optimization/3384: PowerPC -funroll-loops -fbranch-count-reg misoptimization

[Michael Hayes]

Zoltan Hidvegi writes:
 > Thanks for looking at that, but could you please consider the patch in
 > http://gcc.gnu.org/ml/gcc-patches/2001-07/msg01080.html in favor of this
 > one?  This improves a tiny bit on normal non-unrolled loops, and it has
 > some comments.  Also I forgot to remove some misleanding comments in that
 > patch, so I attach below a comment-only patch that should be applied
 > after http://gcc.gnu.org/ml/gcc-patches/2001-07/msg01080.html .

The doloop part of this patch is OK.  

 > Also note that there is still a bug here, that is not new:
 > 
 > main(){unsigned int i; for (i = ~0-39; i < ~0u-1; i += 4);}

 > This is an infinite loop, but it quits when compiled with -O2.  It would
 > be easy to fix for the constant limit case, actually I have some hack for
 > that, but for the variable final value case that would require extra
 > run-time checks for every loop even if they would never overflow, and the
 > standard behavior is undefined anyways, so it probably does not worth
 > fixing it (that was that way even in 2.95).

Do you know what the standard says about these type of loops?  I could
never envisage a case where this behaviour would be deliberate.  When
people want an infinite loop, a non-conditional expression is used for
the loop termination test.

Michael.