partial callgrap and time support
v3 cover partial callgrap and time support
From: Kan Liang <[email protected]>
This patchkit adds the ability to turn off callgraphs and time stamps per event. This in term can reduce sampling overhead and the size of the perf.data.
Changes since V1:
- Break up V1 patches into three patches(parse option changes, partial time support and partial callgrap support).
- Use strings 'fp,dwarf,lbr,no' to identify callchains
- Add test case in parse-events.c
Changes since V2:
- Rebase on 60cd37eb10
Kan Liang (5):
perf,tools: introduce OPT_CALLBACK_SET/OPT_CALLBACK_NOOPT_SET perf/documentation: Add description for period perf,tool: partial time support
perf,tool: partial callgrap support
perf,tests: Add tests to callgrap and time parse
tools/perf/Documentation/perf-record.txt | 14 ++++++++ tools/perf/builtin-record.c | 4 +-- tools/perf/perf.h | 1 + tools/perf/tests/parse-events.c | 28 ++++++++++++++++ tools/perf/util/evsel.c | 56 ++++++++++++++++++++++++++++++-- tools/perf/util/parse-events.c | 25 ++++++++++++++ tools/perf/util/parse-events.h | 3 ++ tools/perf/util/parse-events.l | 3 ++ tools/perf/util/parse-options.c | 2 ++ tools/perf/util/parse-options.h | 4 +++ 10 files changed, 135 insertions(+), 5 deletions(-)
Signature · 1 line
1.8.3.1
RFC PATCH v3 1/5 perf,tools: introduce OPT_CALLBACK_SET/OPT_CALLBACK_NOOPT_SET
From: Kan Liang <[email protected]>
This patch extends the OPT_CALLBACK/OPT_CALLBACK_NOOPT to support 'set' item of struct option. So the perf knows whether an option was set by user.
The new macros are used by call-graph options.
--- tools/perf/builtin-record.c | 4 ++-- tools/perf/perf.h | 1 + tools/perf/util/parse-options.c | 2 ++ tools/perf/util/parse-options.h | 4 ++++ 4 files changed, 9 insertions(+), 2 deletions(-)
tools/perf/builtin-record.c +2 -2
@@ -1018,10 +1018,10 @@ struct option __record_options[] = { record__parse_mmap_pages),OPT_BOOLEAN (0 ,"group" , &record.opts.group,"put the counters into a counter group" ), -OPT_CALLBACK_NOOPT ('g' ,NULL , &record.opts, +OPT_CALLBACK_NOOPT_SET ('g' ,NULL , &record.opts, &record.opts.callgraph_set,NULL ,"enables call-graph recording" , &record_callchain_opt), -OPT_CALLBACK (0 ,"call-graph" , &record.opts, +OPT_CALLBACK_SET (0 ,"call-graph" , &record.opts, &record.opts.callgraph_set,"mode[,dump_size]" , record_callchain_help, &record_parse_callchain_opt), OPT_INCR('v', "verbose", &verbose,
tools/perf/perf.h +1 -0
@@ -52,6 +52,7 @@ struct record_opts {bool sample_weight;bool sample_time;bool sample_time_set; +bool callgraph_set;bool period;bool sample_intr_regs;bool running_time;
tools/perf/util/parse-options.c +2 -0
@@ -140,6 +140,8 @@ static int get_value(struct parse_opt_ctx_t *p,return err;case OPTION_CALLBACK : +if (opt->set) + *(bool *)opt->set = true;if (unset)return (*opt->callback)(opt,NULL ,1 ) ? (-1 ) :0 ;if (opt->flags &PARSE_OPT_NOARG )
tools/perf/util/parse-options.h +4 -0
@@ -132,8 +132,12 @@ struct option { { .type = OPTION_CALLBACK, .short_name = (s), .long_name = (l), .value = (v), .argh = "time", .help = (h), .callback = parse_opt_approxidate_cb }#define OPT_CALLBACK (s, l, v, a, h, f) \ { .type = OPTION_CALLBACK, .short_name = (s), .long_name = (l), .value = (v), (a), .help = (h), .callback = (f) } +#define OPT_CALLBACK_SET (s, l, v, os, a, h, f) \ + { .type = OPTION_CALLBACK, .short_name = (s), .long_name = (l), .value = (v), (a), .help = (h), .callback = (f), .set = check_vtype(os, bool *) }#define OPT_CALLBACK_NOOPT (s, l, v, a, h, f) \ { .type = OPTION_CALLBACK, .short_name = (s), .long_name = (l), .value = (v), (a), .help = (h), .callback = (f), .flags = PARSE_OPT_NOARG } +#define OPT_CALLBACK_NOOPT_SET (s, l, v, os, a, h, f) \ + { .type = OPTION_CALLBACK, .short_name = (s), .long_name = (l), .value = (v), (a), .help = (h), .callback = (f), .flags = PARSE_OPT_NOARG, .set = check_vtype(os, bool *) }#define OPT_CALLBACK_DEFAULT (s, l, v, a, h, f, d) \ { .type = OPTION_CALLBACK, .short_name = (s), .long_name = (l), .value = (v), (a), .help = (h), .callback = (f), .defval = (intptr_t)d, .flags = PARSE_OPT_LASTARG_DEFAULT }#define OPT_CALLBACK_DEFAULT_NOOPT (s, l, v, a, h, f, d) \
Signature · 1 line
1.8.3.1
Re: [PATCH RFC V3 1/5] perf,tools: introduce OPT_CALLBACK_SET/OPT_CALLBACK_NOOPT_SET
Quoted 21 lines · On Wed, Jul 08, 2015 at 04:44:53AM -0400, [email protected] wrote:
From: Kan Liang <[email protected]>
This patch extends the OPT_CALLBACK/OPT_CALLBACK_NOOPT to support 'set' item of struct option. So the perf knows whether an option was set by user.
