Re: [PATCH] Skip ubsan internal fns in tail-merge

[Tom de Vries <Tom_deVries@mentor.com>]

[-- Attachment #1: Type: text/plain, Size: 1714 bytes --]

On 07/03/16 12:32, Jakub Jelinek wrote:
> On Mon, Mar 07, 2016 at 12:26:10PM +0100, Tom de Vries wrote:
>> OK for stage4 trunk if bootstrap and reg-test succeeds?
>
> Ok, with a minor nits.
>
>> 2016-03-07  Tom de Vries  <tom@codesourcery.com>
>>
>> 	PR tree-optimization/70116
>> 	* tree-ssa-tail-merge.c (merge_stmts_p): New function, factored out
>> 	of ...
>> 	(find_duplicate): ... here.  Move merge_stmts_p test to after
>> 	gimple_equal_p test.
>> 	(merge_stmts_p): Return false for ubsan/asan internal functions with
>> 	different gimple_location.
>
> Listing the same function twice is weird, and for a single line short test
> copied from elsewhere, especially if it originally was used twice and now
> just once, I think it is not worth mentioning.  I'd just write:
> (merge_stmts_p): New function.
> (find_duplicate): Use it.  Don't test is_tm_ending here.
>

The only problem I have with that is that the ChangeLog entry doesn't 
mention ubsan/usan internal fns. I've resolved that by elaborating on 
the new function.

>> +	case IFN_UBSAN_NULL:
>> +	case IFN_UBSAN_BOUNDS:
>> +	case IFN_UBSAN_VPTR:
>> +	case IFN_UBSAN_CHECK_ADD:
>> +	case IFN_UBSAN_CHECK_SUB:
>> +	case IFN_UBSAN_CHECK_MUL:
>> +	case IFN_UBSAN_OBJECT_SIZE:
>> +	case IFN_ASAN_CHECK:
>
> This deserves a comment, you should mention what has been mentioned in the
> thread, that for these internal functions gimple_location is like another
> artificial parameter, or explain what it would mean if it attempted to merge
> these stmts with different locations.
>> +	  return gimple_location (stmt1) == gimple_location (stmt2);

Done.

[ And I fixed a missing closing brace error. ]

Currently bootstrapping/reg-testing.

Thanks,
- Tom


[-- Attachment #2: 0001-Skip-ubsan-asan-internal-fns-with-different-location-in-tail-merge.patch --]
[-- Type: text/x-patch, Size: 2534 bytes --]

Skip ubsan/asan internal fns with different location in tail-merge

2016-03-07  Tom de Vries  <tom@codesourcery.com>

	PR tree-optimization/70116
	* tree-ssa-tail-merge.c	(merge_stmts_p): New function, handling
	is_tm_ending stmts and ubsan/asan internal functions.
	(find_duplicate): Use it.  Don't test is_tm_ending here.

---
 gcc/tree-ssa-tail-merge.c | 44 ++++++++++++++++++++++++++++++++++++++------
 1 file changed, 38 insertions(+), 6 deletions(-)

diff --git a/gcc/tree-ssa-tail-merge.c b/gcc/tree-ssa-tail-merge.c
index 5d32790..e95879f 100644
--- a/gcc/tree-ssa-tail-merge.c
+++ b/gcc/tree-ssa-tail-merge.c
@@ -1207,6 +1207,42 @@ gsi_advance_bw_nondebug_nonlocal (gimple_stmt_iterator *gsi, tree *vuse,
     }
 }
 
+/* Return true if equal (in the sense of gimple_equal_p) statements STMT1 and
+   STMT2 are allowed to be merged.  */
+
+static bool
+merge_stmts_p (gimple *stmt1, gimple *stmt2)
+{
+  /* What could be better than this here is to blacklist the bb
+     containing the stmt, when encountering the stmt f.i. in
+     same_succ_hash.  */
+  if (is_tm_ending (stmt1))
+    return false;
+
+  if (is_gimple_call (stmt1)
+      && gimple_call_internal_p (stmt1))
+    switch (gimple_call_internal_fn (stmt1))
+      {
+      case IFN_UBSAN_NULL:
+      case IFN_UBSAN_BOUNDS:
+      case IFN_UBSAN_VPTR:
+      case IFN_UBSAN_CHECK_ADD:
+      case IFN_UBSAN_CHECK_SUB:
+      case IFN_UBSAN_CHECK_MUL:
+      case IFN_UBSAN_OBJECT_SIZE:
+      case IFN_ASAN_CHECK:
+	/* For these internal functions, gimple_location is an implicit
+	   parameter, which will be used explicitly after expansion.
+	   Merging these statements may cause confusing line numbers in
+	   sanitizer messages.  */
+	return gimple_location (stmt1) == gimple_location (stmt2);
+      default:
+	break;
+      }
+
+  return true;
+}
+
 /* Determines whether BB1 and BB2 (members of same_succ) are duplicates.  If so,
    clusters them.  */
 
@@ -1226,14 +1262,10 @@ find_duplicate (same_succ *same_succ, basic_block bb1, basic_block bb2)
       gimple *stmt1 = gsi_stmt (gsi1);
       gimple *stmt2 = gsi_stmt (gsi2);
 
-      /* What could be better than this here is to blacklist the bb
-	 containing the stmt, when encountering the stmt f.i. in
-	 same_succ_hash.  */
-      if (is_tm_ending (stmt1)
-	  || is_tm_ending (stmt2))
+      if (!gimple_equal_p (same_succ, stmt1, stmt2))
 	return;
 
-      if (!gimple_equal_p (same_succ, stmt1, stmt2))
+      if (!merge_stmts_p (stmt1, stmt2))
 	return;
 
       gsi_prev_nondebug (&gsi1);