Chromium Code Reviews| OLD | NEW |
|---|---|
| (Empty) | |
| 1 #!/usr/bin/python | |
| 2 # Copyright 2014 The Chromium Authors. All rights reserved. | |
| 3 # Use of this source code is governed by a BSD-style license that can be | |
| 4 # found in the LICENSE file. | |
| 5 | |
| 6 """Generate a spatial analysis against an arbitrary library. | |
| 7 | |
| 8 To use, build the 'binary_size_java' target. Then run this tool, passing | |
| 9 in the location of the library to be analyzed along with any other options | |
| 10 you desire. | |
| 11 """ | |
| 12 | |
| 13 import fileinput | |
| 14 import optparse | |
| 15 import os | |
| 16 import pprint | |
| 17 import re | |
| 18 import shutil | |
| 19 import subprocess | |
| 20 import sys | |
| 21 import tempfile | |
| 22 import json | |
| 23 | |
| 24 def format_bytes(bytes): | |
| 25 """Pretty-print a number of bytes.""" | |
| 26 if bytes > 1e6: | |
| 27 bytes = bytes / 1.0e6 | |
| 28 return '%.1fm' % bytes | |
| 29 if bytes > 1e3: | |
| 30 bytes = bytes / 1.0e3 | |
| 31 return '%.1fk' % bytes | |
| 32 return str(bytes) | |
| 33 | |
| 34 | |
| 35 def parse_nm(input): | |
|
bulach
2014/01/07 19:38:30
not sure, is this needed?
Andrew Hayden (chromium.org)
2014/01/08 00:56:45
Yes, the bloat script parses nm to convert the str
bulach
2014/01/08 15:04:00
got it... there's some hidden irony in that whilst
| |
| 36 """Parse nm output. | |
| 37 | |
| 38 Argument: an iterable over lines of nm output. | |
| 39 | |
| 40 Yields: (symbol name, symbol type, symbol size, source file path). | |
| 41 Path may be None if nm couldn't figure out the source file. | |
| 42 """ | |
| 43 | |
| 44 # Match lines with size, symbol, optional location, optional discriminator | |
| 45 sym_re = re.compile(r'^[0-9a-f]{8} ([0-9a-f]{8}) (.) ([^\t]+)(?:\t(.*):[\d\? ]+)?.*$') | |
| 46 | |
| 47 # Match lines with addr but no size. | |
| 48 addr_re = re.compile(r'^[0-9a-f]{8} (.) ([^\t]+)(?:\t.*)?$') | |
| 49 # Match lines that don't have an address at all -- typically external symbol s. | |
| 50 noaddr_re = re.compile(r'^ {8} (.) (.*)$') | |
| 51 | |
| 52 for line in input: | |
| 53 line = line.rstrip() | |
| 54 match = sym_re.match(line) | |
| 55 if match: | |
| 56 size, type, sym = match.groups()[0:3] | |
| 57 size = int(size, 16) | |
| 58 type = type.lower() | |
| 59 if type == 'v': | |
| 60 type = 'w' # just call them all weak | |
| 61 if type == 'b': | |
| 62 continue # skip all BSS for now | |
| 63 path = match.group(4) | |
| 64 yield sym, type, size, path | |
| 65 continue | |
| 66 match = addr_re.match(line) | |
| 67 if match: | |
| 68 type, sym = match.groups()[0:2] | |
| 69 # No size == we don't care. | |
| 70 continue | |
| 71 match = noaddr_re.match(line) | |
| 72 if match: | |
| 73 type, sym = match.groups() | |
| 74 if type in ('U', 'w'): | |
| 75 # external or weak symbol | |
| 76 continue | |
| 77 | |
| 78 print >>sys.stderr, 'unparsed:', repr(line) | |
| 79 | |
| 80 | |
| 81 def treeify_syms(symbols): | |
| 82 dirs = {} | |
| 83 for sym, type, size, path in symbols: | |
| 84 if path: | |
| 85 path = os.path.normpath(path) | |
| 86 if path.startswith('/usr/include'): | |
| 87 path = path.replace('/usr/include', 'usrinclude') | |
| 88 elif path.startswith('/'): | |
| 89 path = path[1:] | |
| 90 | |
| 91 parts = None | |
| 92 # TODO: make segmenting by namespace work. | |
| 93 if False and '::' in sym: | |
| 94 if sym.startswith('vtable for '): | |
| 95 sym = sym[len('vtable for '):] | |
| 96 parts = sym.split('::') | |
| 97 parts.append('[vtable]') | |
| 98 else: | |
