|
111
|
1 #!/usr/bin/python
|
|
|
2 # Generate Intel taken branches Linux perf event script for autofdo profiling.
|
|
|
3
|
|
|
4 # Copyright (C) 2016 Free Software Foundation, Inc.
|
|
|
5 #
|
|
|
6 # GCC is free software; you can redistribute it and/or modify it under
|
|
|
7 # the terms of the GNU General Public License as published by the Free
|
|
|
8 # Software Foundation; either version 3, or (at your option) any later
|
|
|
9 # version.
|
|
|
10 #
|
|
|
11 # GCC is distributed in the hope that it will be useful, but WITHOUT ANY
|
|
|
12 # WARRANTY; without even the implied warranty of MERCHANTABILITY or
|
|
|
13 # FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
|
|
|
14 # for more details.
|
|
|
15 #
|
|
|
16 # You should have received a copy of the GNU General Public License
|
|
|
17 # along with GCC; see the file COPYING3. If not see
|
|
|
18 # <http://www.gnu.org/licenses/>. */
|
|
|
19
|
|
|
20 # Run it with perf record -b -e EVENT program ...
|
|
|
21 # The Linux Kernel needs to support the PMU of the current CPU, and
|
|
|
22 # It will likely not work in VMs.
|
|
|
23 # Add --all to print for all cpus, otherwise for current cpu.
|
|
|
24 # Add --script to generate shell script to run correct event.
|
|
|
25 #
|
|
|
26 # Requires internet (https) access. This may require setting up a proxy
|
|
|
27 # with export https_proxy=...
|
|
|
28 #
|
|
|
29 import urllib2
|
|
|
30 import sys
|
|
|
31 import json
|
|
|
32 import argparse
|
|
|
33 import collections
|
|
|
34
|
|
|
35 baseurl = "https://download.01.org/perfmon"
|
|
|
36
|
|
|
37 target_events = (u'BR_INST_RETIRED.NEAR_TAKEN',
|
|
|
38 u'BR_INST_EXEC.TAKEN',
|
|
|
39 u'BR_INST_RETIRED.TAKEN_JCC',
|
|
|
40 u'BR_INST_TYPE_RETIRED.COND_TAKEN')
|
|
|
41
|
|
|
42 ap = argparse.ArgumentParser()
|
|
|
43 ap.add_argument('--all', '-a', help='Print for all CPUs', action='store_true')
|
|
|
44 ap.add_argument('--script', help='Generate shell script', action='store_true')
|
|
|
45 args = ap.parse_args()
|
|
|
46
|
|
|
47 eventmap = collections.defaultdict(list)
|
|
|
48
|
|
|
49 def get_cpu_str():
|
|
|
50 with open('/proc/cpuinfo', 'r') as c:
|
|
|
51 vendor, fam, model = None, None, None
|
|
|
52 for j in c:
|
|
|
53 n = j.split()
|
|
|
54 if n[0] == 'vendor_id':
|
|
|
55 vendor = n[2]
|
|
|
56 elif n[0] == 'model' and n[1] == ':':
|
|
|
57 model = int(n[2])
|
|
|
58 elif n[0] == 'cpu' and n[1] == 'family':
|
|
|
59 fam = int(n[3])
|
|
|
60 if vendor and fam and model:
|
|
|
61 return "%s-%d-%X" % (vendor, fam, model), model
|
|
|
62 return None, None
|
|
|
63
|
|
|
64 def find_event(eventurl, model):
|
|
|
65 print >>sys.stderr, "Downloading", eventurl
|
|
|
66 u = urllib2.urlopen(eventurl)
|
|
|
67 events = json.loads(u.read())
|
|
|
68 u.close()
|
|
|
69
|
|
|
70 found = 0
|
|
|
71 for j in events:
|
|
|
72 if j[u'EventName'] in target_events:
|
|
|
