public inbox for gcc-cvs@sourceware.org
help / color / mirror / Atom feed
* [gcc(refs/users/kubaneko/heads/histogram)] Do not instrument loops with known number of iterations and with no instrumentable exits.
@ 2023-05-01 21:51 Jan Hubicka
0 siblings, 0 replies; only message in thread
From: Jan Hubicka @ 2023-05-01 21:51 UTC (permalink / raw)
To: gcc-cvs
https://gcc.gnu.org/g:b37d3aa26581cc6145ee8be6c82ad81f203f313a
commit b37d3aa26581cc6145ee8be6c82ad81f203f313a
Author: Honza <jh@ryzen3.suse.cz>
Date: Mon May 1 23:50:31 2023 +0200
Do not instrument loops with known number of iterations and with no instrumentable exits.
Diff:
---
gcc/profile.cc | 5 ++++-
gcc/value-prof.cc | 34 ++++++++++++++++++++++++++++++++++
2 files changed, 38 insertions(+), 1 deletion(-)
diff --git a/gcc/profile.cc b/gcc/profile.cc
index d2b8dfcadc1..4e3f74dfe1d 100644
--- a/gcc/profile.cc
+++ b/gcc/profile.cc
@@ -66,6 +66,7 @@ along with GCC; see the file COPYING3. If not see
#include "cfgloop.h"
#include "sreal.h"
#include "file-prefix-map.h"
+#include "tree-scalar-evolution.h"
#include "profile.h"
@@ -1227,7 +1228,8 @@ branch_prob (bool thunk)
total_num_times_called++;
calculate_dominance_info (CDI_DOMINATORS);
- loop_optimizer_init (LOOPS_HAVE_SIMPLE_LATCHES);
+ loop_optimizer_init (LOOPS_NORMAL);
+ scev_initialize ();
flow_call_edges_add (NULL);
add_noreturn_fake_exit_edges ();
@@ -1590,6 +1592,7 @@ branch_prob (bool thunk)
values.release ();
free_edge_list (el);
coverage_end_function (lineno_checksum, cfg_checksum);
+ scev_finalize ();
if (flag_branch_probabilities
&& (profile_status_for_fn (cfun) == PROFILE_READ))
{
diff --git a/gcc/value-prof.cc b/gcc/value-prof.cc
index 5e029c811e2..f85e83388b7 100644
--- a/gcc/value-prof.cc
+++ b/gcc/value-prof.cc
@@ -44,6 +44,8 @@ along with GCC; see the file COPYING3. If not see
#include "builtins.h"
#include "cfgloop.h"
#include "tree-ssa-loop-manip.h"
+#include "tree-ssa-loop.h"
+#include "tree-ssa-loop-niter.h"
/* In this file value profile based optimizations are placed. Currently the
following optimizations are implemented (for more detailed descriptions
@@ -1952,6 +1954,38 @@ gimple_histogram_values_to_profile (function *fun, histogram_values *values)
{
tree var;
gimple_stmt_iterator gsi;
+ edge exit_edge = NULL;
+ auto_vec<edge> exits = get_loop_exit_edges (loop);
+ bool complex_exit = false;
+ bool exit_found = false;
+
+ /* If loop has only exits that can not be instrumented,
+ we can not profile histograms. */
+ for (auto exit : exits)
+ if (exit->flags & EDGE_FAKE)
+ ;
+ else if (exit->flags & EDGE_COMPLEX)
+ complex_exit = true;
+ else
+ {
+ if (!exit_found)
+ exit_edge = exit;
+ else
+ exit_edge = NULL;
+ exit_found = true;
+ break;
+ }
+ if (!exit_found)
+ continue;
+
+ /* If there is only one non-fake exit edge, see if we know number of
+ iterations. */
+ tree niter;
+ edge ex;
+ if (!complex_exit && exit_edge
+ && (niter = find_loop_niter (loop, &ex))
+ && TREE_CODE (niter) == INTEGER_CST)
+ continue;
gsi = gsi_start_bb (loop->latch);
create_iv (build_int_cst_type (get_gcov_type (), 0),
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2023-05-01 21:51 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-05-01 21:51 [gcc(refs/users/kubaneko/heads/histogram)] Do not instrument loops with known number of iterations and with no instrumentable exits Jan Hubicka
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).