The new macros are used by call-graph options.Signed-off-by Kan Liang <[email protected]>--- tools/perf/builtin-record.c | 4 ++-- tools/perf/perf.h | 1 + tools/perf/util/parse-options.c | 2 ++ tools/perf/util/parse-options.h | 4 ++++ 4 files changed, 9 insertions(+), 2 deletions(-)tools/perf/builtin-record.c +0
-0@@ -1018,10 +1018,10 @@ struct option __record_options[] = { record__parse_mmap_pages),
tools/perf/builtin-record.c struct option __record_options[] = { · line 1026
Quoted 10 lines
tools/perf/builtin-record.c +2
-2OPT_BOOLEAN (0 ,"group" , &record.opts.group,"put the counters into a counter group" ), -OPT_CALLBACK_NOOPT ('g' ,NULL , &record.opts, +OPT_CALLBACK_NOOPT_SET ('g' ,NULL , &record.opts, &record.opts.callgraph_set,NULL ,"enables call-graph recording" , &record_callchain_opt), -OPT_CALLBACK (0 ,"call-graph" , &record.opts, +OPT_CALLBACK_SET (0 ,"call-graph" , &record.opts, &record.opts.callgraph_set,"mode[,dump_size]" , record_callchain_help, &record_parse_callchain_opt),
we already provide &record.opts and callback which can set record.opts.callgraph_set.. seems redundant do I miss some other benefit?
jirka
Quoted 46 lines
tools/perf/builtin-record.c +0
-0OPT_INCR('v', "verbose", &verbose,tools/perf/perf.h +1
-0@@ -52,6 +52,7 @@ struct record_opts {bool sample_weight;bool sample_time;bool sample_time_set; +bool callgraph_set;bool period;bool sample_intr_regs;bool running_time;tools/perf/util/parse-options.c +2
-0@@ -140,6 +140,8 @@ static int get_value(struct parse_opt_ctx_t *p,return err;case OPTION_CALLBACK : +if (opt->set) + *(bool *)opt->set = true;if (unset)return (*opt->callback)(opt,NULL ,1 ) ? (-1 ) :0 ;if (opt->flags &PARSE_OPT_NOARG )tools/perf/util/parse-options.h +4
-0@@ -132,8 +132,12 @@ struct option { { .type = OPTION_CALLBACK, .short_name = (s), .long_name = (l), .value = (v), .argh = "time", .help = (h), .callback = parse_opt_approxidate_cb }#define OPT_CALLBACK (s, l, v, a, h, f) \ { .type = OPTION_CALLBACK, .short_name = (s), .long_name = (l), .value = (v), (a), .help = (h), .callback = (f) } +#define OPT_CALLBACK_SET (s, l, v, os, a, h, f) \ + { .type = OPTION_CALLBACK, .short_name = (s), .long_name = (l), .value = (v), (a), .help = (h), .callback = (f), .set = check_vtype(os, bool *) }#define OPT_CALLBACK_NOOPT (s, l, v, a, h, f) \ { .type = OPTION_CALLBACK, .short_name = (s), .long_name = (l), .value = (v), (a), .help = (h), .callback = (f), .flags = PARSE_OPT_NOARG } +#define OPT_CALLBACK_NOOPT_SET (s, l, v, os, a, h, f) \ + { .type = OPTION_CALLBACK, .short_name = (s), .long_name = (l), .value = (v), (a), .help = (h), .callback = (f), .flags = PARSE_OPT_NOARG, .set = check_vtype(os, bool *) }#define OPT_CALLBACK_DEFAULT (s, l, v, a, h, f, d) \ { .type = OPTION_CALLBACK, .short_name = (s), .long_name = (l), .value = (v), (a), .help = (h), .callback = (f), .defval = (intptr_t)d, .flags = PARSE_OPT_LASTARG_DEFAULT }#define OPT_CALLBACK_DEFAULT_NOOPT (s, l, v, a, h, f, d) \Signature · 1 line
1.8.3.1
RFC PATCH v3 2/5 perf/documentation: Add description for period
In reply to [email protected]From: Kan Liang <[email protected]>
'period' param is not defined in
/sys/bus/event_sources/devices/<pmu>/format/*
--- tools/perf/Documentation/perf-record.txt | 8 ++++++++ 1 file changed, 8 insertions(+)
tools/perf/Documentation/perf-record.txt +8 -0
@@ -45,6 +45,14 @@ OPTIONS param1 and param2 are defined as formats for the PMU in: /sys/bus/event_sources/devices/<pmu>/format/* + There are also some params which are not defined in .../<pmu>/format/*. + These params can be used to set event defaults. + Here is a list of the params. + - 'period': Set event sampling period + + Note: If user explicitly sets options which conflict with the params, + the value set by the params will be overridden. + - a hardware breakpoint event in the form of '\mem:addr[/len][:access]' where addr is the address in memory you want to break in. Access is the memory access type (read, write, execute) it can
Signature · 1 line
1.8.3.1
Re: [PATCH RFC V3 2/5] perf/documentation: Add description for period
Quoted 15 lines · On Wed, Jul 08, 2015 at 04:44:54AM -0400, [email protected] wrote:
From: Kan Liang <[email protected]>
'period' param is not defined in
/sys/bus/event_sources/devices/<pmu>/format/*Signed-off-by Kan Liang <[email protected]>--- tools/perf/Documentation/perf-record.txt | 8 ++++++++ 1 file changed, 8 insertions(+)tools/perf/Documentation/perf-record.txt +0
-0
tools/perf/Documentation/perf-record.txt OPTIONS · line 51
Quoted 7 lines
tools/perf/Documentation/perf-record.txt +4
-0@@ -45,6 +45,14 @@ OPTIONS param1 and param2 are defined as formats for the PMU in: /sys/bus/event_sources/devices/<pmu>/format/* + There are also some params which are not defined in .../<pmu>/format/*. + These params can be used to set event defaults. + Here is a list of the params. + - 'period': Set event sampling period
while at it, could you please add doc also for 'name'?