| 99 parts = sym.split('::') | |
| 100 parts[0] = '::' + parts[0] | |
| 101 elif path and '/' in path: | |
| 102 parts = path.split('/') | |
| 103 | |
| 104 if parts: | |
| 105 key = parts.pop() | |
| 106 tree = dirs | |
| 107 try: | |
| 108 for part in parts: | |
| 109 assert part != '', path | |
| 110 if part not in tree: | |
| 111 tree[part] = {} | |
| 112 tree = tree[part] | |
| 113 tree[key] = tree.get(key, 0) + size | |
| 114 except: | |
| 115 print >>sys.stderr, sym, parts, key | |
| 116 raise | |
| 117 else: | |
| 118 key = 'symbols without paths' | |
| 119 if key not in dirs: | |
| 120 dirs[key] = {} | |
| 121 tree = dirs[key] | |
| 122 subkey = 'misc' | |
| 123 if (sym.endswith('::__FUNCTION__') or | |
| 124 sym.endswith('::__PRETTY_FUNCTION__')): | |
| 125 subkey = '__FUNCTION__' | |
| 126 elif sym.startswith('CSWTCH.'): | |
| 127 subkey = 'CSWTCH' | |
| 128 elif '::' in sym: | |
| 129 subkey = sym[0:sym.find('::') + 2] | |
| 130 #else: | |
| 131 # print >>sys.stderr, 'unbucketed (no path?):', sym, type, size, path | |
| 132 tree[subkey] = tree.get(subkey, 0) + size | |
| 133 return dirs | |
| 134 | |
| 135 | |
| 136 def jsonify_tree(tree, name): | |
| 137 children = [] | |
| 138 total = 0 | |
| 139 files = 0 | |
| 140 | |
| 141 for key, val in tree.iteritems(): | |
| 142 if isinstance(val, dict): | |
| 143 subtree = jsonify_tree(val, key) | |
| 144 total += subtree['data']['$area'] | |
| 145 children.append(subtree) | |
| 146 else: | |
| 147 total += val | |
| 148 children.append({ | |
| 149 'name': key + ' ' + format_bytes(val), | |
| 150 'data': { '$area': val } | |
| 151 }) | |
| 152 | |
| 153 children.sort(key=lambda child: -child['data']['$area']) | |
| 154 | |
| 155 return { | |
| 156 'name': name + ' ' + format_bytes(total), | |
| 157 'data': { | |
| 158 '$area': total, | |
| 159 }, | |
| 160 'children': children, | |
| 161 } | |
| 162 | |
| 163 | |
| 164 def dump_nm(infile, outfile): | |
| 165 dirs = treeify_syms(parse_nm(infile)) | |
| 166 out = sys.stdout | |
| 167 if outfile is not None: | |
| 168 out = open(outfile, 'w') | |
| 169 out.write('var kTree = ' + json.dumps(jsonify_tree(dirs, '/'), indent=2)) | |
| 170 out.flush() | |
| 171 if outfile is not None: | |
| 172 out.close() | |
| 173 | |
| 174 | |
| 175 def run_pa2l(outfile, library, arch, threads, verbose=False): | |
| 176 """Run a parallel addr2line processing engine to dump and resolve symbols""" | |
| 177 out_dir = os.getenv('CHROMIUM_OUT_DIR', 'out') | |
| 178 buildtype = os.getenv('BUILDTYPE', 'Release') | |
| 179 classpath = out_dir + '/' + buildtype + '/lib.java/binary_size_java.jar' | |
| 180 cmd = ['java', | |
| 181 '-classpath', classpath, | |
| 182 'org.chromium.tools.binary_size.ParallelAddress2Line', | |
| 183 '--disambiguate', | |
| 184 '--outfile', outfile, | |
| 185 '--library', library, | |
| 186 '--threads', threads] | |
| 187 if verbose is True: | |
| 188 cmd.append('--verbose') | |
| 189 if arch == 'android-arm': | |
| 190 cmd.extend([ | |
| 191 '--nm', 'third_party/android_tools/ndk/toolchains/arm-linux- androideabi-4.7/prebuilt/linux-x86_64/bin/arm-linux-androideabi-nm', | |
| 192 '--addr2line', 'third_party/android_tools/ndk/toolchains/arm -linux-androideabi-4.7/prebuilt/linux-x86_64/bin/arm-linux-androideabi-addr2line ', | |
| 193 ]) | |
| 194 elif arch == 'android-mips': | |
| 195 cmd.extend([ | |
| 196 '--nm', 'third_party/android_tools/ndk/toolchains/mipsel-lin ux-android-4.7/prebuilt/linux-x86_64/bin/mipsel-linux-android-nm', | |
