Re: [PATCH] ipa: Self-DCE of uses of removed call LHSs (PR 108007)

[Jan Hubicka <hubicka@ucw.cz>]

> >> PR 108007 is another manifestation where we rely on DCE to clean-up
> >> after IPA-SRA and if the user explicitely switches DCE off, IPA-SRA
> >> can leave behind statements which are fed uninitialized values and
> >> trap, even though their results are themselves never used.
> >>
> >> I have already fixed this for unused parameters in callees, this bug
> >> shows that almost the same thing can happen for removed returns, on
> >> the side of callers.  This means that the issue has to be fixed
> >> elsewhere, in call redirection.  This patch adds a function which
> >> recursivewly looks for uses of operations fed specific SSA names and
> >> removes them all.
> >>
> >> That would have been easy if it wasn't for debug statements during
> >> tree-inline (from which call redirection is also invoked).  Debug
> >> statements are decoupled from the rest at this point and iterating
> >> over uses of SSAs does not bring them up.  During tree-inline they are
> >> handled especially at the end, I assume in order to make sure that
> >> relative ordering of UIDs are the same with and without debug info.
> >>
> >> This means that during tree-inline we need to make a hash of killed
> >> SSAs, that we already have in copy_body_data, available to the
> >> function making the purging.  So the patch duly does also that, making
> >> the interface slightly ugly.
> >>
> >> Bootstrapped and tested on x86_64-linux.  OK for master?  (I am not sure
> >> the problem is grave enough to warrant backporting to release branches
> >> but can do that as well if people think I should.)
> >>
> >> Thanks,
> >>
> >> Martin
> >>
> >>
> >> gcc/ChangeLog:
> >>
> >> 2023-05-11  Martin Jambor  <mjambor@suse.cz>
> >>
> >> 	PR ipa/108007
> >> 	* cgraph.h (cgraph_edge): Add a parameter to
> >> 	redirect_call_stmt_to_callee.
> >> 	* ipa-param-manipulation.h (ipa_param_adjustments): Added a
> >> 	parameter to modify_call.
> >> 	* cgraph.cc (cgraph_edge::redirect_call_stmt_to_callee): New
> >> 	parameter killed_ssas, pass it to padjs->modify_call.
> >> 	* ipa-param-manipulation.cc (purge_transitive_uses): New function.
> >> 	(ipa_param_adjustments::modify_call): New parameter killed_ssas.
> >> 	Instead of substitutin uses, invoke purge_transitive_uses.  If
> >> 	hash of killed SSAs has not been provided, create a temporary one
> >> 	and release SSAs that have been added to it.
> >> 	* tree-inline.cc (redirect_all_calls): Create
> >> 	id->killed_new_ssa_names earlier, pass it to edge redirection,
> >> 	adjust a comment.
> >> 	(copy_body): Release SSAs in id->killed_new_ssa_names.
> >>
> >> gcc/testsuite/ChangeLog:
> >>
> >> 2023-05-11  Martin Jambor  <mjambor@suse.cz>
> >>
> >> 	PR ipa/108007
> >> 	* gcc.dg/ipa/pr108007.c: New test.
> >> ---
> >>  gcc/cgraph.cc                       | 10 +++-
> >>  gcc/cgraph.h                        |  9 ++-
> >>  gcc/ipa-param-manipulation.cc       | 85 +++++++++++++++++++++--------
> >>  gcc/ipa-param-manipulation.h        |  3 +-
> >>  gcc/testsuite/gcc.dg/ipa/pr108007.c | 32 +++++++++++
> >>  gcc/tree-inline.cc                  | 28 ++++++----
> >>  6 files changed, 129 insertions(+), 38 deletions(-)
> >>  create mode 100644 gcc/testsuite/gcc.dg/ipa/pr108007.c
> >>
> >> +/* Remove all statements that use NAME and transitively those that use the
> >> +   result of such statements.  KILLED_SSAS contains the SSA_NAMEs that are
> >> +   already being or have been processed and new ones need to be added to it.
> >> +   The funtction only has to process situations handled by
> >> +   ssa_name_only_returned_p in ipa-sra.cc with the exception that it can assume
> >> +   it must never reach a use in a return statement.  */
> >> +
> >> +static void
> >> +purge_transitive_uses (tree name, hash_set <tree> *killed_ssas)
> >> +{
> >> +  imm_use_iterator imm_iter;
> >> +  gimple *stmt;
> >> +
> >> +  FOR_EACH_IMM_USE_STMT (stmt, imm_iter, name)
> >> +    {
> >> +      if (gimple_debug_bind_p (stmt))
> >> +	{
> >> +	  /* When runing within tree-inline, we will never end up here but
> >> +	     adding the SSAs to killed_ssas will do the trick in this case and
> >> +	     the respective debug statements will get reset. */
> >> +
> >> +	  gimple_debug_bind_reset_value (stmt);
> >> +	  update_stmt (stmt);
> >> +	  continue;
> >> +	}
> >> +
> >> +      tree lhs = NULL_TREE;
> >> +      if (is_gimple_assign (stmt))
> >> +	lhs = gimple_assign_lhs (stmt);
> >> +      else if (gimple_code (stmt) == GIMPLE_PHI)
> >> +	lhs = gimple_phi_result (stmt);
> >> +      gcc_assert (lhs
> >> +		  && (TREE_CODE (lhs) == SSA_NAME)
> >> +		  && !gimple_vdef (stmt));
> >> +
> >> +      if (!killed_ssas->contains (lhs))
> >> +	{
> >> +	  killed_ssas->add (lhs);
> >> +	  purge_transitive_uses (lhs, killed_ssas);

SSA graph may be deep so this may cause stack overflow, so I think we
should use worklist here (it is also easy to do).

OK with that change.
Honza