Support Intel uncore event lists
This adds uncore support on top of the recently merged JSON event list infrastructure for core events. Uncore is everything outside the core, including memory controllers, PCI, interconnect etc.
Uncore is more complicated to handle than core events because it uses many duplicated PMUs, which leads to long event lists and verbose duplicated outputs.
In fact previously it was nearly unusable for many cases without special tools to generate event list and aggregate data (such as
https://github.com/andikleen/pmu-tools/tree/master/ucevent)
With this patchkit we add:
- Basic support for uncore events in JSON events
- Support aliases that get duplicated over many PMUs transparently
- Support summing up duplicated PMUs per socket
- Support extending the perf stat builtin metrics with simple expressions specified in the event list.
So far mainly servers are supported. Also this is not using full event lists (which are full of very obscure events) but only for a smaller subset of curated useful and understandable metrics.
The actual event lists are not posted, but available atgit://git.kernel.org/pub/scm/linux/kernel/git/ak/linux-misc perf/intel-uncore-json-files-3
The code is available heregit://git.kernel.org/pub/scm/linux/kernel/git/ak/linux-misc perf/builtin-json-25
v1: Initial post
v2: Address review feedback. See changelog in commits.
v3: Repost. Rebase to latest tree.
v4: Rebase. Change DividedBy to generic simple expression parser. Fix refactoring problem that broke git bisect.
v5: Address review feedback. See changelog in commits. Add acks.
-Andi
PATCH 1/10 perf, tools: Parse eventcode as number in jevents
From: Andi Kleen <[email protected]>
The next patch needs to modify event code. Previously eventcode was just passed through as a string. Now parse it as a number.
v2: Don't special case 0
--- tools/perf/pmu-events/jevents.c | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-)
tools/perf/pmu-events/jevents.c +9 -1
@@ -135,7 +135,6 @@ static struct field {const char *field;const char *kernel; } fields[] = { - { "EventCode", "event=" }, { "UMask", "umask=" }, { "CounterMask", "cmask=" }, { "Invert", "inv=" },
@@ -343,6 +342,7 @@ int json_events(const char *fn,jsmntok_t *tokens, *tok;int i, j, len;char *map; +char buf[128 ];if (!fn)return -ENOENT ;
@@ -356,6 +356,7 @@ int json_events(const char *fn,char *event =NULL , *desc =NULL , *name =NULL ;char *long_desc =NULL ;char *extra_desc =NULL ; +unsigned long long eventcode =0 ;struct msrmap *msr =NULL ;jsmntok_t *msrval =NULL ;jsmntok_t *precise =NULL ;
@@ -376,6 +377,11 @@ int json_events(const char *fn, nz = !json_streq (map, val,"0" );if (match_field (map, field, nz, &event, val)) {/* ok */ + }else if (json_streq (map, field,"EventCode" )) { +char *code =NULL ; +addfield (map, &code,"" ,"" , val); + eventcode |=strtoul (code,NULL ,0 ); +free (code); }else if (json_streq (map, field,"EventName" )) {addfield (map, &name,"" ,"" , val); }else if (json_streq (map, field,"BriefDescription" )) {
@@ -410,6 +416,8 @@ int json_events(const char *fn,addfield (map, &extra_desc," " ,"(Precise event)" ,NULL ); } +snprintf (buf,sizeof buf,"event=%#llx" , eventcode); +addfield (map, &event,"," , buf,NULL );if (desc && extra_desc)addfield (map, &desc," " , extra_desc,NULL );if (long_desc && extra_desc)
Signature · 1 line
2.9.3
tip-bot for Andi Kleen
merge bot[tip:perf/core] perf jevents: Parse eventcode as number
perf jevents: Parse eventcode as number
Commit-ID: d581141970ef3965c1624960fa928d765afd8a3e Gitweb: http://git.kernel.org/tip/d581141970ef3965c1624960fa928d765afd8a3e Author: Andi Kleen <[email protected]> AuthorDate: Fri, 27 Jan 2017 18:03:36 -0800 Committer: Arnaldo Carvalho de Melo <[email protected]> CommitDate: Wed, 8 Feb 2017 08:55:02 -0300
perf jevents: Parse eventcode as number
The next patch needs to modify event code. Previously eventcode was just passed through as a string. Now parse it as a number.
v2: Don't special case 0
--- tools/perf/pmu-events/jevents.c | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-)
tools/perf/pmu-events/jevents.c +9 -1
@@ -135,7 +135,6 @@ static struct field {const char *field;const char *kernel; } fields[] = { - { "EventCode", "event=" }, { "UMask", "umask=" }, { "CounterMask", "cmask=" }, { "Invert", "inv=" },
@@ -343,6 +342,7 @@ int json_events(const char *fn,jsmntok_t *tokens, *tok;int i, j, len;char *map; +char buf[128 ];if (!fn)return -ENOENT ;
@@ -356,6 +356,7 @@ int json_events(const char *fn,char *event =NULL , *desc =NULL , *name =NULL ;char *long_desc =NULL ;char *extra_desc =NULL ; +unsigned long long eventcode =0 ;struct msrmap *msr =NULL ;jsmntok_t *msrval =NULL ;jsmntok_t *precise =NULL ;
@@ -376,6 +377,11 @@ int json_events(const char *fn, nz = !json_streq (map, val,"0" );if (match_field (map, field, nz, &event, val)) {/* ok */ + }else if (json_streq (map, field,"EventCode" )) { +char *code =NULL ; +addfield (map, &code,"" ,"" , val); + eventcode |=strtoul (code,NULL ,0 ); +free (code); }else if (json_streq (map, field,"EventName" )) {addfield (map, &name,"" ,"" , val); }else if (json_streq (map, field,"BriefDescription" )) {
@@ -410,6 +416,8 @@ int json_events(const char *fn,addfield (map, &extra_desc," " ,"(Precise event)" ,NULL ); } +snprintf (buf,sizeof buf,"event=%#llx" , eventcode); +addfield (map, &event,"," , buf,NULL );if (desc && extra_desc)addfield (map, &desc," " , extra_desc,NULL );if (long_desc && extra_desc)
PATCH 2/10 perf, tools: Add support for parsing uncore json files
In reply to Andi KleenFrom: Andi Kleen <[email protected]>
Handle the Unit field, which is needed to find the right PMU for an event. We call it "pmu" and convert it to the perf pmu name with an uncore prefix.
Handle the ExtSel field, which just extends the event mask with an additional bit.
Handle the Filter field which adds parameters to the main event to configure filtering.
Handle the Unit field which declares the unit the values should be scaled too (similar to what the kernel exports)
Set up the perpkg field for uncore events so that perf knows they are per package (similar to what the kernel exports)
Then output the fields into the pmu-events data structures which are compiled into perf.
Filter out zero fields, except for the event itself.
v2: Fix compilation. Add uncore_ prefix at pre-processing time. Move eventcode change to separate patch.
v3: Remove extra __maybe_unused
v4: dont duplicate aliases for cpu pmu events
--- tools/perf/pmu-events/jevents.c | 74 +++++++++++++++++++++++++++++++++++--- tools/perf/pmu-events/jevents.h | 4 ++- tools/perf/pmu-events/pmu-events.h | 3 ++ tools/perf/util/pmu.c | 30 +++++++++++----- 4 files changed, 98 insertions(+), 13 deletions(-)
tools/perf/pmu-events/jevents.c +70 -4
@@ -188,6 +188,27 @@ static struct msrmap *lookup_msr(char *map, jsmntok_t *val)return NULL ; } +static struct map { +const char *json; +const char *perf; +} unit_to_pmu[] = { + {"CBO" ,"uncore_cbox" }, + {"QPI LL" ,"uncore_qpi" }, + {"SBO" ,"uncore_sbox" }, + {} +}; + +static const char *field_to_perf (struct map *table,char *map,jsmntok_t *val) +{ +int i; + +for (i =0 ; table[i].json; i++) { +if (json_streq (map, val, table[i].json)) +return table[i].perf; + } +return NULL ; +} +#define EXPECT (e, t, m) do { if (!(e)) { \ jsmntok_t *loc = (t); \ if (!(t)->start && (t) > tokens) \
@@ -269,7 +290,8 @@ static void print_events_table_prefix(FILE *fp, const char *tblname) }static int print_events_table_entry (void *data,char *name,char *event, -char *desc,char *long_desc) +char *desc,char *long_desc, +char *pmu,char *unit,char *perpkg) {struct perf_entry_data *pd = data;FILE *outfp = pd->outfp;
@@ -287,7 +309,12 @@ static int print_events_table_entry(void *data, char *name, char *event,fprintf (outfp,"\t.topic = \"%s\",\n" , topic);if (long_desc && long_desc[0 ])fprintf (outfp,"\t.long_desc = \"%s\",\n" , long_desc); - +if (pmu) +fprintf (outfp,"\t.pmu = \"%s\",\n" , pmu); +if (unit) +fprintf (outfp,"\t.unit = \"%s\",\n" , unit); +if (perpkg) +fprintf (outfp,"\t.perpkg = \"%s\",\n" , perpkg);fprintf (outfp,"},\n" );return 0 ;
@@ -334,7 +361,8 @@ static char *real_event(const char *name, char *event)/* Call func with each event in the json file */ int json_events (const char *fn,int (*func)(void *data,char *name,char *event,char *desc, -char *long_desc), +char *long_desc, +char *pmu,char *unit,char *perpkg),void *data) {int err = -EIO ;
@@ -356,6 +384,10 @@ int json_events(const char *fn,char *event =NULL , *desc =NULL , *name =NULL ;char *long_desc =NULL ;char *extra_desc =NULL ; +char *pmu =NULL ; +char *filter =NULL ; +char *perpkg =NULL ; +char *unit =NULL ;unsigned long long eventcode =0 ;struct msrmap *msr =NULL ;jsmntok_t *msrval =NULL ;
@@ -382,6 +414,11 @@ int json_events(const char *fn,addfield (map, &code,"" ,"" , val); eventcode |=strtoul (code,NULL ,0 );free (code); + }else if (json_streq (map, field,"ExtSel" )) { +char *code =NULL ; +addfield (map, &code,"" ,"" , val); + eventcode |=strtoul (code,NULL ,0 ) <<21 ; +free (code); }else if (json_streq (map, field,"EventName" )) {addfield (map, &name,"" ,"" , val); }else if (json_streq (map, field,"BriefDescription" )) {
@@ -405,6 +442,28 @@ int json_events(const char *fn,addfield (map, &extra_desc,". " ," Supports address when precise" ,NULL ); + }else if (json_streq (map, field,"Unit" )) { +const char *ppmu; +char *s; + + ppmu =field_to_perf (unit_to_pmu, map, val); +if (ppmu) { + pmu =strdup (ppmu); + }else { +if (!pmu) + pmu =strdup ("uncore_" ); +addfield (map, &pmu,"" ,"" , val); +for (s = pmu; *s; s++) + *s =tolower (*s); + } +addfield (map, &desc,". " ,"Unit: " ,NULL ); +addfield (map, &desc,"" , pmu,NULL ); + }else if (json_streq (map, field,"Filter" )) { +addfield (map, &filter,"" ,"" , val); + }else if (json_streq (map, field,"ScaleUnit" )) { +addfield (map, &unit,"" ,"" , val); + }else if (json_streq (map, field,"PerPkg" )) { +addfield (map, &perpkg,"" ,"" , val); }/* ignore unknown fields */ }
@@ -422,16 +481,23 @@ int json_events(const char *fn,addfield (map, &desc," " , extra_desc,NULL );if (long_desc && extra_desc)addfield (map, &long_desc," " , extra_desc,NULL ); +if (filter) +addfield (map, &event,"," , filter,NULL );if (msr !=NULL )addfield (map, &event,"," , msr->pname, msrval);fixname (name); - err =func (data, name,real_event (name, event), desc, long_desc); + err =func (data, name,real_event (name, event), desc, long_desc, + pmu, unit, perpkg);free (event);free (desc);free (name);free (long_desc);free (extra_desc); +free (pmu); +free (filter); +free (perpkg); +free (unit);if (err)break ; tok += j;
tools/perf/pmu-events/jevents.h +3 -1
@@ -3,7 +3,9 @@int json_events (const char *fn,int (*func)(void *data,char *name,char *event,char *desc, -char *long_desc), +char *long_desc, +char *pmu, +char *unit,char *perpkg),void *data);char *get_cpu_str (void );
tools/perf/pmu-events/pmu-events.h +3 -0
@@ -10,6 +10,9 @@ struct pmu_event {const char *desc;const char *topic;const char *long_desc; +const char *pmu; +const char *unit; +const char *perpkg; };/*
tools/perf/util/pmu.c +22 -8
