perf tools: Add non-architectural event aliases
v2 cover perf tools: Add non-architectural event aliases
hi,
adding support to use non-architectural events in perf via name, to be able to use them like:
$ perf stat -e 'cpu/IDQ.ALL_DSB_CYCLES/' ... $ perf stat -e 'cpu/L2_STORE_LOCK_RQSTS.MISS/' ... ...
The perf list command displays current non-arch. events available:
$ perf list ... cpu/AGU_BYPASS_CANCEL.COUNT/ [Kernel PMU event] cpu/ARITH.FPU_DIV/ [Kernel PMU event] cpu/ARITH.FPU_DIV_ACTIVE/ [Kernel PMU event] cpu/BACLEARS.ANY/ [Kernel PMU event] ...
This version is dependent on precise changes:
http://marc.info/?l=linux-kernel&m=135922143117888&w=2
First version of this RFC is here:
http://marc.info/?l=linux-kernel&m=135575146128720&w=2
This version provides support for non-architectural events aliasses being defined in sigle files, like for SandyBridge:
$ cat arch/x86/events/intel/snb | head -5 name=AGU_BYPASS_CANCEL.COUNT,event=0xb6,umask=0x1 name=ARITH.FPU_DIV_ACTIVE,event=0x14,umask=0x1 name=ARITH.FPU_DIV,event=0x14,umask=0x10401 name=BACLEARS.ANY,event=0xe6,umask=0x1f ...
So far this is still under testing and only SandyBridge and IvyBridge events were added. I have no problem to add the rest of the archs. when the shape of this is agreed.
Available also at:
git://git.kernel.org/pub/scm/linux/kernel/git/jolsa/linux.git
perf/nonarch9
any comments are welcome,
jirka
---
Andi Kleen (1):
perf tools: List kernel supplied event aliases in perf list v2
Jiri Olsa (7):
perf tools: Add '.' as part of the event 'name' token perf tools: Change perf_pmu__new_alias function interface perf tools: Add name term processing for alias
perf tools: Add pmu interface to parse single file of aliases perf tools: Add support to include non architectural event aliases perf tools: Add non arch events for SandyBridge microarchitecture perf tools: Add non arch events for IvyBridge micro architecture
tools/perf/Documentation/perf-list.txt | 4 +- tools/perf/arch/x86/Makefile | 11 ++++ tools/perf/arch/x86/events/intel/ivb | 248 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ tools/perf/arch/x86/events/intel/snb | 274 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ tools/perf/arch/x86/util/pmu.c | 125 ++++++++++++++++++++++++++++++++++++++++++ tools/perf/builtin-list.c | 3 + tools/perf/util/parse-events.c | 24 ++------ tools/perf/util/parse-events.l | 4 +- tools/perf/util/pmu.c | 158 ++++++++++++++++++++++++++++++++++++++++++++++++----- tools/perf/util/pmu.h | 9 +++ 10 files changed, 824 insertions(+), 36 deletions(-)
PATCH v2 1/8 perf tools: Add '.' as part of the event 'name' token
Adding '.' to be recognized as valid part of the event 'name' token. Upcoming non architectural events use '.' as part of the name.
--- tools/perf/util/parse-events.l | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
tools/perf/util/parse-events.l +2 -2
@@ -80,8 +80,8 @@ event [^,{}/]+ num_dec [0-9]+ num_hex 0x[a-fA-F0-9]+ num_raw_hex [a-fA-F0-9]+ -name [a-zA-Z_*?][a-zA-Z0-9_*?]* -name_minus [a-zA-Z_*?][a-zA-Z0-9\-_*?]* +name [a-zA-Z_*?][a-zA-Z0-9_*?\.]* +name_minus [a-zA-Z_*?][a-zA-Z0-9\-_*?\.]* modifier_event [ukhpGH]+ modifier_bp [rwx]{1,3}
Signature · 1 line
1.7.11.7
Re: [PATCH 1/8] perf tools: Add '.' as part of the event 'name' token
Jiri,
I don't see part 0/8 of this series. Did you send it to me too?
I have some comments about it. I don't see why create something from scratch when I have been developing a library (libpfm4) that takes care of that and that is already used by many tool developers. That library can be linked with perf and provide full symbolic events + all the modifiers. The library is portable and supports all existing archs. It can also be used by self-monitoring apps.
You're introducing yet another event table to maintain. And believe me this is a lot of work to maintain this.
I don't understand why not use this existing library.
Quoted 36 lines · Jiri Olsa
Adding '.' to be recognized as valid part of the event 'name' token. Upcoming non architectural events use '.' as part of the name.
Signed-off-by Jiri Olsa <[email protected]>Cc Corey Ashford <[email protected]>Cc Frederic Weisbecker <[email protected]>Cc Ingo Molnar <[email protected]>Cc Namhyung Kim <[email protected]>Cc Paul Mackerras <[email protected]>Cc Peter Zijlstra <[email protected]>Cc Arnaldo Carvalho de Melo <[email protected]>Cc Andi Kleen <[email protected]>Cc Stephane Eranian <[email protected]>--- tools/perf/util/parse-events.l | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)tools/perf/util/parse-events.l +2
-3@@ -80,8 +80,8 @@ event [^,{}/]+ num_dec [0-9]+ num_hex 0x[a-fA-F0-9]+ num_raw_hex [a-fA-F0-9]+ -name [a-zA-Z_*?][a-zA-Z0-9_*?]* -name_minus [a-zA-Z_*?][a-zA-Z0-9\-_*?]* +name [a-zA-Z_*?][a-zA-Z0-9_*?\.]* +name_minus [a-zA-Z_*?][a-zA-Z0-9\-_*?\.]* modifier_event [ukhpGH]+ modifier_bp [rwx]{1,3} --1.7.11.7
Re: [PATCH 1/8] perf tools: Add '.' as part of the event 'name' token
Quoted 11 lines · Stephane Eranian
Jiri,
I don't see part 0/8 of this series. Did you send it to me too?
I have some comments about it. I don't see why create something from scratch when I have been developing a library (libpfm4) that takes care of that and that is already used by many tool developers. That library can be linked with perf and provide full symbolic events + all the modifiers. The library is portable and supports all existing archs. It can also be used by self-monitoring apps.
Quoted 5 lines
You're introducing yet another event table to maintain. And believe me this is a lot of work to maintain this.
I don't understand why not use this existing library.
I meant to add that I think it would be more productive if we (you and I) were to work on the library to extend it with external text-based event tables that could be used by perf either directly or thru the libpfm4 interface.
Quoted 38 lines · Jiri Olsa
Quoted 36 lines
Adding '.' to be recognized as valid part of the event 'name' token. Upcoming non architectural events use '.' as part of the name.
