On Wed, 24 Aug 2022, Jakub Jelinek wrote:
> Hi!
>
> This patch fixes 2 __builtin_unreachable/__builtin_trap related issues.
> One (first hunk) is that CDDCE happily removes calls to .TRAP ()
> internal-fn as useless. The problem is that the internal-fn is
> ECF_CONST | ECF_NORETURN, doesn't have lhs and so DCE thinks it doesn't
> have side-effects and removes it. __builtin_unreachable which has
> the same ECF_* flags works fine, as since PR44485 we implicitly add
> ECF_LOOPING_CONST_OR_PURE to ECF_CONST | ECF_NORETURN builtins, but
> do it in flags_from_decl_or_type which isn't called for internal-fns.
> As IFN_TRAP is the only ifn with such flags, it seems easier to
> add it explicitly.
>
> The other issue (which on the testcase can be seen only with the
> first bug unfixed) is that execute_fixup_cfg can add a __builtin_trap
> which needs vops, but nothing adds it and it can appear in many passes
> which don't have corresponding TODO_update_ssa_only_virtuals etc.
> Fixed similarly as last time but emitting ifn there instead.
>
> Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
I still disagree with this in general - but well, seems the ship
has sailed.
Do we have to repeat this pattern elsewhere? It seems we have
build_builtin_unreachable (not using .IFN_TRAP) but no corresponding
GIMPLE variant? Maybe have builtin_cfn_unreachable () that returns
BUILT_IN_UNREACHABLE or IFN_TRAP and use gimple_build () with that
combined_fn argument? See below
> 2022-08-24 Jakub Jelinek <ja...@redhat.com>
>
> PR tree-optimization/106099
> * internal-fn.def (TRAP): Add ECF_LOOPING_CONST_OR_PURE flag.
> * tree-cfg.cc (execute_fixup_cfg): Add IFN_TRAP instead of
> __builtin_trap to avoid the need of vops.
>
> * gcc.dg/pr106099.c: New test.
>
> --- gcc/internal-fn.def.jj 2022-07-28 12:43:12.876295553 +0200
> +++ gcc/internal-fn.def 2022-08-23 14:21:49.559364691 +0200
> @@ -458,7 +458,8 @@ DEF_INTERNAL_FN (SPACESHIP, ECF_CONST |
>
> /* __builtin_trap created from/for __builtin_unreachable. */
> DEF_INTERNAL_FN (TRAP, ECF_CONST | ECF_LEAF | ECF_NORETURN
> - | ECF_NOTHROW | ECF_COLD, NULL)
> + | ECF_NOTHROW | ECF_COLD | ECF_LOOPING_CONST_OR_PURE,
> + NULL)
>
> #undef DEF_INTERNAL_INT_FN
> #undef DEF_INTERNAL_FLT_FN
> --- gcc/tree-cfg.cc.jj 2022-07-26 10:32:23.998267698 +0200
> +++ gcc/tree-cfg.cc 2022-08-23 12:36:03.818856327 +0200
> @@ -9879,10 +9879,16 @@ execute_fixup_cfg (void)
> if (stmt && is_gimple_call (stmt))
> gimple_call_set_ctrl_altering (stmt, false);
> tree fndecl = builtin_decl_unreachable ();
> - stmt = gimple_build_call (fndecl, 0);
> + if (DECL_FUNCTION_CODE (fndecl) != BUILT_IN_TRAP)
> + stmt = gimple_build_call (fndecl, 0);
> + else
> + /* Instead of __builtin_trap use .TRAP, so that it doesn't
> + need vops. */
> + stmt = gimple_build_call_internal (IFN_TRAP, 0);
> gimple_stmt_iterator gsi = gsi_last_bb (bb);
So
combined_fn cfn = builtin_cfn_unreachable ();
gimple_build (&gsi, false, GSI_NEW_STMT, cfn, void_type_node);
?
> gsi_insert_after (&gsi, stmt, GSI_NEW_STMT);
> - if (!cfun->after_inlining)
> + if (!cfun->after_inlining
> + && DECL_FUNCTION_CODE (fndecl) != BUILT_IN_TRAP)
> {
> gcall *call_stmt = dyn_cast <gcall *> (stmt);
> node->create_edge (cgraph_node::get_create (fndecl),
> --- gcc/testsuite/gcc.dg/pr106099.c.jj 2022-08-23 14:30:51.992057144
> +0200
> +++ gcc/testsuite/gcc.dg/pr106099.c 2022-08-23 14:29:04.271508337 +0200
> @@ -0,0 +1,10 @@
> +/* PR tree-optimization/106099 */
> +/* { dg-do compile } */
> +/* { dg-options "-O -fharden-compares -fno-tree-forwprop -fno-tree-ch
> -fno-tree-dominator-opts -fno-tree-ccp -funreachable-traps
> --param=scev-max-expr-size=1" } */
> +
> +void
> +foo (void)
> +{
> + for (unsigned i = 0; i == 0; i++)
> + __builtin_printf ("%d", i);
> +}
>
> Jakub
>
>
--
Richard Biener <rguent...@suse.de>
SUSE Software Solutions Germany GmbH, Frankenstrasse 146, 90461 Nuernberg,
Germany; GF: Ivo Totev, Andrew Myers, Andrew McDonald, Boudien Moerman;
HRB 36809 (AG Nuernberg)