thanks,
jirka
Quoted 10 lines
tools/perf/Documentation/perf-record.txt +4
-0+ + Note: If user explicitly sets options which conflict with the params, + the value set by the params will be overridden. + - a hardware breakpoint event in the form of '\mem:addr[/len][:access]' where addr is the address in memory you want to break in. Access is the memory access type (read, write, execute) it canSignature · 1 line
1.8.3.1
Re: [PATCH RFC V3 2/5] perf/documentation: Add description for period
Em Tue, Jul 14, 2015 at 08:55:18AM +0200, Jiri Olsa escreveu:
Quoted 19 lines · On Wed, Jul 08, 2015 at 04:44:54AM -0400, [email protected] wrote:
Quoted 18 lines
From: Kan Liang <[email protected]>
'period' param is not defined in
/sys/bus/event_sources/devices/<pmu>/format/*Signed-off-by Kan Liang <[email protected]>--- tools/perf/Documentation/perf-record.txt | 8 ++++++++ 1 file changed, 8 insertions(+)tools/perf/Documentation/perf-record.txt +0
-0@@ -45,6 +45,14 @@ OPTIONS param1 and param2 are defined as formats for the PMU in: /sys/bus/event_sources/devices/<pmu>/format/*
Quoted 6 lines
tools/perf/Documentation/perf-record.txt OPTIONS · line 51
Quoted 4 lines
tools/perf/Documentation/perf-record.txt +4
-0+ There are also some params which are not defined in .../<pmu>/format/*. + These params can be used to set event defaults. + Here is a list of the params. + - 'period': Set event sampling periodAcked-by Jiri Olsa <[email protected]>
Thanks, applied.
Quoted 20 lines
while at it, could you please add doc also for 'name'?
thanks,
jirkaQuoted 10 lines
tools/perf/Documentation/perf-record.txt +4
-0+ + Note: If user explicitly sets options which conflict with the params, + the value set by the params will be overridden. + - a hardware breakpoint event in the form of '\mem:addr[/len][:access]' where addr is the address in memory you want to break in. Access is the memory access type (read, write, execute) it canSignature · 1 line
1.8.3.1--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
tip-bot for Kan Liang
merge bot[tip:perf/core] perf record: Document setting '-e pmu/period=N/' in man page
In reply to [email protected]perf record: Document setting '-e pmu/period=N/' in man page
Commit-ID: 3d5d68aabf4ff47cfb447ad90448ce50531be7f5 Gitweb: http://git.kernel.org/tip/3d5d68aabf4ff47cfb447ad90448ce50531be7f5 Author: Kan Liang <[email protected]> AuthorDate: Wed, 8 Jul 2015 04:44:54 -0400 Committer: Arnaldo Carvalho de Melo <[email protected]> CommitDate: Tue, 14 Jul 2015 07:52:00 -0300
perf record: Document setting '-e pmu/period=N/' in man page
The 'period' param is not defined in /sys/bus/event_sources/devices/<pmu>/format/*, but can be used, document it.
--- tools/perf/Documentation/perf-record.txt | 8 ++++++++ 1 file changed, 8 insertions(+)
tools/perf/Documentation/perf-record.txt +8 -0
@@ -45,6 +45,14 @@ OPTIONS param1 and param2 are defined as formats for the PMU in: /sys/bus/event_sources/devices/<pmu>/format/* + There are also some params which are not defined in .../<pmu>/format/*. + These params can be used to set event defaults. + Here is a list of the params. + - 'period': Set event sampling period + + Note: If user explicitly sets options which conflict with the params, + the value set by the params will be overridden. + - a hardware breakpoint event in the form of '\mem:addr[/len][:access]' where addr is the address in memory you want to break in. Access is the memory access type (read, write, execute) it can
RFC PATCH v3 3/5 perf,tool: partial time support
In reply to [email protected]From: Kan Liang <[email protected]>
When multiple events are sampled it may not be needed to collect fine grained time stamps on all events. The sample sites are usually nearby. It's enough to have time stamps on the regular reference events. This patchkit adds the ability to turn off time stamps per event. This in term can reduce sampling overhead and the size of the perf.data.
--- tools/perf/Documentation/perf-record.txt | 4 +++- tools/perf/util/evsel.c | 32 ++++++++++++++++++++++++++++++-- tools/perf/util/parse-events.c | 7 +++++++ tools/perf/util/parse-events.h | 1 + tools/perf/util/parse-events.l | 1 + 5 files changed, 42 insertions(+), 3 deletions(-)
tools/perf/Documentation/perf-record.txt +3 -1
@@ -49,7 +49,9 @@ OPTIONS These params can be used to set event defaults. Here is a list of the params. - 'period': Set event sampling period - + - 'time': Disable/enable time stamping. Acceptable values are 1 for + enabling time stamping. 0 for disabling time stamping. + The default is 1. Note: If user explicitly sets options which conflict with the params, the value set by the params will be overridden.
