[PATCH] Fix up norm2 simplification (PR middle-end/88074)

[Thomas Koenig]

Hi Jakub,

I checked the patch together with Richard's (by which I assume you
mean https://gcc.gnu.org/bugzilla/attachment.cgi?id=45052 ), and
thinks looked good.

So, the Fortran part of this is OK.

However, we should really also do power-of-two scaling
for the runtime method.

Also, we seem to have a lot of issues with IEEE flags
when calculating NORM2, this would also need to be
addressed.

Regards

	Thomas