Signed-off-by Jiri Olsa <[email protected]>Cc Corey Ashford <[email protected]>Cc Frederic Weisbecker <[email protected]>Cc Ingo Molnar <[email protected]>Cc Namhyung Kim <[email protected]>Cc Paul Mackerras <[email protected]>Cc Peter Zijlstra <[email protected]>Cc Arnaldo Carvalho de Melo <[email protected]>Cc Andi Kleen <[email protected]>Cc Stephane Eranian <[email protected]>--- tools/perf/util/parse-events.l | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)tools/perf/util/parse-events.l +2
-3@@ -80,8 +80,8 @@ event [^,{}/]+ num_dec [0-9]+ num_hex 0x[a-fA-F0-9]+ num_raw_hex [a-fA-F0-9]+ -name [a-zA-Z_*?][a-zA-Z0-9_*?]* -name_minus [a-zA-Z_*?][a-zA-Z0-9\-_*?]* +name [a-zA-Z_*?][a-zA-Z0-9_*?\.]* +name_minus [a-zA-Z_*?][a-zA-Z0-9\-_*?\.]* modifier_event [ukhpGH]+ modifier_bp [rwx]{1,3} --1.7.11.7
Re: [PATCH 1/8] perf tools: Add '.' as part of the event 'name' token
Quoted 15 lines · Stephane Eranian
Quoted 14 lines · Stephane Eranian
Jiri,
I don't see part 0/8 of this series. Did you send it to me too?
I have some comments about it. I don't see why create something from scratch when I have been developing a library (libpfm4) that takes care of that and that is already used by many tool developers. That library can be linked with perf and provide full symbolic events + all the modifiers. The library is portable and supports all existing archs. It can also be used by self-monitoring apps.
You're introducing yet another event table to maintain. And believe me this is a lot of work to maintain this.
Quoted 7 lines
Quoted 2 lines
I don't understand why not use this existing library.
I meant to add that I think it would be more productive if we (you and I) were to work on the library to extend it with external text-based event tables that could be used by perf either directly or thru the libpfm4 interface.
perf is intentionally external file free and does not (fundamentally) depend on external libraries either, other than core system libraries.
There are several advantages to that:
- there is no version skew and no design/maintenance friction
- 'upgrading' perf between similar boxes is as simple as copying the perf binary
- it's self-sufficient
- there's no real advantage of external text files versus internal text files (i.e. text tables within the source code), while there are several disadvantages
So as long as Jiri is happy to maintain these tables I think it's a superior solution (even if the tables start out incomplete), from the perf project's perspective.
Thanks,
Ingo
Re: [PATCH 1/8] perf tools: Add '.' as part of the event 'name' token
Quoted 7 lines · Ingo Molnar
Quoted 4 lines · Stephane Eranian
Quoted 3 lines · Stephane Eranian
Jiri,
I don't see part 0/8 of this series. Did you send it to me too?
oops, it went to lkml only, here it is:
http://marc.info/?l=linux-kernel&m=135923067520675&w=2
Quoted 27 lines
Quoted 11 lines
Quoted 11 lines
I have some comments about it. I don't see why create something from scratch when I have been developing a library (libpfm4) that takes care of that and that is already used by many tool developers. That library can be linked with perf and provide full symbolic events + all the modifiers. The library is portable and supports all existing archs. It can also be used by self-monitoring apps.
You're introducing yet another event table to maintain. And believe me this is a lot of work to maintain this.
Quoted 7 lines
Quoted 2 lines
I don't understand why not use this existing library.
I meant to add that I think it would be more productive if we (you and I) were to work on the library to extend it with external text-based event tables that could be used by perf either directly or thru the libpfm4 interface.
perf is intentionally external file free and does not (fundamentally) depend on external libraries either, other than core system libraries.
There are several advantages to that:
- there is no version skew and no design/maintenance friction
Quoted 8 lines
- 'upgrading' perf between similar boxes is as simple as copying the perf binary
- it's self-sufficient
- there's no real advantage of external text files versus internal text files (i.e. text tables within the source code), while there are several disadvantages
When you say 'internal'.. do you mean like inside binary internal?
Currently this file gets installed with perf and is loaded in run time based on the detected architecture.
Quoted 4 lines
So as long as Jiri is happy to maintain these tables I think it's a superior solution (even if the tables start out incomplete), from the perf project's perspective.
Actually what I did here was that I wrote a tool that sucked and get all the events from libpfm4 in the alias perf format.
It took some libpfm4 changes as well, so I did not post the tool source. But given that libpfm4 is not going in perf, this might be answer.. having tool to keep us in sync with libpfm4 and run it once in a while.. ;-)
jirka
Re: [PATCH 1/8] perf tools: Add '.' as part of the event 'name' token
In reply to Ingo MolnarQuoted 21 lines · Ingo Molnar
Quoted 18 lines · Stephane Eranian
Quoted 16 lines · Stephane Eranian
Jiri,
I don't see part 0/8 of this series. Did you send it to me too?
I have some comments about it. I don't see why create something from scratch when I have been developing a library (libpfm4) that takes care of that and that is already used by many tool developers. That library can be linked with perf and provide full symbolic events + all the modifiers. The library is portable and supports all existing archs. It can also be used by self-monitoring apps.
You're introducing yet another event table to maintain. And believe me this is a lot of work to maintain this.
I don't understand why not use this existing library.
Quoted 9 lines
Quoted 4 lines
I meant to add that I think it would be more productive if we (you and I) were to work on the library to extend it with external text-based event tables that could be used by perf either directly or thru the libpfm4 interface.
perf is intentionally external file free and does not (fundamentally) depend on external libraries either, other than core system libraries.
Perf does use external files, otherwise there would not be a need for what's in libexec.
As far as I know, perf is constantly adding dependencies to other libraries. For instance, nowadays, it does not compile if you don't have libnuma, libdw, and the likes. libpfm4 is similar in nature.
The goal of the library is to provide access to event table in a portable manner for tools and self-monitoring programs (on which they are many). Users don't have to have to change event names, event modifier syntax when they go from one tool to another. Or even when they switch from one OS to another. That's what libpfm4 provides and the scope goes beyond the perf tool.
Quoted 4 lines
There are several advantages to that:
- there is no version skew and no design/maintenance friction
There is rarely an version problem with event tables.
Quoted 3 lines
- 'upgrading' perf between similar boxes is as simple as copying the perf binary
Not quite.
Quoted 6 lines
- it's self-sufficient
- there's no real advantage of external text files versus internal text files (i.e. text tables within the source code), while there are several disadvantages
libpfm4 currently has the tables internally hardcoded.
Quoted 4 lines
So as long as Jiri is happy to maintain these tables I think it's a superior solution (even if the tables start out incomplete), from the perf project's perspective.
Adding the option for users to link with libpfm4 was no big deal in my mind. You are opposed to that. That's your choice. I don't buy your arguments for this patch series and will continue to provide a patch to link perf and libpfm4 for users who are interested.
I would have appreciated some discussion on this instead, once again, work was done behind closed door and thrown at my face. The worst being that as far I can see in the patches, no proper credits to libpfm4 contributors is even given for providing the raw information that allowed you to build your event tables without too much sweat! Building electronic event tables from vendor specs is a lot harder than it seems, but you may not know this.
Re: [PATCH 1/8] perf tools: Add '.' as part of the event 'name' token
Quoted 5 lines · Stephane Eranian
Quoted 4 lines · Ingo Molnar
Quoted 1 line · Stephane Eranian
On Mon, Jan 28, 2013 at 9:52 PM, Stephane Eranian <[email protected]> wrote:
SNIP
Quoted 8 lines
Adding the option for users to link with libpfm4 was no big deal in my mind. You are opposed to that. That's your choice. I don't buy your arguments for this patch series and will continue to provide a patch to link perf and libpfm4 for users who are interested.