tools/perf/util/evsel.c +30 -2
@@ -619,10 +619,37 @@ void perf_evsel__config(struct perf_evsel *evsel, struct record_opts *opts)struct perf_event_attr *attr = &evsel->attr;int track = evsel->tracking;bool per_cpu = opts->target.default_per_cpu && !opts->target.per_thread; +bool sample_time = opts->sample_time; attr->sample_id_all = perf_missing_features.sample_id_all ?0 :1 ; attr->inherit = !opts->no_inherit; +/* +* If user doesn't explicitly set time option, +* let event attribute decide. +*/ + +if (!opts->sample_time_set) { +if (attr->sample_type &PERF_SAMPLE_TIME ) + sample_time = true; +else + sample_time = false; + } + +/* +* Event parsing doesn't check the availability +* Clear the bit which event parsing may be set. +* Let following code check and reset if available +* +* Also, the sample size may be caculated mistakenly, +* because event parsing may set the PERF_SAMPLE_TIME. +* Remove the size which add in perf_evsel__init +*/ +if (attr->sample_type &PERF_SAMPLE_TIME ) { + attr->sample_type &= ~PERF_SAMPLE_TIME ; + evsel->sample_size -=sizeof (u64); + } +perf_evsel__set_sample_bit (evsel,IP );perf_evsel__set_sample_bit (evsel,TID );
@@ -705,14 +732,15 @@ void perf_evsel__config(struct perf_evsel *evsel, struct record_opts *opts)/* * When the user explicitely disabled time don't force it here. */ -if (opts->sample_time && +if (sample_time && (!perf_missing_features.sample_id_all && (!opts->no_inherit ||target__has_cpu (&opts->target) || per_cpu || opts->sample_time_set)))perf_evsel__set_sample_bit (evsel,TIME );if (opts->raw_samples && !evsel->no_aux_samples) { -perf_evsel__set_sample_bit (evsel,TIME ); +if (sample_time) +perf_evsel__set_sample_bit (evsel,TIME );perf_evsel__set_sample_bit (evsel,RAW );perf_evsel__set_sample_bit (evsel,CPU ); }
tools/perf/util/parse-events.c +7 -0
@@ -598,6 +598,13 @@ do { \* attr->branch_sample_type = term->val.num; */ break ; +case PARSE_EVENTS__TERM_TYPE_TIME : +CHECK_TYPE_VAL (NUM ); +if (term->val.num >1 ) +return -EINVAL ; +if (term->val.num ==1 ) + attr->sample_type |=PERF_SAMPLE_TIME ; +break ;case PARSE_EVENTS__TERM_TYPE_NAME :CHECK_TYPE_VAL (STR );break ;
tools/perf/util/parse-events.h +1 -0
@@ -62,6 +62,7 @@ enum {PARSE_EVENTS__TERM_TYPE_NAME ,PARSE_EVENTS__TERM_TYPE_SAMPLE_PERIOD ,PARSE_EVENTS__TERM_TYPE_BRANCH_SAMPLE_TYPE , + PARSE_EVENTS__TERM_TYPE_TIME, }; struct parse_events_term {
tools/perf/util/parse-events.l +1 -0
@@ -183,6 +183,7 @@ config2 { return term(yyscanner, PARSE_EVENTS__TERM_TYPE_CONFIG2); } name { return term(yyscanner, PARSE_EVENTS__TERM_TYPE_NAME); } period { return term(yyscanner, PARSE_EVENTS__TERM_TYPE_SAMPLE_PERIOD); } branch_type { return term(yyscanner, PARSE_EVENTS__TERM_TYPE_BRANCH_SAMPLE_TYPE); } +time { return term(yyscanner, PARSE_EVENTS__TERM_TYPE_TIME); } , { return ','; } "/" { BEGIN(INITIAL); return '/'; } {name_minus} { return str(yyscanner, PE_NAME); }
Signature · 1 line
1.8.3.1
Namhyung Kim
maintainerRe: [PATCH RFC V3 3/5] perf,tool: partial time support
Hi,
(CC-ing Adrian)
Quoted 7 lines · On Wed, Jul 08, 2015 at 04:44:55AM -0400, [email protected] wrote:
From: Kan Liang <[email protected]>
When multiple events are sampled it may not be needed to collect fine grained time stamps on all events. The sample sites are usually nearby. It's enough to have time stamps on the regular reference events. This patchkit adds the ability to turn off time stamps per event. This in term can reduce sampling overhead and the size of the perf.data.
So this patch makes the PERF_SAMPLE_TIME bit set or not independently, right? But AFAIK we sometimes just use first evsel for checking sample_type value, especially for evlist->id_pos. I'm not sure it'll work for all cases of mixed time/notime events..
Thanks,
Namhyung
Quoted 130 lines
Signed-off-by Kan Liang <[email protected]>--- tools/perf/Documentation/perf-record.txt | 4 +++- tools/perf/util/evsel.c | 32 ++++++++++++++++++++++++++++++-- tools/perf/util/parse-events.c | 7 +++++++ tools/perf/util/parse-events.h | 1 + tools/perf/util/parse-events.l | 1 + 5 files changed, 42 insertions(+), 3 deletions(-)tools/perf/Documentation/perf-record.txt +3
-1@@ -49,7 +49,9 @@ OPTIONS These params can be used to set event defaults. Here is a list of the params. - 'period': Set event sampling period - + - 'time': Disable/enable time stamping. Acceptable values are 1 for + enabling time stamping. 0 for disabling time stamping. + The default is 1. Note: If user explicitly sets options which conflict with the params, the value set by the params will be overridden.tools/perf/util/evsel.c +30
-2@@ -619,10 +619,37 @@ void perf_evsel__config(struct perf_evsel *evsel, struct record_opts *opts)struct perf_event_attr *attr = &evsel->attr;int track = evsel->tracking;bool per_cpu = opts->target.default_per_cpu && !opts->target.per_thread; +bool sample_time = opts->sample_time; attr->sample_id_all = perf_missing_features.sample_id_all ?0 :1 ; attr->inherit = !opts->no_inherit; +/* +* If user doesn't explicitly set time option, +* let event attribute decide. +*/ + +if (!opts->sample_time_set) { +if (attr->sample_type &PERF_SAMPLE_TIME ) + sample_time = true; +else + sample_time = false; + } + +/* +* Event parsing doesn't check the availability +* Clear the bit which event parsing may be set. +* Let following code check and reset if available +* +* Also, the sample size may be caculated mistakenly, +* because event parsing may set the PERF_SAMPLE_TIME. +* Remove the size which add in perf_evsel__init +*/ +if (attr->sample_type &PERF_SAMPLE_TIME ) { + attr->sample_type &= ~PERF_SAMPLE_TIME ; + evsel->sample_size -=sizeof (u64); + } +perf_evsel__set_sample_bit (evsel,IP );perf_evsel__set_sample_bit (evsel,TID );@@ -705,14 +732,15 @@ void perf_evsel__config(struct perf_evsel *evsel, struct record_opts *opts)/* * When the user explicitely disabled time don't force it here. */ -if (opts->sample_time && +if (sample_time && (!perf_missing_features.sample_id_all && (!opts->no_inherit ||target__has_cpu (&opts->target) || per_cpu || opts->sample_time_set)))perf_evsel__set_sample_bit (evsel,TIME );if (opts->raw_samples && !evsel->no_aux_samples) { -perf_evsel__set_sample_bit (evsel,TIME ); +if (sample_time) +perf_evsel__set_sample_bit (evsel,TIME );perf_evsel__set_sample_bit (evsel,RAW );perf_evsel__set_sample_bit (evsel,CPU ); }tools/perf/util/parse-events.c +7
