1 #!/usr/bin/env python3 1 #!/usr/bin/env python3
2 # SPDX-License-Identifier: GPL-2.0 2 # SPDX-License-Identifier: GPL-2.0
3 # 3 #
4 # Run a perf script command multiple times in 4 # Run a perf script command multiple times in parallel, using perf script
5 # options --cpu and --time so that each job pr 5 # options --cpu and --time so that each job processes a different chunk
6 # of the data. 6 # of the data.
7 # 7 #
8 # Copyright (c) 2024, Intel Corporation. 8 # Copyright (c) 2024, Intel Corporation.
9 9
10 import subprocess 10 import subprocess
11 import argparse 11 import argparse
12 import pathlib 12 import pathlib
13 import shlex 13 import shlex
14 import time 14 import time
15 import copy 15 import copy
16 import sys 16 import sys
17 import os 17 import os
18 import re 18 import re
19 19
20 glb_prog_name = "parallel-perf.py" 20 glb_prog_name = "parallel-perf.py"
21 glb_min_interval = 10.0 21 glb_min_interval = 10.0
22 glb_min_samples = 64 22 glb_min_samples = 64
23 23
24 class Verbosity(): 24 class Verbosity():
25 25
26 def __init__(self, quiet=False, verbos 26 def __init__(self, quiet=False, verbose=False, debug=False):
27 self.normal = True 27 self.normal = True
28 self.verbose = verbose 28 self.verbose = verbose
29 self.debug = debug 29 self.debug = debug
30 self.self_test = True 30 self.self_test = True
31 if self.debug: 31 if self.debug:
32 self.verbose = True 32 self.verbose = True
33 if self.verbose: 33 if self.verbose:
34 quiet = False 34 quiet = False
35 if quiet: 35 if quiet:
36 self.normal = False 36 self.normal = False
37 37
38 # Manage work (Start/Wait/Kill), as represente 38 # Manage work (Start/Wait/Kill), as represented by a subprocess.Popen command
39 class Work(): 39 class Work():
40 40
41 def __init__(self, cmd, pipe_to, outpu 41 def __init__(self, cmd, pipe_to, output_dir="."):
42 self.popen = None 42 self.popen = None
43 self.consumer = None 43 self.consumer = None
44 self.cmd = cmd 44 self.cmd = cmd
45 self.pipe_to = pipe_to 45 self.pipe_to = pipe_to
46 self.output_dir = output_dir 46 self.output_dir = output_dir
47 self.cmdout_name = f"{output_d 47 self.cmdout_name = f"{output_dir}/cmd.txt"
48 self.stdout_name = f"{output_d 48 self.stdout_name = f"{output_dir}/out.txt"
49 self.stderr_name = f"{output_d 49 self.stderr_name = f"{output_dir}/err.txt"
50 50
51 def Command(self): 51 def Command(self):
52 sh_cmd = [ shlex.quote(x) for 52 sh_cmd = [ shlex.quote(x) for x in self.cmd ]
53 return " ".join(self.cmd) 53 return " ".join(self.cmd)
54 54
55 def Stdout(self): 55 def Stdout(self):
56 return open(self.stdout_name, 56 return open(self.stdout_name, "w")
57 57
58 def Stderr(self): 58 def Stderr(self):
59 return open(self.stderr_name, 59 return open(self.stderr_name, "w")
60 60
61 def CreateOutputDir(self): 61 def CreateOutputDir(self):
62 pathlib.Path(self.output_dir). 62 pathlib.Path(self.output_dir).mkdir(parents=True, exist_ok=True)
63 63
64 def Start(self): 64 def Start(self):
65 if self.popen: 65 if self.popen:
66 return 66 return
67 self.CreateOutputDir() 67 self.CreateOutputDir()
68 with open(self.cmdout_name, "w 68 with open(self.cmdout_name, "w") as f:
69 f.write(self.Command() 69 f.write(self.Command())
70 f.write("\n") 70 f.write("\n")
71 stdout = self.Stdout() 71 stdout = self.Stdout()
72 stderr = self.Stderr() 72 stderr = self.Stderr()
73 if self.pipe_to: 73 if self.pipe_to:
74 self.popen = subproces 74 self.popen = subprocess.Popen(self.cmd, stdout=subprocess.PIPE, stderr=stderr)
75 args = shlex.split(sel 75 args = shlex.split(self.pipe_to)
76 self.consumer = subpro 76 self.consumer = subprocess.Popen(args, stdin=self.popen.stdout, stdout=stdout, stderr=stderr)
77 else: 77 else:
78 self.popen = subproces 78 self.popen = subprocess.Popen(self.cmd, stdout=stdout, stderr=stderr)
79 79
80 def RemoveEmptyErrFile(self): 80 def RemoveEmptyErrFile(self):
81 if os.path.exists(self.stderr_ 81 if os.path.exists(self.stderr_name):
82 if os.path.getsize(sel 82 if os.path.getsize(self.stderr_name) == 0:
83 os.unlink(self 83 os.unlink(self.stderr_name)
84 84
85 def Errors(self): 85 def Errors(self):
86 if os.path.exists(self.stderr_ 86 if os.path.exists(self.stderr_name):
87 if os.path.getsize(sel 87 if os.path.getsize(self.stderr_name) != 0:
88 return [ f"Non 88 return [ f"Non-empty error file {self.stderr_name}" ]
89 return [] 89 return []
90 90
91 def TidyUp(self): 91 def TidyUp(self):
92 self.RemoveEmptyErrFile() 92 self.RemoveEmptyErrFile()
93 93
94 def RawPollWait(self, p, wait): 94 def RawPollWait(self, p, wait):
95 if wait: 95 if wait:
96 return p.wait() 96 return p.wait()
97 return p.poll() 97 return p.poll()
98 98
99 def Poll(self, wait=False): 99 def Poll(self, wait=False):
100 if not self.popen: 100 if not self.popen:
101 return None 101 return None
102 result = self.RawPollWait(self 102 result = self.RawPollWait(self.popen, wait)
103 if self.consumer: 103 if self.consumer:
104 res = result 104 res = result
105 result = self.RawPollW 105 result = self.RawPollWait(self.consumer, wait)
106 if result != None and 106 if result != None and res == None:
107 self.popen.kil 107 self.popen.kill()
108 result = None 108 result = None
109 elif result == 0 and r 109 elif result == 0 and res != None and res != 0:
110 result = res 110 result = res
111 if result != None: 111 if result != None:
112 self.TidyUp() 112 self.TidyUp()
113 return result 113 return result
114 114
115 def Wait(self): 115 def Wait(self):
116 return self.Poll(wait=True) 116 return self.Poll(wait=True)
117 117
118 def Kill(self): 118 def Kill(self):
119 if not self.popen: 119 if not self.popen:
120 return 120 return
121 self.popen.kill() 121 self.popen.kill()
122 if self.consumer: 122 if self.consumer:
123 self.consumer.kill() 123 self.consumer.kill()
124 124
125 def KillWork(worklist, verbosity): 125 def KillWork(worklist, verbosity):
126 for w in worklist: 126 for w in worklist:
127 w.Kill() 127 w.Kill()
128 for w in worklist: 128 for w in worklist:
129 w.Wait() 129 w.Wait()
130 130
131 def NumberOfCPUs(): 131 def NumberOfCPUs():
132 return os.sysconf("SC_NPROCESSORS_ONLN 132 return os.sysconf("SC_NPROCESSORS_ONLN")
133 133
134 def NanoSecsToSecsStr(x): 134 def NanoSecsToSecsStr(x):
135 if x == None: 135 if x == None:
136 return "" 136 return ""
137 x = str(x) 137 x = str(x)
138 if len(x) < 10: 138 if len(x) < 10:
139 x = "0" * (10 - len(x)) + x 139 x = "0" * (10 - len(x)) + x
140 return x[:len(x) - 9] + "." + x[-9:] 140 return x[:len(x) - 9] + "." + x[-9:]
141 141
142 def InsertOptionAfter(cmd, option, after): 142 def InsertOptionAfter(cmd, option, after):
143 try: 143 try:
144 pos = cmd.index(after) 144 pos = cmd.index(after)
145 cmd.insert(pos + 1, option) 145 cmd.insert(pos + 1, option)
146 except: 146 except:
147 cmd.append(option) 147 cmd.append(option)
148 148
149 def CreateWorkList(cmd, pipe_to, output_dir, c 149 def CreateWorkList(cmd, pipe_to, output_dir, cpus, time_ranges_by_cpu):
