bug with biv increment -> giv transformation

Joern Rennecke amylaar@cygnus.co.uk
Thu Feb 18 15:51:00 GMT 1999


> The code for setting biv_dead_after_loop was apparently copied from the
> code for setting bl->eliminable, except that you dropped the final_biv_value
> check.  Why?  It would be useful to document in a comment that the code is a
> duplicate of code taken from elswhere, and to explain why the final_biv_value
> check isn't here.

That would be a chicken and egg problem - iteration_info wants the information
about givs to be already set up.  Besides, I don't see how this would
gain us much, since the number of iterations can't be determined if there
is a random jump out of the loop.

However, there is something here that I don't understand: I can't find
where iteration_info guards against this scenario.


More information about the Gcc-patches mailing list