| 197 '--addr2line', 'third_party/android_tools/ndk/toolchains/mip sel-linux-android-4.7/prebuilt/linux-x86_64/bin/mipsel-linux-android-addr2line', | |
| 198 ]) | |
| 199 elif arch == 'android-x86': | |
| 200 cmd.extend([ | |
| 201 '--nm', 'third_party/android_tools/ndk/toolchains/x86-4.7/pr ebuilt/linux-x86_64/bin/i686-linux-android-nm' | |
| 202 '--addr2line', 'third_party/android_tools/ndk/toolchains/x86 -4.7/prebuilt/linux-x86_64/bin/i686-linux-android-addr2line', | |
| 203 ]) | |
| 204 # else, use whatever is in PATH (don't pass --nm or --addr2line) | |
| 205 | |
| 206 if verbose: | |
| 207 print cmd | |
| 208 | |
| 209 return_code = subprocess.call(cmd) | |
| 210 if return_code: | |
| 211 raise RuntimeError('Failed to run ParallelAddress2Line: returned ' + str (return_code)) | |
| 212 | |
| 213 usage="""%prog [options] | |
| 214 | |
| 215 Runs a spatial analysis on a given library, looking up the source locations of | |
| 216 its symbols and calculating how much space each directory, source file, and so | |
| 217 on is taking. The result is a report that can be used to pinpoint sources of | |
| 218 large portions of the binary, etceteras. | |
| 219 | |
| 220 Under normal circumstances, you only need to pass two arguments, thusly: | |
| 221 | |
| 222 %prog --library /path/to/library --destdir /path/to/output | |
| 223 | |
| 224 In this mode, the program will dump the symbols from the specified library and | |
| 225 map those symbols back to source locations, producing a web-based report in the | |
| 226 specified output directory. | |
| 227 | |
| 228 Other options are available via '--help'. | |
| 229 """ | |
| 230 parser = optparse.OptionParser(usage=usage) | |
| 231 parser.add_option('--nm-in', dest='nm_in', metavar='PATH', | |
| 232 help='if specified, use nm input from <path> instead of ' | |
| 233 'generating it. Note that source locations should be present ' | |
| 234 'in the file; i.e., no addr2line symbol lookups will be ' | |
| 235 'performed when this option is specified. Mutually exclusive ' | |
| 236 'with --library.') | |
| 237 parser.add_option('--destdir', metavar='PATH', | |
| 238 help='write output to the specified directory. An HTML ' | |
| 239 'report is generated here along with supporting files; any ' | |
| 240 'existing report will be overwritten.') | |
| 241 parser.add_option('--library', metavar='PATH', | |
| 242 help='if specified, process symbols in the library at the ' | |
| 243 'specified path. Mutually exclusive with --nm-in.') | |
| 244 parser.add_option('--arch', | |
| 245 help='the architecture that the library is targeted to. ' | |
| 246 'Currently supports the following: ' | |
| 247 'host-native, android-arm, android-mips, android-x86.' | |
| 248 'the default is host-native. This determines ' | |
| 249 'what nm/addr2line binaries are used. When host-native is ' | |
| 250 'chosen (the default), the program will use whichever ' | |
| 251 'nm/addr2line binaries are on the PATH. This is appropriate ' | |
| 252 'when you are analyzing a binary by and for your computer. ' | |
| 253 'This argument is only valid when using --library.') | |
| 254 parser.add_option('--pa2l-threads', dest='threads', | |
| 255 help='number of threads to use for the parallel addr2line ' | |
| 256 'processing pool; defaults to 1. More threads greatly ' | |
| 257 'improve throughput but eat RAM like popcorn, and take ' | |
| 258 'several gigabytes each. Start low and ramp this number up ' | |
| 259 'until your machine begins to struggle with RAM.' | |
| 260 'This argument is only valid when using --library.') | |
