[Bug rtl-optimization/24194] emit_input_reload_insns secondary reload handling is unsafe

[amylaar at spamcop dot net]

------- Comment #2 from amylaar at spamcop dot net  2006-07-20 17:57 -------
Subject: Re:  emit_input_reload_insns secondary
        reload handling is unsafe

Quoting pinskia at gcc dot gnu dot org <gcc-bugzilla@gcc.gnu.org>:

>
>
> ------- Comment #1 from pinskia at gcc dot gnu dot org  2006-07-20 11:19
> -------
> Isn't this really fixed by the TARGET_SECONDARY_RELOAD target hook?

reload_adjust_reg_for_temp checks only the first hard register for suitability
wrt. NEW_CLASS.
I.e. we check TEST_HARD_REG_BIT (reg_class_contents[(int) new_class], regno) ,
but for a two-hard-reg register we'd also have to check
TEST_HARD_REG_BIT (reg_class_contents[(int) new_class], regno) .

AFAICT the other issues in this PR have been addresssed.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24194