@@ -229,11 +229,13 @@ static int perf_pmu__parse_snapshot(struct perf_pmu_alias *alias, }static int __perf_pmu__new_alias (struct list_head *list,char *dir,char *name, -char *desc,char *val,char *long_desc, -char *topic) +char *desc,char *val, +char *long_desc,char *topic, +char *unit,char *perpkg) {struct perf_pmu_alias *alias;int ret; +int num; alias =malloc (sizeof (*alias));if (!alias)
@@ -267,7 +269,12 @@ static int __perf_pmu__new_alias(struct list_head *list, char *dir, char *name, alias->long_desc = long_desc ?strdup (long_desc) : desc ?strdup (desc) :NULL ; alias->topic = topic ?strdup (topic) :NULL ; - +if (unit) { +if (convert_scale (unit, &unit, &alias->scale) <0 ) +return -1 ; +snprintf (alias->unit,sizeof (alias->unit),"%s" , unit); + } + alias->per_pkg = perpkg &&sscanf (perpkg,"%d" , &num) ==1 && num ==1 ;list_add_tail (&alias->list, list);return 0 ;
@@ -284,7 +291,8 @@ static int perf_pmu__new_alias(struct list_head *list, char *dir, char *name, FI buf[ret] =0 ; -return __perf_pmu__new_alias (list, dir, name,NULL , buf,NULL ,NULL ); +return __perf_pmu__new_alias (list, dir, name,NULL , buf,NULL ,NULL ,NULL , +NULL ); }static inline bool pmu_alias_info_file (char *name)
@@ -504,7 +512,7 @@ char * __weak get_cpuid_str(void)* to the current running CPU. Then, add all PMU events from that table * as aliases. */ -static void pmu_add_cpu_aliases (struct list_head *head) +static void pmu_add_cpu_aliases (struct list_head *head,const char *name) {int i;struct pmu_events_map *map;
@@ -540,14 +548,21 @@ static void pmu_add_cpu_aliases(struct list_head *head)*/ i =0 ;while (1 ) { +const char *pname; + pe = &map->table[i++];if (!pe->name)break ; + pname = pe->pmu ? pe->pmu :"cpu" ; +if (strncmp (pname, name,strlen (pname))) +continue ; +/* need type casts to override 'const' */ __perf_pmu__new_alias (head,NULL , (char *)pe->name, (char *)pe->desc, (char *)pe->event, - (char *)pe->long_desc, (char *)pe->topic); + (char *)pe->long_desc, (char *)pe->topic, + (char *)pe->unit, (char *)pe->perpkg); } out:
@@ -578,8 +593,7 @@ static struct perf_pmu *pmu_lookup(const char *name)if (pmu_aliases (name, &aliases))return NULL ; -if (!strcmp (name,"cpu" )) -pmu_add_cpu_aliases (&aliases); +pmu_add_cpu_aliases (&aliases, name);if (pmu_type (name, &type))return NULL ;
Signature · 1 line
2.9.3
tip-bot for Andi Kleen
merge bot[tip:perf/core] perf jevents: Add support for parsing uncore json files
perf jevents: Add support for parsing uncore json files
Commit-ID: fedb2b518239cbc00abcf0d200e0be8436251c11 Gitweb: http://git.kernel.org/tip/fedb2b518239cbc00abcf0d200e0be8436251c11 Author: Andi Kleen <[email protected]> AuthorDate: Fri, 27 Jan 2017 18:03:37 -0800 Committer: Arnaldo Carvalho de Melo <[email protected]> CommitDate: Wed, 8 Feb 2017 08:55:03 -0300
perf jevents: Add support for parsing uncore json files
Handle the "Unit" field, which is needed to find the right PMU for an event. We call it "pmu" and convert it to the perf pmu name with an uncore prefix.
Handle the "ExtSel" field, which just extends the event mask with an additional bit.
Handle the "Filter" field which adds parameters to the main event to configure filtering.
Handle the "Unit" field which declares the unit the values should be scaled too (similar to what the kernel exports)
Set up the "perpkg" field for uncore events so that perf knows they are per package (similar to what the kernel exports)
Then output the fields into the pmu-events data structures which are compiled into perf.
Filter out zero fields, except for the event itself.
v2: Fix compilation. Add uncore_ prefix at pre-processing time.
Move eventcode change to separate patch.
v3: Remove extra __maybe_unused
v4: dont duplicate aliases for cpu pmu events
--- tools/perf/pmu-events/jevents.c | 74 +++++++++++++++++++++++++++++++++++--- tools/perf/pmu-events/jevents.h | 4 ++- tools/perf/pmu-events/pmu-events.h | 3 ++ tools/perf/util/pmu.c | 30 +++++++++++----- 4 files changed, 98 insertions(+), 13 deletions(-)
tools/perf/pmu-events/jevents.c +70 -4
@@ -188,6 +188,27 @@ static struct msrmap *lookup_msr(char *map, jsmntok_t *val)return NULL ; } +static struct map { +const char *json; +const char *perf; +} unit_to_pmu[] = { + {"CBO" ,"uncore_cbox" }, + {"QPI LL" ,"uncore_qpi" }, + {"SBO" ,"uncore_sbox" }, + {} +}; + +static const char *field_to_perf (struct map *table,char *map,jsmntok_t *val) +{ +int i; + +for (i =0 ; table[i].json; i++) { +if (json_streq (map, val, table[i].json)) +return table[i].perf; + } +return NULL ; +} +#define EXPECT (e, t, m) do { if (!(e)) { \ jsmntok_t *loc = (t); \ if (!(t)->start && (t) > tokens) \
@@ -269,7 +290,8 @@ static void print_events_table_prefix(FILE *fp, const char *tblname) }static int print_events_table_entry (void *data,char *name,char *event, -char *desc,char *long_desc) +char *desc,char *long_desc, +char *pmu,char *unit,char *perpkg) {struct perf_entry_data *pd = data;FILE *outfp = pd->outfp;
@@ -287,7 +309,12 @@ static int print_events_table_entry(void *data, char *name, char *event,fprintf (outfp,"\t.topic = \"%s\",\n" , topic);if (long_desc && long_desc[0 ])fprintf (outfp,"\t.long_desc = \"%s\",\n" , long_desc); - +if (pmu) +fprintf (outfp,"\t.pmu = \"%s\",\n" , pmu); +if (unit) +fprintf (outfp,"\t.unit = \"%s\",\n" , unit); +if (perpkg) +fprintf (outfp,"\t.perpkg = \"%s\",\n" , perpkg);fprintf (outfp,"},\n" );return 0 ;
@@ -334,7 +361,8 @@ static char *real_event(const char *name, char *event)/* Call func with each event in the json file */ int json_events (const char *fn,int (*func)(void *data,char *name,char *event,char *desc, -char *long_desc), +char *long_desc, +char *pmu,char *unit,char *perpkg),void *data) {int err = -EIO ;
@@ -356,6 +384,10 @@ int json_events(const char *fn,char *event =NULL , *desc =NULL , *name =NULL ;char *long_desc =NULL ;char *extra_desc =NULL ; +char *pmu =NULL ; +char *filter =NULL ; +char *perpkg =NULL ; +char *unit =NULL ;unsigned long long eventcode =0 ;struct msrmap *msr =NULL ;jsmntok_t *msrval =NULL ;
@@ -382,6 +414,11 @@ int json_events(const char *fn,addfield (map, &code,"" ,"" , val); eventcode |=strtoul (code,NULL ,0 );free (code); + }else if (json_streq (map, field,"ExtSel" )) { +char *code =NULL ; +addfield (map, &code,"" ,"" , val); + eventcode |=strtoul (code,NULL ,0 ) <<21 ; +free (code); }else if (json_streq (map, field,"EventName" )) {addfield (map, &name,"" ,"" , val); }else if (json_streq (map, field,"BriefDescription" )) {
@@ -405,6 +442,28 @@ int json_events(const char *fn,addfield (map, &extra_desc,". " ," Supports address when precise" ,NULL ); + }else if (json_streq (map, field,"Unit" )) { +const char *ppmu; +char *s; + + ppmu =field_to_perf (unit_to_pmu, map, val); +if (ppmu) { + pmu =strdup (ppmu); + }else { +if (!pmu) + pmu =strdup ("uncore_" ); +addfield (map, &pmu,"" ,"" , val); +for (s = pmu; *s; s++) + *s =tolower (*s); + } +addfield (map, &desc,". " ,"Unit: " ,NULL ); +addfield (map, &desc,"" , pmu,NULL ); + }else if (json_streq (map, field,"Filter" )) { +addfield (map, &filter,"" ,"" , val); + }else if (json_streq (map, field,"ScaleUnit" )) { +addfield (map, &unit,"" ,"" , val); + }else if (json_streq (map, field,"PerPkg" )) { +addfield (map, &perpkg,"" ,"" , val); }/* ignore unknown fields */ }
@@ -422,16 +481,23 @@ int json_events(const char *fn,addfield (map, &desc," " , extra_desc,NULL );if (long_desc && extra_desc)addfield (map, &long_desc," " , extra_desc,NULL ); +if (filter) +addfield (map, &event,"," , filter,NULL );if (msr !=NULL )addfield (map, &event,"," , msr->pname, msrval);fixname (name); - err =func (data, name,real_event (name, event), desc, long_desc); + err =func (data, name,real_event (name, event), desc, long_desc, + pmu, unit, perpkg);free (event);free (desc);free (name);free (long_desc);free (extra_desc); +free (pmu); +free (filter); +free (perpkg); +free (unit);if (err)break ; tok += j;
tools/perf/pmu-events/jevents.h +3 -1
@@ -3,7 +3,9 @@int json_events (const char *fn,int (*func)(void *data,char *name,char *event,char *desc, -char *long_desc), +char *long_desc, +char *pmu, +char *unit,char *perpkg),void *data);char *get_cpu_str (void );
tools/perf/pmu-events/pmu-events.h +3 -0
@@ -10,6 +10,9 @@ struct pmu_event {const char *desc;const char *topic;const char *long_desc; +const char *pmu; +const char *unit; +const char *perpkg; };/*
tools/perf/util/pmu.c +22 -8
@@ -229,11 +229,13 @@ static int perf_pmu__parse_snapshot(struct perf_pmu_alias *alias, }static int __perf_pmu__new_alias (struct list_head *list,char *dir,char *name, -char *desc,char *val,char *long_desc, -char *topic) +char *desc,char *val, +char *long_desc,char *topic, +char *unit,char *perpkg) {struct perf_pmu_alias *alias;int ret; +int num; alias =malloc (sizeof (*alias));if (!alias)
@@ -267,7 +269,12 @@ static int __perf_pmu__new_alias(struct list_head *list, char *dir, char *name, alias->long_desc = long_desc ?strdup (long_desc) : desc ?strdup (desc) :NULL ; alias->topic = topic ?strdup (topic) :NULL ; - +if (unit) { +if (convert_scale (unit, &unit, &alias->scale) <0 ) +return -1 ; +snprintf (alias->unit,sizeof (alias->unit),"%s" , unit); + } + alias->per_pkg = perpkg &&sscanf (perpkg,"%d" , &num) ==1 && num ==1 ;list_add_tail (&alias->list, list);return 0 ;
@@ -284,7 +291,8 @@ static int perf_pmu__new_alias(struct list_head *list, char *dir, char *name, FI buf[ret] =0 ; -return __perf_pmu__new_alias (list, dir, name,NULL , buf,NULL ,NULL ); +return __perf_pmu__new_alias (list, dir, name,NULL , buf,NULL ,NULL ,NULL , +NULL ); }static inline bool pmu_alias_info_file (char *name)
@@ -504,7 +512,7 @@ char * __weak get_cpuid_str(void)* to the current running CPU. Then, add all PMU events from that table * as aliases. */ -static void pmu_add_cpu_aliases (struct list_head *head) +static void pmu_add_cpu_aliases (struct list_head *head,const char *name) {int i;struct pmu_events_map *map;
@@ -540,14 +548,21 @@ static void pmu_add_cpu_aliases(struct list_head *head)*/ i =0 ;while (1 ) { +const char *pname; + pe = &map->table[i++];if (!pe->name)break ; + pname = pe->pmu ? pe->pmu :"cpu" ; +if (strncmp (pname, name,strlen (pname))) +continue ; +/* need type casts to override 'const' */ __perf_pmu__new_alias (head,NULL , (char *)pe->name, (char *)pe->desc, (char *)pe->event, - (char *)pe->long_desc, (char *)pe->topic); + (char *)pe->long_desc, (char *)pe->topic, + (char *)pe->unit, (char *)pe->perpkg); } out:
@@ -578,8 +593,7 @@ static struct perf_pmu *pmu_lookup(const char *name)if (pmu_aliases (name, &aliases))return NULL ; -if (!strcmp (name,"cpu" )) -pmu_add_cpu_aliases (&aliases); +pmu_add_cpu_aliases (&aliases, name);if (pmu_type (name, &type))return NULL ;
PATCH 3/10 perf, tools: Support per pmu json aliases
In reply to Andi KleenFrom: Andi Kleen <[email protected]>
Add support for registering json aliases per PMU. Any alias with an unit matching the prefix is registered to the PMU. Uncore has multiple instances of most units, so all these aliases get registered for each individual PMU (this is important later to run the event on every instance of the PMU).
