[committed] Minor cleanup/prep in DOM

Message ID 44815a60-2cd4-9408-64a9-d718163bca71@ventanamicro.com
State New, archived
Headers
Series [committed] Minor cleanup/prep in DOM |

Commit Message

Jeff Law Sept. 30, 2022, 11:05 p.m. UTC
  It's a bit weird that free_dom_edge_info leaves a dangling pointer in 
e->aux.  Not sure what I was thinking.


There's two callers.  One wipes e->aux immediately after the call, the 
other attaches a newly created object immediately after the call.  So we 
can wipe e->aux within the call and simplify one of the two call sites.

This is preparatory work for a minor optimization where we want to 
detect another class of edge equivalences in DOM (until something better 
is available) and either attach them an existing edge_info structure or 
create a new one if one doesn't currently exist for a given edge.

Bootstrapped and regression tested on x86_64.  Installing on the trunk.


Jeff
commit fbd95c027edcc169cc3b40806375fbabc08500e0
Author: Jeff Law <jeffreyalaw@gmail.com>
Date:   Fri Sep 30 18:59:24 2022 -0400

    Minor cleanup/prep in DOM
    
    It's a bit weird that free_dom_edge_info leaves a dangling pointer in e->aux.
    Not sure what I was thinking.
    
    There's two callers.  One wipes e->aux immediately after the call, the other
    attaches a newly created object immediately after the call.  So we can wipe
    e->aux within the call and simplify one of the two call sites.
    
    This is preparatory work for a minor optimization where we want to detect
    another class of edge equivalences in DOM (until something better is available)
    and either attach them an existing edge_info structure or create a new one if
    one doesn't currently exist for a given edge.
    
    gcc/
            * tree-ssa-dom.cc (free_dom_edge_info): Clear e->aux too.
            (free_all_edge_infos): Do not clear e->aux here.
  

Comments

H.J. Lu Oct. 1, 2022, 12:07 a.m. UTC | #1
On Fri, Sep 30, 2022 at 4:06 PM Jeff Law <jlaw@ventanamicro.com> wrote:
>
>
> It's a bit weird that free_dom_edge_info leaves a dangling pointer in
> e->aux.  Not sure what I was thinking.
>
>
> There's two callers.  One wipes e->aux immediately after the call, the
> other attaches a newly created object immediately after the call.  So we
> can wipe e->aux within the call and simplify one of the two call sites.
>
> This is preparatory work for a minor optimization where we want to
> detect another class of edge equivalences in DOM (until something better
> is available) and either attach them an existing edge_info structure or
> create a new one if one doesn't currently exist for a given edge.
>
> Bootstrapped and regression tested on x86_64.  Installing on the trunk.
>

I got


/export/gnu/import/git/sources/gcc/gcc/tree-ssa-dom.cc: In function
‘void record_edge_info(basic_block)’:
/export/gnu/import/git/sources/gcc/gcc/tree-ssa-dom.cc:689:27: error:
‘dst’ was not declared in this scope; did you mean ‘dse’?
  689 |                       if (dst == PHI_ARG_DEF (phi, !alternative))
      |                           ^~~
      |                           dse
In file included from /export/gnu/import/git/sources/gcc/gcc/gimple-ssa.h:24,
                 from /export/gnu/import/git/sources/gcc/gcc/ssa.h:27,
                 from /export/gnu/import/git/sources/gcc/gcc/tree-ssa-dom.cc:28:
/export/gnu/import/git/sources/gcc/gcc/tree-ssa-dom.cc:689:47: error:
‘phi’ was not declared in this scope; did you mean ‘gphi’?
  689 |                       if (dst == PHI_ARG_DEF (phi, !alternative))
      |                                               ^~~
/export/gnu/import/git/sources/gcc/gcc/tree-ssa-operands.h:82:54:
note: in definition of macro ‘PHI_ARG_DEF’
   82 | #define PHI_ARG_DEF(PHI, I)     gimple_phi_arg_def ((PHI), (I))
      |                                                      ^~~
make: *** [Makefile:1146: tree-ssa-dom.o] Error 1
  