-0@@ -598,6 +598,13 @@ do { \* attr->branch_sample_type = term->val.num; */ break ; +case PARSE_EVENTS__TERM_TYPE_TIME : +CHECK_TYPE_VAL (NUM ); +if (term->val.num >1 ) +return -EINVAL ; +if (term->val.num ==1 ) + attr->sample_type |=PERF_SAMPLE_TIME ; +break ;case PARSE_EVENTS__TERM_TYPE_NAME :CHECK_TYPE_VAL (STR );break ;tools/perf/util/parse-events.h +1
-0@@ -62,6 +62,7 @@ enum {PARSE_EVENTS__TERM_TYPE_NAME ,PARSE_EVENTS__TERM_TYPE_SAMPLE_PERIOD ,PARSE_EVENTS__TERM_TYPE_BRANCH_SAMPLE_TYPE , + PARSE_EVENTS__TERM_TYPE_TIME, }; struct parse_events_term {tools/perf/util/parse-events.l +1
-0@@ -183,6 +183,7 @@ config2 { return term(yyscanner, PARSE_EVENTS__TERM_TYPE_CONFIG2); } name { return term(yyscanner, PARSE_EVENTS__TERM_TYPE_NAME); } period { return term(yyscanner, PARSE_EVENTS__TERM_TYPE_SAMPLE_PERIOD); } branch_type { return term(yyscanner, PARSE_EVENTS__TERM_TYPE_BRANCH_SAMPLE_TYPE); } +time { return term(yyscanner, PARSE_EVENTS__TERM_TYPE_TIME); } , { return ','; } "/" { BEGIN(INITIAL); return '/'; } {name_minus} { return str(yyscanner, PE_NAME); }Signature · 1 line
1.8.3.1
RE: [PATCH RFC V3 3/5] perf,tool: partial time support
Quoted 15 lines
Hi,
(CC-ing Adrian)
Quoted 7 lines · On Wed, Jul 08, 2015 at 04:44:55AM -0400, [email protected] wrote:
From: Kan Liang <[email protected]>
When multiple events are sampled it may not be needed to collect fine grained time stamps on all events. The sample sites are usually nearby. It's enough to have time stamps on the regular reference events. This patchkit adds the ability to turn off time stamps per event. This in term can reduce sampling overhead and the size of the perf.data.
Quoted 5 lines
So this patch makes the
PERF_SAMPLE_TIMEbit set or not independently, right? But AFAIK we sometimes just use first evsel for checkingsample_typevalue, especially forevlist->id_pos. I'm not sure it'll work for all cases of mixed time/notime events..
If the sample_type is different, IDENTIFIER will be set to make sure id is in a fixed position. So we don't need to worry about evlist->id_pos here.
I think the only unsupported case should be that the kernel doesn't support IDENTIFIER. If so, under my test, perf record will print warning " failed to process type". Perf report will also print warning " non matching sample_type". So the user can switch back to full time support.
If you think the warning is not good enough, we can show specific warning and error out when non IDENTIFIER support is detected.
Thanks,
Kan ���{.n�+�������+%�����ݶ��w��{.n�+����{��G�����{ay�ʇڙ�,j��f���h��������z_�(�階�ݢj"���m������G������?����&���~��iO��z��v�^�m����������?�I�
Namhyung Kim
maintainerRe: [PATCH RFC V3 3/5] perf,tool: partial time support
Quoted 15 lines · Kan
Quoted 9 lines · On Wed, Jul 08, 2015 at 04:44:55AM -0400, [email protected] wrote:
Quoted 7 lines
From: Kan Liang <[email protected]>
When multiple events are sampled it may not be needed to collect fine grained time stamps on all events. The sample sites are usually nearby. It's enough to have time stamps on the regular reference events. This patchkit adds the ability to turn off time stamps per event. This in term can reduce sampling overhead and the size of the perf.data.
Quoted 5 lines
So this patch makes the
PERF_SAMPLE_TIMEbit set or not independently, right? But AFAIK we sometimes just use first evsel for checkingsample_typevalue, especially forevlist->id_pos. I'm not sure it'll work for all cases of mixed time/notime events..
Quoted 8 lines
If the
sample_typeis different, IDENTIFIER will be set to make sure id is in a fixed position. So we don't need to worry aboutevlist->id_poshere.I think the only unsupported case should be that the kernel doesn't support IDENTIFIER. If so, under my test, perf record will print warning " failed to process type". Perf report will also print warning " non matching
sample_type". So the user can switch back to full time support.
Ah, ok then.
Quoted 2 lines
If you think the warning is not good enough, we can show specific warning and error out when non IDENTIFIER support is detected.
Well, I think it's good enough. :)
However, perf_evlist__id_hdr_size() seems to be affected also. Are you sure it's okay too?
Thanks,
Namhyung
RE: [PATCH RFC V3 3/5] perf,tool: partial time support
Quoted 6 lines
However,
perf_evlist__id_hdr_size()seems to be affected also. Are you sure it's okay too?
Yes.id_hdr_size is calculated according to the first event. It impacts perf tool generated COMM, MMAP and FORK events.
But all of these events have same and position fixed IDENTIFIER, 0. When parsing the events by perf_evlist__event2evsel, id 0 always return first event.
So only first event will be involved to parse perf tool generated COMM, MMAP and FORK events. TIME setting from other events doesn’t impact.
Thanks,
Kan ���{.n�+�������+%�����ݶ��w��{.n�+����{��G�����{ay�ʇڙ�,j��f���h��������z_�(�階�ݢj"���m������G������?����&���~��iO��z��v�^�m����������?�I�
Namhyung Kim
maintainerRe: [PATCH RFC V3 3/5] perf,tool: partial time support
Hi,
Quoted 12 lines · Kan
Quoted 3 lines
However,
perf_evlist__id_hdr_size()seems to be affected also. Are you sure it's okay too?Yes.
id_hdr_sizeis calculated according to the first event. It impacts perf tool generated COMM, MMAP and FORK events.
But all of these events have same and position fixed IDENTIFIER, 0. When parsing the events byperf_evlist__event2evsel, id 0 always return first event.