| 261 parser.add_option('-v', dest='verbose', action='store_true', | |
| 262 help='be verbose, printing lots of status information.') | |
| 263 parser.add_option('--nm-out', dest='nm_out', | |
| 264 help='keep the nm output file, and store it at the specified ' | |
| 265 'path. This is useful if you want to see the fully processed ' | |
| 266 'nm output after the symbols have been mapped to source ' | |
| 267 'locations. By default, a tempfile is used and is deleted ' | |
| 268 'when the program terminates.' | |
| 269 'This argument is only valid when using --library.') | |
| 270 opts, args = parser.parse_args() | |
| 271 | |
| 272 if ((not opts.library) and (not opts.nm_in)) or (opts.library and opts.nm_in): | |
| 273 parser.error('exactly one of --library or --nm-in is required') | |
| 274 if (opts.nm_in): | |
| 275 if opts.threads: | |
| 276 print >> sys.stderr, ('WARNING: --pa2l-threads has no effect ' | |
| 277 'when used with --nm-in') | |
| 278 if opts.arch: | |
| 279 print >> sys.stderr, ('WARNING: --arch has no effect ' | |
| 280 'when used with --nm-in') | |
| 281 if not opts.destdir: | |
| 282 parser.error('--destdir is required argument') | |
| 283 if not opts.threads: | |
| 284 opts.threads = 1 | |
| 285 if not opts.arch: | |
| 286 opts.arch = 'host-native' | |
| 287 | |
| 288 if opts.arch not in ['host-native', 'android-arm', | |
| 289 'android-mips', 'android-x86']: | |
| 290 parser.error('arch must be one of ' | |
| 291 '[host-native,android-arm,android-mips,android-x86]') | |
| 292 | |
| 293 nm_in = opts.nm_in | |
| 294 temp_file = None | |
| 295 if nm_in is None: | |
| 296 if opts.nm_out is None: | |
| 297 temp_file = tempfile.NamedTemporaryFile(prefix='binary_size_nm', delete= False) | |
| 298 nm_in = temp_file.name | |
| 299 else: | |
| 300 nm_in = opts.nm_out | |
| 301 | |
| 302 if opts.verbose: | |
| 303 print 'Running parallel addr2line, dumping symbols to ' + nm_in; | |
| 304 run_pa2l(outfile=nm_in, | |
| 305 library=opts.library, | |
| 306 arch=opts.arch, | |
| 307 threads=opts.threads, | |
| 308 verbose=(opts.verbose is True)) | |
| 309 elif opts.verbose: | |
| 310 print 'Using nm input from ' + nm_in | |
| 311 | |
| 312 if not os.path.exists(opts.destdir): | |
| 313 os.makedirs(opts.destdir, 0755) | |
| 314 | |
| 315 jspath = opts.destdir + '/treemap-dump.js' | |
| 316 nmfile = open(nm_in, 'r') | |
| 317 dump_nm(nmfile, jspath) | |
| 318 if not os.path.exists(opts.destdir + '/webtreemap.js'): | |
| 319 url = 'https://github.com/martine/webtreemap/archive/gh-pages.zip' | |
| 320 tmpdir = tempfile.mkdtemp('binary_size') | |
| 321 try: | |
| 322 cmd = ['wget', '-O', tmpdir + '/webtreemap.zip', url] | |
| 323 return_code = subprocess.call(cmd) | |
| 324 if return_code: | |
| 325 raise RuntimeError('Failed to download: returned ' + str(return_code )) | |
| 326 cmd = ['unzip', '-o', tmpdir + '/webtreemap.zip', '-d', tmpdir] | |
| 327 return_code = subprocess.call(cmd) | |
| 328 if return_code: | |
| 329 raise RuntimeError('Failed to unzip: returned ' + str(return_code)) | |
| 330 | |
| 331 shutil.move(tmpdir + '/webtreemap-gh-pages/COPYING', opts.destdir) | |
| 332 shutil.move(tmpdir + '/webtreemap-gh-pages/webtreemap.js', opts.destdir) | |
| 333 shutil.move(tmpdir + '/webtreemap-gh-pages/webtreemap.css', opts.destdir ) | |
| 334 finally: | |
| 335 shutil.rmtree(tmpdir, ignore_errors=True) | |
| 336 shutil.copy('tools/binary_size/template/index.html', opts.destdir) | |
| 337 if opts.verbose: | |
| 338 print 'Report saved to ' + opts.destdir + '/index.html' | |
| OLD | NEW |