Jeff Law Oct. 1, 2022, 12:50 a.m. UTC | #2
On 9/30/22 18:07, H.J. Lu wrote:
> On Fri, Sep 30, 2022 at 4:06 PM Jeff Law <jlaw@ventanamicro.com> wrote:
>>
>> It's a bit weird that free_dom_edge_info leaves a dangling pointer in
>> e->aux.  Not sure what I was thinking.
>>
>>
>> There's two callers.  One wipes e->aux immediately after the call, the
>> other attaches a newly created object immediately after the call.  So we
>> can wipe e->aux within the call and simplify one of the two call sites.
>>
>> This is preparatory work for a minor optimization where we want to
>> detect another class of edge equivalences in DOM (until something better
>> is available) and either attach them an existing edge_info structure or
>> create a new one if one doesn't currently exist for a given edge.
>>
>> Bootstrapped and regression tested on x86_64.  Installing on the trunk.
>>
> I got

[ ... ]

Pushed wrong version.  Sorry about that.  Fixed by attached patch which 
just moves the test down to its proper position.


Jeff
commit aa360fbf68b11e54017e8fa5b1bdb87ce7c19188
Author: Jeff Law <jeffreyalaw@gmail.com>
Date:   Fri Sep 30 20:46:04 2022 -0400

    Install correct patch version.
    
    gcc/
            * tree-ssa-dom.cc (record_edge_info): Install correct version of
            patch.

diff --git a/gcc/tree-ssa-dom.cc b/gcc/tree-ssa-dom.cc
index 8d8312ca350..e6b8dace5e9 100644
--- a/gcc/tree-ssa-dom.cc
+++ b/gcc/tree-ssa-dom.cc
@@ -684,11 +684,6 @@ record_edge_info (basic_block bb)
 		       !gsi_end_p (gsi);
 		       gsi_next (&gsi))
 		    {
-		      /* If the other alternative is the same as the result,
-			 then this is a degenerate and can be ignored.  */
-		      if (dst == PHI_ARG_DEF (phi, !alternative))
-			continue;
-
 		      /* Now get the EDGE_INFO class so we can append
 			 it to our list.  We want the successor edge
 			 where the destination is not the source of
@@ -697,6 +692,11 @@ record_edge_info (basic_block bb)
 		      tree src = PHI_ARG_DEF (phi, alternative);
 		      tree dst = PHI_RESULT (phi);
 
+		      /* If the other alternative is the same as the result,
+			 then this is a degenerate and can be ignored.  */
+		      if (dst == PHI_ARG_DEF (phi, !alternative))
+			continue;
+
 		      if (EDGE_SUCC (bb, 0)->dest
 			  != EDGE_PRED (bb, !alternative)->src)
 			edge_info = (class edge_info *)EDGE_SUCC (bb, 0)->aux;
  

Patch

diff --git a/gcc/tree-ssa-dom.cc b/gcc/tree-ssa-dom.cc
index 84bef798f52..fa43dbe6c44 100644
--- a/gcc/tree-ssa-dom.cc
+++ b/gcc/tree-ssa-dom.cc
@@ -393,7 +393,8 @@  edge_info::record_simple_equiv (tree lhs, tree rhs)
     simple_equivalences.safe_push (equiv_pair (lhs, rhs));
 }
 
-/* Free the edge_info data attached to E, if it exists.  */
+/* Free the edge_info data attached to E, if it exists and
+   clear e->aux.  */
 
 void
 free_dom_edge_info (edge e)
@@ -402,6 +403,7 @@  free_dom_edge_info (edge e)
 
   if (edge_info)
     delete edge_info;
+  e->aux = NULL;
 }
 
 /* Free all EDGE_INFO structures associated with edges in the CFG.
@@ -420,10 +422,7 @@  free_all_edge_infos (void)
   FOR_EACH_BB_FN (bb, cfun)
     {
       FOR_EACH_EDGE (e, ei, bb->preds)
-        {
-	  free_dom_edge_info (e);
-	  e->aux = NULL;
-	}
+	free_dom_edge_info (e);
     }
 }