150 max_len = len(str(cpus[-1])) 150 max_len = len(str(cpus[-1]))
151 cpu_dir_fmt = f"cpu-%.{max_len}u" 151 cpu_dir_fmt = f"cpu-%.{max_len}u"
152 worklist = [] 152 worklist = []
153 pos = 0 153 pos = 0
154 for cpu in cpus: 154 for cpu in cpus:
155 if cpu >= 0: 155 if cpu >= 0:
156 cpu_dir = os.path.join 156 cpu_dir = os.path.join(output_dir, cpu_dir_fmt % cpu)
157 cpu_option = f"--cpu={ 157 cpu_option = f"--cpu={cpu}"
158 else: 158 else:
159 cpu_dir = output_dir 159 cpu_dir = output_dir
160 cpu_option = None 160 cpu_option = None
161 161
162 tr_dir_fmt = "time-range" 162 tr_dir_fmt = "time-range"
163 163
164 if len(time_ranges_by_cpu) > 1 164 if len(time_ranges_by_cpu) > 1:
165 time_ranges = time_ran 165 time_ranges = time_ranges_by_cpu[pos]
166 tr_dir_fmt += f"-{pos} 166 tr_dir_fmt += f"-{pos}"
167 pos += 1 167 pos += 1
168 else: 168 else:
169 time_ranges = time_ran 169 time_ranges = time_ranges_by_cpu[0]
170 170
171 max_len = len(str(len(time_ran 171 max_len = len(str(len(time_ranges)))
172 tr_dir_fmt += f"-%.{max_len}u" 172 tr_dir_fmt += f"-%.{max_len}u"
173 173
174 i = 0 174 i = 0
175 for r in time_ranges: 175 for r in time_ranges:
176 if r == [None, None]: 176 if r == [None, None]:
177 time_option = 177 time_option = None
178 work_output_di 178 work_output_dir = cpu_dir
179 else: 179 else:
180 time_option = 180 time_option = "--time=" + NanoSecsToSecsStr(r[0]) + "," + NanoSecsToSecsStr(r[1])
181 work_output_di 181 work_output_dir = os.path.join(cpu_dir, tr_dir_fmt % i)
182 i += 1 182 i += 1
183 work_cmd = list(cmd) 183 work_cmd = list(cmd)
184 if time_option != None 184 if time_option != None:
185 InsertOptionAf 185 InsertOptionAfter(work_cmd, time_option, "script")
186 if cpu_option != None: 186 if cpu_option != None:
187 InsertOptionAf 187 InsertOptionAfter(work_cmd, cpu_option, "script")
188 w = Work(work_cmd, pip 188 w = Work(work_cmd, pipe_to, work_output_dir)
189 worklist.append(w) 189 worklist.append(w)
190 return worklist 190 return worklist
191 191
192 def DoRunWork(worklist, nr_jobs, verbosity): 192 def DoRunWork(worklist, nr_jobs, verbosity):
193 nr_to_do = len(worklist) 193 nr_to_do = len(worklist)
194 not_started = list(worklist) 194 not_started = list(worklist)
195 running = [] 195 running = []
196 done = [] 196 done = []
197 chg = False 197 chg = False
198 while True: 198 while True:
199 nr_done = len(done) 199 nr_done = len(done)
200 if chg and verbosity.normal: 200 if chg and verbosity.normal:
201 nr_run = len(running) 201 nr_run = len(running)
202 print(f"\rThere are {n 202 print(f"\rThere are {nr_to_do} jobs: {nr_done} completed, {nr_run} running", flush=True, end=" ")
203 if verbosity.verbose: 203 if verbosity.verbose:
204 print() 204 print()
205 chg = False 205 chg = False
206 if nr_done == nr_to_do: 206 if nr_done == nr_to_do:
207 break 207 break
208 while len(running) < nr_jobs a 208 while len(running) < nr_jobs and len(not_started):
209 w = not_started.pop(0) 209 w = not_started.pop(0)
210 running.append(w) 210 running.append(w)
211 if verbosity.verbose: 211 if verbosity.verbose:
212 print("Startin 212 print("Starting:", w.Command())
213 w.Start() 213 w.Start()
214 chg = True 214 chg = True
215 if len(running): 215 if len(running):
216 time.sleep(0.1) 216 time.sleep(0.1)
217 finished = [] 217 finished = []
218 not_finished = [] 218 not_finished = []
219 while len(running): 219 while len(running):
220 w = running.pop(0) 220 w = running.pop(0)
221 r = w.Poll() 221 r = w.Poll()
222 if r == None: 222 if r == None:
223 not_finished.a 223 not_finished.append(w)
224 continue 224 continue
225 if r == 0: 225 if r == 0:
226 if verbosity.v 226 if verbosity.verbose:
227 print( 227 print("Finished:", w.Command())
228 finished.appen 228 finished.append(w)
229 chg = True 229 chg = True
230 continue 230 continue
231 if verbosity.normal an 231 if verbosity.normal and not verbosity.verbose:
232 print() 232 print()
233 print("Job failed!\n 233 print("Job failed!\n return code:", r, "\n command: ", w.Command())
234 if w.pipe_to: 234 if w.pipe_to:
235 print(" pip 235 print(" piped to: ", w.pipe_to)
236 print("Killing outstan 236 print("Killing outstanding jobs")
237 KillWork(not_finished, 237 KillWork(not_finished, verbosity)
238 KillWork(running, verb 238 KillWork(running, verbosity)
239 return False 239 return False
240 running = not_finished 240 running = not_finished
241 done += finished 241 done += finished
242 errorlist = [] 242 errorlist = []
243 for w in worklist: 243 for w in worklist:
244 errorlist += w.Errors() 244 errorlist += w.Errors()
245 if len(errorlist): 245 if len(errorlist):
246 print("Errors:") 246 print("Errors:")
247 for e in errorlist: 247 for e in errorlist:
248 print(e) 248 print(e)
249 elif verbosity.normal: 249 elif verbosity.normal:
250 print("\r"," "*50, "\rAll jobs 250 print("\r"," "*50, "\rAll jobs finished successfully", flush=True)
251 return True 251 return True
252 252
253 def RunWork(worklist, nr_jobs=NumberOfCPUs(), 253 def RunWork(worklist, nr_jobs=NumberOfCPUs(), verbosity=Verbosity()):
254 try: 254 try:
255 return DoRunWork(worklist, nr_ 255 return DoRunWork(worklist, nr_jobs, verbosity)
256 except: 256 except:
257 for w in worklist: 257 for w in worklist:
258 w.Kill() 258 w.Kill()
259 raise 259 raise
260 return True 260 return True
261 261
262 def ReadHeader(perf, file_name): 262 def ReadHeader(perf, file_name):
263 return subprocess.Popen([perf, "script 263 return subprocess.Popen([perf, "script", "--header-only", "--input", file_name], stdout=subprocess.PIPE).stdout.read().decode("utf-8")
264 264
265 def ParseHeader(hdr): 265 def ParseHeader(hdr):
266 result = {} 266 result = {}
267 lines = hdr.split("\n") 267 lines = hdr.split("\n")
268 for line in lines: 268 for line in lines:
269 if ":" in line and line[0] == 269 if ":" in line and line[0] == "#":
270 pos = line.index(":") 270 pos = line.index(":")
271 name = line[1:pos-1].s 271 name = line[1:pos-1].strip()
272 value = line[pos+1:].s 272 value = line[pos+1:].strip()
273 if name in result: 273 if name in result:
274 orig_name = na 274 orig_name = name
275 nr = 2 275 nr = 2
276 while True: 276 while True:
277 name = 277 name = f"{orig_name} {nr}"
278 if nam 278 if name not in result:
279 279 break
280 nr += 280 nr += 1
281 result[name] = value 281 result[name] = value
282 return result 282 return result
283 283
284 def HeaderField(hdr_dict, hdr_fld): 284 def HeaderField(hdr_dict, hdr_fld):
285 if hdr_fld not in hdr_dict: 285 if hdr_fld not in hdr_dict:
286 raise Exception(f"'{hdr_fld}' 286 raise Exception(f"'{hdr_fld}' missing from header information")
287 return hdr_dict[hdr_fld] 287 return hdr_dict[hdr_fld]
288 288
289 # Represent the position of an option within a 289 # Represent the position of an option within a command string
290 # and provide the option value and/or remove t 290 # and provide the option value and/or remove the option
