[patch, fortran] Fix PR 42131, improvement in do loops

[Richard Henderson]

On 12/01/2009 09:01 AM, Janne Blomqvist wrote:
> So, any preferences?

Install the patch.  The proper place to optimize between these
two forms is in the target's "mov<mode>cc" pattern, not in the
front end.

File an optimization bug, if you like, against the x86 backend
for how we render this if conversion.  Though as you note it may
be difficult to pull this change out of the noise.


r~