To avoid printing the events multiple times in perf list filter out duplicated events during printing.
v2: Rely on uncore_ prefix already in unit
v3: Document why calls were reordered
--- tools/perf/util/pmu.c | 13 +++++++++---- 1 file changed, 9 insertions(+), 4 deletions(-)
tools/perf/util/pmu.c +9 -4
@@ -590,14 +590,16 @@ static struct perf_pmu *pmu_lookup(const char *name)if (pmu_format (name, &format))return NULL ; -if (pmu_aliases (name, &aliases)) +/* +* Check the type first to avoid unnecessary work. +*/ +if (pmu_type (name, &type))return NULL ; -pmu_add_cpu_aliases (&aliases, name); - -if (pmu_type (name, &type)) +if (pmu_aliases (name, &aliases))return NULL ; +pmu_add_cpu_aliases (&aliases, name); pmu =zalloc (sizeof (*pmu));if (!pmu)return NULL ;
@@ -1195,6 +1197,9 @@ void print_pmu_events(const char *event_glob, bool name_only, bool quiet_flag, len = j;qsort (aliases, len,sizeof (struct sevent ), cmp_sevent);for (j =0 ; j < len; j++) { +/* Skip duplicates */ +if (j >0 && !strcmp (aliases[j].name, aliases[j -1 ].name)) +continue ;if (name_only) {printf ("%s " , aliases[j].name);continue ;
Signature · 1 line
2.9.3
tip-bot for Andi Kleen
merge bot[tip:perf/core] perf pmu: Support per pmu json aliases
perf pmu: Support per pmu json aliases
Commit-ID: 15b22ed369aa23ef4d083ffb9621650c353d3ddd Gitweb: http://git.kernel.org/tip/15b22ed369aa23ef4d083ffb9621650c353d3ddd Author: Andi Kleen <[email protected]> AuthorDate: Fri, 27 Jan 2017 18:03:38 -0800 Committer: Arnaldo Carvalho de Melo <[email protected]> CommitDate: Wed, 8 Feb 2017 08:55:03 -0300
perf pmu: Support per pmu json aliases
Add support for registering json aliases per PMU. Any alias with an unit matching the prefix is registered to the PMU. Uncore has multiple instances of most units, so all these aliases get registered for each individual PMU (this is important later to run the event on every instance of the PMU).
To avoid printing the events multiple times in perf list filter out duplicated events during printing.
v2: Rely on uncore_ prefix already in unit
v3: Document why calls were reordered
--- tools/perf/util/pmu.c | 13 +++++++++---- 1 file changed, 9 insertions(+), 4 deletions(-)
tools/perf/util/pmu.c +9 -4
@@ -590,14 +590,16 @@ static struct perf_pmu *pmu_lookup(const char *name)if (pmu_format (name, &format))return NULL ; -if (pmu_aliases (name, &aliases)) +/* +* Check the type first to avoid unnecessary work. +*/ +if (pmu_type (name, &type))return NULL ; -pmu_add_cpu_aliases (&aliases, name); - -if (pmu_type (name, &type)) +if (pmu_aliases (name, &aliases))return NULL ; +pmu_add_cpu_aliases (&aliases, name); pmu =zalloc (sizeof (*pmu));if (!pmu)return NULL ;
@@ -1195,6 +1197,9 @@ void print_pmu_events(const char *event_glob, bool name_only, bool quiet_flag, len = j;qsort (aliases, len,sizeof (struct sevent ), cmp_sevent);for (j =0 ; j < len; j++) { +/* Skip duplicates */ +if (j >0 && !strcmp (aliases[j].name, aliases[j -1 ].name)) +continue ;if (name_only) {printf ("%s " , aliases[j].name);continue ;
PATCH 4/10 perf, tools: Support event aliases for non cpu// pmus
In reply to Andi KleenFrom: Andi Kleen <[email protected]>
The code for handling pmu aliases without specifying the PMU hardcoded only supported the cpu PMU.
This patch extends it to work for all PMUs. We always duplicate the event for all PMUs that have an matching alias. This allows to automatically expand an alias for all instances of a PMU (so for example you can monitor all cache boxes with a single event)
--- tools/perf/util/parse-events.c | 46 ++++++++++++++++++++++++------------------ tools/perf/util/parse-events.y | 32 ++++++++++++++++++++++------- 2 files changed, 51 insertions(+), 27 deletions(-)
tools/perf/util/parse-events.c +26 -20
@@ -1504,35 +1504,41 @@ static void perf_pmu__parse_init(void)struct perf_pmu_alias *alias;int len =0 ; - pmu =perf_pmu__find ("cpu" ); -if ((pmu ==NULL ) ||list_empty (&pmu->aliases)) { + pmu =NULL ; +while ((pmu =perf_pmu__scan (pmu)) !=NULL ) { +list_for_each_entry (alias, &pmu->aliases, list) { +if (strchr (alias->name,'-' )) + len++; + len++; + } + } + +if (len ==0 ) { perf_pmu_events_list_num =-1 ;return ; } -list_for_each_entry (alias, &pmu->aliases, list) { -if (strchr (alias->name,'-' )) - len++; - len++; - } perf_pmu_events_list =malloc (sizeof (struct perf_pmu_event_symbol ) * len);if (!perf_pmu_events_list)return ; perf_pmu_events_list_num = len; len =0 ; -list_for_each_entry (alias, &pmu->aliases, list) { -struct perf_pmu_event_symbol *p = perf_pmu_events_list + len; -char *tmp =strchr (alias->name,'-' ); - -if (tmp !=NULL ) { -SET_SYMBOL (strndup (alias->name, tmp - alias->name), -PMU_EVENT_SYMBOL_PREFIX ); - p++; -SET_SYMBOL (strdup (++tmp),PMU_EVENT_SYMBOL_SUFFIX ); - len +=2 ; - }else { -SET_SYMBOL (strdup (alias->name),PMU_EVENT_SYMBOL ); - len++; + pmu =NULL ; +while ((pmu =perf_pmu__scan (pmu)) !=NULL ) { +list_for_each_entry (alias, &pmu->aliases, list) { +struct perf_pmu_event_symbol *p = perf_pmu_events_list + len; +char *tmp =strchr (alias->name,'-' ); + +if (tmp !=NULL ) { +SET_SYMBOL (strndup (alias->name, tmp - alias->name), +PMU_EVENT_SYMBOL_PREFIX ); + p++; +SET_SYMBOL (strdup (++tmp),PMU_EVENT_SYMBOL_SUFFIX ); + len +=2 ; + }else { +SET_SYMBOL (strdup (alias->name),PMU_EVENT_SYMBOL ); + len++; + } } } qsort(perf_pmu_events_list, len,
tools/perf/util/parse-events.y +25 -7
@@ -12,6 +12,7 @@ #include <linux/list.h> #include <linux/types.h> #include "util.h" +#include "pmu.h" #include "parse-events.h" #include "parse-events-bison.h"
@@ -236,15 +237,32 @@ PE_KERNEL_PMU_EVENT sep_dc struct list_head *head; struct parse_events_term *term; struct list_head *list; + struct perf_pmu *pmu = NULL; + int ok = 0; - ALLOC_LIST(head); - ABORT_ON(parse_events_term__num(&term, PARSE_EVENTS__TERM_TYPE_USER, - $1, 1, &@1, NULL)); - list_add_tail(&term->list, head); - + /* Add it for all PMUs that support the alias */ ALLOC_LIST(list); - ABORT_ON(parse_events_add_pmu(data, list, "cpu", head)); - parse_events_terms__delete(head); + while ((pmu = perf_pmu__scan(pmu)) != NULL) { + struct perf_pmu_alias *alias; + + list_for_each_entry(alias, &pmu->aliases, list) { + if (!strcasecmp(alias->name, $1)) { + ALLOC_LIST(head); + ABORT_ON(parse_events_term__num(&term, PARSE_EVENTS__TERM_TYPE_USER, + $1, 1, &@1, NULL)); + list_add_tail(&term->list, head); + + if (!parse_events_add_pmu(data, list, + pmu->name, head)) { + ok++; + } + + parse_events_terms__delete(head); + } + } + } + if (!ok) + YYABORT; $$ = list; } |
Signature · 1 line
2.9.3
tip-bot for Andi Kleen
merge bot[tip:perf/core] perf pmu: Support event aliases for non cpu// pmus
perf pmu: Support event aliases for non cpu// pmus
Commit-ID: 231bb2aa32498cbebef1306889a02114e9dfc934 Gitweb: http://git.kernel.org/tip/231bb2aa32498cbebef1306889a02114e9dfc934 Author: Andi Kleen <[email protected]> AuthorDate: Fri, 27 Jan 2017 18:03:39 -0800 Committer: Arnaldo Carvalho de Melo <[email protected]> CommitDate: Wed, 8 Feb 2017 08:55:04 -0300
perf pmu: Support event aliases for non cpu// pmus
The code for handling pmu aliases without specifying the PMU hardcoded only supported the cpu PMU.
