Reviewers: jh_suse.cz, davidxl, bjanakiraman_google.com,
Message:
Jan, please take a look and provide some feedback.
Thanks,
Description:
Added better handling of outdated profiles.
gcc dumps profile information in .gcda files that contain the function
id and
the corresponding counter information. The function cfg and line number
checksums are also stored along with the ids. When the program source
changes
between when the instrumentation run is done vs. when the -fprofile-use
is
passed, gcc will have functions that do not have a valid profile. These
functions could be hot and may not be inlined because the summary of the
profile
is still valid (sum_max, etc.). This can lead to a situation where the
performance with -fprofile-use is lower than with -fno-profile-use.
It is easy to have a situation where the profile is outdated. A single
line
change in the header file can potentially renumber all functions and
none of the
ids will match the profile generated from the old source.
This patch makes the performance degradation in the outdated case much
less
severe. It adds checks that the profile has been read for a certain
function
before using the counter values (which are 0 in the case where the
profile is
invalid).
Without this patch we see up to 15% performance degradation by using
outdated
profiles vs. not using any profile information for Chrome. This patch
decreases
that degradation down to (0% to 4%).
Please review this at http://codereview.appspot.com/5989046/
Affected files:
M gcc/ipa-inline.c
M gcc/predict.c
M gcc/profile.c
M gcc/tree.h
From 88d8094b6408b38d05241ecf574a51edabeabb3a Mon Sep 17 00:00:00 2001
From: Ahmad Sharif asha...@chromium.org
Date: Wed, 4 Apr 2012 19:09:23 -0700
Subject: [PATCH] Added better handling of outdated profiles.
gcc dumps profile information in .gcda files that contain the function id
and
the corresponding counter information. The function cfg and line number
checksums are also stored along with the ids. When the program source
changes
between when the instrumentation run is done vs. when the -fprofile-use is
passed, gcc will have functions that do not have a valid profile. These
functions could be hot and may not be inlined because the summary of the
profile
is still valid (sum_max, etc.). This can lead to a situation where the
performance with -fprofile-use is lower than with -fno-profile-use.
It is easy to have a situation where the profile is outdated. A single line
change in the header file can potentially renumber all functions and none
of the
ids will match the profile generated from the old source.
This patch makes the performance degradation in the outdated case much less
severe. It adds checks that the profile has been read for a certain function
before using the counter values (which are 0 in the case where the profile
is
invalid).
Without this patch we see up to 15% performance degradation by using
outdated
profiles vs. not using any profile information for Chrome. This patch
decreases
that degradation down to (0% to 4%).
---
gcc/ipa-inline.c |4 +++-
gcc/predict.c| 13 +
gcc/profile.c|2 +-
gcc/tree.h |5 +
4 files changed, 18 insertions(+), 6 deletions(-)
Index: gcc/ipa-inline.c
diff --git a/gcc/ipa-inline.c b/gcc/ipa-inline.c
index d7ccf68..fbc6c4a 100644
--- a/gcc/ipa-inline.c
+++ b/gcc/ipa-inline.c
@@ -780,7 +780,9 @@ edge_badness (struct cgraph_edge *edge, bool dump)
The fraction is upside down, becuase on edge counts and time beneits
the bounds are known. Edge growth is essentially unlimited. */
- else if (max_count)
+ else if (max_count
+DECL_READ_PROFILE_P(edge-caller-decl)
+DECL_READ_PROFILE_P(edge-callee-decl))
{
int relbenefit = relative_time_benefit (callee_info, edge,
time_growth);
badness =
Index: gcc/predict.c
diff --git a/gcc/predict.c b/gcc/predict.c
index c12b45f..94af954 100644
--- a/gcc/predict.c
+++ b/gcc/predict.c
@@ -114,7 +114,8 @@ static inline bool
maybe_hot_frequency_p (int freq)
{
struct cgraph_node *node = cgraph_get_node (current_function_decl);
- if (!profile_info || !flag_branch_probabilities)
+ if (!profile_info || !flag_branch_probabilities
+ || !DECL_READ_PROFILE_P(current_function_decl))
{
if (node-frequency == NODE_FREQUENCY_UNLIKELY_EXECUTED)
return false;
@@ -162,6 +163,7 @@ bool
cgraph_maybe_hot_edge_p (struct cgraph_edge *edge)
{
if (profile_info flag_branch_probabilities
+ DECL_READ_PROFILE_P(edge-caller-decl)
(edge-count
= profile_info-sum_max / PARAM_VALUE (HOT_BB_COUNT_FRACTION)))
return false;
@@ -202,9 +204,11 @@ maybe_hot_edge_p (edge e)
bool
probably_never_executed_bb_p (const_basic_block bb)
{
- if (profile_info flag_branch_probabilities)
+ if (profile_info flag_branch_probabilities
+ DECL_READ_PROFILE_P(current_function_decl))
return ((bb-count + profile_info-runs / 2) /
