Re: [PATCH] Indirect-call topn targets profiler (instrumentation)

[Rong Xu <xur at google dot com>]

On Mon, Oct 6, 2014 at 12:21 PM, Jan Hubicka <hubicka@ucw.cz> wrote:
>> On Mon, Oct 6, 2014 at 11:22 AM, Jan Hubicka <hubicka@ucw.cz> wrote:
>> > Rong,
>> > Would be possible to use topn profiler to get resonale histograms for switch
>> > expansion, too?  In that case it may make sense to have value version of it as well.
>>

Yes. That's doable. I'm just not sure if top 2 entries are enough. I
can see from your example that it will be useful for BBs having
multiple case stmt.

>> The underlying topn_profiler can be used the same way as the current
>> one_value profiler for different scenarios. However, I wonder if it is
>> useful for switch value profiling -- the arc profiling can do the
>> same.
>
> Well, for switch expansion one really wants a histogram of values across the switch
> values for the decision tree balancing.  But I guess this profiler does not fit
> that very well.
>
> Consider cases like
>
> switch (a)
> {
>    case 3:
>    case 5:
>    case 7:
>         prime();
>    case 2:
>    case 4:
>    case 6:
>         even();
>   default:
>         bogus ();
> }
>
> Arc profiling does tell you how often a is prime or even, but it does not help you
> to balance the tree, because the value ranges are iinterlacing
>>
>> David
>>
>> >
>> > Otherwise the patch seems OK.  I would implement it myself by introducing separate
>> > variables holding the topn profiler declaration, but can not think of downsides of
>> > doing it your way.
>> >
>> > The topn profiler should be better on determining the dominating target, so perhaps
>> > you could look into hooking it into ipa-profile.c.  Extending speculative edges for
>> > multiple targets ought to be also quite easy - we need to update speculative_call_info
>> > to collect edges annotated to a given call stmt into a vector instead of taking the
>> > two references it does now.
>> >
>> > It would be also very nice to update it to handle case where all the edges are direct
>> > so we can use it tospeculatively inlinine functions that can be interposed at runtime.

Yes. I'm looking at this too. I actually ported the fdo-use part of
the patch and it passes spec2006. But it performs the transformation
using the old way (in gimple_ic_transform()) -- so I did not send out
that patch.
I'm trying to do this in the new way (in ipa-profile()). In any case,
that will be a separated patch.

Also, the switch expansion profile should be also a separated patch.

Is it ok to commit these two patches now?