I would have appreciated some discussion on this instead, once again, work was done behind closed door and thrown at my face.
Not sure why you think it was 'closed door'.. I needed some way to access those events and sent RFC for it with you being on CC list.
Quoted 3 lines
The worst being that as far I can see in the patches, no proper credits to libpfm4 contributors is even given for providing the raw information that allowed you to build your event tables without too
Right, sorry for that.. I'll add it.
Quoted 2 lines
much sweat! Building electronic event tables from vendor specs is a lot harder than it seems, but you may not know this.
jirka
Peter Zijlstra
maintainerRe: [PATCH 1/8] perf tools: Add '.' as part of the event 'name' token
In reply to Stephane EranianQuoted 3 lines
I meant to add that I think it would be more productive if we (you and I) were to work on the library to extend it with external text-based event tables that could be used by perf either directly or thru the libpfm4 interface.
That would be very useful indeed.. I think we talked about this before. Also wasn't there some vendor support for this idea as well?
I think we want a format that is relatively simple to parse and human readable (which excludes XML on both counts ;-).
The main objection of using libpfm4 has always been that its primarily something else. It also does the event listing thing, but that's only a small part of it.
But if we can split out that part and use it in both projects that would be best I think.
Re: [PATCH 1/8] perf tools: Add '.' as part of the event 'name' token
Quoted 10 lines · Peter Zijlstra
Quoted 3 lines
I meant to add that I think it would be more productive if we (you and I) were to work on the library to extend it with external text-based event tables that could be used by perf either directly or thru the libpfm4 interface.
That would be very useful indeed.. I think we talked about this before. Also wasn't there some vendor support for this idea as well?
I think we want a format that is relatively simple to parse and human readable (which excludes XML on both counts ;-).
Quoted 6 lines
The main objection of using libpfm4 has always been that its primarily something else. It also does the event listing thing, but that's only a small part of it.
But if we can split out that part and use it in both projects that would be best I think.
If that libpfm reference library moved to tools/lib/ like tools/lib/traceevents/ then that would be a nice, useful solution indeed.
Thanks,
Ingo
PATCH v2 2/8 perf tools: Change perf_pmu__new_alias function interface
In reply to Jiri OlsaChanging perf_pmu__new_alias interface not to work directly with FILE object, so it can be reused by other code paths coming in shortly.
--- tools/perf/util/pmu.c | 23 +++++++++++++---------- 1 file changed, 13 insertions(+), 10 deletions(-)
tools/perf/util/pmu.c +13 -10
@@ -95,23 +95,17 @@ static int pmu_format(char *name, struct list_head *format)return 0 ; } -static int perf_pmu__new_alias (struct list_head *list,char *name,FILE *file) +static int perf_pmu__new_alias (struct list_head *list,char *name,char *data) {struct perf_pmu_alias *alias; -char buf[256 ];int ret; - ret =fread (buf,1 ,sizeof (buf), file); -if (ret ==0 ) -return -EINVAL ; - buf[ret] =0 ; - alias =malloc (sizeof (*alias));if (!alias)return -ENOMEM ;INIT_LIST_HEAD (&alias->terms); - ret =parse_events_terms (&alias->terms, buf); + ret =parse_events_terms (&alias->terms, data);if (ret) {free (alias);return ret;
@@ -137,7 +131,7 @@ static int pmu_aliases_parse(char *dir, struct list_head *head)return -EINVAL ;while (!ret && (evt_ent =readdir (event_dir))) { -char path[PATH_MAX ]; +char path[PATH_MAX ], buf[256 ];char *name = evt_ent->d_name;FILE *file;
@@ -150,7 +144,16 @@ static int pmu_aliases_parse(char *dir, struct list_head *head) file =fopen (path,"r" );if (!file)break ; - ret =perf_pmu__new_alias (head, name, file); + + ret =fread (buf,1 ,sizeof (buf), file); +if (ret ==0 ) { + ret = -EINVAL ; +break ; + } + + buf[ret] =0 ; + + ret =perf_pmu__new_alias (head, name, buf);fclose (file); }
Signature · 1 line
1.7.11.7
PATCH v2 3/8 perf tools: Add name term processing for alias
In reply to Jiri OlsaAdding support for name term being specified within the alias definition and gives the name for the alias.
Alias could be now defined like:
name=BR_MISP_EXEC.ALL_BRANCHES,event=0x89,umask=0xff
It'll be handy for having single file with multiple alias definitions.
--- tools/perf/util/parse-events.c | 19 ++----------------- tools/perf/util/pmu.c | 35 +++++++++++++++++++++++++++++++++++ tools/perf/util/pmu.h | 3 +++ 3 files changed, 40 insertions(+), 17 deletions(-)
tools/perf/util/parse-events.c +2 -17
@@ -616,22 +616,6 @@ int parse_events_add_numeric(struct list_head **list, int *idx,return add_event (list, idx, &attr,NULL ); } -static int parse_events__is_name_term (struct parse_events_term *term) -{ -return term->type_term ==PARSE_EVENTS__TERM_TYPE_NAME ; -} - -static char *pmu_event_name (struct list_head *head_terms) -{ -struct parse_events_term *term; - -list_for_each_entry (term, head_terms, list) -if (parse_events__is_name_term (term)) -return term->val.str; - -return NULL ; -} -int parse_events_add_pmu (struct list_head **list,int *idx,char *name,struct list_head *head_config) {
@@ -656,7 +640,8 @@ int parse_events_add_pmu(struct list_head **list, int *idx,if (perf_pmu__config (pmu, &attr, head_config))return -EINVAL ; -return __add_event (list, idx, &attr,pmu_event_name (head_config), +return __add_event (list, idx, &attr, +perf_pmu__event_name (head_config, false), pmu->cpus); }
tools/perf/util/pmu.c +35 -0
@@ -95,6 +95,27 @@ static int pmu_format(char *name, struct list_head *format)return 0 ; } +static int parse_events__is_name_term (struct parse_events_term *term) +{ +return term->type_term ==PARSE_EVENTS__TERM_TYPE_NAME ; +} + +char *perf_pmu__event_name (struct list_head *head_terms,bool remove) +{ +struct parse_events_term *term; + +list_for_each_entry (term, head_terms, list) +if (parse_events__is_name_term (term)) { +if (remove) { +list_del (&term->list); +free (term); + } +return term->val.str; + } + +return NULL ; +} +static int perf_pmu__new_alias (struct list_head *list,char *name,char *data) {struct perf_pmu_alias *alias;
