Re: [PATCH][vectorizer][2/2] PR 65951: Hook up mult synthesis logic into vectorisation of mult-by-constant

[Marc Glisse <marc dot glisse at inria dot fr>]

On Tue, 14 Jun 2016, Kyrill Tkachov wrote:

> On 14/06/16 08:04, Marc Glisse wrote:
> > On Mon, 13 Jun 2016, Kyrill Tkachov wrote:
> >
> > > The new function vect_synth_mult_by_constant that does all the hard work 
> > > is very similar in structure to expand_mult_const from expmed.c but it 
> > > operates on gimple SSA rather than RTL.
> >
> > I think in some cases this transformation creates undefined overflow from 
> > well-defined code. While this is fine in RTL, I would expect to need a cast 
> > to unsigned_type_for in gimple for some of the algorithms.
> > (not completely sure, just something to look at)
>
> Would casting the initial argument to the unsigned type using fold_convert, 
> performing the algorithm and then
> casting the final result back to signed if necessary at the end work?

I don't know if fold_convert is the best interface for that, but yes, that 
should work. It is probably only needed in a few cases (if we multiply by 
a negative divisor of INT_MIN, or if the outermost operation is a 
subtraction, for instance), I don't know if detecting those is worth the 
trouble or we should always cast (when !TYPE_OVERFLOW_WRAPS).

--
Marc Glisse