291 class OptPos(): 291 class OptPos():
292 292
293 def Init(self, opt_element=-1, value_e 293 def Init(self, opt_element=-1, value_element=-1, opt_pos=-1, value_pos=-1, error=None):
294 self.opt_element = opt_element 294 self.opt_element = opt_element # list element that contains option
295 self.value_element = value_ele 295 self.value_element = value_element # list element that contains option value
296 self.opt_pos = opt_pos 296 self.opt_pos = opt_pos # string position of option
297 self.value_pos = value_pos 297 self.value_pos = value_pos # string position of value
298 self.error = error 298 self.error = error # error message string
299 299
300 def __init__(self, args, short_name, l 300 def __init__(self, args, short_name, long_name, default=None):
301 self.args = list(args) 301 self.args = list(args)
302 self.default = default 302 self.default = default
303 n = 2 + len(long_name) 303 n = 2 + len(long_name)
304 m = len(short_name) 304 m = len(short_name)
305 pos = -1 305 pos = -1
306 for opt in args: 306 for opt in args:
307 pos += 1 307 pos += 1
308 if m and opt[:2] == f" 308 if m and opt[:2] == f"-{short_name}":
309 if len(opt) == 309 if len(opt) == 2:
310 if pos 310 if pos + 1 < len(args):
311 311 self.Init(pos, pos + 1, 0, 0)
312 else: 312 else:
313 313 self.Init(error = f"-{short_name} option missing value")
314 else: 314 else:
315 self.I 315 self.Init(pos, pos, 0, 2)
316 return 316 return
317 if opt[:n] == f"--{lon 317 if opt[:n] == f"--{long_name}":
318 if len(opt) == 318 if len(opt) == n:
319 if pos 319 if pos + 1 < len(args):
320 320 self.Init(pos, pos + 1, 0, 0)
321 else: 321 else:
322 322 self.Init(error = f"--{long_name} option missing value")
323 elif opt[n] == 323 elif opt[n] == "=":
324 self.I 324 self.Init(pos, pos, 0, n + 1)
325 else: 325 else:
326 self.I 326 self.Init(error = f"--{long_name} option expected '='")
327 return 327 return
328 if m and opt[:1] == "- 328 if m and opt[:1] == "-" and opt[:2] != "--" and short_name in opt:
329 ipos = opt.ind 329 ipos = opt.index(short_name)
330 if "-" in opt[ 330 if "-" in opt[1:]:
331 hpos = 331 hpos = opt[1:].index("-")
332 if hpo 332 if hpos < ipos:
333 333 continue
334 if ipos + 1 == 334 if ipos + 1 == len(opt):
335 if pos 335 if pos + 1 < len(args):
336 336 self.Init(pos, pos + 1, ipos, 0)
337 else: 337 else:
338 338 self.Init(error = f"-{short_name} option missing value")
339 else: 339 else:
340 self.I 340 self.Init(pos, pos, ipos, ipos + 1)
341 return 341 return
342 self.Init() 342 self.Init()
343 343
344 def Value(self): 344 def Value(self):
345 if self.opt_element >= 0: 345 if self.opt_element >= 0:
346 if self.opt_element != 346 if self.opt_element != self.value_element:
347 return self.ar 347 return self.args[self.value_element]
348 else: 348 else:
349 return self.ar 349 return self.args[self.value_element][self.value_pos:]
350 return self.default 350 return self.default
351 351
352 def Remove(self, args): 352 def Remove(self, args):
353 if self.opt_element == -1: 353 if self.opt_element == -1:
354 return 354 return
355 if self.opt_element != self.va 355 if self.opt_element != self.value_element:
356 del args[self.value_el 356 del args[self.value_element]
357 if self.opt_pos: 357 if self.opt_pos:
358 args[self.opt_element] 358 args[self.opt_element] = args[self.opt_element][:self.opt_pos]
359 else: 359 else:
360 del args[self.opt_elem 360 del args[self.opt_element]
361 361
362 def DetermineInputFileName(cmd): 362 def DetermineInputFileName(cmd):
363 p = OptPos(cmd, "i", "input", "perf.da 363 p = OptPos(cmd, "i", "input", "perf.data")
364 if p.error: 364 if p.error:
365 raise Exception(f"perf command 365 raise Exception(f"perf command {p.error}")
366 file_name = p.Value() 366 file_name = p.Value()
367 if not os.path.exists(file_name): 367 if not os.path.exists(file_name):
368 raise Exception(f"perf command 368 raise Exception(f"perf command input file '{file_name}' not found")
369 return file_name 369 return file_name
370 370
371 def ReadOption(args, short_name, long_name, er 371 def ReadOption(args, short_name, long_name, err_prefix, remove=False):
372 p = OptPos(args, short_name, long_name 372 p = OptPos(args, short_name, long_name)
373 if p.error: 373 if p.error:
374 raise Exception(f"{err_prefix} 374 raise Exception(f"{err_prefix}{p.error}")
375 value = p.Value() 375 value = p.Value()
376 if remove: 376 if remove:
377 p.Remove(args) 377 p.Remove(args)
378 return value 378 return value
379 379
380 def ExtractOption(args, short_name, long_name, 380 def ExtractOption(args, short_name, long_name, err_prefix):
381 return ReadOption(args, short_name, lo 381 return ReadOption(args, short_name, long_name, err_prefix, True)
382 382
383 def ReadPerfOption(args, short_name, long_name 383 def ReadPerfOption(args, short_name, long_name):
384 return ReadOption(args, short_name, lo 384 return ReadOption(args, short_name, long_name, "perf command ")
385 385
386 def ExtractPerfOption(args, short_name, long_n 386 def ExtractPerfOption(args, short_name, long_name):
387 return ExtractOption(args, short_name, 387 return ExtractOption(args, short_name, long_name, "perf command ")
388 388
389 def PerfDoubleQuickCommands(cmd, file_name): 389 def PerfDoubleQuickCommands(cmd, file_name):
390 cpu_str = ReadPerfOption(cmd, "C", "cp 390 cpu_str = ReadPerfOption(cmd, "C", "cpu")
391 time_str = ReadPerfOption(cmd, "", "ti 391 time_str = ReadPerfOption(cmd, "", "time")
392 # Use double-quick sampling to determi 392 # Use double-quick sampling to determine trace data density
393 times_cmd = ["perf", "script", "--ns", 393 times_cmd = ["perf", "script", "--ns", "--input", file_name, "--itrace=qqi"]
394 if cpu_str != None and cpu_str != "": 394 if cpu_str != None and cpu_str != "":
395 times_cmd.append(f"--cpu={cpu_ 395 times_cmd.append(f"--cpu={cpu_str}")
396 if time_str != None and time_str != "" 396 if time_str != None and time_str != "":
397 times_cmd.append(f"--time={tim 397 times_cmd.append(f"--time={time_str}")
398 cnts_cmd = list(times_cmd) 398 cnts_cmd = list(times_cmd)
399 cnts_cmd.append("-Fcpu") 399 cnts_cmd.append("-Fcpu")
400 times_cmd.append("-Fcpu,time") 400 times_cmd.append("-Fcpu,time")
401 return cnts_cmd, times_cmd 401 return cnts_cmd, times_cmd
402 402
403 class CPUTimeRange(): 403 class CPUTimeRange():
404 def __init__(self, cpu): 404 def __init__(self, cpu):
405 self.cpu = cpu 405 self.cpu = cpu
406 self.sample_cnt = 0 406 self.sample_cnt = 0
407 self.time_ranges = None 407 self.time_ranges = None
408 self.interval = 0 408 self.interval = 0
409 self.interval_remaining = 0 409 self.interval_remaining = 0
410 self.remaining = 0 410 self.remaining = 0
411 self.tr_pos = 0 411 self.tr_pos = 0
412 412
413 def CalcTimeRangesByCPU(line, cpu, cpu_time_ra 413 def CalcTimeRangesByCPU(line, cpu, cpu_time_ranges, max_time):
414 cpu_time_range = cpu_time_ranges[cpu] 414 cpu_time_range = cpu_time_ranges[cpu]
