PR43902 patch: Widening multiply-accumulate

[Richard Guenther]

On Wed, Jun 23, 2010 at 3:42 PM, Steven Bosscher <stevenb.gcc@gmail.com> wrote:
> On 6/23/10, Bernd Schmidt <bernds@codesourcery.com> wrote:
>>> Your support for ternary gimple is far from complete - I'm not sure
>>> we want to have this half-supported state (though I guess I don't
>>> care too much and definitely like that we start on it rather than
>>> using more single rhss).
>>
>> I'm thinking it makes little sense to try to convert other passes if
>> there is no way at the moment to test these modifications.
>
> You could move up the pass that creates the new gimple in the
> pipeline, and see what happens?

Coverage wouldn't be great I suppose, but fixing up the ICEs that
happen with the existing testcases would be indeed nice (though
can be done as a followup as well).

> It seems to me that adding this new gimple only in later passes breaks
> pass reordering plugins (like Milepost/ICI).

They have to be careful anyway with random re-orderings.

Richard.

> Ciao!
> Steven
>