public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
* [PATCH] lto/106334 - relax assert during WPA tree merging
@ 2022-07-19  8:30 Richard Biener
  0 siblings, 0 replies; 3+ messages in thread
From: Richard Biener @ 2022-07-19  8:30 UTC (permalink / raw)
  To: gcc-patches

The dwarf2out map of tree to symbol + offset is populated too early
when streaming in trees so that when WPA tree merging decides to
recycle them the mapping prevails and if we are unlucky the same
address is used for another tree with a symbol + offset DIE to
record.  The following mitigates the resulting ICE by relaxing the
assert, allowing re-use of a slot during WPA.  Delaying the register
would be better but it's already somewhat hairy and uglifying this
further doesn't look too important right now.

Bootstrap / regtest running on x86_64-unknown-linux-gnu.

	* dwarf2out.cc (dwarf2out_register_external_die): Allow
	map entry re-use during WPA.
---
 gcc/dwarf2out.cc | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/gcc/dwarf2out.cc b/gcc/dwarf2out.cc
index e3920c898f5..3ac39c1a5b0 100644
--- a/gcc/dwarf2out.cc
+++ b/gcc/dwarf2out.cc
@@ -6069,7 +6069,11 @@ dwarf2out_register_external_die (tree decl, const char *sym,
 
   if (!external_die_map)
     external_die_map = hash_map<tree, sym_off_pair>::create_ggc (1000);
-  gcc_checking_assert (!external_die_map->get (decl));
+  /* When we do tree merging during WPA we can end up re-using GC memory
+     as there's currently no way to unregister external DIEs.  Ideally
+     we'd register them only after merging finished but allowing override
+     here is easiest.  See PR106334.  */
+  gcc_checking_assert (flag_wpa || !external_die_map->get (decl));
   sym_off_pair p = { IDENTIFIER_POINTER (get_identifier (sym)), off };
   external_die_map->put (decl, p);
 }
-- 
2.35.3

^ permalink raw reply	[flat|nested] 3+ messages in thread

* [PATCH] lto/106334 - relax assert during WPA tree merging
@ 2022-07-19  8:30 Richard Biener
  0 siblings, 0 replies; 3+ messages in thread
From: Richard Biener @ 2022-07-19  8:30 UTC (permalink / raw)
  To: gcc-patches

The dwarf2out map of tree to symbol + offset is populated too early
when streaming in trees so that when WPA tree merging decides to
recycle them the mapping prevails and if we are unlucky the same
address is used for another tree with a symbol + offset DIE to
record.  The following mitigates the resulting ICE by relaxing the
assert, allowing re-use of a slot during WPA.  Delaying the register
would be better but it's already somewhat hairy and uglifying this
further doesn't look too important right now.

Bootstrap / regtest running on x86_64-unknown-linux-gnu.

	* dwarf2out.cc (dwarf2out_register_external_die): Allow
	map entry re-use during WPA.
---
 gcc/dwarf2out.cc | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/gcc/dwarf2out.cc b/gcc/dwarf2out.cc
index e3920c898f5..3ac39c1a5b0 100644
--- a/gcc/dwarf2out.cc
+++ b/gcc/dwarf2out.cc
@@ -6069,7 +6069,11 @@ dwarf2out_register_external_die (tree decl, const char *sym,
 
   if (!external_die_map)
     external_die_map = hash_map<tree, sym_off_pair>::create_ggc (1000);
-  gcc_checking_assert (!external_die_map->get (decl));
+  /* When we do tree merging during WPA we can end up re-using GC memory
+     as there's currently no way to unregister external DIEs.  Ideally
+     we'd register them only after merging finished but allowing override
+     here is easiest.  See PR106334.  */
+  gcc_checking_assert (flag_wpa || !external_die_map->get (decl));
   sym_off_pair p = { IDENTIFIER_POINTER (get_identifier (sym)), off };
   external_die_map->put (decl, p);
 }
-- 
2.35.3

^ permalink raw reply	[flat|nested] 3+ messages in thread

* [PATCH] lto/106334 - relax assert during WPA tree merging
@ 2022-07-19  8:30 Richard Biener
  0 siblings, 0 replies; 3+ messages in thread
From: Richard Biener @ 2022-07-19  8:30 UTC (permalink / raw)
  To: gcc-patches

The dwarf2out map of tree to symbol + offset is populated too early
when streaming in trees so that when WPA tree merging decides to
recycle them the mapping prevails and if we are unlucky the same
address is used for another tree with a symbol + offset DIE to
record.  The following mitigates the resulting ICE by relaxing the
assert, allowing re-use of a slot during WPA.  Delaying the register
would be better but it's already somewhat hairy and uglifying this
further doesn't look too important right now.

Bootstrap / regtest running on x86_64-unknown-linux-gnu.

	* dwarf2out.cc (dwarf2out_register_external_die): Allow
	map entry re-use during WPA.
---
 gcc/dwarf2out.cc | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/gcc/dwarf2out.cc b/gcc/dwarf2out.cc
index e3920c898f5..3ac39c1a5b0 100644
--- a/gcc/dwarf2out.cc
+++ b/gcc/dwarf2out.cc
@@ -6069,7 +6069,11 @@ dwarf2out_register_external_die (tree decl, const char *sym,
 
   if (!external_die_map)
     external_die_map = hash_map<tree, sym_off_pair>::create_ggc (1000);
-  gcc_checking_assert (!external_die_map->get (decl));
+  /* When we do tree merging during WPA we can end up re-using GC memory
+     as there's currently no way to unregister external DIEs.  Ideally
+     we'd register them only after merging finished but allowing override
+     here is easiest.  See PR106334.  */
+  gcc_checking_assert (flag_wpa || !external_die_map->get (decl));
   sym_off_pair p = { IDENTIFIER_POINTER (get_identifier (sym)), off };
   external_die_map->put (decl, p);
 }
-- 
2.35.3

^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2022-07-19  8:30 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-07-19  8:30 [PATCH] lto/106334 - relax assert during WPA tree merging Richard Biener
2022-07-19  8:30 Richard Biener
2022-07-19  8:30 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).