This patch extends it to work for all PMUs. We always duplicate the event for all PMUs that have an matching alias. This allows to automatically expand an alias for all instances of a PMU (so for example you can monitor all cache boxes with a single event)
--- tools/perf/util/parse-events.c | 46 ++++++++++++++++++++++++------------------ tools/perf/util/parse-events.y | 32 ++++++++++++++++++++++------- 2 files changed, 51 insertions(+), 27 deletions(-)
tools/perf/util/parse-events.c +26 -20
@@ -1504,35 +1504,41 @@ static void perf_pmu__parse_init(void)struct perf_pmu_alias *alias;int len =0 ; - pmu =perf_pmu__find ("cpu" ); -if ((pmu ==NULL ) ||list_empty (&pmu->aliases)) { + pmu =NULL ; +while ((pmu =perf_pmu__scan (pmu)) !=NULL ) { +list_for_each_entry (alias, &pmu->aliases, list) { +if (strchr (alias->name,'-' )) + len++; + len++; + } + } + +if (len ==0 ) { perf_pmu_events_list_num =-1 ;return ; } -list_for_each_entry (alias, &pmu->aliases, list) { -if (strchr (alias->name,'-' )) - len++; - len++; - } perf_pmu_events_list =malloc (sizeof (struct perf_pmu_event_symbol ) * len);if (!perf_pmu_events_list)return ; perf_pmu_events_list_num = len; len =0 ; -list_for_each_entry (alias, &pmu->aliases, list) { -struct perf_pmu_event_symbol *p = perf_pmu_events_list + len; -char *tmp =strchr (alias->name,'-' ); - -if (tmp !=NULL ) { -SET_SYMBOL (strndup (alias->name, tmp - alias->name), -PMU_EVENT_SYMBOL_PREFIX ); - p++; -SET_SYMBOL (strdup (++tmp),PMU_EVENT_SYMBOL_SUFFIX ); - len +=2 ; - }else { -SET_SYMBOL (strdup (alias->name),PMU_EVENT_SYMBOL ); - len++; + pmu =NULL ; +while ((pmu =perf_pmu__scan (pmu)) !=NULL ) { +list_for_each_entry (alias, &pmu->aliases, list) { +struct perf_pmu_event_symbol *p = perf_pmu_events_list + len; +char *tmp =strchr (alias->name,'-' ); + +if (tmp !=NULL ) { +SET_SYMBOL (strndup (alias->name, tmp - alias->name), +PMU_EVENT_SYMBOL_PREFIX ); + p++; +SET_SYMBOL (strdup (++tmp),PMU_EVENT_SYMBOL_SUFFIX ); + len +=2 ; + }else { +SET_SYMBOL (strdup (alias->name),PMU_EVENT_SYMBOL ); + len++; + } } } qsort(perf_pmu_events_list, len,
tools/perf/util/parse-events.y +25 -7
@@ -12,6 +12,7 @@ #include <linux/list.h> #include <linux/types.h> #include "util.h" +#include "pmu.h" #include "parse-events.h" #include "parse-events-bison.h"
@@ -236,15 +237,32 @@ PE_KERNEL_PMU_EVENT sep_dc struct list_head *head; struct parse_events_term *term; struct list_head *list; + struct perf_pmu *pmu = NULL; + int ok = 0; - ALLOC_LIST(head); - ABORT_ON(parse_events_term__num(&term, PARSE_EVENTS__TERM_TYPE_USER, - $1, 1, &@1, NULL)); - list_add_tail(&term->list, head); - + /* Add it for all PMUs that support the alias */ ALLOC_LIST(list); - ABORT_ON(parse_events_add_pmu(data, list, "cpu", head)); - parse_events_terms__delete(head); + while ((pmu = perf_pmu__scan(pmu)) != NULL) { + struct perf_pmu_alias *alias; + + list_for_each_entry(alias, &pmu->aliases, list) { + if (!strcasecmp(alias->name, $1)) { + ALLOC_LIST(head); + ABORT_ON(parse_events_term__num(&term, PARSE_EVENTS__TERM_TYPE_USER, + $1, 1, &@1, NULL)); + list_add_tail(&term->list, head); + + if (!parse_events_add_pmu(data, list, + pmu->name, head)) { + ok++; + } + + parse_events_terms__delete(head); + } + } + } + if (!ok) + YYABORT; $$ = list; } |
PATCH 5/10 perf, tools: Add debug support for outputing alias string
In reply to Andi KleenFrom: Andi Kleen <[email protected]>
For debugging and testing it is useful to see the converted alias string. Add support to perf stat/record and perf list to print the alias conversion. The text string is saved in the alias structure. For perf stat/record it is folded into the normal -v. For perf list -v was taken, so we use --debug.
Before:
% perf list
...
cache:
l1d.replacement
[L1D data line replacements]
l1d_pend_miss.fb_full
[Cycles a demand request was blocked due to Fill Buffers inavailability]
After
% perf list --debug
...
cache:
l1d.replacement
[L1D data line replacements]
cpu/umask=0x1,period=2000003,event=0x51/
l1d_pend_miss.fb_full
[Cycles a demand request was blocked due to Fill Buffers inavailability]
cpu/umask=0x2,period=2000003,cmask=1,event=0x48/
--- tools/perf/builtin-list.c | 3 +++ tools/perf/util/parse-events.y | 3 +++ tools/perf/util/pmu.c | 8 ++++++++ tools/perf/util/pmu.h | 1 + 4 files changed, 15 insertions(+)
tools/perf/builtin-list.c +3 -0
@@ -14,6 +14,7 @@#include "util/parse-events.h" #include "util/cache.h" #include "util/pmu.h" +#include "util/debug.h" #include <subcmd/parse-options.h> static bool desc_flag = true;
@@ -29,6 +30,8 @@ int cmd_list(int argc, const char **argv, const char *prefix __maybe_unused) "Print extra event descriptions. --no-desc to not print."),OPT_BOOLEAN ('v' ,"long-desc" , &long_desc_flag,"Print longer event descriptions." ), +OPT_INCR (0 ,"debug" , &verbose, +"Enable debugging output" ),OPT_END () }; const char * const list_usage[] = {
tools/perf/util/parse-events.y +3 -0
@@ -13,6 +13,7 @@ #include <linux/types.h> #include "util.h" #include "pmu.h" +#include "debug.h" #include "parse-events.h" #include "parse-events-bison.h"
@@ -254,6 +255,8 @@ PE_KERNEL_PMU_EVENT sep_dc if (!parse_events_add_pmu(data, list, pmu->name, head)) { + pr_debug("%s -> %s/%s/\n", $1, + pmu->name, alias->str); ok++; }
tools/perf/util/pmu.c +8 -0
@@ -275,6 +275,8 @@ static int __perf_pmu__new_alias(struct list_head *list, char *dir, char *name,snprintf (alias->unit,sizeof (alias->unit),"%s" , unit); } alias->per_pkg = perpkg &&sscanf (perpkg,"%d" , &num) ==1 && num ==1 ; + alias->str =strdup (val); +list_add_tail (&alias->list, list);return 0 ;
@@ -1087,6 +1089,8 @@ struct sevent {char *name;char *desc;char *topic; +char *str; +char *pmu; };static int cmp_sevent (const void *a,const void *b)
@@ -1183,6 +1187,8 @@ void print_pmu_events(const char *event_glob, bool name_only, bool quiet_flag, aliases[j].desc = long_desc ? alias->long_desc : alias->desc; aliases[j].topic = alias->topic; + aliases[j].str = alias->str; + aliases[j].pmu = pmu->name; j++; } if (pmu->selectable &&
@@ -1217,6 +1223,8 @@ void print_pmu_events(const char *event_glob, bool name_only, bool quiet_flag,printf ("%*s" ,8 ,"[" );wordwrap (aliases[j].desc,8 , columns,0 );printf ("]\n" ); +if (verbose) +printf ("%*s%s/%s/\n" ,8 ,"" , aliases[j].pmu, aliases[j].str); }else printf (" %-50s [Kernel PMU event]\n" , aliases[j].name); printed++;
tools/perf/util/pmu.h +1 -0
@@ -43,6 +43,7 @@ struct perf_pmu_alias {char *desc;char *long_desc;char *topic; +char *str;struct list_head terms;/* HEAD struct parse_events_term -> list */ struct list_head list;/* ELEM */ char unit[UNIT_MAX_LEN +1 ];
Signature · 1 line
2.9.3
tip-bot for Andi Kleen
merge bot[tip:perf/core] perf list: Add debug support for outputing alias string
perf list: Add debug support for outputing alias string
Commit-ID: f23610245c1aa0e912476e642bd5107d04122230 Gitweb: http://git.kernel.org/tip/f23610245c1aa0e912476e642bd5107d04122230 Author: Andi Kleen <[email protected]> AuthorDate: Fri, 27 Jan 2017 18:03:40 -0800 Committer: Arnaldo Carvalho de Melo <[email protected]> CommitDate: Wed, 8 Feb 2017 08:55:04 -0300
perf list: Add debug support for outputing alias string
For debugging and testing it is useful to see the converted alias string. Add support to perf stat/record and perf list to print the alias conversion. The text string is saved in the alias structure. For perf stat/record it is folded into the normal -v. For perf list -v was taken, so we use --debug.
Before:
% perf list
...
cache:
l1d.replacement
[L1D data line replacements]
l1d_pend_miss.fb_full
[Cycles a demand request was blocked due to Fill Buffers inavailability]
After
% perf list --debug
...
cache:
l1d.replacement
[L1D data line replacements]
cpu/umask=0x1,period=2000003,event=0x51/
l1d_pend_miss.fb_full
[Cycles a demand request was blocked due to Fill Buffers inavailability]
cpu/umask=0x2,period=2000003,cmask=1,event=0x48/
--- tools/perf/builtin-list.c | 3 +++ tools/perf/util/parse-events.y | 3 +++ tools/perf/util/pmu.c | 8 ++++++++ tools/perf/util/pmu.h | 1 + 4 files changed, 15 insertions(+)
tools/perf/builtin-list.c +3 -0
@@ -14,6 +14,7 @@#include "util/parse-events.h" #include "util/cache.h" #include "util/pmu.h" +#include "util/debug.h" #include <subcmd/parse-options.h> static bool desc_flag = true;
@@ -29,6 +30,8 @@ int cmd_list(int argc, const char **argv, const char *prefix __maybe_unused) "Print extra event descriptions. --no-desc to not print."),OPT_BOOLEAN ('v' ,"long-desc" , &long_desc_flag,"Print longer event descriptions." ), +OPT_INCR (0 ,"debug" , &verbose, +"Enable debugging output" ),OPT_END () }; const char * const list_usage[] = {
tools/perf/util/parse-events.y +3 -0
@@ -13,6 +13,7 @@ #include <linux/types.h> #include "util.h" #include "pmu.h" +#include "debug.h" #include "parse-events.h" #include "parse-events-bison.h"
@@ -254,6 +255,8 @@ PE_KERNEL_PMU_EVENT sep_dc if (!parse_events_add_pmu(data, list, pmu->name, head)) { + pr_debug("%s -> %s/%s/\n", $1, + pmu->name, alias->str); ok++; }
tools/perf/util/pmu.c +8 -0
@@ -275,6 +275,8 @@ static int __perf_pmu__new_alias(struct list_head *list, char *dir, char *name,snprintf (alias->unit,sizeof (alias->unit),"%s" , unit); } alias->per_pkg = perpkg &&sscanf (perpkg,"%d" , &num) ==1 && num ==1 ; + alias->str =strdup (val); +list_add_tail (&alias->list, list);return 0 ;
@@ -1087,6 +1089,8 @@ struct sevent {char *name;char *desc;char *topic; +char *str; +char *pmu; };static int cmp_sevent (const void *a,const void *b)
@@ -1183,6 +1187,8 @@ void print_pmu_events(const char *event_glob, bool name_only, bool quiet_flag, aliases[j].desc = long_desc ? alias->long_desc : alias->desc; aliases[j].topic = alias->topic; + aliases[j].str = alias->str; + aliases[j].pmu = pmu->name; j++; } if (pmu->selectable &&
@@ -1217,6 +1223,8 @@ void print_pmu_events(const char *event_glob, bool name_only, bool quiet_flag,printf ("%*s" ,8 ,"[" );wordwrap (aliases[j].desc,8 , columns,0 );printf ("]\n" ); +if (verbose) +printf ("%*s%s/%s/\n" ,8 ,"" , aliases[j].pmu, aliases[j].str); }else printf (" %-50s [Kernel PMU event]\n" , aliases[j].name); printed++;
tools/perf/util/pmu.h +1 -0
@@ -43,6 +43,7 @@ struct perf_pmu_alias {char *desc;char *long_desc;char *topic; +char *str;struct list_head terms;/* HEAD struct parse_events_term -> list */ struct list_head list;/* ELEM */ char unit[UNIT_MAX_LEN +1 ];
PATCH 6/10 perf, tools: Collapse identically named events in perf stat
In reply to Andi KleenFrom: Andi Kleen <[email protected]>
The uncore PMU has a lot of duplicated PMUs for different subsystems. When expanding an uncore alias we usually end up with a large number of identically named aliases, which makes perf stat output difficult to read.