415 cpu_time_range.remaining -= 1 415 cpu_time_range.remaining -= 1
416 cpu_time_range.interval_remaining -= 1 416 cpu_time_range.interval_remaining -= 1
417 if cpu_time_range.remaining == 0: 417 if cpu_time_range.remaining == 0:
418 cpu_time_range.time_ranges[cpu 418 cpu_time_range.time_ranges[cpu_time_range.tr_pos][1] = max_time
419 return 419 return
420 if cpu_time_range.interval_remaining = 420 if cpu_time_range.interval_remaining == 0:
421 time = TimeVal(line[1][:-1], 0 421 time = TimeVal(line[1][:-1], 0)
422 time_ranges = cpu_time_range.t 422 time_ranges = cpu_time_range.time_ranges
423 time_ranges[cpu_time_range.tr_ 423 time_ranges[cpu_time_range.tr_pos][1] = time - 1
424 time_ranges.append([time, max_ 424 time_ranges.append([time, max_time])
425 cpu_time_range.tr_pos += 1 425 cpu_time_range.tr_pos += 1
426 cpu_time_range.interval_remain 426 cpu_time_range.interval_remaining = cpu_time_range.interval
427 427
428 def CountSamplesByCPU(line, cpu, cpu_time_rang 428 def CountSamplesByCPU(line, cpu, cpu_time_ranges):
429 try: 429 try:
430 cpu_time_ranges[cpu].sample_cn 430 cpu_time_ranges[cpu].sample_cnt += 1
431 except: 431 except:
432 print("exception") 432 print("exception")
433 print("cpu", cpu) 433 print("cpu", cpu)
434 print("len(cpu_time_ranges)", 434 print("len(cpu_time_ranges)", len(cpu_time_ranges))
435 raise 435 raise
436 436
437 def ProcessCommandOutputLines(cmd, per_cpu, fn 437 def ProcessCommandOutputLines(cmd, per_cpu, fn, *x):
438 # Assume CPU number is at beginning of 438 # Assume CPU number is at beginning of line and enclosed by []
439 pat = re.compile(r"\s*\[[0-9]+\]") 439 pat = re.compile(r"\s*\[[0-9]+\]")
440 p = subprocess.Popen(cmd, stdout=subpr 440 p = subprocess.Popen(cmd, stdout=subprocess.PIPE)
441 while True: 441 while True:
442 line = p.stdout.readline() 442 line = p.stdout.readline()
443 if line: 443 if line:
444 line = line.decode("ut 444 line = line.decode("utf-8")
445 if pat.match(line): 445 if pat.match(line):
446 line = line.sp 446 line = line.split()
447 if per_cpu: 447 if per_cpu:
448 # Assu 448 # Assumes CPU number is enclosed by []
449 cpu = 449 cpu = int(line[0][1:-1])
450 else: 450 else:
451 cpu = 451 cpu = 0
452 fn(line, cpu, 452 fn(line, cpu, *x)
453 else: 453 else:
454 break 454 break
455 p.wait() 455 p.wait()
456 456
457 def IntersectTimeRanges(new_time_ranges, time_ 457 def IntersectTimeRanges(new_time_ranges, time_ranges):
458 pos = 0 458 pos = 0
459 new_pos = 0 459 new_pos = 0
460 # Can assume len(time_ranges) != 0 and 460 # Can assume len(time_ranges) != 0 and len(new_time_ranges) != 0
461 # Note also, there *must* be at least 461 # Note also, there *must* be at least one intersection.
462 while pos < len(time_ranges) and new_p 462 while pos < len(time_ranges) and new_pos < len(new_time_ranges):
463 # new end < old start => no in 463 # new end < old start => no intersection, remove new
464 if new_time_ranges[new_pos][1] 464 if new_time_ranges[new_pos][1] < time_ranges[pos][0]:
465 del new_time_ranges[ne 465 del new_time_ranges[new_pos]
466 continue 466 continue
467 # new start > old end => no in 467 # new start > old end => no intersection, check next
468 if new_time_ranges[new_pos][0] 468 if new_time_ranges[new_pos][0] > time_ranges[pos][1]:
469 pos += 1 469 pos += 1
470 if pos < len(time_rang 470 if pos < len(time_ranges):
471 continue 471 continue
472 # no next, so remove r 472 # no next, so remove remaining
473 while new_pos < len(ne 473 while new_pos < len(new_time_ranges):
474 del new_time_r 474 del new_time_ranges[new_pos]
475 return 475 return
476 # Found an intersection 476 # Found an intersection
477 # new start < old start => adj 477 # new start < old start => adjust new start = old start
478 if new_time_ranges[new_pos][0] 478 if new_time_ranges[new_pos][0] < time_ranges[pos][0]:
479 new_time_ranges[new_po 479 new_time_ranges[new_pos][0] = time_ranges[pos][0]
480 # new end > old end => keep th 480 # new end > old end => keep the overlap, insert the remainder
481 if new_time_ranges[new_pos][1] 481 if new_time_ranges[new_pos][1] > time_ranges[pos][1]:
482 r = [ time_ranges[pos] 482 r = [ time_ranges[pos][1] + 1, new_time_ranges[new_pos][1] ]
483 new_time_ranges[new_po 483 new_time_ranges[new_pos][1] = time_ranges[pos][1]
484 new_pos += 1 484 new_pos += 1
485 new_time_ranges.insert 485 new_time_ranges.insert(new_pos, r)
486 continue 486 continue
487 # new [start, end] is within o 487 # new [start, end] is within old [start, end]
488 new_pos += 1 488 new_pos += 1
489 489
490 def SplitTimeRangesByTraceDataDensity(time_ran 490 def SplitTimeRangesByTraceDataDensity(time_ranges, cpus, nr, cmd, file_name, per_cpu, min_size, min_interval, verbosity):
491 if verbosity.normal: 491 if verbosity.normal:
492 print("\rAnalyzing...", flush= 492 print("\rAnalyzing...", flush=True, end=" ")
493 if verbosity.verbose: 493 if verbosity.verbose:
494 print() 494 print()
495 cnts_cmd, times_cmd = PerfDoubleQuickC 495 cnts_cmd, times_cmd = PerfDoubleQuickCommands(cmd, file_name)
496 496
497 nr_cpus = cpus[-1] + 1 if per_cpu else 497 nr_cpus = cpus[-1] + 1 if per_cpu else 1
498 if per_cpu: 498 if per_cpu:
499 nr_cpus = cpus[-1] + 1 499 nr_cpus = cpus[-1] + 1
500 cpu_time_ranges = [ CPUTimeRan 500 cpu_time_ranges = [ CPUTimeRange(cpu) for cpu in range(nr_cpus) ]
501 else: 501 else:
502 nr_cpus = 1 502 nr_cpus = 1
503 cpu_time_ranges = [ CPUTimeRan 503 cpu_time_ranges = [ CPUTimeRange(-1) ]
504 504
505 if verbosity.debug: 505 if verbosity.debug:
506 print("nr_cpus", nr_cpus) 506 print("nr_cpus", nr_cpus)
507 print("cnts_cmd", cnts_cmd) 507 print("cnts_cmd", cnts_cmd)
508 print("times_cmd", times_cmd) 508 print("times_cmd", times_cmd)
509 509
510 # Count the number of "double quick" s 510 # Count the number of "double quick" samples per CPU
511 ProcessCommandOutputLines(cnts_cmd, pe 511 ProcessCommandOutputLines(cnts_cmd, per_cpu, CountSamplesByCPU, cpu_time_ranges)
512 512
513 tot = 0 513 tot = 0
514 mx = 0 514 mx = 0
515 for cpu_time_range in cpu_time_ranges: 515 for cpu_time_range in cpu_time_ranges:
516 cnt = cpu_time_range.sample_cn 516 cnt = cpu_time_range.sample_cnt
517 tot += cnt 517 tot += cnt
518 if cnt > mx: 518 if cnt > mx:
519 mx = cnt 519 mx = cnt
520 if verbosity.debug: 520 if verbosity.debug:
521 print("cpu:", cpu_time 521 print("cpu:", cpu_time_range.cpu, "sample_cnt", cnt)
522 522
523 if min_size < 1: 523 if min_size < 1:
524 min_size = 1 524 min_size = 1
525 525
526 if mx < min_size: 526 if mx < min_size:
527 # Too little data to be worth 527 # Too little data to be worth splitting
528 if verbosity.debug: 528 if verbosity.debug:
529 print("Too little data 529 print("Too little data to split by time")
530 if nr == 0: 530 if nr == 0:
531 nr = 1 531 nr = 1
532 return [ SplitTimeRangesIntoN( 532 return [ SplitTimeRangesIntoN(time_ranges, nr, min_interval) ]
533 533
534 if nr: 534 if nr:
535 divisor = nr 535 divisor = nr
