diff mbox

Improve verification of loop->latch in verify_loop_structure

Message ID 56547D49.5040609@mentor.com
State New
Headers show

Commit Message

Tom de Vries Nov. 24, 2015, 3:07 p.m. UTC 
[ was: Re: [PATCH] Check NULL loop->latch in verify_loop_structure ]

On 23/11/15 11:28, Richard Biener wrote:
> On Mon, 23 Nov 2015, Tom de Vries wrote:
>
>> Hi,
>>
>> In verify_loop_structure, we stop checking the latch once we find that it's
>> NULL.
>>
>> This patch tries a bit harder:
>> - if !LOOPS_MAY_HAVE_MULTIPLE_LATCHES, we don't allow a NULL latch
>> - if LOOPS_MAY_HAVE_MULTIPLE_LATCHES, we check that indeed there's no
>>    single loop latch.
>>
>> As a consequence of adding this check, I needed to fix expand_omp_for_generic,
>> which missed an initialization of a loop latch.
>>
>> Bootstrapped and reg-tested on x86_64.
>>
>> OK for stage3 trunk?
>
> You miss to catch the case where loop->latch is non-NULL but there
> are multiple latches, so I think the patch can be improved.
> That case is more important for correctness (passes
> seeing ->latch non-NULL assume a single latch).
>

Updated according to comment.

Bootstrapped and reg-tested on x86_64.

OK for stage3 trunk?

Thanks,
- Tom

Comments

Richard Biener Nov. 25, 2015, 10:44 a.m. UTC | #1 
On Tue, 24 Nov 2015, Tom de Vries wrote:

> [ was: Re: [PATCH] Check NULL loop->latch in verify_loop_structure ]
> 
> On 23/11/15 11:28, Richard Biener wrote:
> > On Mon, 23 Nov 2015, Tom de Vries wrote:
> > 
> > > Hi,
> > > 
> > > In verify_loop_structure, we stop checking the latch once we find that
> > > it's
> > > NULL.
> > > 
> > > This patch tries a bit harder:
> > > - if !LOOPS_MAY_HAVE_MULTIPLE_LATCHES, we don't allow a NULL latch
> > > - if LOOPS_MAY_HAVE_MULTIPLE_LATCHES, we check that indeed there's no
> > >    single loop latch.
> > > 
> > > As a consequence of adding this check, I needed to fix
> > > expand_omp_for_generic,
> > > which missed an initialization of a loop latch.
> > > 
> > > Bootstrapped and reg-tested on x86_64.
> > > 
> > > OK for stage3 trunk?
> > 
> > You miss to catch the case where loop->latch is non-NULL but there
> > are multiple latches, so I think the patch can be improved.
> > That case is more important for correctness (passes
> > seeing ->latch non-NULL assume a single latch).
> > 
> 
> Updated according to comment.
> 
> Bootstrapped and reg-tested on x86_64.
> 
> OK for stage3 trunk?

Ok.

Thanks,
Richard.
Alan Lawrence Nov. 26, 2015, 11:33 a.m. UTC | #2 
This caused an ICE compiling value.c from gdb on
aarch64-none-linux-gnu; the testcase, after preprocessing on aarch64,
ICEs on both aarch64 and x86_64, but is about 1MB - I'm working on
reducing that down to something small enough to post...

$ ./gcc/xgcc -B ./gcc -O2 -g value.c
../../binutils-gdb/gdb/value.c: In function ‘show_convenience’:
../../binutils-gdb/gdb/value.c:2615:1: error: loop 3’s latch is missing
../../binutils-gdb/gdb/value.c:2615:1: internal compiler error: in
verify_loop_structure, at cfgloop.c:1669
0x71e653 verify_loop_structure()
        /work/alalaw01/src2/gcc/gcc/cfgloop.c:1669
0x97c6ae checking_verify_loop_structure
        /work/alalaw01/src2/gcc/gcc/cfgloop.h:325
0x97c6ae loop_optimizer_init(unsigned int)
        /work/alalaw01/src2/gcc/gcc/loop-init.c:106
0x97c78a rtl_loop_init
        /work/alalaw01/src2/gcc/gcc/loop-init.c:398
0x97c78a execute
        /work/alalaw01/src2/gcc/gcc/loop-init.c:425

--Alan
Alan Lawrence Nov. 26, 2015, 11:54 a.m. UTC | #3 
Here's a reduced testcase (reduced to the point of generating lots of
warnings, I'm compiling with -O2 -w, on x86_64):

struct __jmp_buf_tag
  {
  };
typedef struct __jmp_buf_tag sigjmp_buf[1];
extern struct cmd_list_element *showlist;
struct internalvar
{
  struct internalvar *next;
};
static struct internalvar *internalvars;
struct internalvar *
create_internalvar (const char *name)
{
  struct internalvar *var = ((struct internalvar *) xmalloc (sizeof
(struct internalvar)));
  internalvars = var;
}

void
show_convenience (char *ignore, int from_tty)
{
  struct gdbarch *gdbarch = get_current_arch ();
  int varseen = 0;
  for (struct internalvar *var = internalvars; var; var = var->next)
    {
      if (!varseen)
        varseen = 1;
      sigjmp_buf *buf = exceptions_state_mc_init ();
      __sigsetjmp ();
      while (exceptions_state_mc_action_iter ())
        while (exceptions_state_mc_action_iter_1 ())
          ;
    }
  if (!varseen)
      printf_unfiltered ();
}