Automatically sum them up in perf stat, unless --no-merge is specified.
This can be default because only the uncores generally have duplicated aliases. Other PMUs have unique names.
Before:
% perf stat --no-merge -a -e unc_c_llc_lookup.any sleep 1
Performance counter stats for 'system wide':
694,976 Bytes unc_c_llc_lookup.any
706,304 Bytes unc_c_llc_lookup.any
956,608 Bytes unc_c_llc_lookup.any
782,720 Bytes unc_c_llc_lookup.any
605,696 Bytes unc_c_llc_lookup.any
442,816 Bytes unc_c_llc_lookup.any
659,328 Bytes unc_c_llc_lookup.any
509,312 Bytes unc_c_llc_lookup.any
263,936 Bytes unc_c_llc_lookup.any
592,448 Bytes unc_c_llc_lookup.any
672,448 Bytes unc_c_llc_lookup.any
608,640 Bytes unc_c_llc_lookup.any
641,024 Bytes unc_c_llc_lookup.any
856,896 Bytes unc_c_llc_lookup.any
808,832 Bytes unc_c_llc_lookup.any
684,864 Bytes unc_c_llc_lookup.any
710,464 Bytes unc_c_llc_lookup.any
538,304 Bytes unc_c_llc_lookup.any
1.002577660 seconds time elapsedAfter:
% perf stat -a -e unc_c_llc_lookup.any sleep 1
Performance counter stats for 'system wide':
2,685,120 Bytes unc_c_llc_lookup.any
1.002648032 seconds time elapsedv2: Split collect_aliases. Rename alias flag.
v3: Make sure unsupported/not counted is always printed.
--- tools/perf/Documentation/perf-stat.txt | 3 + tools/perf/builtin-stat.c | 145 +++++++++++++++++++++++++++------ tools/perf/util/evsel.h | 1 + 3 files changed, 126 insertions(+), 23 deletions(-)
tools/perf/Documentation/perf-stat.txt +3 -0
@@ -237,6 +237,9 @@ To interpret the results it is usually needed to know on which CPUs the workload runs on. If needed the CPUs can be forced using taskset. +--no-merge:: +Do not merge results from same PMUs. + EXAMPLES --------
tools/perf/builtin-stat.c +122 -23
207 lines not on this page · raw message
tools/perf/util/evsel.h +1 -0
@@ -131,6 +131,7 @@ struct perf_evsel {bool cmdline_group_boundary;struct list_head config_terms;int bpf_fd; +bool merged_stat; };union u64_swap {
Signature · 1 line
2.9.3
Re: [PATCH 06/10] perf, tools: Collapse identically named events in perf stat
On Fri, Jan 27, 2017 at 06:03:41PM -0800, Andi Kleen wrote:
SNIP
Quoted 19 lines
tools/perf/builtin-stat.c +19
-0+static void collect_all_aliases (struct perf_evsel *counter, +void (*cb)(struct perf_evsel *counter,void *data, +bool first), +void *data) +{ +struct perf_evsel *alias; + + alias =list_prepare_entry (counter, &(evsel_list->entries), node); + list_for_each_entry_continue (alias, &evsel_list->entries, node) { +if (strcmp (perf_evsel__name (alias),perf_evsel__name (counter)) || + alias->scale != counter->scale || + alias->cgrp != counter->cgrp || +strcmp (alias->unit, counter->unit) || +nsec_counter (alias) !=nsec_counter (counter)) +break ; + alias->merged_stat = true; +cb (alias, data, false); + } +}
tools/perf/builtin-stat.c static void aggr_update_shadow(void) · line 1210
Quoted 10 lines
tools/perf/builtin-stat.c +10
-0+ +static void collect_aliases (struct perf_evsel *counter, +void (*cb)(struct perf_evsel *counter,void *data, +bool first), +void *data) +{ +cb (counter, data, true); +if (!no_merge) +collect_all_aliases (counter, cb, data); +}
could you please split this into 2 changes:
- adding adding collect_aliases function with no functional change just changing all the print callers to use it
- adding collect_all_aliases change that actualy change the behaviour and merges the stats
also please change the collect_aliases name to something generic like collect data
thanks,
jirka
PATCH 7/10 perf, tools: Expand PMU events by prefix match
In reply to Andi KleenFrom: Andi Kleen <[email protected]>
When the user specifies a pmu directly, expand it automatically with a prefix match, similar as we do for the normal aliases now.
This allows to specify attributes for duplicated boxes quickly. For example uncore_cbox_{0,6}/.../ can be now specified as cbox/.../ and it gets automatically expanded.
Before
% perf stat -a -e uncore_cbox_0/event=0x35,umask=0x1,filter_opc=0x19C/,\
uncore_cbox_1/event=0x35,umask=0x1,filter_opc=0x19C/,\
uncore_cbox_2/event=0x35,umask=0x1,filter_opc=0x19C/,\
uncore_cbox_3/event=0x35,umask=0x1,filter_opc=0x19C/,\
uncore_cbox_4/event=0x35,umask=0x1,filter_opc=0x19C/,\
uncore_cbox_5/event=0x35,umask=0x1,filter_opc=0x19C/ sleep 1
After
perf stat -a -e cbox/event=0x35,umask=0x1,filter_opc=0x19C/ sleep 1
v2: Handle all bison rules. Move multi add code to separate function. Handle uncore_ prefix correctly.
--- tools/perf/util/parse-events.c | 71 ++++++++++++++++++++++++++++++++++++++++ tools/perf/util/parse-events.h | 8 +++++ tools/perf/util/parse-events.y | 73 +++++++++++++++++------------------------- 3 files changed, 109 insertions(+), 43 deletions(-)
tools/perf/util/parse-events.c +71 -0
@@ -1257,6 +1257,52 @@ int parse_events_add_pmu(struct parse_events_evlist *data,return evsel ?0 : -ENOMEM ; } +int parse_events_multi_pmu_add (struct parse_events_evlist *data, +char *str,struct list_head **listp) +{ +struct list_head *head; +struct parse_events_term *term; +struct list_head *list; +struct perf_pmu *pmu =NULL ; +int ok =0 ; + + *listp =NULL ; +/* Add it for all PMUs that support the alias */ + list =malloc (sizeof (struct list_head )); +if (!list) +return -1 ; +INIT_LIST_HEAD (list); +while ((pmu =perf_pmu__scan (pmu)) !=NULL ) { +struct perf_pmu_alias *alias; + +list_for_each_entry (alias, &pmu->aliases, list) { +if (!strcasecmp (alias->name, str)) { + head =malloc (sizeof (struct list_head )); +if (!head) +return -1 ; +INIT_LIST_HEAD (head); +if (parse_events_term__num (&term,PARSE_EVENTS__TERM_TYPE_USER , + str,1 , &str,NULL ) <0 ) +return -1 ; +list_add_tail (&term->list, head); + +if (!parse_events_add_pmu (data, list, + pmu->name, head)) { +pr_debug ("%s -> %s/%s/\n" , str, + pmu->name, alias->str); + ok++; + } + +parse_events_terms__delete (head); + } + } + } +if (!ok) +return -1 ; + *listp = list; +return 0 ; +} +int parse_events__modifier_group (struct list_head *list,char *event_mod) {
@@ -2406,6 +2452,31 @@ int parse_events_term__clone(struct parse_events_term **new, term->err_term, term->err_val); } +int parse_events_copy_term_list (struct list_head *old, +struct list_head **new) +{ +struct parse_events_term *term, *n; +int ret; + +if (!old) { + *new =NULL ; +return 0 ; + } + + *new =malloc (sizeof (struct list_head )); +if (!*new) +return -ENOMEM ; +INIT_LIST_HEAD (*new); + + list_for_each_entry (term, old, list) { + ret =parse_events_term__clone (&n, term); +if (ret) +return ret; +list_add_tail (&n->list, *new); + } +return 0 ; +} +void parse_events_terms__purge (struct list_head *terms) {struct parse_events_term *term, *h;
tools/perf/util/parse-events.h +8 -0
@@ -164,6 +164,14 @@ int parse_events_add_breakpoint(struct list_head *list, int *idx,int parse_events_add_pmu (struct parse_events_evlist *data,struct list_head *list,char *name,struct list_head *head_config); + +int parse_events_multi_pmu_add (struct parse_events_evlist *data, +char *str, +struct list_head **listp); + +int parse_events_copy_term_list (struct list_head *old, +struct list_head **new); +enum perf_pmu_event_symbol_type perf_pmu__parse_check (const char *name);void parse_events__set_leader (char *name,struct list_head *list);
tools/perf/util/parse-events.y +30 -43
98 lines not on this page · raw message
Signature · 1 line
2.9.3
Re: [PATCH 07/10] perf, tools: Expand PMU events by prefix match
On Fri, Jan 27, 2017 at 06:03:42PM -0800, Andi Kleen wrote:
SNIP
Quoted 25 lines
enum
perf_pmu_event_symbol_typeperf_pmu__parse_check(const char *name);
voidparse_events__set_leader(char *name, struct list_head *list);tools/perf/util/parse-events.y +10
-2@@ -224,68 +224,55 @@ event_pmu: PE_NAME opt_event_config { struct parse_events_evlist *data = _data; - struct list_head *list; + struct list_head *list, *orig_terms, *terms; + + if (parse_events_copy_term_list($2, &orig_terms)) + YYABORT; ALLOC_LIST(list); - ABORT_ON(parse_events_add_pmu(data, list, $1, $2)); + if (parse_events_add_pmu(data, list, $1, $2)) { + struct perf_pmu *pmu = NULL; + int ok = 0; + + while ((pmu = perf_pmu__scan(pmu)) != NULL) { + char *name = pmu->name;
tools/perf/util/parse-events.y event_pmu: · line 248
Quoted 10 lines
tools/perf/util/parse-events.y +10
-0+ + if (!strncmp(name, "uncore_", 7) && + strncmp($1, "uncore_", 7)) + name += 7; + if (!strncmp($1, name, strlen($1))) { + if (parse_events_copy_term_list(orig_terms, &terms)) + YYABORT; + if (!parse_events_add_pmu(data, list, pmu->name, terms)) + ok++; + parse_events_terms__delete(terms);
could you please split this into 2 changes:
- adding parse_events_multi_pmu_add with no functional changes
- uncore box related matching
thanks,
jirka
PATCH 8/10 perf, tools: Add a simple expression parser for JSON
In reply to Andi KleenFrom: Andi Kleen <[email protected]>
Add a simple expression parser good enough to parse JSON relation expressions. The parser is implemented using bison.