536 min_size = 1 536 min_size = 1
537 else: 537 else:
538 divisor = NumberOfCPUs() 538 divisor = NumberOfCPUs()
539 539
540 interval = int(round(tot / divisor, 0) 540 interval = int(round(tot / divisor, 0))
541 if interval < min_size: 541 if interval < min_size:
542 interval = min_size 542 interval = min_size
543 543
544 if verbosity.debug: 544 if verbosity.debug:
545 print("divisor", divisor) 545 print("divisor", divisor)
546 print("min_size", min_size) 546 print("min_size", min_size)
547 print("interval", interval) 547 print("interval", interval)
548 548
549 min_time = time_ranges[0][0] 549 min_time = time_ranges[0][0]
550 max_time = time_ranges[-1][1] 550 max_time = time_ranges[-1][1]
551 551
552 for cpu_time_range in cpu_time_ranges: 552 for cpu_time_range in cpu_time_ranges:
553 cnt = cpu_time_range.sample_cn 553 cnt = cpu_time_range.sample_cnt
554 if cnt == 0: 554 if cnt == 0:
555 cpu_time_range.time_ra 555 cpu_time_range.time_ranges = copy.deepcopy(time_ranges)
556 continue 556 continue
557 # Adjust target interval for C 557 # Adjust target interval for CPU to give approximately equal interval sizes
558 # Determine number of interval 558 # Determine number of intervals, rounding to nearest integer
559 n = int(round(cnt / interval, 559 n = int(round(cnt / interval, 0))
560 if n < 1: 560 if n < 1:
561 n = 1 561 n = 1
562 # Determine interval size, rou 562 # Determine interval size, rounding up
563 d, m = divmod(cnt, n) 563 d, m = divmod(cnt, n)
564 if m: 564 if m:
565 d += 1 565 d += 1
566 cpu_time_range.interval = d 566 cpu_time_range.interval = d
567 cpu_time_range.interval_remain 567 cpu_time_range.interval_remaining = d
568 cpu_time_range.remaining = cnt 568 cpu_time_range.remaining = cnt
569 # Init. time ranges for each C 569 # Init. time ranges for each CPU with the start time
570 cpu_time_range.time_ranges = [ 570 cpu_time_range.time_ranges = [ [min_time, max_time] ]
571 571
572 # Set time ranges so that the same num 572 # Set time ranges so that the same number of "double quick" samples
573 # will fall into each time range. 573 # will fall into each time range.
574 ProcessCommandOutputLines(times_cmd, p 574 ProcessCommandOutputLines(times_cmd, per_cpu, CalcTimeRangesByCPU, cpu_time_ranges, max_time)
575 575
576 for cpu_time_range in cpu_time_ranges: 576 for cpu_time_range in cpu_time_ranges:
577 if cpu_time_range.sample_cnt: 577 if cpu_time_range.sample_cnt:
578 IntersectTimeRanges(cp 578 IntersectTimeRanges(cpu_time_range.time_ranges, time_ranges)
579 579
580 return [cpu_time_ranges[cpu].time_rang 580 return [cpu_time_ranges[cpu].time_ranges for cpu in cpus]
581 581
582 def SplitSingleTimeRangeIntoN(time_range, n): 582 def SplitSingleTimeRangeIntoN(time_range, n):
583 if n <= 1: 583 if n <= 1:
584 return [time_range] 584 return [time_range]
585 start = time_range[0] 585 start = time_range[0]
586 end = time_range[1] 586 end = time_range[1]
587 duration = int((end - start + 1) / n) 587 duration = int((end - start + 1) / n)
588 if duration < 1: 588 if duration < 1:
589 return [time_range] 589 return [time_range]
590 time_ranges = [] 590 time_ranges = []
591 for i in range(n): 591 for i in range(n):
592 time_ranges.append([start, sta 592 time_ranges.append([start, start + duration - 1])
593 start += duration 593 start += duration
594 time_ranges[-1][1] = end 594 time_ranges[-1][1] = end
595 return time_ranges 595 return time_ranges
596 596
597 def TimeRangeDuration(r): 597 def TimeRangeDuration(r):
598 return r[1] - r[0] + 1 598 return r[1] - r[0] + 1
599 599
600 def TotalDuration(time_ranges): 600 def TotalDuration(time_ranges):
601 duration = 0 601 duration = 0
602 for r in time_ranges: 602 for r in time_ranges:
603 duration += TimeRangeDuration( 603 duration += TimeRangeDuration(r)
604 return duration 604 return duration
605 605
606 def SplitTimeRangesByInterval(time_ranges, int 606 def SplitTimeRangesByInterval(time_ranges, interval):
607 new_ranges = [] 607 new_ranges = []
608 for r in time_ranges: 608 for r in time_ranges:
609 duration = TimeRangeDuration(r 609 duration = TimeRangeDuration(r)
610 n = duration / interval 610 n = duration / interval
611 n = int(round(n, 0)) 611 n = int(round(n, 0))
612 new_ranges += SplitSingleTimeR 612 new_ranges += SplitSingleTimeRangeIntoN(r, n)
613 return new_ranges 613 return new_ranges
614 614
615 def SplitTimeRangesIntoN(time_ranges, n, min_i 615 def SplitTimeRangesIntoN(time_ranges, n, min_interval):
616 if n <= len(time_ranges): 616 if n <= len(time_ranges):
617 return time_ranges 617 return time_ranges
618 duration = TotalDuration(time_ranges) 618 duration = TotalDuration(time_ranges)
619 interval = duration / n 619 interval = duration / n
620 if interval < min_interval: 620 if interval < min_interval:
621 interval = min_interval 621 interval = min_interval
622 return SplitTimeRangesByInterval(time_ 622 return SplitTimeRangesByInterval(time_ranges, interval)
623 623
624 def RecombineTimeRanges(tr): 624 def RecombineTimeRanges(tr):
625 new_tr = copy.deepcopy(tr) 625 new_tr = copy.deepcopy(tr)
626 n = len(new_tr) 626 n = len(new_tr)
627 i = 1 627 i = 1
628 while i < len(new_tr): 628 while i < len(new_tr):
629 # if prev end + 1 == cur start 629 # if prev end + 1 == cur start, combine them
630 if new_tr[i - 1][1] + 1 == new 630 if new_tr[i - 1][1] + 1 == new_tr[i][0]:
631 new_tr[i][0] = new_tr[ 631 new_tr[i][0] = new_tr[i - 1][0]
632 del new_tr[i - 1] 632 del new_tr[i - 1]
633 else: 633 else:
634 i += 1 634 i += 1
635 return new_tr 635 return new_tr
636 636
637 def OpenTimeRangeEnds(time_ranges, min_time, m 637 def OpenTimeRangeEnds(time_ranges, min_time, max_time):
638 if time_ranges[0][0] <= min_time: 638 if time_ranges[0][0] <= min_time:
639 time_ranges[0][0] = None 639 time_ranges[0][0] = None
640 if time_ranges[-1][1] >= max_time: 640 if time_ranges[-1][1] >= max_time:
641 time_ranges[-1][1] = None 641 time_ranges[-1][1] = None
642 642
643 def BadTimeStr(time_str): 643 def BadTimeStr(time_str):
644 raise Exception(f"perf command bad tim 644 raise Exception(f"perf command bad time option: '{time_str}'\nCheck also 'time of first sample' and 'time of last sample' in perf script --header-only")
645 645
646 def ValidateTimeRanges(time_ranges, time_str): 646 def ValidateTimeRanges(time_ranges, time_str):
647 n = len(time_ranges) 647 n = len(time_ranges)
648 for i in range(n): 648 for i in range(n):
649 start = time_ranges[i][0] 649 start = time_ranges[i][0]
650 end = time_ranges[i][1] 650 end = time_ranges[i][1]
651 if i != 0 and start <= time_ra 651 if i != 0 and start <= time_ranges[i - 1][1]:
652 BadTimeStr(time_str) 652 BadTimeStr(time_str)
653 if start > end: 653 if start > end:
654 BadTimeStr(time_str) 654 BadTimeStr(time_str)
655 655
656 def TimeVal(s, dflt): 656 def TimeVal(s, dflt):
657 s = s.strip() 657 s = s.strip()
658 if s == "": 658 if s == "":
659 return dflt 659 return dflt
660 a = s.split(".") 660 a = s.split(".")
