[RFC] Fix PR28684

Revital1 Eres ERES@il.ibm.com
Thu Nov 16 16:24:00 GMT 2006



Clint Whaley <whaley@cs.utsa.edu> wrote on 16/11/2006 17:32:12:

> Guys,
>
> >I wasn't sure if general strength reduction was still controlled by the
> >flag, but if it is, we need to mention it, and again the user can
>
> Just to mention, if we need strength reduction, the definition that would
> make me happiest would be that no new flops are introduced by it, but
> some may be removed.  Then, only the people doing things like NaN traps
> need worry.  I.e., 2.0*a <--> a+a and a+b-b -> a would be OK, but
> 3*a -> a+a+a would not be allowed . . .

vectorization does not need strength reduction so far.
I did not see such transformation in the list of transformations
which I though might fit the flag.
The fact that it is not controlled by the flag
was implicitly mentioned in the definition -

An example of a transformation which is not allowed by this flag is
transforming x * 0 to 0 as the result is not IEEE compliant when x
is Inf.

Should I add a comment about it in the definition?

Revital



More information about the Gcc-patches mailing list