On 26 November 2015 at 11:33, Alan Lawrence <alan.lawrence@arm.com> wrote:
> This caused an ICE compiling value.c from gdb on
> aarch64-none-linux-gnu; the testcase, after preprocessing on aarch64,
> ICEs on both aarch64 and x86_64, but is about 1MB - I'm working on
> reducing that down to something small enough to post...
>
> $ ./gcc/xgcc -B ./gcc -O2 -g value.c
> ../../binutils-gdb/gdb/value.c: In function ‘show_convenience’:
> ../../binutils-gdb/gdb/value.c:2615:1: error: loop 3’s latch is missing
> ../../binutils-gdb/gdb/value.c:2615:1: internal compiler error: in
> verify_loop_structure, at cfgloop.c:1669
> 0x71e653 verify_loop_structure()
>         /work/alalaw01/src2/gcc/gcc/cfgloop.c:1669
> 0x97c6ae checking_verify_loop_structure
>         /work/alalaw01/src2/gcc/gcc/cfgloop.h:325
> 0x97c6ae loop_optimizer_init(unsigned int)
>         /work/alalaw01/src2/gcc/gcc/loop-init.c:106
> 0x97c78a rtl_loop_init
>         /work/alalaw01/src2/gcc/gcc/loop-init.c:398
> 0x97c78a execute
>         /work/alalaw01/src2/gcc/gcc/loop-init.c:425
>
> --Alan
Richard Biener Nov. 26, 2015, 12:15 p.m. UTC | #4 
On Thu, 26 Nov 2015, Alan Lawrence wrote:

> This caused an ICE compiling value.c from gdb on
> aarch64-none-linux-gnu; the testcase, after preprocessing on aarch64,
> ICEs on both aarch64 and x86_64, but is about 1MB - I'm working on
> reducing that down to something small enough to post...
> 
> $ ./gcc/xgcc -B ./gcc -O2 -g value.c
> ../../binutils-gdb/gdb/value.c: In function ‘show_convenience’:
> ../../binutils-gdb/gdb/value.c:2615:1: error: loop 3’s latch is missing
> ../../binutils-gdb/gdb/value.c:2615:1: internal compiler error: in
> verify_loop_structure, at cfgloop.c:1669
> 0x71e653 verify_loop_structure()
>         /work/alalaw01/src2/gcc/gcc/cfgloop.c:1669
> 0x97c6ae checking_verify_loop_structure
>         /work/alalaw01/src2/gcc/gcc/cfgloop.h:325
> 0x97c6ae loop_optimizer_init(unsigned int)
>         /work/alalaw01/src2/gcc/gcc/loop-init.c:106
> 0x97c78a rtl_loop_init
>         /work/alalaw01/src2/gcc/gcc/loop-init.c:398
> 0x97c78a execute
>         /work/alalaw01/src2/gcc/gcc/loop-init.c:425

See also PR68549 for why I think this happens "by design".  Thus
I think we need to revert the checking when 
LOOPS_MAY_HAVE_MULTIPLE_LATCHES for now.

Richard.
Tom de Vries Nov. 26, 2015, 2:35 p.m. UTC | #5 
On 26/11/15 13:15, Richard Biener wrote:
> On Thu, 26 Nov 2015, Alan Lawrence wrote:
>
>> This caused an ICE compiling value.c from gdb on
>> aarch64-none-linux-gnu; the testcase, after preprocessing on aarch64,
>> ICEs on both aarch64 and x86_64, but is about 1MB - I'm working on
>> reducing that down to something small enough to post...
>>
>> $ ./gcc/xgcc -B ./gcc -O2 -g value.c
>> ../../binutils-gdb/gdb/value.c: In function ‘show_convenience’:
>> ../../binutils-gdb/gdb/value.c:2615:1: error: loop 3’s latch is missing
>> ../../binutils-gdb/gdb/value.c:2615:1: internal compiler error: in
>> verify_loop_structure, at cfgloop.c:1669
>> 0x71e653 verify_loop_structure()
>>          /work/alalaw01/src2/gcc/gcc/cfgloop.c:1669
>> 0x97c6ae checking_verify_loop_structure
>>          /work/alalaw01/src2/gcc/gcc/cfgloop.h:325
>> 0x97c6ae loop_optimizer_init(unsigned int)
>>          /work/alalaw01/src2/gcc/gcc/loop-init.c:106
>> 0x97c78a rtl_loop_init
>>          /work/alalaw01/src2/gcc/gcc/loop-init.c:398
>> 0x97c78a execute
>>          /work/alalaw01/src2/gcc/gcc/loop-init.c:425
>
> See also PR68549 for why I think this happens "by design".  Thus
> I think we need to revert the checking when
> LOOPS_MAY_HAVE_MULTIPLE_LATCHES for now.