661 if len(a) > 2: 661 if len(a) > 2:
662 raise Exception(f"Bad time val 662 raise Exception(f"Bad time value'{s}'")
663 x = int(a[0]) 663 x = int(a[0])
664 if x < 0: 664 if x < 0:
665 raise Exception("Negative time 665 raise Exception("Negative time not allowed")
666 x *= 1000000000 666 x *= 1000000000
667 if len(a) > 1: 667 if len(a) > 1:
668 x += int((a[1] + "000000000")[ 668 x += int((a[1] + "000000000")[:9])
669 return x 669 return x
670 670
671 def BadCPUStr(cpu_str): 671 def BadCPUStr(cpu_str):
672 raise Exception(f"perf command bad cpu 672 raise Exception(f"perf command bad cpu option: '{cpu_str}'\nCheck also 'nrcpus avail' in perf script --header-only")
673 673
674 def ParseTimeStr(time_str, min_time, max_time) 674 def ParseTimeStr(time_str, min_time, max_time):
675 if time_str == None or time_str == "": 675 if time_str == None or time_str == "":
676 return [[min_time, max_time]] 676 return [[min_time, max_time]]
677 time_ranges = [] 677 time_ranges = []
678 for r in time_str.split(): 678 for r in time_str.split():
679 a = r.split(",") 679 a = r.split(",")
680 if len(a) != 2: 680 if len(a) != 2:
681 BadTimeStr(time_str) 681 BadTimeStr(time_str)
682 try: 682 try:
683 start = TimeVal(a[0], 683 start = TimeVal(a[0], min_time)
684 end = TimeVal(a[1], 684 end = TimeVal(a[1], max_time)
685 except: 685 except:
686 BadTimeStr(time_str) 686 BadTimeStr(time_str)
687 time_ranges.append([start, end 687 time_ranges.append([start, end])
688 ValidateTimeRanges(time_ranges, time_s 688 ValidateTimeRanges(time_ranges, time_str)
689 return time_ranges 689 return time_ranges
690 690
691 def ParseCPUStr(cpu_str, nr_cpus): 691 def ParseCPUStr(cpu_str, nr_cpus):
692 if cpu_str == None or cpu_str == "": 692 if cpu_str == None or cpu_str == "":
693 return [-1] 693 return [-1]
694 cpus = [] 694 cpus = []
695 for r in cpu_str.split(","): 695 for r in cpu_str.split(","):
696 a = r.split("-") 696 a = r.split("-")
697 if len(a) < 1 or len(a) > 2: 697 if len(a) < 1 or len(a) > 2:
698 BadCPUStr(cpu_str) 698 BadCPUStr(cpu_str)
699 try: 699 try:
700 start = int(a[0].strip 700 start = int(a[0].strip())
701 if len(a) > 1: 701 if len(a) > 1:
702 end = int(a[1] 702 end = int(a[1].strip())
703 else: 703 else:
704 end = start 704 end = start
705 except: 705 except:
706 BadCPUStr(cpu_str) 706 BadCPUStr(cpu_str)
707 if start < 0 or end < 0 or end 707 if start < 0 or end < 0 or end < start or end >= nr_cpus:
708 BadCPUStr(cpu_str) 708 BadCPUStr(cpu_str)
709 cpus.extend(range(start, end + 709 cpus.extend(range(start, end + 1))
710 cpus = list(set(cpus)) # Remove duplic 710 cpus = list(set(cpus)) # Remove duplicates
711 cpus.sort() 711 cpus.sort()
712 return cpus 712 return cpus
713 713
714 class ParallelPerf(): 714 class ParallelPerf():
715 715
716 def __init__(self, a): 716 def __init__(self, a):
717 for arg_name in vars(a): 717 for arg_name in vars(a):
718 setattr(self, arg_name 718 setattr(self, arg_name, getattr(a, arg_name))
719 self.orig_nr = self.nr 719 self.orig_nr = self.nr
720 self.orig_cmd = list(self.cmd) 720 self.orig_cmd = list(self.cmd)
721 self.perf = self.cmd[0] 721 self.perf = self.cmd[0]
722 if os.path.exists(self.output_ 722 if os.path.exists(self.output_dir):
723 raise Exception(f"Outp 723 raise Exception(f"Output '{self.output_dir}' already exists")
724 if self.jobs < 0 or self.nr < 724 if self.jobs < 0 or self.nr < 0 or self.interval < 0:
725 raise Exception("Bad o 725 raise Exception("Bad options (negative values): try -h option for help")
726 if self.nr != 0 and self.inter 726 if self.nr != 0 and self.interval != 0:
727 raise Exception("Canno 727 raise Exception("Cannot specify number of time subdivisions and time interval")
728 if self.jobs == 0: 728 if self.jobs == 0:
729 self.jobs = NumberOfCP 729 self.jobs = NumberOfCPUs()
730 if self.nr == 0 and self.inter 730 if self.nr == 0 and self.interval == 0:
731 if self.per_cpu: 731 if self.per_cpu:
732 self.nr = 1 732 self.nr = 1
733 else: 733 else:
734 self.nr = self 734 self.nr = self.jobs
735 735
736 def Init(self): 736 def Init(self):
737 if self.verbosity.debug: 737 if self.verbosity.debug:
738 print("cmd", self.cmd) 738 print("cmd", self.cmd)
739 self.file_name = DetermineInpu 739 self.file_name = DetermineInputFileName(self.cmd)
740 self.hdr = ReadHeader(self.per 740 self.hdr = ReadHeader(self.perf, self.file_name)
741 self.hdr_dict = ParseHeader(se 741 self.hdr_dict = ParseHeader(self.hdr)
742 self.cmd_line = HeaderField(se 742 self.cmd_line = HeaderField(self.hdr_dict, "cmdline")
743 743
744 def ExtractTimeInfo(self): 744 def ExtractTimeInfo(self):
745 self.min_time = TimeVal(Header 745 self.min_time = TimeVal(HeaderField(self.hdr_dict, "time of first sample"), 0)
746 self.max_time = TimeVal(Header 746 self.max_time = TimeVal(HeaderField(self.hdr_dict, "time of last sample"), 0)
747 self.time_str = ExtractPerfOpt 747 self.time_str = ExtractPerfOption(self.cmd, "", "time")
748 self.time_ranges = ParseTimeSt 748 self.time_ranges = ParseTimeStr(self.time_str, self.min_time, self.max_time)
749 if self.verbosity.debug: 749 if self.verbosity.debug:
750 print("time_ranges", s 750 print("time_ranges", self.time_ranges)
751 751
752 def ExtractCPUInfo(self): 752 def ExtractCPUInfo(self):
753 if self.per_cpu: 753 if self.per_cpu:
754 nr_cpus = int(HeaderFi 754 nr_cpus = int(HeaderField(self.hdr_dict, "nrcpus avail"))
755 self.cpu_str = Extract 755 self.cpu_str = ExtractPerfOption(self.cmd, "C", "cpu")
756 if self.cpu_str == Non 756 if self.cpu_str == None or self.cpu_str == "":
757 self.cpus = [ 757 self.cpus = [ x for x in range(nr_cpus) ]
758 else: 758 else:
759 self.cpus = Pa 759 self.cpus = ParseCPUStr(self.cpu_str, nr_cpus)
760 else: 760 else:
761 self.cpu_str = None 761 self.cpu_str = None
762 self.cpus = [-1] 762 self.cpus = [-1]
763 if self.verbosity.debug: 763 if self.verbosity.debug:
764 print("cpus", self.cpu 764 print("cpus", self.cpus)
765 765
766 def IsIntelPT(self): 766 def IsIntelPT(self):
767 return self.cmd_line.find("int 767 return self.cmd_line.find("intel_pt") >= 0
768 768
769 def SplitTimeRanges(self): 769 def SplitTimeRanges(self):
770 if self.IsIntelPT() and self.i 770 if self.IsIntelPT() and self.interval == 0:
771 self.split_time_ranges 771 self.split_time_ranges_for_each_cpu = \
772 SplitTimeRange 772 SplitTimeRangesByTraceDataDensity(self.time_ranges, self.cpus, self.orig_nr,
773 773 self.orig_cmd, self.file_name, self.per_cpu,
774 774 self.min_size, self.min_interval, self.verbosity)
775 elif self.nr: 775 elif self.nr:
776 self.split_time_ranges 776 self.split_time_ranges_for_each_cpu = [ SplitTimeRangesIntoN(self.time_ranges, self.nr, self.min_interval) ]
777 else: 777 else:
778 self.split_time_ranges 778 self.split_time_ranges_for_each_cpu = [ SplitTimeRangesByInterval(self.time_ranges, self.interval) ]
779 779
780 def CheckTimeRanges(self): 780 def CheckTimeRanges(self):
781 for tr in self.split_time_rang 781 for tr in self.split_time_ranges_for_each_cpu:
782 # Re-combined time ran 782 # Re-combined time ranges should be the same
783 new_tr = RecombineTime 783 new_tr = RecombineTimeRanges(tr)
784 if new_tr != self.time 784 if new_tr != self.time_ranges:
785 if self.verbos 785 if self.verbosity.debug:
786 print( 786 print("tr", tr)
787 print( 787 print("new_tr", new_tr)
788 raise Exceptio 788 raise Exception("Self test failed!")
