perf script: Support relative time
authorAndi Kleen <ak@linux.intel.com>
Thu, 14 Mar 2019 22:50:00 +0000 (15:50 -0700)
committerArnaldo Carvalho de Melo <acme@redhat.com>
Tue, 19 Mar 2019 19:52:03 +0000 (16:52 -0300)
When comparing time stamps in 'perf script' traces it can be annoying to
work with the full perf time stamps.

Add a --reltime option that displays time stamps relative to the trace
start to make it easier to read the traces.

Note: not currently supported for --time. Report an error in this
case.

Before:

  % perf script
      swapper 0 [000] 245402.891216:    1 cycles:ppp: ffffffffa0068814 native_write_msr+0x4 ([kernel.kallsyms])
      swapper 0 [000] 245402.891223:    1 cycles:ppp: ffffffffa0068814 native_write_msr+0x4 ([kernel.kallsyms])
      swapper 0 [000] 245402.891227:    5 cycles:ppp: ffffffffa0068814 native_write_msr+0x4 ([kernel.kallsyms])
      swapper 0 [000] 245402.891231:   41 cycles:ppp: ffffffffa0068816 native_write_msr+0x6 ([kernel.kallsyms])
      swapper 0 [000] 245402.891235:  355 cycles:ppp: ffffffffa000dd51 intel_bts_enable_local+0x21 ([kernel.kallsyms])
      swapper 0 [000] 245402.891239: 3084 cycles:ppp: ffffffffa0a0150a end_repeat_nmi+0x48 ([kernel.kallsyms])

After:

  % perf script --reltime

      swapper 0 [000]     0.000000:    1 cycles:ppp: ffffffffa0068814 native_write_msr+0x4 ([kernel.kallsyms])
      swapper 0 [000]     0.000006:    1 cycles:ppp: ffffffffa0068814 native_write_msr+0x4 ([kernel.kallsyms])
      swapper 0 [000]     0.000010:    5 cycles:ppp: ffffffffa0068814 native_write_msr+0x4 ([kernel.kallsyms])
      swapper 0 [000]     0.000014:   41 cycles:ppp: ffffffffa0068816 native_write_msr+0x6 ([kernel.kallsyms])
      swapper 0 [000]     0.000018:  355 cycles:ppp: ffffffffa000dd51 intel_bts_enable_local+0x21 ([kernel.kallsyms])
      swapper 0 [000]     0.000022: 3084 cycles:ppp: ffffffffa0a0150a end_repeat_nmi+0x48 ([kernel.kallsyms])

Committer notes:

Do not use 'time' as the name of a variable, as this breaks the build on
older glibcs:

  cc1: warnings being treated as errors
  builtin-script.c: In function 'perf_sample__fprintf_start':
  builtin-script.c:691: warning: declaration of 'time' shadows a global declaration
  /usr/include/time.h:187: warning: shadowed declaration is here

Signed-off-by: Andi Kleen <ak@linux.intel.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
LPU-Reference: 20190314225002.30108-8-andi@firstfloor.org
Link: https://lkml.kernel.org/n/tip-bpahyi6pr9r399mvihu65fvc@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
tools/perf/Documentation/perf-script.txt
tools/perf/builtin-script.c

index 2e19fd7..9b0d04d 100644 (file)
@@ -380,6 +380,9 @@ include::itrace.txt[]
        Set the maximum number of program blocks to print with brstackasm for
        each sample.
 
+--reltime::
+       Print time stamps relative to trace start.
+
 --per-event-dump::
        Create per event files with a "perf.data.EVENT.dump" name instead of
         printing to stdout, useful, for instance, for generating flamegraphs.
index 2f93d60..61cfd8f 100644 (file)
@@ -53,6 +53,8 @@
 
 static char const              *script_name;
 static char const              *generate_script_lang;
+static bool                    reltime;
+static u64                     initial_time;
 static bool                    debug_mode;
 static u64                     last_timestamp;
 static u64                     nr_unordered;
@@ -686,7 +688,13 @@ static int perf_sample__fprintf_start(struct perf_sample *sample,
        }
 
        if (PRINT_FIELD(TIME)) {
-               nsecs = sample->time;
+               u64 t = sample->time;
+               if (reltime) {
+                       if (!initial_time)
+                               initial_time = sample->time;
+                       t = sample->time - initial_time;
+               }
+               nsecs = t;
                secs = nsecs / NSEC_PER_SEC;
                nsecs -= secs * NSEC_PER_SEC;
 
@@ -694,7 +702,7 @@ static int perf_sample__fprintf_start(struct perf_sample *sample,
                        printed += fprintf(fp, "%5lu.%09llu: ", secs, nsecs);
                else {
                        char sample_time[32];
-                       timestamp__scnprintf_usec(sample->time, sample_time, sizeof(sample_time));
+                       timestamp__scnprintf_usec(t, sample_time, sizeof(sample_time));
                        printed += fprintf(fp, "%12s: ", sample_time);
                }
        }
@@ -3413,6 +3421,7 @@ int cmd_script(int argc, const char **argv)
                     "Set the maximum stack depth when parsing the callchain, "
                     "anything beyond the specified depth will be ignored. "
                     "Default: kernel.perf_event_max_stack or " __stringify(PERF_MAX_STACK_DEPTH)),
+       OPT_BOOLEAN(0, "reltime", &reltime, "Show time stamps relative to start"),
        OPT_BOOLEAN('I', "show-info", &show_full_info,
                    "display extended information from perf.data file"),
        OPT_BOOLEAN('\0', "show-kernel-path", &symbol_conf.show_kernel_path,
@@ -3487,6 +3496,11 @@ int cmd_script(int argc, const char **argv)
                }
        }
 
+       if (script.time_str && reltime) {
+               fprintf(stderr, "Don't combine --reltime with --time\n");
+               return -1;
+       }
+
        if (itrace_synth_opts.callchain &&
            itrace_synth_opts.callchain_sz > scripting_max_stack)
                scripting_max_stack = itrace_synth_opts.callchain_sz;