I'll revert the whole patch for now.

Thanks,
- Tom
diff mbox

Patch

Improve verification of loop->latch in verify_loop_structure

2015-11-22  Tom de Vries  <tom@codesourcery.com>

	* cfgloop.c (find_single_latch): New function, factored out of ...
	(flow_loops_find): ... here.
	(verify_loop_structure): Improve verification of loop->latch.
	* cfgloop.h (find_single_latch): Declare.
	* omp-low.c (expand_omp_for_generic): Initialize latch of orig_loop.

---
 gcc/cfgloop.c | 70 ++++++++++++++++++++++++++++++++++++++++++-----------------
 gcc/cfgloop.h |  1 +
 gcc/omp-low.c |  1 +
 3 files changed, 52 insertions(+), 20 deletions(-)

diff --git a/gcc/cfgloop.c b/gcc/cfgloop.c
index bf00e0e..a6850e3 100644
--- a/gcc/cfgloop.c
+++ b/gcc/cfgloop.c
@@ -388,6 +388,33 @@  bb_loop_header_p (basic_block header)
   return false;
 }
 
+/* Return the latch block for this header block, if it has just a single one.
+   Otherwise, return NULL.  */
+
+basic_block
+find_single_latch (struct loop* loop)
+{
+  basic_block header = loop->header;
+  edge_iterator ei;
+  edge e;
+  basic_block latch = NULL;
+
+  FOR_EACH_EDGE (e, ei, header->preds)
+    {
+      basic_block cand = e->src;
+      if (!flow_bb_inside_loop_p (loop, cand))
+	continue;
+
+      if (latch != NULL)
+	/* More than one latch edge.  */
+	return NULL;
+
+      latch = cand;
+    }
+
+  return latch;
+}
+
 /* Find all the natural loops in the function and save in LOOPS structure and
    recalculate loop_father information in basic block structures.
    If LOOPS is non-NULL then the loop structures for already recorded loops
@@ -482,29 +509,10 @@  flow_loops_find (struct loops *loops)
     {
       struct loop *loop = larray[i];
       basic_block header = loop->header;
-      edge_iterator ei;
-      edge e;
 
       flow_loop_tree_node_add (header->loop_father, loop);
       loop->num_nodes = flow_loop_nodes_find (loop->header, loop);
-
-      /* Look for the latch for this header block, if it has just a
-	 single one.  */
-      FOR_EACH_EDGE (e, ei, header->preds)
-	{
-	  basic_block latch = e->src;
-
-	  if (flow_bb_inside_loop_p (loop, latch))
-	    {
-	      if (loop->latch != NULL)
-		{
-		  /* More than one latch edge.  */
-		  loop->latch = NULL;
-		  break;
-		}
-	      loop->latch = latch;
-	    }
-	}
+      loop->latch = find_single_latch (loop);
     }
 
   return loops;
@@ -1439,6 +1447,28 @@  verify_loop_structure (void)
 	      error ("loop %d%'s latch is not dominated by its header", i);
 	      err = 1;
 	    }
+	  if (find_single_latch (loop) == NULL)
+	    {
+	      error ("loop %d%'s latch is is not the only latch", i);
+	      err = 1;
+	    }
+	}
+      else
+	{
+	  if (loops_state_satisfies_p (LOOPS_MAY_HAVE_MULTIPLE_LATCHES))
+	    {
+	      if (find_single_latch (loop) != NULL)
+		{
+		  error ("loop %d%'s latch is missing", i);
+		  err = 1;
+		}
+	    }
+	  else
+	    {
+	      error ("loop %d%'s latch is missing, and loops may not have"
+		     " multiple latches", i);
+	      err = 1;
+	    }
 	}
       if (loops_state_satisfies_p (LOOPS_HAVE_SIMPLE_LATCHES))
 	{
diff --git a/gcc/cfgloop.h b/gcc/cfgloop.h
index ee73bf9..7faf591 100644
--- a/gcc/cfgloop.h
+++ b/gcc/cfgloop.h
@@ -270,6 +270,7 @@  bool mark_irreducible_loops (void);
 void release_recorded_exits (function *);
 void record_loop_exits (void);
 void rescan_loop_exit (edge, bool, bool);
+basic_block find_single_latch (struct loop*);
 
 /* Loop data structure manipulation/querying.  */
 extern void flow_loop_tree_node_add (struct loop *, struct loop *);
diff --git a/gcc/omp-low.c b/gcc/omp-low.c
index 0d4c6e5..2d782eb 100644
--- a/gcc/omp-low.c
+++ b/gcc/omp-low.c
@@ -8903,6 +8903,7 @@  expand_omp_for_generic (struct omp_region *region,
 	  orig_loop->header = l1_bb;
 	  /* The loop may have multiple latches.  */
 	  add_loop (orig_loop, new_loop);
+	  orig_loop->latch = find_single_latch (orig_loop);
 	}
     }
 }