789 789
790 def OpenTimeRangeEnds(self): 790 def OpenTimeRangeEnds(self):
791 for time_ranges in self.split_ 791 for time_ranges in self.split_time_ranges_for_each_cpu:
792 OpenTimeRangeEnds(time 792 OpenTimeRangeEnds(time_ranges, self.min_time, self.max_time)
793 793
794 def CreateWorkList(self): 794 def CreateWorkList(self):
795 self.worklist = CreateWorkList 795 self.worklist = CreateWorkList(self.cmd, self.pipe_to, self.output_dir, self.cpus, self.split_time_ranges_for_each_cpu)
796 796
797 def PerfDataRecordedPerCPU(self): 797 def PerfDataRecordedPerCPU(self):
798 if "--per-thread" in self.cmd_ 798 if "--per-thread" in self.cmd_line.split():
799 return False 799 return False
800 return True 800 return True
801 801
802 def DefaultToPerCPU(self): 802 def DefaultToPerCPU(self):
803 # --no-per-cpu option takes pr 803 # --no-per-cpu option takes precedence
804 if self.no_per_cpu: 804 if self.no_per_cpu:
805 return False 805 return False
806 if not self.PerfDataRecordedPe 806 if not self.PerfDataRecordedPerCPU():
807 return False 807 return False
808 # Default to per-cpu for Intel 808 # Default to per-cpu for Intel PT data that was recorded per-cpu,
809 # because decoding can be done 809 # because decoding can be done for each CPU separately.
810 if self.IsIntelPT(): 810 if self.IsIntelPT():
811 return True 811 return True
812 return False 812 return False
813 813
814 def Config(self): 814 def Config(self):
815 self.Init() 815 self.Init()
816 self.ExtractTimeInfo() 816 self.ExtractTimeInfo()
817 if not self.per_cpu: 817 if not self.per_cpu:
818 self.per_cpu = self.De 818 self.per_cpu = self.DefaultToPerCPU()
819 if self.verbosity.debug: 819 if self.verbosity.debug:
820 print("per_cpu", self. 820 print("per_cpu", self.per_cpu)
821 self.ExtractCPUInfo() 821 self.ExtractCPUInfo()
822 self.SplitTimeRanges() 822 self.SplitTimeRanges()
823 if self.verbosity.self_test: 823 if self.verbosity.self_test:
824 self.CheckTimeRanges() 824 self.CheckTimeRanges()
825 # Prefer open-ended time range 825 # Prefer open-ended time range to starting / ending with min_time / max_time resp.
826 self.OpenTimeRangeEnds() 826 self.OpenTimeRangeEnds()
827 self.CreateWorkList() 827 self.CreateWorkList()
828 828
829 def Run(self): 829 def Run(self):
830 if self.dry_run: 830 if self.dry_run:
831 print(len(self.worklis 831 print(len(self.worklist),"jobs:")
832 for w in self.worklist 832 for w in self.worklist:
833 print(w.Comman 833 print(w.Command())
834 return True 834 return True
835 result = RunWork(self.worklist 835 result = RunWork(self.worklist, self.jobs, verbosity=self.verbosity)
836 if self.verbosity.verbose: 836 if self.verbosity.verbose:
837 print(glb_prog_name, " 837 print(glb_prog_name, "done")
838 return result 838 return result
839 839
840 def RunParallelPerf(a): 840 def RunParallelPerf(a):
841 pp = ParallelPerf(a) 841 pp = ParallelPerf(a)
842 pp.Config() 842 pp.Config()
843 return pp.Run() 843 return pp.Run()
844 844
845 def Main(args): 845 def Main(args):
846 ap = argparse.ArgumentParser( 846 ap = argparse.ArgumentParser(
847 prog=glb_prog_name, formatter_ 847 prog=glb_prog_name, formatter_class = argparse.RawDescriptionHelpFormatter,
848 description = 848 description =
849 """ 849 """
850 Run a perf script command multiple times in pa 850 Run a perf script command multiple times in parallel, using perf script options
851 --cpu and --time so that each job processes a 851 --cpu and --time so that each job processes a different chunk of the data.
852 """, 852 """,
853 epilog = 853 epilog =
854 """ 854 """
855 Follow the options by '--' and then the perf s 855 Follow the options by '--' and then the perf script command e.g.
856 856
857 $ perf record -a -- sleep 10 857 $ perf record -a -- sleep 10
858 $ parallel-perf.py --nr=4 -- perf scri 858 $ parallel-perf.py --nr=4 -- perf script --ns
859 All jobs finished successfully 859 All jobs finished successfully
860 $ tree parallel-perf-output/ 860 $ tree parallel-perf-output/
861 parallel-perf-output/ 861 parallel-perf-output/
862 ├── time-range-0 862 ├── time-range-0
863 │ ├── cmd.txt 863 │ ├── cmd.txt
864 │ └── out.txt 864 │ └── out.txt
865 ├── time-range-1 865 ├── time-range-1
866 │ ├── cmd.txt 866 │ ├── cmd.txt
867 │ └── out.txt 867 │ └── out.txt
868 ├── time-range-2 868 ├── time-range-2
869 │ ├── cmd.txt 869 │ ├── cmd.txt
870 │ └── out.txt 870 │ └── out.txt
871 └── time-range-3 871 └── time-range-3
872 ├── cmd.txt 872 ├── cmd.txt
873 └── out.txt 873 └── out.txt
874 $ find parallel-perf-output -name cmd. 874 $ find parallel-perf-output -name cmd.txt | sort | xargs grep -H .
875 parallel-perf-output/time-range-0/cmd. 875 parallel-perf-output/time-range-0/cmd.txt:perf script --time=,9466.504461499 --ns
876 parallel-perf-output/time-range-1/cmd. 876 parallel-perf-output/time-range-1/cmd.txt:perf script --time=9466.504461500,9469.005396999 --ns
877 parallel-perf-output/time-range-2/cmd. 877 parallel-perf-output/time-range-2/cmd.txt:perf script --time=9469.005397000,9471.506332499 --ns
878 parallel-perf-output/time-range-3/cmd. 878 parallel-perf-output/time-range-3/cmd.txt:perf script --time=9471.506332500, --ns
879 879
880 Any perf script command can be used, including 880 Any perf script command can be used, including the use of perf script options
881 --dlfilter and --script, so that the benefit o 881 --dlfilter and --script, so that the benefit of running parallel jobs
882 naturally extends to them also. 882 naturally extends to them also.
883 883
884 If option --pipe-to is used, standard output i 884 If option --pipe-to is used, standard output is first piped through that
885 command. Beware, if the command fails (e.g. gr 885 command. Beware, if the command fails (e.g. grep with no matches), it will be
886 considered a fatal error. 886 considered a fatal error.
887 887
888 Final standard output is redirected to files n 888 Final standard output is redirected to files named out.txt in separate
889 subdirectories under the output directory. Sim 889 subdirectories under the output directory. Similarly, standard error is
890 written to files named err.txt. In addition, f 890 written to files named err.txt. In addition, files named cmd.txt contain the
891 corresponding perf script command. After proce 891 corresponding perf script command. After processing, err.txt files are removed
892 if they are empty. 892 if they are empty.
893 893
894 If any job exits with a non-zero exit code, th 894 If any job exits with a non-zero exit code, then all jobs are killed and no
895 more are started. A message is printed if any 895 more are started. A message is printed if any job results in a non-empty
896 err.txt file. 896 err.txt file.
