: vector -> integer/real array (except of EQUIVALENCE).

I agree with the mechanism of using assignment interface. And I've got the 
agreement from xlf, that they will provide the same interface as example 
for users to make use of. By this means we could keep the compatibility of 
data initialization and assignment.

How about the changes in frontend concerning recognition of backend 
builtins? Will this be acceptable in gfotran?

Thanks!
Sa