73 event = "cpu/event=%s,umask=%s/" % (j[u'EventCode'], j[u'UMask'])
|
|
|
74 if u'PEBS' in j and j[u'PEBS'] > 0:
|
|
|
75 event += "p"
|
|
|
76 if args.script:
|
|
|
77 eventmap[event].append(model)
|
|
|
78 else:
|
|
|
79 print j[u'EventName'], "event for model", model, "is", event
|
|
|
80 found += 1
|
|
|
81 return found
|
|
|
82
|
|
|
83 if not args.all:
|
|
|
84 cpu, model = get_cpu_str()
|
|
|
85 if not cpu:
|
|
|
86 sys.exit("Unknown CPU type")
|
|
|
87
|
|
|
88 url = baseurl + "/mapfile.csv"
|
|
|
89 print >>sys.stderr, "Downloading", url
|
|
|
90 u = urllib2.urlopen(url)
|
|
|
91 found = 0
|
|
|
92 cpufound = 0
|
|
|
93 for j in u:
|
|
|
94 n = j.rstrip().split(',')
|
|
|
95 if len(n) >= 4 and (args.all or n[0] == cpu) and n[3] == "core":
|
|
|
96 if args.all:
|
|
|
97 vendor, fam, model = n[0].split("-")
|
|
|
98 model = int(model, 16)
|
|
|
99 cpufound += 1
|
|
|
100 found += find_event(baseurl + n[2], model)
|
|
|
101 u.close()
|
|
|
102
|
|
|
103 if args.script:
|
|
|
104 print '''#!/bin/sh
|
|
|
105 # Profile workload for gcc profile feedback (autofdo) using Linux perf.
|
|
|
106 # Auto generated. To regenerate for new CPUs run
|
|
|
107 # contrib/gen_autofdo_event.py --script --all in gcc source
|
|
|
108
|
|
|
109 # usages:
|
|
|
110 # gcc-auto-profile program (profile program and children)
|
|
|
111 # gcc-auto-profile -a sleep X (profile all for X secs, may need root)
|
|
|
112 # gcc-auto-profile -p PID sleep X (profile PID)
|
|
|
113 # gcc-auto-profile --kernel -a sleep X (profile kernel)
|
|
|
114 # gcc-auto-profile --all -a sleep X (profile kernel and user space)
|
|
|
115
|
|
|
116 # Identify branches taken event for CPU.
|
|
|
117 #
|
|
|
118
|
|
|
119 FLAGS=u
|
|
|
120
|
|
|
121 if [ "$1" = "--kernel" ] ; then
|
|
|
122 FLAGS=k
|
|
|
123 shift
|
|
|
124 fi
|
|
|
125 if [ "$1" = "--all" ] ; then
|
|
|
126 FLAGS=uk
|
|
|
127 shift
|
|
|
128 fi
|
|
|
129
|
|
|
130 if ! grep -q Intel /proc/cpuinfo ; then
|
|
|
131 echo >&2 "Only Intel CPUs supported"
|
|
|
132 exit 1
|
|
|
133 fi
|
|
|
134
|
|
|
135 if grep -q hypervisor /proc/cpuinfo ; then
|
|
|
136 echo >&2 "Warning: branch profiling may not be functional in VMs"
|
|
|
137 fi
|
|
|
138
|
|
|
139 case `egrep -q "^cpu family\s*: 6" /proc/cpuinfo &&
|
|
|
140 egrep "^model\s*:" /proc/cpuinfo | head -n1` in'''
|
|
|
141 for event, mod in eventmap.iteritems():
|
|
|
142 for m in mod[:-1]:
|
|
|
143 print "model*:\ %s|\\" % m
|
|
|
144 print 'model*:\ %s) E="%s$FLAGS" ;;' % (mod[-1], event)
|
|
|
145 print '''*)
|
|
|
146 echo >&2 "Unknown CPU. Run contrib/gen_autofdo_event.py --all --script to update script."
|
|
|
147 exit 1 ;;'''
|
|
|
148 print "esac"
|
|
|
149 print 'exec perf record -e $E -b "$@"'
|
|
|
150
|
|
|
151 if cpufound == 0 and not args.all:
|
|
|
152 sys.exit('CPU %s not found' % cpu)
|
|
|
153
|
|
|
154 if found == 0:
|
|
|
155 sys.exit('Branch event not found')
|