--- tools/perf/tests/Build | 1 + tools/perf/tests/builtin-test.c | 4 ++ tools/perf/tests/expr.c | 48 +++++++++++++ tools/perf/tests/tests.h | 1 + tools/perf/util/Build | 5 ++ tools/perf/util/expr.h | 23 ++++++ tools/perf/util/expr.y | 156 ++++++++++++++++++++++++++++++++++++++++ 7 files changed, 238 insertions(+) create mode 100644 tools/perf/tests/expr.c create mode 100644 tools/perf/util/expr.h create mode 100644 tools/perf/util/expr.y
tools/perf/tests/Build +1 -0
@@ -38,6 +38,7 @@ perf-y += cpumap.o perf-y += stat.o perf-y += event_update.o perf-y += event-times.o +perf-y += expr.o perf-y += backward-ring-buffer.o perf-y += sdt.o perf-y += is_printable_array.o
tools/perf/tests/builtin-test.c +4 -0
@@ -44,6 +44,10 @@ static struct test generic_tests[] = { .func = test__parse_events, }, { + .desc = "Simple expression parser", + .func = test__expr, + }, + { .desc = "PERF_RECORD_* events & perf_sample fields", .func = test__PERF_RECORD, },
tools/perf/tests/expr.c new file +48 -0
48 lines not on this page · raw message
tools/perf/tests/tests.h +1 -0
@@ -62,6 +62,7 @@ int test__sample_parsing(int subtest);int test__keep_tracking (int subtest);int test__parse_no_sample_id_all (int subtest);int test__dwarf_unwind (int subtest); +int test__expr (int subtest);int test__hists_filter (int subtest);int test__mmap_thread_lookup (int subtest);int test__thread_mg_share (int subtest);
tools/perf/util/Build +5 -0
@@ -88,6 +88,7 @@ libperf-y += mem-events.o libperf-y += vsprintf.o libperf-y += drv_configs.o libperf-y += time-utils.o +libperf-y += expr-bison.o libperf-$(CONFIG_LIBBPF) += bpf-loader.o libperf-$(CONFIG_BPF_PROLOGUE) += bpf-prologue.o
@@ -140,6 +141,10 @@ $(OUTPUT)util/parse-events-bison.c: util/parse-events.y $(call rule_mkdir) $(Q)$(call echo-cmd,bison)$(BISON) -v util/parse-events.y -d $(PARSER_DEBUG_BISON) -o $@ -p parse_events_ +$(OUTPUT)util/expr-bison.c: util/expr.y + $(call rule_mkdir) + $(Q)$(call echo-cmd,bison)$(BISON) -v util/expr.y -d $(PARSER_DEBUG_BISON) -o $@ -p expr_ + $(OUTPUT)util/pmu-flex.c: util/pmu.l $(OUTPUT)util/pmu-bison.c $(call rule_mkdir) $(Q)$(call echo-cmd,flex)$(FLEX) -o $@ --header-file=$(OUTPUT)util/pmu-flex.h util/pmu.l
tools/perf/util/expr.h new file +23 -0
@@ -0,0 +1,23 @@ +#ifndef PARSE_CTX_H +#define PARSE_CTX_H 1 + +#define MAX_PARSE_ID 2 + +struct parse_id { +const char *name; +double val; +}; + +struct parse_ctx { +int num_ids; +struct parse_id ids[MAX_PARSE_ID ]; +}; + +void expr_ctx_init (struct parse_ctx *ctx); +void expr_add_id (struct parse_ctx *ctx,const char *id,double val); +#ifndef IN_EXPR_Y +int expr_parse (double *final_val,struct parse_ctx *ctx,const char **pp); +#endif +int expr_find_other (const char *p,const char *one,const char **other); + +#endif
tools/perf/util/expr.y new file +156 -0
156 lines not on this page · raw message
Signature · 1 line
2.9.3
Re: [PATCH 08/10] perf, tools: Add a simple expression parser for JSON
On Fri, Jan 27, 2017 at 06:03:43PM -0800, Andi Kleen wrote:
SNIP
Quoted 17 lines
tools/perf/util/expr.h new file +11
-0@@ -0,0 +1,23 @@ +#ifndef PARSE_CTX_H +#define PARSE_CTX_H 1 + +#define MAX_PARSE_ID 2 + +struct parse_id { +const char *name; +double val; +}; + +struct parse_ctx {
tools/perf/util/expr.h line 21
Quoted 10 lines
tools/perf/util/expr.h new file +10
-0+int num_ids; +struct parse_id ids[MAX_PARSE_ID ]; +}; + +void expr_ctx_init (struct parse_ctx *ctx); +void expr_add_id (struct parse_ctx *ctx,const char *id,double val); +#ifndef IN_EXPR_Y +int expr_parse (double *final_val,struct parse_ctx *ctx,const char **pp); +#endif +int expr_find_other (const char *p,const char *one,const char **other);
it all looks ok, just please follow the interface function name style we try to keep with struct name separated with __
thanks,
jirka
PATCH 9/10 perf, tools: Support MetricExpr header in JSON event list
In reply to Andi KleenFrom: Andi Kleen <[email protected]>
Add support for parsing the MetricExpr header in the JSON event lists and storing them in the alias structure.
Used in the next patch.
v2: Change DividedBy to MetricExpr
--- tools/perf/pmu-events/jevents.c | 18 ++++++++++++++---- tools/perf/pmu-events/jevents.h | 2 +- tools/perf/pmu-events/pmu-events.h | 1 + tools/perf/util/pmu.c | 9 ++++++--- tools/perf/util/pmu.h | 1 + 5 files changed, 23 insertions(+), 8 deletions(-)
tools/perf/pmu-events/jevents.c +14 -4
@@ -291,7 +291,8 @@ static void print_events_table_prefix(FILE *fp, const char *tblname)static int print_events_table_entry (void *data,char *name,char *event,char *desc,char *long_desc, -char *pmu,char *unit,char *perpkg) +char *pmu,char *unit,char *perpkg, +char *metric_expr) {struct perf_entry_data *pd = data;FILE *outfp = pd->outfp;
@@ -315,6 +316,8 @@ static int print_events_table_entry(void *data, char *name, char *event,fprintf (outfp,"\t.unit = \"%s\",\n" , unit);if (perpkg)fprintf (outfp,"\t.perpkg = \"%s\",\n" , perpkg); +if (metric_expr) +fprintf (outfp,"\t.metric_expr = \"%s\",\n" , metric_expr);fprintf (outfp,"},\n" );return 0 ;
@@ -362,7 +365,8 @@ static char *real_event(const char *name, char *event)int json_events (const char *fn,int (*func)(void *data,char *name,char *event,char *desc,char *long_desc, -char *pmu,char *unit,char *perpkg), +char *pmu,char *unit,char *perpkg, +char *metric_expr),void *data) {int err = -EIO ;
@@ -388,6 +392,7 @@ int json_events(const char *fn,char *filter =NULL ;char *perpkg =NULL ;char *unit =NULL ; +char *metric_expr =NULL ;unsigned long long eventcode =0 ;struct msrmap *msr =NULL ;jsmntok_t *msrval =NULL ;
@@ -398,6 +403,7 @@ int json_events(const char *fn,for (j =0 ; j < obj->size; j +=2 ) {jsmntok_t *field, *val;int nz; +char *s; field = tok + j; EXPECT(field->type == JSMN_STRING, tok + j,
@@ -444,7 +450,6 @@ int json_events(const char *fn, NULL); }else if (json_streq (map, field,"Unit" )) {const char *ppmu; -char *s; ppmu =field_to_perf (unit_to_pmu, map, val);if (ppmu) {
@@ -464,6 +469,10 @@ int json_events(const char *fn,addfield (map, &unit,"" ,"" , val); }else if (json_streq (map, field,"PerPkg" )) {addfield (map, &perpkg,"" ,"" , val); + }else if (json_streq (map, field,"MetricExpr" )) { +addfield (map, &metric_expr,"" ,"" , val); +for (s = metric_expr; *s; s++) + *s =tolower (*s); }/* ignore unknown fields */ }
@@ -488,7 +497,7 @@ int json_events(const char *fn,fixname (name); err =func (data, name,real_event (name, event), desc, long_desc, - pmu, unit, perpkg); + pmu, unit, perpkg, metric_expr);free (event);free (desc);free (name);
@@ -498,6 +507,7 @@ int json_events(const char *fn,free (filter);free (perpkg);free (unit); +free (metric_expr);if (err)break ; tok += j;
tools/perf/pmu-events/jevents.h +1 -1
@@ -5,7 +5,7 @@ int json_events(const char *fn,int (*func)(void *data,char *name,char *event,char *desc,char *long_desc,char *pmu, -char *unit,char *perpkg), +char *unit,char *perpkg,char *metric_expr), void *data);char *get_cpu_str (void );
tools/perf/pmu-events/pmu-events.h +1 -0
@@ -13,6 +13,7 @@ struct pmu_event {const char *pmu;const char *unit;const char *perpkg; +const char *metric_expr; };/*
tools/perf/util/pmu.c +6 -3
@@ -231,7 +231,8 @@ static int perf_pmu__parse_snapshot(struct perf_pmu_alias *alias,static int __perf_pmu__new_alias (struct list_head *list,char *dir,char *name,char *desc,char *val,char *long_desc,char *topic, -char *unit,char *perpkg) +char *unit,char *perpkg, +char *dividedby) {struct perf_pmu_alias *alias;int ret;
@@ -265,6 +266,7 @@ static int __perf_pmu__new_alias(struct list_head *list, char *dir, char *name,perf_pmu__parse_snapshot (alias, dir, name); } + alias->dividedby = dividedby ?strdup (dividedby) :NULL ; alias->desc = desc ?strdup (desc) :NULL ; alias->long_desc = long_desc ?strdup (long_desc) : desc ?strdup (desc) :NULL ;
@@ -294,7 +296,7 @@ static int perf_pmu__new_alias(struct list_head *list, char *dir, char *name, FI buf[ret] =0 ;return __perf_pmu__new_alias (list, dir, name,NULL , buf,NULL ,NULL ,NULL , -NULL ); +NULL ,NULL ); }static inline bool pmu_alias_info_file (char *name)
@@ -564,7 +566,8 @@ static void pmu_add_cpu_aliases(struct list_head *head, const char *name)__perf_pmu__new_alias (head,NULL , (char *)pe->name, (char *)pe->desc, (char *)pe->event, (char *)pe->long_desc, (char *)pe->topic, - (char *)pe->unit, (char *)pe->perpkg); + (char *)pe->unit, (char *)pe->perpkg, + (char *)pe->dividedby); } out:
tools/perf/util/pmu.h +1 -0
@@ -50,6 +50,7 @@ struct perf_pmu_alias {double scale;bool per_pkg;bool snapshot; +char *dividedby; };struct perf_pmu *perf_pmu__find (const char *name);
Signature · 1 line
2.9.3
Re: [PATCH 09/10] perf, tools: Support MetricExpr header in JSON event list
On Fri, Jan 27, 2017 at 06:03:44PM -0800, Andi Kleen wrote:
SNIP
Quoted 15 lines
tools/perf/util/pmu.c +2
-1@@ -231,7 +231,8 @@ static int perf_pmu__parse_snapshot(struct perf_pmu_alias *alias,static int __perf_pmu__new_alias (struct list_head *list,char *dir,char *name,char *desc,char *val,char *long_desc,char *topic, -char *unit,char *perpkg) +char *unit,char *perpkg, +char *dividedby) {struct perf_pmu_alias *alias;int ret;
tools/perf/util/pmu.c static int __perf_pmu__new_alias(struct list_head *list, char *dir, char *name, · line 269
Quoted 4 lines
tools/perf/util/pmu.c +1
-0@@ -265,6 +266,7 @@ static int __perf_pmu__new_alias(struct list_head *list, char *dir, char *name,perf_pmu__parse_snapshot (alias, dir, name); } + alias->dividedby = dividedby ?strdup (dividedby) :NULL ;
s/dividedby/metric_expr/ ?