Thanks,

-Rong

>> >
>> > Thanks,
>> > Honza
>> >
>> > Index: gcc/tree-profile.c
>> > ===================================================================
>> > --- gcc/tree-profile.c  (revision 215886)
>> > +++ gcc/tree-profile.c  (working copy)
>> > @@ -56,6 +56,7 @@ along with GCC; see the file COPYING3.  If not see
>> >  #include "target.h"
>> >  #include "tree-cfgcleanup.h"
>> >  #include "tree-nested.h"
>> > +#include "params.h"
>> >
>> >  static GTY(()) tree gcov_type_node;
>> >  static GTY(()) tree tree_interval_profiler_fn;
>> > @@ -101,7 +102,10 @@ init_ic_make_global_vars (void)
>> >      {
>> >        ic_void_ptr_var
>> >         = build_decl (UNKNOWN_LOCATION, VAR_DECL,
>> > -                     get_identifier ("__gcov_indirect_call_callee"),
>> > +                     get_identifier (
>> > +                             (PARAM_VALUE (PARAM_INDIR_CALL_TOPN_PROFILE) ?
>> > +                              "__gcov_indirect_call_topn_callee" :
>> > +                              "__gcov_indirect_call_callee")),
>> >                       ptr_void);
>> >        TREE_PUBLIC (ic_void_ptr_var) = 1;
>> >        DECL_EXTERNAL (ic_void_ptr_var) = 1;
>> > @@ -131,7 +135,10 @@ init_ic_make_global_vars (void)
>> >      {
>> >        ic_gcov_type_ptr_var
>> >         = build_decl (UNKNOWN_LOCATION, VAR_DECL,
>> > -                     get_identifier ("__gcov_indirect_call_counters"),
>> > +                     get_identifier (
>> > +                             (PARAM_VALUE (PARAM_INDIR_CALL_TOPN_PROFILE) ?
>> > +                              "__gcov_indirect_call_topn_counters" :
>> > +                              "__gcov_indirect_call_counters")),
>> >                       gcov_type_ptr);
>> >        TREE_PUBLIC (ic_gcov_type_ptr_var) = 1;
>> >        DECL_EXTERNAL (ic_gcov_type_ptr_var) = 1;
>> > @@ -226,8 +233,10 @@ gimple_init_edge_profiler (void)
>> >                                               ptr_void,
>> >                                               NULL_TREE);
>> >           tree_indirect_call_profiler_fn
>> > -                 = build_fn_decl ("__gcov_indirect_call_profiler_v2",
>> > -                                        ic_profiler_fn_type);
>> > +                 = build_fn_decl ( (PARAM_VALUE (PARAM_INDIR_CALL_TOPN_PROFILE) ?
>> > +                                    "__gcov_indirect_call_topn_profiler":
>> > +                                    "__gcov_indirect_call_profiler_v2"),
>> > +                                  ic_profiler_fn_type);
>> >          }
>> >        TREE_NOTHROW (tree_indirect_call_profiler_fn) = 1;
>> >        DECL_ATTRIBUTES (tree_indirect_call_profiler_fn)
>> > @@ -398,6 +407,12 @@ gimple_gen_ic_profiler (histogram_value value, uns
>> >    gimple_stmt_iterator gsi = gsi_for_stmt (stmt);
>> >    tree ref_ptr = tree_coverage_counter_addr (tag, base);
>> >
>> > +  if ( (PARAM_VALUE (PARAM_INDIR_CALL_TOPN_PROFILE) &&
>> > +        tag == GCOV_COUNTER_V_INDIR) ||
>> > +       (!PARAM_VALUE (PARAM_INDIR_CALL_TOPN_PROFILE) &&
>> > +        tag == GCOV_COUNTER_ICALL_TOPNV))
>> > +    return;
>> > +
>> >    ref_ptr = force_gimple_operand_gsi (&gsi, ref_ptr,
>> >                                       true, NULL_TREE, true, GSI_SAME_STMT);
>> >
>> > @@ -442,8 +457,7 @@ gimple_gen_ic_func_profiler (void)
>> >      stmt1: __gcov_indirect_call_profiler_v2 (profile_id,
>> >                                              &current_function_decl)
>> >     */
>> > -  gsi =
>> > -                                            gsi_after_labels (split_edge (single_succ_edge (ENTRY_BLOCK_PTR_FOR_FN (cfun))));
>> > +  gsi = gsi_after_labels (split_edge (single_succ_edge (ENTRY_BLOCK_PTR_FOR_FN (cfun))));
>> >
>> >    cur_func = force_gimple_operand_gsi (&gsi,
>> >                                        build_addr (current_function_decl,
>> > Index: gcc/value-prof.c
>> > ===================================================================
>> > --- gcc/value-prof.c    (revision 215886)
>> > +++ gcc/value-prof.c    (working copy)
>> > @@ -60,6 +60,7 @@ along with GCC; see the file COPYING3.  If not see
>> >  #include "builtins.h"
