OpenE2K/gcc
2
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

* analyze_brprob: Fix more typos.

Browse Source 
From-SVN: r44263
This commit is contained in:
Andreas Jaeger 2001-07-23 15:21:13 +02:00 committed by Andreas Jaeger
parent 075cc945ad
commit f981f97beb
2 changed files with 21 additions and 17 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
contrib/ChangeLog
View File

@ -1,3 +1,7 @@
2001-07-23 Andreas Jaeger <aj@suse.de>
* analyze_brprob: Fix more typos.
2001-07-23 Andreas Jaeger <aj@suse.de>
* analyze_brprob: Fix typos.

34
contrib/analyze_brprob
View File

@ -1,6 +1,6 @@
#!/usr/bin/awk -f
# Script to analyze experimental results of our branch prediction heuristics
# Contributed by Jan Hubicka, SuSE inc.
# Contributed by Jan Hubicka, SuSE Inc.
# Copyright (C) 2001 Free Software Foundation, Inc.
#
# This file is part of GNU CC.
@ -22,28 +22,28 @@
#
#
# This script is used to calculate two basic properties of the branch prediction
# heuristics - coverage and hitrate. Coverage is number of executions of given
# heuristics - coverage and hitrate. Coverage is number of execution of a given
# branch matched by the heuristics and hitrate is probability that once branch is
# predicted as taken it is really taken.
#
# These values are usefull to determine quality of given heuristics and hitrate
# These values are useful to determine the quality of given heuristics. Hitrate
# may be directly used in predict.c.
#
# Usage:
# step 1: compile and profile your program. You need to use -fprofile-arcs
# Step 1: compile and profile your program. You need to use -fprofile-arcs
# flag to get the profiles
# step 2: Generate log files. The information about given heuristics are
# saved into *.life dumps. You need to pass -df swtich to compiler as well
# as -fbranch-probabilities to get results of profiling noted in the dumps.
# Ensure that there is no "Arc profiling: some edge counts were bad." warnings.
# Step 2: Generate log files. The information about given heuristics are
# saved into *.life dumps. You need to pass the -df switch to the compiler as well
# as -fbranch-probabilities to get the results of profiling noted in the dumps.
# Ensure that there are no "Arc profiling: some edge counts were bad." warnings.
# Keep the -fprofile-arcs switch to ensure that CFGs match.
# step 3: Run this script to concatetation of all *.life files:
# Step 3: Run this script to concatenate all *.life files:
# analyze_brprob `find . -name *.life`
# the information is collected and print once all files are parsed. This
# may take a while.
# Note that script does use bc to perform long arithmetic.
# step 4: Read the results. Basically following table is printed:
# (this is just example from very early stages of branch prediction pass
# Note that the script does use bc to perform long arithmetic.
# Step 4: Read the results. Basically the following table is printed:
# (this is just an example from a very early stage of branch prediction pass
# development, so please don't take these numbers seriously)
#
#HEURISTICS BRANCHES (REL) HITRATE COVERAGE (REL)
@ -57,15 +57,15 @@
#loop branch 51 1.5% 99.26%/ 99.27% 26854 0.3%
#noreturn call 951 27.6% 100.00%/100.00% 1759809 17.6%
#
# The heuristics called "first match" is heuristics used by gcc branch
# The heuristic called "first match" is a heuristic used by GCC branch
# prediction pass and it predicts 89.92% branches correctly.
#
# The quality of heuristics can be rated using both, coverage and hitrate
# parameters. For example "loop branch" heuristics (predicting loopback edge
# as taken) have both very high hitrate and coverage, so it is very usefull.
# as taken) have both very high hitrate and coverage, so it is very useful.
# On the other hand, "exit block" heuristics (predicting exit edges as not
# taken) have good hitrate, but poor coverage, so only 3 branches has been
# predicted. The "loop header" heuristics have problem, since it tends to
# taken) have good hitrate, but poor coverage, so only 3 branches have been
# predicted. The "loop header" heuristic has problems, since it tends to
# misspredict.
#
# The implementation of this script is somewhat brute force. My awk skills
@ -133,7 +133,7 @@ END {
# Heuristics called combined predicts just everything.
maxcounts = longeval (counts["first match"])
maxbranches = branches["first match"]
max = names["firat match"]
max = names["first match"]
printf("HEURISTICS BRANCHES (REL) HITRATE COVERAGE (REL)\n")
for (i = 0; i < nnames ; i++)
{