CC util/pmu.outil/pmu.c: In function ‘pmu_add_cpu_aliases’:util/pmu.c:570:15: error: ‘struct pmu_event’ has no member named ‘dividedby’
(char *)pe->dividedby);
Quoted 13 lines
tools/perf/util/pmu.c +1
-1alias->desc = desc ?strdup (desc) :NULL ; alias->long_desc = long_desc ?strdup (long_desc) : desc ?strdup (desc) :NULL ;@@ -294,7 +296,7 @@ static int perf_pmu__new_alias(struct list_head *list, char *dir, char *name, FI buf[ret] =0 ;return __perf_pmu__new_alias (list, dir, name,NULL , buf,NULL ,NULL ,NULL , -NULL ); +NULL ,NULL ); }static inline bool pmu_alias_info_file (char *name)
tools/perf/util/pmu.c static void pmu_add_cpu_aliases(struct list_head *head, const char *name) · line 570
Quoted 6 lines
tools/perf/util/pmu.c +2
-1@@ -564,7 +566,8 @@ static void pmu_add_cpu_aliases(struct list_head *head, const char *name)__perf_pmu__new_alias (head,NULL , (char *)pe->name, (char *)pe->desc, (char *)pe->event, (char *)pe->long_desc, (char *)pe->topic, - (char *)pe->unit, (char *)pe->perpkg); + (char *)pe->unit, (char *)pe->perpkg, + (char *)pe->dividedby);
ditto
tools/perf/util/pmu.h struct perf_pmu_alias { · line 53
Quoted 12 lines
tools/perf/util/pmu.c +0
-0} out:tools/perf/util/pmu.h +1
-0@@ -50,6 +50,7 @@ struct perf_pmu_alias {double scale;bool per_pkg;bool snapshot; +char *dividedby;
ditto
Quoted 6 lines
tools/perf/util/pmu.h +0
-0};struct perf_pmu *perf_pmu__find (const char *name);Signature · 1 line
2.9.3
PATCH 10/10 perf, tools, stat: Output JSON MetricExpr metric
In reply to Andi KleenFrom: Andi Kleen <[email protected]>
Add generic infrastructure to perf stat to output ratios for "MetricExpr" entries in the event lists. Many events are more useful as ratios than in raw form, typically some count in relation to total ticks.
Transfer the MetricExpr information from the alias to the evsel.
We mark the events that need to be collected for MetricExpr, and also link the events using them with a pointer. The code is careful to always prefer the right event in the same group to minimize multiplexing errors. At the moment only a single relation is supported.
Then add a rblist to the stat shadow code that remembers stats based on the cpu and context.
Then finally update and retrieve and print these values similarly to the existing hardcoded perf metrics. We use the simple expression parser added earlier to evaluate the expression.
Normally we just output the result without further commentary, but for --metric-only this would lead to empty columns. So for this case use the original event as description.
So far there is no attempt to automatically add the MetricExpr event, if it is missing, however we suggest it to the user.
$ perf stat -a -I 1000 -e '{unc_p_clockticks,unc_p_freq_max_os_cycles}'
1.000228813 800,139,950 unc_p_clockticks
1.000228813 789,833,783 unc_p_freq_max_os_cycles # 98.7
2.000654229 800,308,990 unc_p_clockticks
2.000654229 396,214,238 unc_p_freq_max_os_cycles # 49.5$ perf stat -a -I 1000 -e '{unc_p_clockticks,unc_p_freq_max_os_cycles}' --metric-only
1.000206740 48.0
2.000451543 48.1v2: Change from DivideBy to MetricExpr
--- tools/perf/builtin-stat.c | 3 + tools/perf/util/evsel.c | 3 + tools/perf/util/evsel.h | 3 + tools/perf/util/parse-events.c | 1 + tools/perf/util/pmu.c | 8 ++- tools/perf/util/pmu.h | 3 +- tools/perf/util/stat-shadow.c | 152 +++++++++++++++++++++++++++++++++++++++++ tools/perf/util/stat.h | 2 + 8 files changed, 171 insertions(+), 4 deletions(-)
tools/perf/builtin-stat.c +3 -0
@@ -1141,6 +1141,7 @@ static void printout(int id, int nr, struct perf_evsel *counter, double uval, out.print_metric = pm; out.new_line = nl; out.ctx = &os; + out.force_header = false;if (csv_output && !metric_only) {print_noise (counter, noise);
@@ -1478,6 +1479,7 @@ static void print_metric_headers(const char *prefix, bool no_indent) out.ctx = &os; out.print_metric = print_metric_header; out.new_line = new_line_metric; + out.force_header = true; os.evsel = counter; perf_stat__print_shadow_stats(counter, 0, 0,
@@ -2460,6 +2462,7 @@ int cmd_stat(int argc, const char **argv, const char *prefix __maybe_unused) argc =parse_options_subcommand (argc, argv, stat_options, stat_subcommands, (const char **) stat_usage,PARSE_OPT_STOP_AT_NON_OPTION ); +perf_stat__collect_metric_expr (evsel_list);perf_stat__init_shadow_stats ();if (csv_sep) {
tools/perf/util/evsel.c +3 -0
@@ -236,6 +236,9 @@ void perf_evsel__init(struct perf_evsel *evsel, evsel->sample_size =__perf_evsel__sample_size (attr->sample_type);perf_evsel__calc_id_pos (evsel); evsel->cmdline_group_boundary = false; + evsel->metric_expr =NULL ; + evsel->metric_event =NULL ; + evsel->collect_stat = false; }struct perf_evsel *perf_evsel__new_idx (struct perf_event_attr *attr,int idx)
tools/perf/util/evsel.h +3 -0
@@ -132,6 +132,9 @@ struct perf_evsel {struct list_head config_terms;int bpf_fd;bool merged_stat; +const char * metric_expr; +struct perf_evsel *metric_event; +bool collect_stat; };union u64_swap {
tools/perf/util/parse-events.c +1 -0
@@ -1252,6 +1252,7 @@ int parse_events_add_pmu(struct parse_events_evlist *data, evsel->scale = info.scale; evsel->per_pkg = info.per_pkg; evsel->snapshot = info.snapshot; + evsel->metric_expr = info.metric_expr; }return evsel ?0 : -ENOMEM ;
tools/perf/util/pmu.c +5 -3
@@ -232,7 +232,7 @@ static int __perf_pmu__new_alias(struct list_head *list, char *dir, char *name, char *desc, char *val, char *long_desc, char *topic, char *unit, char *perpkg, - char *dividedby) + char *metric_expr) { struct perf_pmu_alias *alias;int ret;
@@ -266,7 +266,7 @@ static int __perf_pmu__new_alias(struct list_head *list, char *dir, char *name,perf_pmu__parse_snapshot (alias, dir, name); } - alias->dividedby = dividedby ?strdup (dividedby) :NULL ; + alias->metric_expr = metric_expr ?strdup (metric_expr) :NULL ; alias->desc = desc ?strdup (desc) :NULL ; alias->long_desc = long_desc ?strdup (long_desc) : desc ?strdup (desc) :NULL ;
@@ -567,7 +567,7 @@ static void pmu_add_cpu_aliases(struct list_head *head, const char *name) (char *)pe->desc, (char *)pe->event, (char *)pe->long_desc, (char *)pe->topic, (char *)pe->unit, (char *)pe->perpkg, - (char *)pe->dividedby); + (char *)pe->metric_expr); } out:
@@ -985,6 +985,7 @@ int perf_pmu__check_alias(struct perf_pmu *pmu, struct list_head *head_terms, info->unit =NULL ; info->scale =0.0 ; info->snapshot = false; + info->metric_expr =NULL ;list_for_each_entry_safe (term, h, head_terms, list) { alias =pmu_find_alias (pmu, term);
@@ -1000,6 +1001,7 @@ int perf_pmu__check_alias(struct perf_pmu *pmu, struct list_head *head_terms,if (alias->per_pkg) info->per_pkg = true; + info->metric_expr = alias->metric_expr;list_del (&term->list);free (term);
tools/perf/util/pmu.h +2 -1
@@ -31,6 +31,7 @@ struct perf_pmu {struct perf_pmu_info {const char *unit; +const char *metric_expr;double scale;bool per_pkg;bool snapshot;
@@ -50,7 +51,7 @@ struct perf_pmu_alias {double scale;bool per_pkg;bool snapshot; -char *dividedby; +char *metric_expr; };struct perf_pmu *perf_pmu__find (const char *name);
tools/perf/util/stat-shadow.c +152 -0
201 lines not on this page · raw message
tools/perf/util/stat.h +2 -0
@@ -85,11 +85,13 @@ struct perf_stat_output_ctx {void *ctx;print_metric_t print_metric;new_line_t new_line; +bool force_header; };void perf_stat__print_shadow_stats (struct perf_evsel *evsel,double avg,int cpu,struct perf_stat_output_ctx *out); +void perf_stat__collect_metric_expr (struct perf_evlist *);int perf_evlist__alloc_stats (struct perf_evlist *evlist,bool alloc_raw);void perf_evlist__free_stats (struct perf_evlist *evlist);
Signature · 1 line
2.9.3
Re: [PATCH 10/10] perf, tools, stat: Output JSON MetricExpr metric
Quoted 24 lines · Andi Kleen
From: Andi Kleen <[email protected]>
Add generic infrastructure to perf stat to output ratios for "MetricExpr" entries in the event lists. Many events are more useful as ratios than in raw form, typically some count in relation to total ticks.
Transfer the MetricExpr information from the alias to the evsel.
We mark the events that need to be collected for MetricExpr, and also link the events using them with a pointer. The code is careful to always prefer the right event in the same group to minimize multiplexing errors. At the moment only a single relation is supported.
Then add a rblist to the stat shadow code that remembers stats based on the cpu and context.
Then finally update and retrieve and print these values similarly to the existing hardcoded perf metrics. We use the simple expression parser added earlier to evaluate the expression.
Normally we just output the result without further commentary, but for --metric-only this would lead to empty columns. So for this case use the original event as description.
Quoted 8 lines
So far there is no attempt to automatically add the MetricExpr event, if it is missing, however we suggest it to the user.
$ perf stat -a -I 1000 -e '{unc_p_clockticks,unc_p_freq_max_os_cycles}' 1.000228813 800,139,950 unc_p_clockticks 1.000228813 789,833,783 unc_p_freq_max_os_cycles # 98.7 2.000654229 800,308,990 unc_p_clockticks 2.000654229 396,214,238 unc_p_freq_max_os_cycles # 49.5
[jolsa@krava perf]$ ./perf stat -a -I 1000 -e '{unc_p_clockticks,unc_p_freq_max_os_cycles}' invalid or unsupported event: '{unc_p_clockticks,unc_p_freq_max_os_cycles}' Run 'perf list' for a list of valid events
could you show an example of the MetricExpr?
it's part of the event record, what if you wanted to have 2 or more metrics defined for event?
who defines those expressions?
what if you dont provide the necessary events needed for the expression?
shouldnt we do it the other way around? like pick an expression we are interested in and perf will configure the needful..