>> >  #include "tree-nested.h"
>> >  #include "hash-set.h"
>> > +#include "params.h"
>> >
>> >  /* In this file value profile based optimizations are placed.  Currently the
>> >     following optimizations are implemented (for more detailed descriptions
>> > @@ -359,6 +360,22 @@ dump_histogram_value (FILE *dump_file, histogram_v
>> >        }
>> >        fprintf (dump_file, ".\n");
>> >        break;
>> > +    case HIST_TYPE_INDIR_CALL_TOPN:
>> > +      fprintf (dump_file, "Indirect call topn ");
>> > +      if (hist->hvalue.counters)
>> > +       {
>> > +           int i;
>> > +
>> > +           fprintf (dump_file, "accu:%"PRId64, hist->hvalue.counters[0]);
>> > +           for (i = 1; i < (GCOV_ICALL_TOPN_VAL << 2); i += 2)
>> > +             {
>> > +               fprintf (dump_file, " target:%"PRId64 " value:%"PRId64,
>> > +                       (int64_t) hist->hvalue.counters[i],
>> > +                       (int64_t) hist->hvalue.counters[i+1]);
>> > +             }
>> > +        }
>> > +      fprintf (dump_file, ".\n");
>> > +      break;
>> >      case HIST_TYPE_MAX:
>> >        gcc_unreachable ();
>> >     }
>> > @@ -432,9 +449,14 @@ stream_in_histogram_value (struct lto_input_block
>> >           break;
>> >
>> >         case HIST_TYPE_IOR:
>> > -  case HIST_TYPE_TIME_PROFILE:
>> > +        case HIST_TYPE_TIME_PROFILE:
>> >           ncounters = 1;
>> >           break;
>> > +
>> > +        case HIST_TYPE_INDIR_CALL_TOPN:
>> > +          ncounters = (GCOV_ICALL_TOPN_VAL << 2) + 1;
>> > +          break;
>> > +
>> >         case HIST_TYPE_MAX:
>> >           gcc_unreachable ();
>> >         }
>> > @@ -1920,8 +1942,12 @@ gimple_indirect_call_to_profile (gimple stmt, hist
>> >
>> >    values->reserve (3);
>> >
>> > -  values->quick_push (gimple_alloc_histogram_value (cfun, HIST_TYPE_INDIR_CALL,
>> > -                                                   stmt, callee));
>> > +  values->quick_push (gimple_alloc_histogram_value (
>> > +                        cfun,
>> > +                        PARAM_VALUE (PARAM_INDIR_CALL_TOPN_PROFILE) ?
>> > +                          HIST_TYPE_INDIR_CALL_TOPN :
>> > +                          HIST_TYPE_INDIR_CALL,
>> > +                       stmt, callee));
>> >
>> >    return;
>> >  }
>> > @@ -2011,9 +2037,9 @@ gimple_find_values_to_profile (histogram_values *v
>> >           hist->n_counters = 3;
>> >           break;
>> >
>> > -  case HIST_TYPE_TIME_PROFILE:
>> > -    hist->n_counters = 1;
>> > -    break;
>> > +        case HIST_TYPE_TIME_PROFILE:
>> > +          hist->n_counters = 1;
>> > +          break;
>> >
>> >         case HIST_TYPE_AVERAGE:
>> >           hist->n_counters = 2;
>> > @@ -2023,6 +2049,10 @@ gimple_find_values_to_profile (histogram_values *v
>> >           hist->n_counters = 1;
>> >           break;
>> >
>> > +        case HIST_TYPE_INDIR_CALL_TOPN:
>> > +          hist->n_counters = GCOV_ICALL_TOPN_NCOUNTS;
>> > +          break;
>> > +
>> >         default:
>> >           gcc_unreachable ();
>> >         }
>> > Index: gcc/value-prof.h
>> > ===================================================================
>> > --- gcc/value-prof.h    (revision 215886)
>> > +++ gcc/value-prof.h    (working copy)
>> > @@ -35,6 +35,8 @@ enum hist_type
>> >    HIST_TYPE_AVERAGE,   /* Compute average value (sum of all values).  */
>> >    HIST_TYPE_IOR,       /* Used to compute expected alignment.  */
>> >    HIST_TYPE_TIME_PROFILE, /* Used for time profile */
>> > +  HIST_TYPE_INDIR_CALL_TOPN, /* Tries to identify the top N most frequently
>> > +                                called functions in indirect call.  */
>> >    HIST_TYPE_MAX
>> >  };
>> >
>> > Index: gcc/profile.c
>> > ===================================================================
>> > --- gcc/profile.c       (revision 215886)
>> > +++ gcc/profile.c       (working copy)
>> > @@ -183,6 +183,7 @@ instrument_values (histogram_values values)
>> >           break;
>> >
>> >         case HIST_TYPE_INDIR_CALL:
>> > +       case HIST_TYPE_INDIR_CALL_TOPN:
>> >           gimple_gen_ic_profiler (hist, t, 0);
>> >           break;
>> >