On Thu, 25 Nov 2021, Jan Hubicka wrote:
> Hi,
> this pach removes gimple_call_cahin checkin ref_maybe_used_by_call_p that
> disables check for CONST functions. I suppose it was meant to allow consts
> to read variables from the static chain but this is not what other places
> do. The testcase:
>
> int
> main()
> {
> int a =0;
> __attribute__ ((noinline,const))
> int reta ()
> {
> return a;
> }
> int val = reta();
> a=1;
> return val+reta ();
> }
>
> Gets optimized to single call of reta since at least gcc 4.1.
>
> LTO bootstrapped and regtested x86_64-linux all languages. OK?
I suppose at some point it broke.
But yes, I agree, thus OK.
Thanks,
Richard.
> * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Do not check
> gimple_call_call when treating const functions.
>
> diff --git a/gcc/tree-ssa-alias.c b/gcc/tree-ssa-alias.c
> index cd6a0b2f67b..3c253e2843f 100644
> --- a/gcc/tree-ssa-alias.c
> +++ b/gcc/tree-ssa-alias.c
> @@ -2743,9 +2743,7 @@ ref_maybe_used_by_call_p_1 (gcall *call, ao_ref *ref,
> bool tbaa_p)
> unsigned i;
> int flags = gimple_call_flags (call);
>
> - /* Const functions without a static chain do not implicitly use memory. */
> - if (!gimple_call_chain (call)
> - && (flags & (ECF_CONST|ECF_NOVOPS)))
> + if (flags & (ECF_CONST|ECF_NOVOPS))
> goto process_args;
>
> /* A call that is not without side-effects might involve volatile
>
--
Richard Biener <rguent...@suse.de>
SUSE Software Solutions Germany GmbH, Maxfeldstrasse 5, 90409 Nuernberg,
Germany; GF: Ivo Totev; HRB 36809 (AG Nuernberg)
