Your change of September 11, 1998
Richard Kenner
kenner@vlsi1.ultra.nyu.edu
Tue Jan 16 03:59:00 GMT 2001
Actually, this begs the question of why we have call_value at all?
If we distinguished between death via set and death via clobber,
it might matter. Flow doesn't though. Perhaps reorg and its
mildly broken forward-scanning analysis cares?
Having thought about this overnight, the issue isn't *death*, but *life*.
It indicates which registers are live after the return and that is
useful information, which means we have to get it right.
More information about the Gcc
mailing list