Strength Reduction on x86

[Mark Horsburgh]

It appears that with the Cygnus ia32 backend being folded back into main
gcc sources the -fstrength-reduce bugs has reared it's ugly head again. I
would normally post the code that provoked it here, but it is in a large
amount of code machine generated by the Berkeley ICSI Sather compiler so
I'd rather not try to track down exactly where the bug is causing trouble.

Does anyone with a better memory than me remember a simple test case that
would provoke the old strength reduction bug? Or, better yet, does anyone
remember what was required to fix it?

Regards, Mark