@@ -111,6 +132,20 @@ static int perf_pmu__new_alias(struct list_head *list, char *name, char *data)return ret; } +/* +* Use NAME term to get alias name. In case there's no name +* at all, bail out. In case we find NAME term, remove it +* not to mangle with event term name. +*/ +if (!name) { + name =perf_pmu__event_name (&alias->terms, true); + +if (!name) { +free (alias); +return -EINVAL ; + } + } + alias->name =strdup (name);list_add_tail (&alias->list, list);return 0 ;
tools/perf/util/pmu.h +3 -0
@@ -3,6 +3,7 @@#include <linux/bitops.h> #include <linux/perf_event.h> +#include <stdbool.h> enum {PERF_PMU_FORMAT_VALUE_CONFIG ,
@@ -40,5 +41,7 @@ int perf_pmu__format_parse(char *dir, struct list_head *head);struct perf_pmu *perf_pmu__scan (struct perf_pmu *pmu); +char *perf_pmu__event_name (struct list_head *head_terms,bool remove); +int perf_pmu__test (void );#endif /* __PMU_H */
Signature · 1 line
1.7.11.7
PATCH v2 4/8 perf tools: Add pmu interface to parse single file of aliases
In reply to Jiri OlsaAdding pmu_aliases_parse_multi function to parse and add aliases from single file. The file format follows the alias format, each line for single alias. Each line must contains 'name' term, like:
name=BR_MISP_EXEC.ALL_BRANCHES,event=0x89,umask=0xff name=BR_MISP_EXEC.COND,event=0x89,umask=0x1
--- tools/perf/util/pmu.c | 23 +++++++++++++++++++++++ tools/perf/util/pmu.h | 2 ++ 2 files changed, 25 insertions(+)
tools/perf/util/pmu.c +23 -0
@@ -146,11 +146,34 @@ static int perf_pmu__new_alias(struct list_head *list, char *name, char *data) } } +pr_debug3 ("adding alias '%s' => '%s'\n" , name, data); + alias->name =strdup (name);list_add_tail (&alias->list, list);return 0 ; } +int pmu_aliases_parse_multi (char *path,struct list_head *head) +{ +FILE *file; +char *data =NULL ; +size_t len; +int ret = -EINVAL ; + + file =fopen (path,"r" ); +if (!file) +return -EINVAL ; + +while ((getline (&data, &len, file)) !=-1 ) { + ret =perf_pmu__new_alias (head,NULL , data); +free (data); + data =NULL ; + } + +fclose (file); +return ret; +} +/* * Process all the sysfs attributes located under the directory * specified in 'dir' parameter.
tools/perf/util/pmu.h +2 -0
@@ -43,5 +43,7 @@ struct perf_pmu *perf_pmu__scan(struct perf_pmu *pmu);char *perf_pmu__event_name (struct list_head *head_terms,bool remove); +int pmu_aliases_parse_multi (char *path,struct list_head *head); +int perf_pmu__test (void );#endif /* __PMU_H */
Signature · 1 line
1.7.11.7
PATCH v2 5/8 perf tools: Add support to include non architectural event aliases
In reply to Jiri OlsaAdding support to parse non architectural event aliases for given cpu. These aliases will be provided as single files parsed by pmu_aliases_parse_multi function.
--- tools/perf/arch/x86/Makefile | 7 +++ tools/perf/arch/x86/util/pmu.c | 121 +++++++++++++++++++++++++++++++++++++++++ tools/perf/util/pmu.c | 16 ++++-- tools/perf/util/pmu.h | 2 + 4 files changed, 141 insertions(+), 5 deletions(-) create mode 100644 tools/perf/arch/x86/util/pmu.c
tools/perf/arch/x86/Makefile +7 -0
@@ -6,3 +6,10 @@ ifndef NO_LIBUNWIND LIB_OBJS += $(OUTPUT)arch/$(ARCH)/util/unwind.o endif LIB_OBJS += $(OUTPUT)arch/$(ARCH)/util/header.o +LIB_OBJS += $(OUTPUT)arch/$(ARCH)/util/pmu.o + +$(OUTPUT)$(OUTPUT)arch/$(ARCH)/util/pmu.o: $(OUTPUT)arch/$(ARCH)/util/pmu.c $(OUTPUT)PERF-CFLAGS + $(QUIET_CC)$(CC) -o $@ -c $(ALL_CFLAGS) \ + '-DPERF_EXEC_PATH="$(perfexecdir_SQ)"' \ + '-DPREFIX="$(prefix_SQ)"' \ + $<
tools/perf/arch/x86/util/pmu.c new file +121 -0
121 lines not on this page · raw message
tools/perf/util/pmu.c +11 -5
@@ -219,6 +219,13 @@ static int pmu_aliases_parse(char *dir, struct list_head *head)return ret; } +__attribute__((weak)) +int arch_pmu_aliases (char *name__maybe_unused , +struct list_head *head__maybe_unused ) +{ +return 0 ; +} +/* * Reading the pmu event aliases definition, which should be located at: * /sys/bus/event_source/devices/<dev>/events as sysfs group attributes.
@@ -236,13 +243,12 @@ static int pmu_aliases(char *name, struct list_head *head)snprintf (path,PATH_MAX ,"%s/bus/event_source/devices/%s/events" , sysfs, name); -if (stat (path, &st) <0 ) -return 0 ;/* no error if 'events' does not exist */ - -if (pmu_aliases_parse (path, head)) +/* no error if 'events' does not exist */ +if (!stat (path, &st) && +pmu_aliases_parse (path, head))return -1 ; -return 0 ; +return arch_pmu_aliases (name, head); } static int pmu_alias_terms(struct perf_pmu_alias *alias,
tools/perf/util/pmu.h +2 -0
@@ -45,5 +45,7 @@ char *perf_pmu__event_name(struct list_head *head_terms, bool remove);int pmu_aliases_parse_multi (char *path,struct list_head *head); +int arch_pmu_aliases (char *name,struct list_head *head); +int perf_pmu__test (void );#endif /* __PMU_H */
Signature · 1 line
1.7.11.7
PATCH v2 6/8 perf tools: Add non arch events for SandyBridge microarchitecture
In reply to Jiri OlsaAdding non architectural event aliases for SandyBridge micro architecture.
--- tools/perf/arch/x86/Makefile | 4 + tools/perf/arch/x86/events/intel/snb | 274 +++++++++++++++++++++++++++++++++++ tools/perf/arch/x86/util/pmu.c | 4 +- 3 files changed, 281 insertions(+), 1 deletion(-) create mode 100644 tools/perf/arch/x86/events/intel/snb
tools/perf/arch/x86/Makefile +4 -0
@@ -13,3 +13,7 @@ $(OUTPUT)$(OUTPUT)arch/$(ARCH)/util/pmu.o: $(OUTPUT)arch/$(ARCH)/util/pmu.c $(OU '-DPERF_EXEC_PATH="$(perfexecdir_SQ)"' \ '-DPREFIX="$(prefix_SQ)"' \ $< + +install: + $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/events/x86/intel/' + $(INSTALL) -m 644 arch/$(ARCH)/events/intel/* $(DESTDIR_SQ)$(perfexec_instdir_SQ)/events/x86/intel/
tools/perf/arch/x86/events/intel/snb new file +274 -0
274 lines not on this page · raw message
tools/perf/arch/x86/util/pmu.c +3 -1
@@ -56,8 +56,10 @@ static int intel_aliases(struct list_head *head, unsigned model)return -1 ;switch (model) { +case 42 :/* Sandy Bridge (Core i7 26xx, 25xx) */ +case 45 :/* Sandy Bridge EP */ +ADD_ALIASES ("snb" );default : -ADD_ALIASES ("" );break ; }
Signature · 1 line
1.7.11.7
PATCH v2 7/8 perf tools: Add non arch events for IvyBridge micro architecture
In reply to Jiri OlsaAdding non architectural event aliases for IvyBridge micro architecture.