So only first event will be involved to parse perf tool generated COMM, MMAP and FORK events. TIME setting from other events doesn’t impact.
Good. Thanks for the explanation..
Thanks,
Namhyung
Re: [PATCH RFC V3 3/5] perf,tool: partial time support
In reply to [email protected]Quoted 7 lines · On Wed, Jul 08, 2015 at 04:44:55AM -0400, [email protected] wrote:
From: Kan Liang <[email protected]>
When multiple events are sampled it may not be needed to collect fine grained time stamps on all events. The sample sites are usually nearby. It's enough to have time stamps on the regular reference events. This patchkit adds the ability to turn off time stamps per event. This in term can reduce sampling overhead and the size of the perf.data.
[PATCH RFC V3 3/5] perf,tool: partial time support
for a minute I was wondering why's the patch not full and just partial ;-)
'per event' might sound better
SNIP
Quoted 28 lines
tools/perf/util/evsel.c +16
-0@@ -619,10 +619,37 @@ void perf_evsel__config(struct perf_evsel *evsel, struct record_opts *opts)struct perf_event_attr *attr = &evsel->attr;int track = evsel->tracking;bool per_cpu = opts->target.default_per_cpu && !opts->target.per_thread; +bool sample_time = opts->sample_time; attr->sample_id_all = perf_missing_features.sample_id_all ?0 :1 ; attr->inherit = !opts->no_inherit; +/* +* If user doesn't explicitly set time option, +* let event attribute decide. +*/ + +if (!opts->sample_time_set) { +if (attr->sample_type &PERF_SAMPLE_TIME ) + sample_time = true; +else + sample_time = false; + } + +/* +* Event parsing doesn't check the availability +* Clear the bit which event parsing may be set.
tools/perf/util/evsel.c void perf_evsel__config(struct perf_evsel *evsel, struct record_opts *opts) · line 651
Quoted 10 lines
tools/perf/util/evsel.c +10
-0+* Let following code check and reset if available +* +* Also, the sample size may be caculated mistakenly, +* because event parsing may set the PERF_SAMPLE_TIME. +* Remove the size which add in perf_evsel__init +*/ +if (attr->sample_type &PERF_SAMPLE_TIME ) { + attr->sample_type &= ~PERF_SAMPLE_TIME ; + evsel->sample_size -=sizeof (u64); + }
so we have TIME enabled by default, like:
[jolsa@krava perf]$ ./perf record ls
...
[jolsa@krava perf]$ ./perf evlist -v
cycles: size: 112, { sample_period, sample_freq }: 4000, sample_type:
IP|TID|TIME|PERIOD, disabled: 1, inherit: 1, mmap: 1, comm: 1, freq: 1,enable_on_exec: 1, task: 1, sample_id_all: 1, exclude_guest: 1, mmap2:
1, comm_exec: 1
this patch disables it by default and there might be people out there who actualy care.
I wonder you should instead disable it completely via --no-timestamp option and enable it via 'time' term for specific events, like:
$ perf record --no-timestamp -e 'cpu/...time/,cpu/.../'ls
or disable timestamps globaly if 'time' term is detected in any event definition..?
jirka
Re: [PATCH RFC V3 3/5] perf,tool: partial time support
In reply to [email protected]Quoted 2 lines · On Wed, Jul 08, 2015 at 04:44:55AM -0400, [email protected] wrote:
From: Kan Liang <[email protected]>
SNIP
Quoted 4 lines
tools/perf/util/evsel.c +4
-0+ +/* +* Event parsing doesn't check the availability +* Clear the bit which event parsing may be set.
tools/perf/util/evsel.c void perf_evsel__config(struct perf_evsel *evsel, struct record_opts *opts) · line 651
Quoted 10 lines
tools/perf/util/evsel.c +10
-0+* Let following code check and reset if available +* +* Also, the sample size may be caculated mistakenly, +* because event parsing may set the PERF_SAMPLE_TIME. +* Remove the size which add in perf_evsel__init +*/ +if (attr->sample_type &PERF_SAMPLE_TIME ) { + attr->sample_type &= ~PERF_SAMPLE_TIME ; + evsel->sample_size -=sizeof (u64); + }
why not use perf_evsel__reset_sample_bit?
Quoted 8 lines
tools/perf/util/evsel.c +1
-0+perf_evsel__set_sample_bit (evsel,IP );perf_evsel__set_sample_bit (evsel,TID );@@ -705,14 +732,15 @@ void perf_evsel__config(struct perf_evsel *evsel, struct record_opts *opts)/* * When the user explicitely disabled time don't force it here. */
SNIP
Quoted 5 lines
tools/perf/util/parse-events.c +0
-0
tools/perf/util/parse-events.c do { \ · line 606
Quoted 9 lines
tools/perf/util/parse-events.c +6
-0@@ -598,6 +598,13 @@ do { \* attr->branch_sample_type = term->val.num; */ break ; +case PARSE_EVENTS__TERM_TYPE_TIME : +CHECK_TYPE_VAL (NUM ); +if (term->val.num >1 ) +return -EINVAL ; +if (term->val.num ==1 ) + attr->sample_type |=PERF_SAMPLE_TIME ;
why not use perf_evsel__set_sample_bit?
tools/perf/util/parse-events.c do { \ · line 610
Quoted 4 lines
tools/perf/util/parse-events.c +1
-0+break ;case PARSE_EVENTS__TERM_TYPE_NAME :CHECK_TYPE_VAL (STR );break ;
SNIP
jirka
RFC PATCH v3 4/5 perf,tool: partial callgrap support
In reply to [email protected]From: Kan Liang <[email protected]>
When multiple events are sampled it may not be needed to collect callgraphs for all of them. The sample sites are usually nearby, and it's enough to collect the callgraphs on a reference event (such as precise cycles or precise instructions).
This patchkit adds the ability to turn off callgraphs and time stamps per event. This in term can reduce sampling overhead and the size of the perf.data. Furthermore, it makes collecting back traces and timestamps possible when PEBS threshold > 1, which significantly reducing the sampling overhead especially for frequently occurring events
(https://lkml.org/lkml/2015/5/10/196). For example, A slower event with a larger period collects back traces/timestamps. Other more events run fast with multi-pebs. The time stamps from the slower events can be used to order the faster events. Their backtraces can give the user enough hint to find the right spot.
