From 249083758d18f86be080a87d0ec8630ad58c1f30 Mon Sep 17 00:00:00 2001 From: "J\"orn Rennecke" Date: Sat, 22 Nov 2003 14:18:48 +0000 Subject: flow.c (update_life_info): Amend comment about when a register can become dead. * flow.c (update_life_info): Amend comment about when a register can become dead. From-SVN: r73838 --- gcc/flow.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'gcc/flow.c') diff --git a/gcc/flow.c b/gcc/flow.c index 7ad7a80..0dffb46 100644 --- a/gcc/flow.c +++ b/gcc/flow.c @@ -594,6 +594,9 @@ verify_local_live_at_start (regset new_live_at_start, basic_block bb) generates subregs of a multi-word pseudo, current life analysis will lose the kill. So we _can_ have a pseudo go live. How irritating. + It is also not true when a peephole decides that it doesn't need one + or more of the inputs. + Including PROP_REG_INFO does not properly refresh regs_ever_live unless the caller resets it to zero. */ -- cgit v1.1