public inbox for gcc-cvs@sourceware.org
help / color / mirror / Atom feed
* [gcc r13-3161] tree-optimization/107153 - autopar SSA update issue
@ 2022-10-07 12:27 Richard Biener
  0 siblings, 0 replies; only message in thread
From: Richard Biener @ 2022-10-07 12:27 UTC (permalink / raw)
  To: gcc-cvs

https://gcc.gnu.org/g:89228e3985c5cdf6be58a3b5b1afcad91e9e3422

commit r13-3161-g89228e3985c5cdf6be58a3b5b1afcad91e9e3422
Author: Richard Biener <rguenther@suse.de>
Date:   Fri Oct 7 10:28:56 2022 +0200

    tree-optimization/107153 - autopar SSA update issue
    
    autopar performs insertion of stores, eventually requiring a
    virtual loop PHI and assorted LC PHI adjustments which we intend
    to do once after the pass finishes.  But we also perform intermediate
    update_ssa after loop duplication which can lose this fact.  The
    following forces renaming of the virtual operand before the final
    SSA update to fix that.  It also removes the explicit update_ssa
    call from the gimple_duplicate_sese_tail utility as has been done
    for all other such utilities and instead performs the SSA update
    from autopar.
    
            PR tree-optimization/107153
            * tree-cfg.cc (gimple_duplicate_sese_tail): Do not update
            SSA form here.
            * tree-parloops.cc (gen_parallel_loop): Update SSA form
            after to-exit-first transform, no PHI insertion is necessary.
            (pass_parallelize_loops::execute): Force re-write of the
            virtual operand SSA web.
    
            * gcc.dg/autopar/pr107153.c: New testcase.

Diff:
---
 gcc/testsuite/gcc.dg/autopar/pr107153.c | 17 +++++++++++++++++
 gcc/tree-cfg.cc                         |  2 --
 gcc/tree-parloops.cc                    |  5 +++++
 3 files changed, 22 insertions(+), 2 deletions(-)

diff --git a/gcc/testsuite/gcc.dg/autopar/pr107153.c b/gcc/testsuite/gcc.dg/autopar/pr107153.c
new file mode 100644
index 00000000000..2391a674d63
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/autopar/pr107153.c
@@ -0,0 +1,17 @@
+/* { dg-do compile } */
+/* { dg-options "-O1 -floop-parallelize-all -ftree-parallelize-loops=2 -fno-tree-dominator-opts" } */
+
+void
+foo (int x, int y)
+{
+  int i;
+
+  for (i = 0; i < 2; i++)
+    {
+    }
+
+  while (x)
+    if (!y)
+      while (y)
+        ++y;
+}
diff --git a/gcc/tree-cfg.cc b/gcc/tree-cfg.cc
index 09fa7b64584..42f40f17c56 100644
--- a/gcc/tree-cfg.cc
+++ b/gcc/tree-cfg.cc
@@ -6926,8 +6926,6 @@ gimple_duplicate_sese_tail (edge entry, edge exit,
   /* Anything that is outside of the region, but was dominated by something
      inside needs to update dominance info.  */
   iterate_fix_dominators (CDI_DOMINATORS, doms, false);
-  /* Update the SSA web.  */
-  update_ssa (TODO_update_ssa);
 
   if (free_region_copy)
     free (region_copy);
diff --git a/gcc/tree-parloops.cc b/gcc/tree-parloops.cc
index e4a148b5365..e680d97dd04 100644
--- a/gcc/tree-parloops.cc
+++ b/gcc/tree-parloops.cc
@@ -3131,6 +3131,7 @@ gen_parallel_loop (class loop *loop,
 	 to the exit of the loop.  */
       transform_to_exit_first_loop (loop, reduction_list, nit);
     }
+  update_ssa (TODO_update_ssa_no_phi);
 
   /* Generate initializations for reductions.  */
   if (!reduction_list->is_empty ())
@@ -4215,6 +4216,10 @@ pass_parallelize_loops::execute (function *fun)
 
       checking_verify_loop_structure ();
 
+      /* ???  Intermediate SSA updates with no PHIs might have lost
+	 the virtual operand renaming needed by separate_decls_in_region,
+	 make sure to rename them again.  */
+      mark_virtual_operands_for_renaming (fun);
       update_ssa (TODO_update_ssa);
       if (in_loop_pipeline)
 	rewrite_into_loop_closed_ssa (NULL, 0);

^ permalink raw reply	[flat|nested] only message in thread

only message in thread, other threads:[~2022-10-07 12:27 UTC | newest]

Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-10-07 12:27 [gcc r13-3161] tree-optimization/107153 - autopar SSA update issue Richard Biener

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).