* [PATCH] tree-optimization/109524 - ICE with VRP edge removal
@ 2023-04-17 8:41 Richard Biener
0 siblings, 0 replies; 2+ messages in thread
From: Richard Biener @ 2023-04-17 8:41 UTC (permalink / raw)
To: gcc-patches; +Cc: Jakub Jelinek
VRP queues edges to process late for updating global ranges for
__builtin_unreachable. But this interferes with edge removal
from substitute_and_fold. The following deals with this by
looking up the edge with source/dest block indices which do not
become stale.
For GCC 14 we probably want to refactor substitute_and_fold but
that doesn't seem appropriate at this stage.
Bootstrapped and tested on x86_64-unknown-linux-gnu, OK?
Thanks,
Richard.
PR tree-optimization/109524
* tree-vrp.cc (remove_unreachable::m_list): Change to a
vector of pairs of block indices.
(remove_unreachable::maybe_register_block): Adjust.
(remove_unreachable::remove_and_update_globals): Likewise.
Deal with removed blocks.
* g++.dg/pr109524.C: New testcase.
---
gcc/testsuite/g++.dg/pr109524.C | 41 +++++++++++++++++++++++++++++++++
gcc/tree-vrp.cc | 13 +++++++----
2 files changed, 50 insertions(+), 4 deletions(-)
create mode 100644 gcc/testsuite/g++.dg/pr109524.C
diff --git a/gcc/testsuite/g++.dg/pr109524.C b/gcc/testsuite/g++.dg/pr109524.C
new file mode 100644
index 00000000000..a560839cd5b
--- /dev/null
+++ b/gcc/testsuite/g++.dg/pr109524.C
@@ -0,0 +1,41 @@
+// { dg-do compile }
+// { dg-require-effective-target c++11 }
+// { dg-options "-O3 -fno-tree-forwprop -fnon-call-exceptions -fno-tree-ccp -fno-tree-fre" }
+
+struct nn;
+void f();
+int *m;
+struct _Guard {
+ ~_Guard() {
+ if (_M_guarded)
+ __builtin_unreachable();
+ }
+ nn *_M_guarded;
+};
+struct nn {
+ int * _M_dataplus;
+ nn(nn &)
+ {
+ f();
+ _Guard __guard;
+ m = _M_dataplus;
+ }
+ nn(){}
+};
+ void hnn(nn *a)
+ {
+ f();
+ _Guard __guard;
+ m = a->_M_dataplus;
+ }
+bool gg();
+static inline nn
+hh(nn str) {
+ if (gg())
+ return str;
+ __builtin_unreachable();
+}
+void h() {
+
+ hh({});
+}
diff --git a/gcc/tree-vrp.cc b/gcc/tree-vrp.cc
index be7d06f565c..f4d484526c7 100644
--- a/gcc/tree-vrp.cc
+++ b/gcc/tree-vrp.cc
@@ -75,7 +75,7 @@ public:
~remove_unreachable () { m_list.release (); }
void maybe_register_block (basic_block bb);
bool remove_and_update_globals (bool final_p);
- vec<edge> m_list;
+ vec<std::pair<int, int> > m_list;
gimple_ranger &m_ranger;
};
@@ -103,9 +103,9 @@ remove_unreachable::maybe_register_block (basic_block bb)
return;
if (un0)
- m_list.safe_push (e1);
+ m_list.safe_push (std::make_pair (e1->src->index, e1->dest->index));
else
- m_list.safe_push (e0);
+ m_list.safe_push (std::make_pair (e0->src->index, e0->dest->index));
}
// Process the edges in the list, change the conditions and removing any
@@ -132,7 +132,12 @@ remove_unreachable::remove_and_update_globals (bool final_p)
auto_bitmap all_exports;
for (i = 0; i < m_list.length (); i++)
{
- edge e = m_list[i];
+ auto eb = m_list[i];
+ basic_block src = BASIC_BLOCK_FOR_FN (cfun, eb.first);
+ basic_block dest = BASIC_BLOCK_FOR_FN (cfun, eb.second);
+ if (!src || !dest)
+ continue;
+ edge e = find_edge (src, dest);
gimple *s = gimple_outgoing_range_stmt_p (e->src);
gcc_checking_assert (gimple_code (s) == GIMPLE_COND);
bool lhs_p = TREE_CODE (gimple_cond_lhs (s)) == SSA_NAME;
--
2.35.3
^ permalink raw reply [flat|nested] 2+ messages in thread
[parent not found: <39683.123041704413900324@us-mta-593.us.mimecast.lan>]
* Re: [PATCH] tree-optimization/109524 - ICE with VRP edge removal
[not found] <39683.123041704413900324@us-mta-593.us.mimecast.lan>
@ 2023-04-17 8:47 ` Jakub Jelinek
0 siblings, 0 replies; 2+ messages in thread
From: Jakub Jelinek @ 2023-04-17 8:47 UTC (permalink / raw)
To: Richard Biener; +Cc: gcc-patches
On Mon, Apr 17, 2023 at 08:41:38AM +0000, Richard Biener wrote:
> VRP queues edges to process late for updating global ranges for
> __builtin_unreachable. But this interferes with edge removal
> from substitute_and_fold. The following deals with this by
> looking up the edge with source/dest block indices which do not
> become stale.
>
> For GCC 14 we probably want to refactor substitute_and_fold but
> that doesn't seem appropriate at this stage.
>
> Bootstrapped and tested on x86_64-unknown-linux-gnu, OK?
>
> Thanks,
> Richard.
>
> PR tree-optimization/109524
> * tree-vrp.cc (remove_unreachable::m_list): Change to a
> vector of pairs of block indices.
> (remove_unreachable::maybe_register_block): Adjust.
> (remove_unreachable::remove_and_update_globals): Likewise.
> Deal with removed blocks.
>
> * g++.dg/pr109524.C: New testcase.
Ok, thanks.
> --- a/gcc/tree-vrp.cc
> +++ b/gcc/tree-vrp.cc
> @@ -75,7 +75,7 @@ public:
> ~remove_unreachable () { m_list.release (); }
> void maybe_register_block (basic_block bb);
> bool remove_and_update_globals (bool final_p);
> - vec<edge> m_list;
> + vec<std::pair<int, int> > m_list;
We are in C++11, you could use >> instead of > > here I think.
Jakub
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2023-04-17 8:48 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-04-17 8:41 [PATCH] tree-optimization/109524 - ICE with VRP edge removal Richard Biener
[not found] <39683.123041704413900324@us-mta-593.us.mimecast.lan>
2023-04-17 8:47 ` Jakub Jelinek
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).