> Am 04.04.2025 um 19:35 schrieb Jakub Jelinek <ja...@redhat.com>:
>
> Hi!
>
> Before my PR119376 r15-9145 changes, suitable_for_tail_call_opt_p would
> return the same value in the same caller, regardless of the calls in it.
> If it fails, the caller clears opt_tailcalls which is a reference and
> therefore shared by all calls in the caller and we only do tail recursion,
> all non-recursive or tail recursion non-optimizable calls are not
> tail call optimized.
>
> For musttail calls we want to allow address taken parameters, but the
> r15-9145 change effectively resulted in the behavior where if there
> are just musttail calls considered, they will be tail call optimized,
> and if there are also other tail call candidates (without musttail),
> we clear opt_tailcall and then error out on all the musttail calls.
>
> The following patch fixes that by moving the address taken parameter
> discovery from suitable_for_tail_call_opt_p to its single caller.
> If there are addressable parameters, if !cfun->has_musttail it will
> work as before, disable all tail calls in the caller but possibly
> allow tail recursions. If cfun->has_musttail, it will set a new
> bool automatic flag and reject non-tail recursions. This way musttail
> calls can be still accepted and normal tail call candidates rejected
> (and tail recursions accepted).
>
> Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
Ok
Richard
> 2025-04-04 Jakub Jelinek <ja...@redhat.com>
>
> PR tree-optimization/119616
> * tree-tailcall.cc (suitable_for_tail_call_opt_p): Move checking
> for addressable parameters from here ...
> (find_tail_calls): ... here. If cfun->has_musttail, don't clear
> opt_tailcalls for it, instead set a local flag and punt if we can't
> tail recurse optimize it.
>
> * c-c++-common/pr119616.c: New test.
>
> --- gcc/tree-tailcall.cc.jj 2025-04-04 08:42:32.880764473 +0200
> +++ gcc/tree-tailcall.cc 2025-04-04 09:50:25.008575240 +0200
> @@ -165,8 +165,6 @@ suitable_for_tail_opt_p (gcall *call, bo
> static bool
> suitable_for_tail_call_opt_p (gcall *call, bool diag_musttail)
> {
> - tree param;
> -
> /* alloca (until we have stack slot life analysis) inhibits
> sibling call optimizations, but not tail recursion. */
> if (cfun->calls_alloca)
> @@ -204,18 +202,6 @@ suitable_for_tail_call_opt_p (gcall *cal
> return false;
> }
>
> - /* ??? It is OK if the argument of a function is taken in some cases,
> - but not in all cases. See PR15387 and PR19616. Revisit for 4.1. */
> - if (!diag_musttail || !gimple_call_must_tail_p (call))
> - for (param = DECL_ARGUMENTS (current_function_decl);
> - param; param = DECL_CHAIN (param))
> - if (TREE_ADDRESSABLE (param))
> - {
> - maybe_error_musttail (call, _("address of caller arguments taken"),
> - diag_musttail);
> - return false;
> - }
> -
> if (diag_musttail
> && gimple_call_must_tail_p (call)
> && warn_musttail_local_addr)
> @@ -565,6 +551,7 @@ find_tail_calls (basic_block bb, struct
> basic_block abb;
> size_t idx;
> tree var;
> + bool only_tailr = false;
>
> if (!single_succ_p (bb)
> && (EDGE_COUNT (bb->succs) || !cfun->has_musttail || !diag_musttail))
> @@ -660,6 +647,25 @@ find_tail_calls (basic_block bb, struct
> if (!suitable_for_tail_call_opt_p (call, diag_musttail))
> opt_tailcalls = false;
>
> + /* ??? It is OK if the argument of a function is taken in some cases,
> + but not in all cases. See PR15387 and PR19616. Revisit for 4.1. */
> + if (!diag_musttail || !gimple_call_must_tail_p (call))
> + for (param = DECL_ARGUMENTS (current_function_decl);
> + param; param = DECL_CHAIN (param))
> + if (TREE_ADDRESSABLE (param))
> + {
> + maybe_error_musttail (call, _("address of caller arguments taken"),
> + diag_musttail);
> + /* If current function has musttail calls, we can't disable tail
> + calls altogether for the whole caller, because those might be
> + actually fine. So just punt if this exact call is not
> + a tail recursion. */
> + if (cfun->has_musttail)
> + only_tailr = true;
> + else
> + opt_tailcalls = false;
> + }
> +
> /* If the LHS of our call is not just a simple register or local
> variable, we can't transform this into a tail or sibling call.
> This situation happens, in (e.g.) "*p = foo()" where foo returns a
> @@ -794,6 +800,9 @@ find_tail_calls (basic_block bb, struct
> tail_recursion = true;
> }
>
> + if (only_tailr && !tail_recursion)
> + return;
> +
> /* Compute live vars if not computed yet. */
> if (live_vars == NULL)
> {
> --- gcc/testsuite/c-c++-common/pr119616.c.jj 2025-04-04 09:49:03.359588201
> +0200
> +++ gcc/testsuite/c-c++-common/pr119616.c 2025-04-04 09:48:35.488592628
> +0200
> @@ -0,0 +1,23 @@
> +/* PR tree-optimization/119616 */
> +/* { dg-do compile { target external_musttail } } */
> +/* { dg-options "-O2" } */
> +
> +int foo (int *);
> +int bar (int);
> +
> +int
> +baz (int x)
> +{
> + if (!x)
> + [[gnu::musttail]] return bar (x);
> + return foo (&x);
> +}
> +
> +int
> +qux (int x)
> +{
> + if (!x)
> + [[gnu::musttail]] return bar (x);
> + foo (&x);
> + return 1;
> +}
>
> Jakub
>
