Autoincrement examples

[Joern Rennecke]

> But I believe we can attack the problems separately -- ie, we can improve
> autoinc detection separately from making the loop optimizer generate code that
> is more autoinc friendly.   The biggest obstacle in doing that is deciding
> precisely what sequences we want to detect for autoinc opts and which ones
> we're going to depend on loop generating autoinc friendly code.

Sorry, but for the example I just posted there is no chance for decent
autoinc generation after the current mainline loop optimizer reduced
tons of givs separately.
We may discuss the patches separately, but we have to test the autoinc patches
with the loop patches in place to get meaningful results.

With loop & regmove patches, the example I posted takes 62 insns for the
SH.  With loop & flow patches, 64.  With the flow patches alone, 185.