On Tue, May 03, 2016 at 03:51:11PM +0200, Bernd Schmidt wrote:
> If that is all ok and we just have a really odd cfg, then your patch is ok
> with a better comment as to which situations can cause the problem.
Thanks, committed rev 235825.
diff --git a/gcc/ira.c b/gcc/ira.c
index a38e67e..269a190 100644
--- a/gcc/ira.c
+++ b/gcc/ira.c
@@ -3742,6 +3742,22 @@ combine_and_move_insns (void)
if (use_insn == BB_HEAD (use_bb))
BB_HEAD (use_bb) = new_insn;
+ /* We know regno dies in use_insn, but inside a loop
+ REG_DEAD notes might be missing when def_insn was in
+ another basic block. However, when we move def_insn into
+ this bb we'll definitely get a REG_DEAD note and reload
+ will see the death. It's possible that update_equiv_regs
+ set up an equivalence referencing regno for a reg set by
+ use_insn, when regno was seen as non-local. Now that
+ regno is local to this block, and dies, such an
+ equivalence is invalid. */
+ if (find_reg_note (use_insn, REG_EQUIV, NULL_RTX))
+ {
+ rtx set = single_set (use_insn);
+ if (set && REG_P (SET_DEST (set)))
+ no_equiv (SET_DEST (set), set, NULL);
+ }
+
ira_reg_equiv[regno].init_insns
= gen_rtx_INSN_LIST (VOIDmode, new_insn, NULL_RTX);
bitmap_set_bit (cleared_regs, regno);
--
Alan Modra
Australia Development Lab, IBM