Quoted 4 lines
$ perf stat -a -I 1000 -e '{unc_p_clockticks,unc_p_freq_max_os_cycles}' --metric-only
1.000206740 48.0
2.000451543 48.1how does user know what those number stand for?
is there a way for user to display the metric expression?
it's still feels to me like a hack without much concept behind
jirka
Re: [PATCH 10/10] perf, tools, stat: Output JSON MetricExpr metric
Quoted 34 lines · Jiri Olsa
Quoted 24 lines · Andi Kleen
From: Andi Kleen <[email protected]>
Add generic infrastructure to perf stat to output ratios for "MetricExpr" entries in the event lists. Many events are more useful as ratios than in raw form, typically some count in relation to total ticks.
Transfer the MetricExpr information from the alias to the evsel.
We mark the events that need to be collected for MetricExpr, and also link the events using them with a pointer. The code is careful to always prefer the right event in the same group to minimize multiplexing errors. At the moment only a single relation is supported.
Then add a rblist to the stat shadow code that remembers stats based on the cpu and context.
Then finally update and retrieve and print these values similarly to the existing hardcoded perf metrics. We use the simple expression parser added earlier to evaluate the expression.
Normally we just output the result without further commentary, but for --metric-only this would lead to empty columns. So for this case use the original event as description.
Quoted 8 lines
So far there is no attempt to automatically add the MetricExpr event, if it is missing, however we suggest it to the user.
$ perf stat -a -I 1000 -e '{unc_p_clockticks,unc_p_freq_max_os_cycles}' 1.000228813 800,139,950 unc_p_clockticks 1.000228813 789,833,783 unc_p_freq_max_os_cycles # 98.7 2.000654229 800,308,990 unc_p_clockticks 2.000654229 396,214,238 unc_p_freq_max_os_cycles # 49.5
Quoted 5 lines
[jolsa@krava perf]$ ./perf stat -a -I 1000 -e '{unc_p_clockticks,unc_p_freq_max_os_cycles}' invalid or unsupported event: '{unc_p_clockticks,unc_p_freq_max_os_cycles}' Run 'perf list' for a list of valid events
could you show an example of the MetricExpr?
It's in the event list branch
https://git.kernel.org/cgit/linux/kernel/git/ak/linux-misc.git/log/?h=perf/intel-uncore-json-files-3
All the metrics currently do is just the same as DividedBy earlier:
generate a percentage out of a count, typically based on clock ticks.
+ "MetricExpr": "(UNC_M_POWER_CHANNEL_PPD / UNC_M_CLOCKTICKS) *
100.",
Quoted 2 lines
it's part of the event record, what if you wanted to have 2 or more metrics defined for event?
Would need multiple copies of the event.
Quoted 2 lines
who defines those expressions?
It's metrics used internally by Intel.
Quoted 2 lines
what if you dont provide the necessary events needed for the expression?
Then perf prints a warning suggesting the events.
It's currently a TODO to add them automatically. Could be added, but the patch was already complex, so I didn't add it.
It's somewhat complicated because you would need to avoid duplicates and have to handle groups correctly. perf stat doesn't have the necessarily knowledge to fully understand the constraints on groups.
Then the extra event may not fit into the group, and it seemed saner to let the user decide what to do then, instead of generating a possible unschedulable group.
So it's not that easy to do it automatically.
Quoted 6 lines
Quoted 4 lines
$ perf stat -a -I 1000 -e '{unc_p_clockticks,unc_p_freq_max_os_cycles}' --metric-only 1.000206740 48.0 2.000451543 48.1how does user know what those number stand for?
I assume it was obvious enough that it is the percentage. Could add something more to the event description.
Quoted 2 lines
is there a way for user to display the metric expression?
Only in the source, but could be added to perf list -v
For most users metrics are much more useful than raw numbers. I think they would rather consider raw numbers to be a "hack with no concept"
-Andi
Re: [PATCH 10/10] perf, tools, stat: Output JSON MetricExpr metric
On Wed, Feb 08, 2017 at 01:51:05PM -0800, Andi Kleen wrote:
SNIP
Quoted 16 lines
Quoted 2 lines
could you show an example of the MetricExpr?
It's in the event list branch
https://git.kernel.org/cgit/linux/kernel/git/ak/linux-misc.git/log/?h=perf/intel-uncore-json-files-3
All the metrics currently do is just the same as DividedBy earlier:
generate a percentage out of a count, typically based on clock ticks.+ "MetricExpr": "(
UNC_M_POWER_CHANNEL_PPD/UNC_M_CLOCKTICKS) *
100.",Quoted 1 line
Quoted 3 lines
Quoted 1 line
it's part of the event record, what if you wanted to have 2 or more metrics defined for event?
Would need multiple copies of the event.
this ...
Quoted 19 lines
Quoted 3 lines
who defines those expressions?
It's metrics used internally by Intel.
Quoted 2 lines
what if you dont provide the necessary events needed for the expression?
Then perf prints a warning suggesting the events.
It's currently a TODO to add them automatically. Could be added, but the patch was already complex, so I didn't add it.
It's somewhat complicated because you would need to avoid duplicates and have to handle groups correctly. perf stat doesn't have the necessarily knowledge to fully understand the constraints on groups.
Quoted 5 lines
Then the extra event may not fit into the group, and it seemed saner to let the user decide what to do then, instead of generating a possible unschedulable group.
So it's not that easy to do it automatically.
and this makes me think, that this is not the right approach
adding extra copy of an event when you want to add new expression?
why can't we have another list/file of those expressions from which point we could point and configure events we need
jirka
Re: [PATCH 10/10] perf, tools, stat: Output JSON MetricExpr metric
Quoted 3 lines · Jiri Olsa
and this makes me think, that this is not the right approach
adding extra copy of an event when you want to add new expression?
I don't want to add new expressions.
I don't even need arbitrary expressions, just DividedBy to get percentages, you just forced me to do the expressions.
Quoted 1 line
why can't we have another list/file of those expressions
The last time I proposed separate files Ingo vetoed it. He wanted everything built in.
Quoted 1 line
from which point we could point and configure events we need
If you want full flexibility you can use your perf stat report approach, or what most people do is to just run a script/spreadsheet over the the -x; output. This all continues to work.
This is just a minimum approach to provide some convenience integrated with the event list to provide something similar as the built in expressions in stat-shadow.
It's not trying to build the great perf scripting language.
-Andi
Re: [PATCH 10/10] perf, tools, stat: Output JSON MetricExpr metric
Quoted 7 lines · Andi Kleen
Quoted 3 lines · Jiri Olsa
and this makes me think, that this is not the right approach
adding extra copy of an event when you want to add new expression?
I don't want to add new expressions.
Quoted 8 lines
I don't even need arbitrary expressions, just DividedBy to get percentages, you just forced me to do the expressions.
Quoted 1 line
why can't we have another list/file of those expressions
The last time I proposed separate files Ingo vetoed it. He wanted everything built in.
sure, he veto it for event files.. expressions could be built in same way as we have events now
Quoted 11 lines
Quoted 1 line
from which point we could point and configure events we need
If you want full flexibility you can use your perf stat report approach, or what most people do is to just run a script/spreadsheet over the the -x; output. This all continues to work.
This is just a minimum approach to provide some convenience integrated with the event list to provide something similar as the built in expressions in stat-shadow.
It's not trying to build the great perf scripting language.
yea I understand that but can't ack that based on the points I descibed in my other email
jirka
Re: [PATCH 10/10] perf, tools, stat: Output JSON MetricExpr metric
Quoted 5 lines · Jiri Olsa
Quoted 2 lines
The last time I proposed separate files Ingo vetoed it. He wanted everything built in.
sure, he veto it for event files.. expressions could be built in same way as we have events now
That's exactly what I implemented. The expression is part of the JSON file, which seems like the logical place.
You just want it in a separate file in the source?
Quoted 7 lines
Quoted 7 lines
Quoted 2 lines
from which point we could point and configure events we need
If you want full flexibility you can use your perf stat report approach, or what most people do is to just run a script/spreadsheet over the the -x; output. This all continues to work.
Quoted 8 lines
Quoted 5 lines
This is just a minimum approach to provide some convenience integrated with the event list to provide something similar as the built in expressions in stat-shadow.
It's not trying to build the great perf scripting language.
yea I understand that but can't ack that based on the points I descibed in my other email
So what are the crucial points that prevent you?
- You want better column descriptions?
I suppose could add another field for this.
- You want multiple expressions per event
(even though they are not needed today)?
It could be implemented, but seems like unnecessary complexity and overengineering to me at this point. If nobody ever uses it would it be time spent well? If someone really uses it they could add the support at that time.
- You want automatic group creation?
It has nasty corner cases.
It would be possible to build something that works for simple cases.
Anything I missed?
-Andi
Re: [PATCH 10/10] perf, tools, stat: Output JSON MetricExpr metric
Quoted 35 lines · Andi Kleen
Quoted 5 lines · Jiri Olsa
Quoted 2 lines
The last time I proposed separate files Ingo vetoed it. He wanted everything built in.
sure, he veto it for event files.. expressions could be built in same way as we have events now
That's exactly what I implemented. The expression is part of the JSON file, which seems like the logical place.
You just want it in a separate file in the source?
Quoted 7 lines
Quoted 7 lines
Quoted 2 lines
from which point we could point and configure events we need
If you want full flexibility you can use your perf stat report approach, or what most people do is to just run a script/spreadsheet over the the -x; output. This all continues to work.
Quoted 8 lines
Quoted 5 lines
This is just a minimum approach to provide some convenience integrated with the event list to provide something similar as the built in expressions in stat-shadow.
It's not trying to build the great perf scripting language.
yea I understand that but can't ack that based on the points I descibed in my other email
So what are the crucial points that prevent you?
- You want better column descriptions?
I suppose could add another field for this.
Quoted 8 lines
- You want multiple expressions per event
(even though they are not needed today)?It could be implemented, but seems like unnecessary complexity and overengineering to me at this point. If nobody ever uses it would it be time spent well? If someone really uses it they could add the support at that time.
ok, I checked optimization doc and you might be right about this one, I couldn't find an example that'd prove you wrong ;-)
however I dont think expression should be annonymous number tied to an event. For example there's this one:
B.7.9.2
Fast Synchronization Penalty
50. Locked Operations Impact: (L1D_CACHE_LOCK_DURATION + 20 * L1D_CACHE_LOCK.MESI) / CPU_CLK_UNHALTED.CORE * 100
Fast synchronization is frequently implemented usin g locked memory accesses. A high value for Locked Operations Impact indicates that locked operations us ed in the workload have high penalty. The latency of a locked operation depends on the location of the
data: L1 data cache, L2 cache, other core cache or memory.
There's a name and description.
which event will pick this expression? L1D_CACHE_LOCK_DURATION or L1D_CACHE_LOCK?
How about we add:
"MetricExpr": "(L1D_CACHE_LOCK_DURATION + 20 * L1D_CACHE_LOCK.MESI) / CPU_CLK_UNHALTED.CORE * 100" "MetricExprName": "Fast Synchronization Penalty" "MetricExprDoc": "Fast synchronization is frequently implemented... "
then it could be part of the event record and we can display it like use the name in column and doc in the perf list
We could append number or something (MetricExpr2..) if there'll be multiple expressions in some case.. or something like that
Quoted 6 lines
- You want automatic group creation?
It has nasty corner cases.
It would be possible to build something that works for simple cases.
that would be nice.. when user says perf stat -e 'Fast Synchronization Penalty' ... it'd create necessary events for him.. or some shorted name probably
jirka
Re: [PATCH 10/10] perf, tools, stat: Output JSON MetricExpr metric
In reply to Andi KleenOn Wed, Feb 08, 2017 at 01:51:05PM -0800, Andi Kleen wrote:
SNIP
Quoted 5 lines
Only in the source, but could be added to perf list -v
For most users metrics are much more useful than raw numbers. I think they would rather consider raw numbers to be a "hack with no concept"
well, we can always fix that by showing the column header, but that's not what I meant
jirka