Here are some examples and test results.
1. Comparing the elapsed time and perf.data size from "kernbench -M -H".
The test command for FULL callgrap and time support.
"perf record -e '{cpu/cpu-cycles,period=100000/,cpu/instructions,period=20000/p}' --call-graph fp --time"
The test command for PARTIAL callgrap and time support.
"perf record -e '{cpu/cpu-cycles,callgraph=fp,time,period=100000/,
cpu/instructions,callgraph=no,time=0,period=20000/p}'"
The elapsed time for FULL is 24.3 Sec, while for PARTIAL is 16.9 Sec. The perf.data size for FULL is 22.1 Gb, while for PARTIAL is 12.4 Gb.
2. Comparing the perf.data size and callgraph results.
The test command for FULL callgrap and time support.
"perf record -e '{cpu/cpu-cycles,period=100000/pp,cpu/instructions,period=20000/p}' --call-graph fp -- ./tchain_edit"
The test command for PARTIAL callgrap and time support.
"perf record -e '{cpu/cpu-cycles,callgraph=fp,time,period=100000/pp,
cpu/instructions,callgraph=no,time=0,period=20000/p}'
-- ./tchain_edit"
The perf.data size for FULL is 43.2 MB, while for PARTIAL is 21.1 MB. The callgraph is roughly the same.
The callgraph from FULL
# Samples: 87K of event 'cpu/cpu-cycles,callgraph=fp,time,period=100000/pp' # Event count (approx.): 8760000000
#
# Children Self Command Shared Object Symbol
# ........ ........ ........... ..................
..........................................
#
99.98% 0.00% tchain_edit libc-2.15.so [.]
__libc_start_main
|
---__libc_start_main
99.97% 0.00% tchain_edit tchain_edit [.] main
|
---main
__libc_start_main
99.97% 0.00% tchain_edit tchain_edit [.] f1
|
---f1
main
__libc_start_main
99.85% 87.01% tchain_edit tchain_edit [.] f3
|
---f3
|
|--99.74%-- f2
| f1
| main
| __libc_start_main
--0.26%-- [...]
99.71% 0.12% tchain_edit tchain_edit [.] f2
|
---f2
f1
main
__libc_start_main The callgraph from PARTIAL
# Samples: 417K of event 'cpu/instructions,callgraph=no,time=0,period=20000/p' # Event count (approx.): 8346980000
#
# Children Self Command Shared Object Symbol
# ........ ........ ........... ................
..........................................
#
98.82% 0.00% tchain_edit libc-2.15.so [.]
__libc_start_main
|
---__libc_start_main
98.82% 0.00% tchain_edit tchain_edit [.] main
|
---main
__libc_start_main
98.82% 0.00% tchain_edit tchain_edit [.] f1
|
---f1
main
__libc_start_main
98.82% 98.28% tchain_edit tchain_edit [.] f3
|
---f3
|
|--0.53%-- f2
| f1
| main
| __libc_start_main
|
|--0.01%-- f1
| main
| __libc_start_main
--99.46%-- [...]
97.63% 0.03% tchain_edit tchain_edit [.] f2
|
---f2
f1
main
__libc_start_main
7.13% 0.03% tchain_edit [kernel.vmlinux] [k] do_nmi
|
---do_nmi
end_repeat_nmi
f3
f2
f1
main
__libc_start_main--- tools/perf/Documentation/perf-record.txt | 4 ++++ tools/perf/util/evsel.c | 24 +++++++++++++++++++++++- tools/perf/util/parse-events.c | 18 ++++++++++++++++++ tools/perf/util/parse-events.h | 2 ++ tools/perf/util/parse-events.l | 2 ++ 5 files changed, 49 insertions(+), 1 deletion(-)
tools/perf/Documentation/perf-record.txt +4 -0
@@ -52,6 +52,10 @@ OPTIONS - 'time': Disable/enable time stamping. Acceptable values are 1 for enabling time stamping. 0 for disabling time stamping. The default is 1. + - 'callgraph': Disable/enable callgraph. Acceptable str are "fp" for + FP mode, "dwarf" for DWARF mode, "lbr" for LBR mode and + "no" for disable callgraph. + - 'stack_size': user stack size for dwarf mode Note: If user explicitly sets options which conflict with the params, the value set by the params will be overridden.
tools/perf/util/evsel.c +23 -1
@@ -620,6 +620,7 @@ void perf_evsel__config(struct perf_evsel *evsel, struct record_opts *opts)int track = evsel->tracking;bool per_cpu = opts->target.default_per_cpu && !opts->target.per_thread;bool sample_time = opts->sample_time; +bool callgraph = callchain_param.enabled; attr->sample_id_all = perf_missing_features.sample_id_all ?0 :1 ; attr->inherit = !opts->no_inherit;
@@ -636,6 +637,23 @@ void perf_evsel__config(struct perf_evsel *evsel, struct record_opts *opts) sample_time = false; } +if (!opts->callgraph_set) { +if (attr->sample_type &PERF_SAMPLE_CALLCHAIN ) { + callgraph = true; +if (attr->sample_type &PERF_SAMPLE_STACK_USER ) { + callchain_param.record_mode =CALLCHAIN_DWARF ; +if (attr->sample_stack_user) + callchain_param.dump_size = attr->sample_stack_user; +else + callchain_param.dump_size =8192 ; + }else if (attr->sample_type &PERF_SAMPLE_BRANCH_STACK ) + callchain_param.record_mode =CALLCHAIN_LBR ; +else + callchain_param.record_mode =CALLCHAIN_FP ; + }else + callgraph = false; + } +/* * Event parsing doesn't check the availability * Clear the bit which event parsing may be set.