897 897
898 There is a separate output subdirectory for ea 898 There is a separate output subdirectory for each time range. If the --per-cpu
899 option is used, these are further grouped unde 899 option is used, these are further grouped under cpu-n subdirectories, e.g.
900 900
901 $ parallel-perf.py --per-cpu --nr=2 -- 901 $ parallel-perf.py --per-cpu --nr=2 -- perf script --ns --cpu=0,1
902 All jobs finished successfully 902 All jobs finished successfully
903 $ tree parallel-perf-output 903 $ tree parallel-perf-output
904 parallel-perf-output/ 904 parallel-perf-output/
905 ├── cpu-0 905 ├── cpu-0
906 │ ├── time-range-0 906 │ ├── time-range-0
907 │ │ ├── cmd.txt 907 │ │ ├── cmd.txt
908 │ │ └── out.txt 908 │ │ └── out.txt
909 │ └── time-range-1 909 │ └── time-range-1
910 │ ├── cmd.txt 910 │ ├── cmd.txt
911 │ └── out.txt 911 │ └── out.txt
912 └── cpu-1 912 └── cpu-1
913 ├── time-range-0 913 ├── time-range-0
914 │ ├── cmd.txt 914 │ ├── cmd.txt
915 │ └── out.txt 915 │ └── out.txt
916 └── time-range-1 916 └── time-range-1
917 ├── cmd.txt 917 ├── cmd.txt
918 └── out.txt 918 └── out.txt
919 $ find parallel-perf-output -name cmd. 919 $ find parallel-perf-output -name cmd.txt | sort | xargs grep -H .
920 parallel-perf-output/cpu-0/time-range- 920 parallel-perf-output/cpu-0/time-range-0/cmd.txt:perf script --cpu=0 --time=,9469.005396999 --ns
921 parallel-perf-output/cpu-0/time-range- 921 parallel-perf-output/cpu-0/time-range-1/cmd.txt:perf script --cpu=0 --time=9469.005397000, --ns
922 parallel-perf-output/cpu-1/time-range- 922 parallel-perf-output/cpu-1/time-range-0/cmd.txt:perf script --cpu=1 --time=,9469.005396999 --ns
923 parallel-perf-output/cpu-1/time-range- 923 parallel-perf-output/cpu-1/time-range-1/cmd.txt:perf script --cpu=1 --time=9469.005397000, --ns
924 924
925 Subdivisions of time range, and cpus if the -- 925 Subdivisions of time range, and cpus if the --per-cpu option is used, are
926 expressed by the --time and --cpu perf script 926 expressed by the --time and --cpu perf script options respectively. If the
927 supplied perf script command has a --time opti 927 supplied perf script command has a --time option, then that time range is
928 subdivided, otherwise the time range given by 928 subdivided, otherwise the time range given by 'time of first sample' to
929 'time of last sample' is used (refer perf scri 929 'time of last sample' is used (refer perf script --header-only). Similarly, the
930 supplied perf script command may provide a --c 930 supplied perf script command may provide a --cpu option, and only those CPUs
931 will be processed. 931 will be processed.
932 932
933 To prevent time intervals becoming too small, 933 To prevent time intervals becoming too small, the --min-interval option can
934 be used. 934 be used.
935 935
936 Note there is special handling for processing 936 Note there is special handling for processing Intel PT traces. If an interval is
937 not specified and the perf record command cont 937 not specified and the perf record command contained the intel_pt event, then the
938 time range will be subdivided in order to prod 938 time range will be subdivided in order to produce subdivisions that contain
939 approximately the same amount of trace data. T 939 approximately the same amount of trace data. That is accomplished by counting
940 double-quick (--itrace=qqi) samples, and choos 940 double-quick (--itrace=qqi) samples, and choosing time ranges that encompass
941 approximately the same number of samples. In t 941 approximately the same number of samples. In that case, time ranges may not be
942 the same for each CPU processed. For Intel PT, 942 the same for each CPU processed. For Intel PT, --per-cpu is the default, but
943 that can be overridden by --no-per-cpu. Note, 943 that can be overridden by --no-per-cpu. Note, for Intel PT, double-quick
944 decoding produces 1 sample for each PSB synchr 944 decoding produces 1 sample for each PSB synchronization packet, which in turn
945 come after a certain number of bytes output, d 945 come after a certain number of bytes output, determined by psb_period (refer
946 perf Intel PT documentation). The minimum numb 946 perf Intel PT documentation). The minimum number of double-quick samples that
947 will define a time range can be set by the --m 947 will define a time range can be set by the --min_size option, which defaults to
948 64. 948 64.
949 """) 949 """)
950 ap.add_argument("-o", "--output-dir", 950 ap.add_argument("-o", "--output-dir", default="parallel-perf-output", help="output directory (default 'parallel-perf-output')")
951 ap.add_argument("-j", "--jobs", type=i 951 ap.add_argument("-j", "--jobs", type=int, default=0, help="maximum number of jobs to run in parallel at one time (default is the number of CPUs)")
952 ap.add_argument("-n", "--nr", type=int 952 ap.add_argument("-n", "--nr", type=int, default=0, help="number of time subdivisions (default is the number of jobs)")
953 ap.add_argument("-i", "--interval", ty 953 ap.add_argument("-i", "--interval", type=float, default=0, help="subdivide the time range using this time interval (in seconds e.g. 0.1 for a tenth of a second)")
954 ap.add_argument("-c", "--per-cpu", act 954 ap.add_argument("-c", "--per-cpu", action="store_true", help="process data for each CPU in parallel")
955 ap.add_argument("-m", "--min-interval" 955 ap.add_argument("-m", "--min-interval", type=float, default=glb_min_interval, help=f"minimum interval (default {glb_min_interval} seconds)")
956 ap.add_argument("-p", "--pipe-to", hel 956 ap.add_argument("-p", "--pipe-to", help="command to pipe output to (optional)")
957 ap.add_argument("-N", "--no-per-cpu", 957 ap.add_argument("-N", "--no-per-cpu", action="store_true", help="do not process data for each CPU in parallel")
958 ap.add_argument("-b", "--min_size", ty 958 ap.add_argument("-b", "--min_size", type=int, default=glb_min_samples, help="minimum data size (for Intel PT in PSBs)")
959 ap.add_argument("-D", "--dry-run", act 959 ap.add_argument("-D", "--dry-run", action="store_true", help="do not run any jobs, just show the perf script commands")
960 ap.add_argument("-q", "--quiet", actio 960 ap.add_argument("-q", "--quiet", action="store_true", help="do not print any messages except errors")
961 ap.add_argument("-v", "--verbose", act 961 ap.add_argument("-v", "--verbose", action="store_true", help="print more messages")
962 ap.add_argument("-d", "--debug", actio 962 ap.add_argument("-d", "--debug", action="store_true", help="print debugging messages")
963 cmd_line = list(args) 963 cmd_line = list(args)
964 try: 964 try:
965 split_pos = cmd_line.index("-- 965 split_pos = cmd_line.index("--")
966 cmd = cmd_line[split_pos + 1:] 966 cmd = cmd_line[split_pos + 1:]
967 args = cmd_line[:split_pos] 967 args = cmd_line[:split_pos]
968 except: 968 except:
969 cmd = None 969 cmd = None
970 args = cmd_line 970 args = cmd_line
971 a = ap.parse_args(args=args[1:]) 971 a = ap.parse_args(args=args[1:])
972 a.cmd = cmd 972 a.cmd = cmd
973 a.verbosity = Verbosity(a.quiet, a.ver 973 a.verbosity = Verbosity(a.quiet, a.verbose, a.debug)
974 try: 974 try:
975 if a.cmd == None: 975 if a.cmd == None:
976 if len(args) <= 1: 976 if len(args) <= 1:
977 ap.print_help( 977 ap.print_help()
978 return True 978 return True
979 raise Exception("Comma 979 raise Exception("Command line must contain '--' before perf command")
980 return RunParallelPerf(a) 980 return RunParallelPerf(a)
981 except Exception as e: 981 except Exception as e:
982 print("Fatal error: ", str(e)) 982 print("Fatal error: ", str(e))
983 if a.debug: 983 if a.debug:
984 raise 984 raise
985 return False 985 return False
986 986
987 if __name__ == "__main__": 987 if __name__ == "__main__":
988 if not Main(sys.argv): 988 if not Main(sys.argv):
989 sys.exit(1) 989 sys.exit(1)
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.