--- tools/perf/arch/x86/events/intel/ivb | 248 +++++++++++++++++++++++++++++++++++ tools/perf/arch/x86/util/pmu.c | 2 + 2 files changed, 250 insertions(+) create mode 100644 tools/perf/arch/x86/events/intel/ivb
tools/perf/arch/x86/events/intel/ivb new file +248 -0
@@ -0,0 +1,248 @@ +name=ARITH.FPU_DIV_ACTIVE,event=0x14,umask=0x1 +name=ARITH.FPU_DIV,event=0x14,umask=0x10401 +name=BACLEARS.ANY,event=0xe6,umask=0x1f +name=BR_INST_EXEC.NONTAKEN_COND,event=0x88,umask=0x41 +name=BR_INST_EXEC.TAKEN_COND,event=0x88,umask=0x81 +name=BR_INST_EXEC.NONTAKEN_DIRECT_JUMP,event=0x88,umask=0x42 +name=BR_INST_EXEC.TAKEN_DIRECT_JUMP,event=0x88,umask=0x82 +name=BR_INST_EXEC.NONTAKEN_INDIRECT_JUMP_NON_CALL_RET,event=0x88,umask=0x44 +name=BR_INST_EXEC.TAKEN_INDIRECT_JUMP_NON_CALL_RET,event=0x88,umask=0x84 +name=BR_INST_EXEC.TAKEN_NEAR_RETURN,event=0x88,umask=0x88 +name=BR_INST_EXEC.TAKEN_DIRECT_NEAR_CALL,event=0x88,umask=0x90 +name=BR_INST_EXEC.TAKEN_INDIRECT_NEAR_CALL,event=0x88,umask=0xa0 +name=BR_INST_EXEC.ALL_BRANCHES,event=0x88,umask=0xff +name=BR_INST_EXEC.ALL_COND,event=0x88,umask=0xc1 +name=BR_INST_EXEC.ANY_COND,event=0x88,umask=0xc1 +name=BR_INST_EXEC.ANY_INDIRECT_JUMP_NON_CALL_RET,event=0x88,umask=0xc4 +name=BR_INST_EXEC.ANY_DIRECT_NEAR_CALL,event=0x88,umask=0xd0 +name=BR_INST_RETIRED.ALL_BRANCHES,event=0xc4,umask=0x4,precise +name=BR_INST_RETIRED.COND,event=0xc4,umask=0x1,precise +name=BR_INST_RETIRED.FAR_BRANCH,event=0xc4,umask=0x40,precise +name=BR_INST_RETIRED.NEAR_CALL,event=0xc4,umask=0x2,precise +name=BR_INST_RETIRED.NEAR_RETURN,event=0xc4,umask=0x8,precise +name=BR_INST_RETIRED.NEAR_TAKEN,event=0xc4,umask=0x20,precise +name=BR_INST_RETIRED.NOT_TAKEN,event=0xc4,umask=0x10,precise +name=BR_MISP_EXEC.NONTAKEN_COND,event=0x89,umask=0x41 +name=BR_MISP_EXEC.TAKEN_COND,event=0x89,umask=0x81 +name=BR_MISP_EXEC.NONTAKEN_INDIRECT_JUMP_NON_CALL_RET,event=0x89,umask=0x44 +name=BR_MISP_EXEC.TAKEN_INDIRECT_JUMP_NON_CALL_RET,event=0x89,umask=0x84 +name=BR_MISP_EXEC.NONTAKEN_NEAR_RETURN,event=0x89,umask=0x48 +name=BR_MISP_EXEC.TAKEN_NEAR_RETURN,event=0x89,umask=0x88 +name=BR_MISP_EXEC.NONTAKEN_DIRECT_NEAR_CALL,event=0x89,umask=0x50 +name=BR_MISP_EXEC.TAKEN_DIRECT_NEAR_CALL,event=0x89,umask=0x90 +name=BR_MISP_EXEC.NONTAKEN_INDIRECT_NEAR_CALL,event=0x89,umask=0x60 +name=BR_MISP_EXEC.TAKEN_INDIRECT_NEAR_CALL,event=0x89,umask=0xa0 +name=BR_MISP_EXEC.ANY_COND,event=0x89,umask=0xc1 +name=BR_MISP_EXEC.ANY_RETURN_NEAR,event=0x89,umask=0xc8 +name=BR_MISP_EXEC.ANY_DIRECT_NEAR_CALL,event=0x89,umask=0xd0 +name=BR_MISP_EXEC.ANY_INDIRECT_JUMP_NON_CALL_RET,event=0x89,umask=0xc4 +name=BR_MISP_EXEC.ALL_BRANCHES,event=0x89,umask=0xff +name=BR_MISP_RETIRED.ALL_BRANCHES,event=0xc5,umask=0x4,precise +name=BR_MISP_RETIRED.COND,event=0xc5,umask=0x1,precise +name=BR_MISP_RETIRED.NEAR_CALL,event=0xc5,umask=0x2,precise +name=BR_MISP_RETIRED.NOT_TAKEN,event=0xc5,umask=0x10,precise +name=BR_MISP_RETIRED.TAKEN,event=0xc5,umask=0x20,precise +name=LOCK_CYCLES.SPLIT_LOCK_UC_LOCK_DURATION,event=0x63,umask=0x1 +name=LOCK_CYCLES.CACHE_LOCK_DURATION,event=0x63,umask=0x2 +name=CPL_CYCLES.RING0,event=0x5c,umask=0x1 +name=CPL_CYCLES.RING0_TRANS,event=0x5c,umask=0x10401 +name=CPL_CYCLES.RING123,event=0x5c,umask=0x2 +name=CPU_CLK_UNHALTED.REF_P,event=0x3c,umask=0x1 +name=CPU_CLK_UNHALTED.REF_XCLK,event=0x3c,umask=0x1 +name=CPU_CLK_UNHALTED.THREAD_P,event=0x3c,umask=0x0 +name=DSB2MITE_SWITCHES.COUNT,event=0xab,umask=0x1 +name=DSB2MITE_SWITCHES.PENALTY_CYCLES,event=0xab,umask=0x2 +name=DSB_FILL.EXCEED_DSB_LINES,event=0xac,umask=0x8 +name=DTLB_LOAD_MISSES.DEMAND_LD_MISS_CAUSES_A_WALK,event=0x8,umask=0x81 +name=DTLB_LOAD_MISSES.DEMAND_LD_WALK_COMPLETED,event=0x8,umask=0x82 +name=DTLB_LOAD_MISSES.DEMAND_LD_WALK_DURATION,event=0x8,umask=0x84 +name=DTLB_STORE_MISSES.MISS_CAUSES_A_WALK,event=0x49,umask=0x1 +name=DTLB_STORE_MISSES.CAUSES_A_WALK,event=0x49,umask=0x1 +name=DTLB_STORE_MISSES.STLB_HIT,event=0x49,umask=0x10 +name=DTLB_STORE_MISSES.WALK_COMPLETED,event=0x49,umask=0x2 +name=DTLB_STORE_MISSES.WALK_DURATION,event=0x49,umask=0x4 +name=FP_ASSIST.ANY,event=0xca,umask=0x1e +name=FP_ASSIST.SIMD_INPUT,event=0xca,umask=0x10 +name=FP_ASSIST.SIMD_OUTPUT,event=0xca,umask=0x8 +name=FP_ASSIST.X87_INPUT,event=0xca,umask=0x4 +name=FP_ASSIST.X87_OUTPUT,event=0xca,umask=0x2 +name=HW_INTERRUPTS.RECEIVED,event=0xcb,umask=0x1 +name=ICACHE.MISSES,event=0x80,umask=0x2 +name=IDQ.EMPTY,event=0x79,umask=0x2 +name=IDQ.MITE_UOPS,event=0x79,umask=0x4 +name=IDQ.DSB_UOPS,event=0x79,umask=0x8 +name=IDQ.MS_DSB_UOPS,event=0x79,umask=0x10 +name=IDQ.MS_MITE_UOPS,event=0x79,umask=0x20 +name=IDQ.MS_UOPS,event=0x79,umask=0x30 +name=IDQ.MITE_UOPS_CYCLES,event=0x79,umask=0x10004 +name=IDQ.DSB_UOPS_CYCLES,event=0x79,umask=0x10008 +name=IDQ.MS_DSB_UOPS_CYCLES,event=0x79,umask=0x10010 +name=IDQ.MS_MITE_UOPS_CYCLES,event=0x79,umask=0x10020 +name=IDQ.MS_UOPS_CYCLES,event=0x79,umask=0x10030 +name=IDQ.ALL_DSB_UOPS,event=0x79,umask=0x18 +name=IDQ.ALL_DSB_CYCLES,event=0x79,umask=0x10018 +name=IDQ.ALL_DSB_CYCLES_4_UOPS,event=0x79,umask=0x40018 +name=IDQ.ALL_MITE_UOPS,event=0x79,umask=0x24 +name=IDQ.ALL_MITE_CYCLES,event=0x79,umask=0x10024 +name=IDQ.ANY_UOPS,event=0x79,umask=0x3c +name=IDQ.MS_DSB_UOPS_OCCUR,event=0x79,umask=0x10410 +name=IDQ_UOPS_NOT_DELIVERED.CORE,event=0x9c,umask=0x1 +name=ILD_STALL.LCP,event=0x87,umask=0x1 +name=ILD_STALL.IQ_FULL,event=0x87,umask=0x4 +name=INST_RETIRED.ANY_P,event=0xc0,umask=0x0 +name=INST_RETIRED.ALL,event=0xc0,umask=0x1,precise +name=INST_RETIRED.PREC_DIST,event=0xc0,umask=0x1,precise +name=ITLB.ITLB_FLUSH,event=0xae,umask=0x1 +name=ITLB.FLUSH,event=0xae,umask=0x1 +name=ITLB_MISSES.MISS_CAUSES_A_WALK,event=0x85,umask=0x1 +name=ITLB_MISSES.CAUSES_A_WALK,event=0x85,umask=0x1 +name=ITLB_MISSES.STLB_HIT,event=0x85,umask=0x10 +name=ITLB_MISSES.WALK_COMPLETED,event=0x85,umask=0x2 +name=ITLB_MISSES.WALK_DURATION,event=0x85,umask=0x4 +name=L1D.REPLACEMENT,event=0x51,umask=0x1 +name=MOVE_ELIMINATION.INT_NOT_ELIMINATED,event=0x58,umask=0x1 +name=MOVE_ELIMINATION.SIMD_NOT_ELIMINATED,event=0x58,umask=0x2 +name=MOVE_ELIMINATION.INT_ELIMINATED,event=0x58,umask=0x4 +name=MOVE_ELIMINATION.SIMD_ELIMINATED,event=0x58,umask=0x2 +name=L1D_PEND_MISS.OCCURRENCES,event=0x48,umask=0x10401 +name=L1D_PEND_MISS.EDGE,event=0x48,umask=0x10401 +name=L1D_PEND_MISS.PENDING,event=0x48,umask=0x1 +name=L1D_PEND_MISS.PENDING_CYCLES,event=0x48,umask=0x10001 +name=L2_L1D_WB_RQSTS.HIT_E,event=0x28,umask=0x4 +name=L2_L1D_WB_RQSTS.HIT_M,event=0x28,umask=0x8 +name=L2_L1D_WB_RQSTS.MISS,event=0x28,umask=0x1 +name=L2_L1D_WB_RQSTS.ALL,event=0x28,umask=0xf +name=L2_LINES_IN.ANY,event=0xf1,umask=0x7 +name=L2_LINES_IN.ALL,event=0xf1,umask=0x7 +name=L2_LINES_IN.E,event=0xf1,umask=0x4 +name=L2_LINES_IN.I,event=0xf1,umask=0x1 +name=L2_LINES_IN.S,event=0xf1,umask=0x2 +name=L2_LINES_OUT.DEMAND_CLEAN,event=0xf2,umask=0x1 +name=L2_LINES_OUT.DEMAND_DIRTY,event=0xf2,umask=0x2 +name=L2_LINES_OUT.PREFETCH_CLEAN,event=0xf2,umask=0x4 +name=L2_LINES_OUT.PF_CLEAN,event=0xf2,umask=0x4 +name=L2_LINES_OUT.PREFETCH_DIRTY,event=0xf2,umask=0x8 +name=L2_LINES_OUT.PF_DIRTY,event=0xf2,umask=0x8 +name=L2_LINES_OUT.DIRTY_ANY,event=0xf2,umask=0xa +name=L2_LINES_OUT.DIRTY_ALL,event=0xf2,umask=0xa +name=L2_RQSTS.ALL_CODE_RD,event=0x24,umask=0x30 +name=L2_RQSTS.CODE_RD_HIT,event=0x24,umask=0x10 +name=L2_RQSTS.CODE_RD_MISS,event=0x24,umask=0x20 +name=L2_RQSTS.ALL_DEMAND_DATA_RD,event=0x24,umask=0x3 +name=L2_RQSTS.DEMAND_DATA_RD_HIT,event=0x24,umask=0x1 +name=L2_RQSTS.ALL_PF,event=0x24,umask=0xc0 +name=L2_RQSTS.PF_HIT,event=0x24,umask=0x40 +name=L2_RQSTS.PF_MISS,event=0x24,umask=0x80 +name=L2_RQSTS.ALL_RFO,event=0x24,umask=0xc +name=L2_RQSTS.RFO_HIT,event=0x24,umask=0x4 +name=L2_RQSTS.RFO_MISS,event=0x24,umask=0x8 +name=L2_STORE_LOCK_RQSTS.MISS,event=0x27,umask=0x1 +name=L2_STORE_LOCK_RQSTS.HIT_M,event=0x27,umask=0x8 +name=L2_STORE_LOCK_RQSTS.ALL,event=0x27,umask=0xf +name=L2_TRANS.ALL,event=0xf0,umask=0x80 +name=L2_TRANS.CODE_RD,event=0xf0,umask=0x4 +name=L2_TRANS.L1D_WB,event=0xf0,umask=0x10 +name=L2_TRANS.DMND_DATA_RD,event=0xf0,umask=0x1 +name=L2_TRANS.L2_FILL,event=0xf0,umask=0x20 +name=L2_TRANS.L2_WB,event=0xf0,umask=0x40 +name=L2_TRANS.ALL_PREFETCH,event=0xf0,umask=0x8 +name=L2_TRANS.ALL_PF,event=0xf0,umask=0x8 +name=L2_TRANS.RFO,event=0xf0,umask=0x2 +name=LD_BLOCKS.STORE_FORWARD,event=0x3,umask=0x2 +name=LD_BLOCKS_PARTIAL.ADDRESS_ALIAS,event=0x7,umask=0x1 +name=LOAD_HIT_PRE.HW_PF,event=0x4c,umask=0x2 +name=LOAD_HIT_PRE.SW_PF,event=0x4c,umask=0x1 +name=L3_LAT_CACHE.MISS,event=0x2e,umask=0x41 +name=L3_LAT_CACHE.REFERENCE,event=0x2e,umask=0x4f +name=MACHINE_CLEARS.MASKMOV,event=0xc3,umask=0x20 +name=MACHINE_CLEARS.MEMORY_ORDERING,event=0xc3,umask=0x2 +name=MACHINE_CLEARS.SMC,event=0xc3,umask=0x4 +name=MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HIT,event=0xd2,umask=0x2,precise +name=MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HITM,event=0xd2,umask=0x4,precise +name=MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_MISS,event=0xd2,umask=0x1,precise +name=MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_NONE,event=0xd2,umask=0x8,precise +name=MEM_LOAD_LLC_HIT_RETIRED.XSNP_HIT,event=0xd2,umask=0x2,precise +name=MEM_LOAD_LLC_HIT_RETIRED.XSNP_HITM,event=0xd2,umask=0x4,precise +name=MEM_LOAD_LLC_HIT_RETIRED.XSNP_MISS,event=0xd2,umask=0x1,precise +name=MEM_LOAD_LLC_HIT_RETIRED.XSNP_NONE,event=0xd2,umask=0x8,precise +name=MEM_LOAD_UOPS_LLC_MISS_RETIRED.LOCAL_DRAM,event=0xd3,umask=0x1,precise +name=MEM_LOAD_UOPS_RETIRED.HIT_LFB,event=0xd1,umask=0x40,precise +name=MEM_LOAD_UOPS_RETIRED.L1_HIT,event=0xd1,umask=0x1,precise +name=MEM_LOAD_UOPS_RETIRED.L2_HIT,event=0xd1,umask=0x2,precise +name=MEM_LOAD_UOPS_RETIRED.L3_HIT,event=0xd1,umask=0x4,precise +name=MEM_LOAD_RETIRED.HIT_LFB,event=0xd1,umask=0x40,precise +name=MEM_LOAD_RETIRED.L1_HIT,event=0xd1,umask=0x1,precise +name=MEM_LOAD_RETIRED.L2_HIT,event=0xd1,umask=0x2,precise +name=MEM_LOAD_RETIRED.L3_HIT,event=0xd1,umask=0x4,precise +name=MEM_TRANS_RETIRED.LATENCY_ABOVE_THRESHOLD,event=0xcd,umask=0x1,precise +name=MEM_TRANS_RETIRED.PRECISE_STORE,event=0xcd,umask=0x2,precise +name=MEM_UOPS_RETIRED.ALL_LOADS,event=0xd0,umask=0x81,precise +name=MEM_UOPS_RETIRED.ANY_LOADS,event=0xd0,umask=0x81,precise +name=MEM_UOPS_RETIRED.ALL_STORES,event=0xd0,umask=0x82,precise +name=MEM_UOPS_RETIRED.ANY_STORES,event=0xd0,umask=0x82,precise +name=MEM_UOPS_RETIRED.LOCK_LOADS,event=0xd0,umask=0x21,precise +name=MEM_UOPS_RETIRED.LOCK_STORES,event=0xd0,umask=0x22,precise +name=MEM_UOPS_RETIRED.SPLIT_LOADS,event=0xd0,umask=0x41,precise +name=MEM_UOPS_RETIRED.SPLIT_STORES,event=0xd0,umask=0x42,precise +name=MEM_UOPS_RETIRED.STLB_MISS_LOADS,event=0xd0,umask=0x11,precise +name=MEM_UOPS_RETIRED.STLB_MISS_STORES,event=0xd0,umask=0x12,precise +name=MEM_UOP_RETIRED.ALL_LOADS,event=0xd0,umask=0x81,precise +name=MEM_UOP_RETIRED.ANY_LOADS,event=0xd0,umask=0x81,precise +name=MEM_UOP_RETIRED.ALL_STORES,event=0xd0,umask=0x82,precise +name=MEM_UOP_RETIRED.ANY_STORES,event=0xd0,umask=0x82,precise +name=MEM_UOP_RETIRED.LOCK_LOADS,event=0xd0,umask=0x21,precise +name=MEM_UOP_RETIRED.LOCK_STORES,event=0xd0,umask=0x22,precise +name=MEM_UOP_RETIRED.SPLIT_LOADS,event=0xd0,umask=0x41,precise +name=MEM_UOP_RETIRED.SPLIT_STORES,event=0xd0,umask=0x42,precise +name=MEM_UOP_RETIRED.STLB_MISS_LOADS,event=0xd0,umask=0x11,precise +name=MEM_UOP_RETIRED.STLB_MISS_STORES,event=0xd0,umask=0x12,precise +name=MISALIGN_MEM_REF.LOADS,event=0x5,umask=0x1 +name=MISALIGN_MEM_REF.STORES,event=0x5,umask=0x2 +name=OFFCORE_REQUESTS.ALL_DATA_RD,event=0xb0,umask=0x8 +name=OFFCORE_REQUESTS.ALL_DATA_READ,event=0xb0,umask=0x8 +name=OFFCORE_REQUESTS.DEMAND_CODE_RD,event=0xb0,umask=0x2 +name=OFFCORE_REQUESTS.DEMAND_DATA_RD,event=0xb0,umask=0x1 +name=OFFCORE_REQUESTS.DEMAND_RFO,event=0xb0,umask=0x4 +name=OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD_CYCLES,event=0x60,umask=0x10008 +name=OFFCORE_REQUESTS_OUTSTANDING.DEMAND_CODE_RD_CYCLES,event=0x60,umask=0x10002 +name=OFFCORE_REQUESTS_OUTSTANDING.DEMAND_DATA_RD_CYCLES,event=0x60,umask=0x10001 +name=OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD,event=0x60,umask=0x8 +name=OFFCORE_REQUESTS_OUTSTANDING.DEMAND_CODE_RD,event=0x60,umask=0x2 +name=OFFCORE_REQUESTS_OUTSTANDING.DEMAND_DATA_RD,event=0x60,umask=0x1 +name=OFFCORE_REQUESTS_OUTSTANDING.DEMAND_RFO,event=0x60,umask=0x4 +name=OFFCORE_REQUESTS_OUTSTANDING.DEMAND_RFO_CYCLES,event=0x60,umask=0x10004 +name=OTHER_ASSISTS.ITLB_MISS_RETIRED,event=0xc1,umask=0x2 +name=OTHER_ASSISTS.AVX_TO_SSE,event=0xc1,umask=0x10 +name=OTHER_ASSISTS.SSE_TO_AVX,event=0xc1,umask=0x20 +name=OTHER_ASSISTS.AVX_STORE,event=0xc1,umask=0x8 +name=RESOURCE_STALLS.ANY,event=0xa2,umask=0x1 +name=RESOURCE_STALLS.RS,event=0xa2,umask=0x4 +name=RESOURCE_STALLS.SB,event=0xa2,umask=0x8 +name=RESOURCE_STALLS.ROB,event=0xa2,umask=0x10 +name=ROB_MISC_EVENTS.LBR_INSERTS,event=0xcc,umask=0x20 +name=RS_EVENTS.EMPTY_CYCLES,event=0x5e,umask=0x1 +name=TLB_ACCESS.LOAD_STLB_HIT,event=0x5f,umask=0x1 +name=TLB_FLUSH.DTLB_THREAD,event=0xbd,umask=0x1 +name=TLB_FLUSH.STLB_ANY,event=0xbd,umask=0x20 +name=UOPS_EXECUTED.CORE,event=0xb1,umask=0x2 +name=UOPS_DISPATCHED_PORT.PORT_0,event=0xa1,umask=0x1 +name=UOPS_DISPATCHED_PORT.PORT_1,event=0xa1,umask=0x2 +name=UOPS_DISPATCHED_PORT.PORT_2_LD,event=0xa1,umask=0x4 +name=UOPS_DISPATCHED_PORT.PORT_2_STA,event=0xa1,umask=0x8 +name=UOPS_DISPATCHED_PORT.PORT_2,event=0xa1,umask=0xc +name=UOPS_DISPATCHED_PORT.PORT_3_LD,event=0xa1,umask=0x10 +name=UOPS_DISPATCHED_PORT.PORT_3_STA,event=0xa1,umask=0x20 +name=UOPS_DISPATCHED_PORT.PORT_3,event=0xa1,umask=0x30 +name=UOPS_DISPATCHED_PORT.PORT_4,event=0xa1,umask=0x40 +name=UOPS_DISPATCHED_PORT.PORT_5,event=0xa1,umask=0x80 +name=UOPS_ISSUED.ANY,event=0xe,umask=0x1 +name=UOPS_ISSUED.CORE_STALL_CYCLES,event=0xe,umask=0x1a001 +name=UOPS_ISSUED.STALL_CYCLES,event=0xe,umask=0x18001 +name=UOPS_ISSUED.FLAGS_MERGE,event=0xe,umask=0x10 +name=UOPS_ISSUED.SLOW_LEA,event=0xe,umask=0x20 +name=UOPS_ISSUED.SINGLE_MUL,event=0xe,umask=0x40 +name=UOPS_RETIRED.ALL,event=0xc2,umask=0x1,precise +name=UOPS_RETIRED.ANY,event=0xc2,umask=0x1,precise +name=UOPS_RETIRED.RETIRE_SLOTS,event=0xc2,umask=0x2,precise +name=UOPS_RETIRED.STALL_CYCLES,event=0xc2,umask=0x18001,precise +name=UOPS_RETIRED.TOTAL_CYCLES,event=0xc2,umask=0x100001,precise
tools/perf/arch/x86/util/pmu.c +2 -0
@@ -59,6 +59,8 @@ static int intel_aliases(struct list_head *head, unsigned model)case 42 :/* Sandy Bridge (Core i7 26xx, 25xx) */ case 45 :/* Sandy Bridge EP */ ADD_ALIASES ("snb" ); +case 58 :/* IvyBridge */ +ADD_ALIASES ("ivb" );default :break ; }
Signature · 1 line
1.7.11.7
PATCH v2 8/8 perf tools: List kernel supplied event aliases in perf list v2
In reply to Jiri OlsaFrom: Andi Kleen <[email protected]>
List the kernel supplied pmu event aliases in perf list
It's better when the users can actually see them.
v2: Fix pattern matching
--- tools/perf/Documentation/perf-list.txt | 4 ++- tools/perf/builtin-list.c | 3 ++ tools/perf/util/parse-events.c | 5 ++- tools/perf/util/pmu.c | 61 ++++++++++++++++++++++++++++++++++ tools/perf/util/pmu.h | 2 ++ 5 files changed, 73 insertions(+), 2 deletions(-)
tools/perf/Documentation/perf-list.txt +3 -1
@@ -8,7 +8,7 @@ perf-list - List all symbolic event types SYNOPSIS -------- [verse] -'perf list' [hw|sw|cache|tracepoint|event_glob] +'perf list' [hw|sw|cache|tracepoint|pmu|event_glob] DESCRIPTION -----------
@@ -104,6 +104,8 @@ To limit the list use: 'subsys_glob:event_glob' to filter by tracepoint subsystems such as sched, block, etc. +. 'pmu' to print the kernel supplied PMU events. + . If none of the above is matched, it will apply the supplied glob to all events, printing the ones that match.
tools/perf/builtin-list.c +3 -0
@@ -13,6 +13,7 @@#include "util/parse-events.h" #include "util/cache.h" +#include "util/pmu.h" int cmd_list (int argc,const char **argv,const char *prefix__maybe_unused ) {
@@ -37,6 +38,8 @@ int cmd_list(int argc, const char **argv, const char *prefix __maybe_unused)else if (strcmp (argv[i],"cache" ) ==0 ||strcmp (argv[i],"hwcache" ) ==0 )print_hwcache_events (NULL , false); +else if (strcmp (argv[i],"pmu" ) ==0 ) +print_pmu_events (NULL , false);else if (strcmp (argv[i],"--raw-dump" ) ==0 )print_events (NULL , true);else {
tools/perf/util/parse-events.c +4 -1
@@ -1088,6 +1088,8 @@ int print_hwcache_events(const char *event_glob, bool name_only) } } +if (printed) +printf ("\n" );return printed; }
@@ -1142,11 +1144,12 @@ void print_events(const char *event_glob, bool name_only)print_hwcache_events (event_glob, name_only); +print_pmu_events (event_glob, name_only); +if (event_glob !=NULL )return ;if (!name_only) { -printf ("\n" );printf (" %-50s [%s]\n" ,"rNNN" , event_type_descriptors[PERF_TYPE_RAW ]);
tools/perf/util/pmu.c +61 -0
64 lines not on this page · raw message
tools/perf/util/pmu.h +2 -0
@@ -47,5 +47,7 @@ int pmu_aliases_parse_multi(char *path, struct list_head *head);int arch_pmu_aliases (char *name,struct list_head *head); +void print_pmu_events (const char *event_glob,bool name_only); +int perf_pmu__test (void );#endif /* __PMU_H */
Signature · 1 line
1.7.11.7