@@ -650,6 +668,10 @@ void perf_evsel__config(struct perf_evsel *evsel, struct record_opts *opts) evsel->sample_size -=sizeof (u64); } + attr->sample_type &= ~(PERF_SAMPLE_CALLCHAIN | +PERF_SAMPLE_STACK_USER | +PERF_SAMPLE_BRANCH_STACK ); +perf_evsel__set_sample_bit (evsel,IP );perf_evsel__set_sample_bit (evsel,TID );
@@ -715,7 +737,7 @@ void perf_evsel__config(struct perf_evsel *evsel, struct record_opts *opts)if (perf_evsel__is_function_event (evsel)) evsel->attr.exclude_callchain_user =1 ; -if (callchain_param.enabled && !evsel->no_aux_samples) +if (callgraph && !evsel->no_aux_samples)perf_evsel__config_callgraph (evsel, opts);if (opts->sample_intr_regs) {
tools/perf/util/parse-events.c +18 -0
@@ -19,6 +19,7 @@#include "thread_map.h" #include "cpumap.h" #include "asm/bug.h" +#include "callchain.h" #define MAX_NAME_LEN 100
@@ -605,6 +606,23 @@ do { \if (term->val.num ==1 ) attr->sample_type |=PERF_SAMPLE_TIME ;break ; +case PARSE_EVENTS__TERM_TYPE_CALLGRAPH : +CHECK_TYPE_VAL (STR ); +if (!strcmp (term->val.str,"fp" )) + attr->sample_type |=PERF_SAMPLE_CALLCHAIN ; +else if (!strcmp (term->val.str,"dwarf" )) + attr->sample_type |=PERF_SAMPLE_CALLCHAIN | +PERF_SAMPLE_STACK_USER ; +else if (!strcmp (term->val.str,"lbr" )) + attr->sample_type |=PERF_SAMPLE_CALLCHAIN | +PERF_SAMPLE_BRANCH_STACK ; +else if (strcmp (term->val.str,"no" )) +return -EINVAL ; +break ; +case PARSE_EVENTS__TERM_TYPE_STACKSIZE : +CHECK_TYPE_VAL (NUM ); + attr->sample_stack_user = term->val.num; +break ;case PARSE_EVENTS__TERM_TYPE_NAME :CHECK_TYPE_VAL (STR );break ;
tools/perf/util/parse-events.h +2 -0
@@ -63,6 +63,8 @@ enum {PARSE_EVENTS__TERM_TYPE_SAMPLE_PERIOD ,PARSE_EVENTS__TERM_TYPE_BRANCH_SAMPLE_TYPE ,PARSE_EVENTS__TERM_TYPE_TIME , +PARSE_EVENTS__TERM_TYPE_CALLGRAPH , + PARSE_EVENTS__TERM_TYPE_STACKSIZE, }; struct parse_events_term {
tools/perf/util/parse-events.l +2 -0
@@ -184,6 +184,8 @@ name { return term(yyscanner, PARSE_EVENTS__TERM_TYPE_NAME); } period { return term(yyscanner, PARSE_EVENTS__TERM_TYPE_SAMPLE_PERIOD); } branch_type { return term(yyscanner, PARSE_EVENTS__TERM_TYPE_BRANCH_SAMPLE_TYPE); } time { return term(yyscanner, PARSE_EVENTS__TERM_TYPE_TIME); } +callgraph { return term(yyscanner, PARSE_EVENTS__TERM_TYPE_CALLGRAPH); } +stack_size { return term(yyscanner, PARSE_EVENTS__TERM_TYPE_TIME); } , { return ','; } "/" { BEGIN(INITIAL); return '/'; } {name_minus} { return str(yyscanner, PE_NAME); }
Signature · 1 line
1.8.3.1
RFC PATCH v3 5/5 perf,tests: Add tests to callgrap and time parse
In reply to [email protected]From: Kan Liang <[email protected]>
Add tests in tests/parse-events.c to check callgrap and time option
--- tools/perf/tests/parse-events.c | 28 ++++++++++++++++++++++++++++ 1 file changed, 28 insertions(+)
tools/perf/tests/parse-events.c +28 -0
@@ -471,6 +471,29 @@ static int test__checkevent_pmu_name(struct perf_evlist *evlist)return 0 ; } +static int test__checkevent_pmu_partial_time_callgraph (struct perf_evlist *evlist) +{ +struct perf_evsel *evsel =perf_evlist__first (evlist); + +/* cpu/config=1,callgraph=fp,time,period=100000/ */ +TEST_ASSERT_VAL ("wrong number of entries" ,2 == evlist->nr_entries); +TEST_ASSERT_VAL ("wrong type" ,PERF_TYPE_RAW == evsel->attr.type); +TEST_ASSERT_VAL ("wrong config" ,1 == evsel->attr.config); +TEST_ASSERT_VAL ("wrong period" ,100000 == evsel->attr.sample_period); +TEST_ASSERT_VAL ("wrong callgraph" ,PERF_SAMPLE_CALLCHAIN & evsel->attr.sample_type); +TEST_ASSERT_VAL ("wrong time" ,PERF_SAMPLE_TIME & evsel->attr.sample_type); + +/* cpu/config=2,callgraph=no,time=0,period=2000/ */ + evsel =perf_evsel__next (evsel); +TEST_ASSERT_VAL ("wrong type" ,PERF_TYPE_RAW == evsel->attr.type); +TEST_ASSERT_VAL ("wrong config" ,2 == evsel->attr.config); +TEST_ASSERT_VAL ("wrong period" ,2000 == evsel->attr.sample_period); +TEST_ASSERT_VAL ("wrong callgraph" , !(PERF_SAMPLE_CALLCHAIN & evsel->attr.sample_type)); +TEST_ASSERT_VAL ("wrong time" , !(PERF_SAMPLE_TIME & evsel->attr.sample_type)); + +return 0 ; +} +static int test__checkevent_pmu_events (struct perf_evlist *evlist) {struct perf_evsel *evsel =perf_evlist__first (evlist);
@@ -1547,6 +1570,11 @@ static struct evlist_test test__events_pmu[] = { .check = test__checkevent_pmu_name, .id = 1, }, + { + .name = "cpu/config=1,callgraph=fp,time,period=100000/,cpu/config=2,callgraph=no,time=0,period=2000/", + .check = test__checkevent_pmu_partial_time_callgraph, + .id = 2, + }, };struct terms_test {
Signature · 1 line
1.8.3.1