Re: [PATCH v9 1/7] perf pmu: "Compat" supports regular expression matching identifiers

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 




在 2023/9/22 下午10:58, Ian Rogers 写道:
> On Fri, Sep 22, 2023 at 1:19 AM Jing Zhang <renyu.zj@xxxxxxxxxxxxxxxxx> wrote:
>>
>>
>>
>> 在 2023/9/21 上午2:36, Ian Rogers 写道:
>>> On Mon, Sep 18, 2023 at 4:52 AM Jing Zhang <renyu.zj@xxxxxxxxxxxxxxxxx> wrote:
>>>>
>>>> The jevent "Compat" is used for uncore PMU alias or metric definitions.
>>>>
>>>> The same PMU driver has different PMU identifiers due to different
>>>> hardware versions and types, but they may have some common PMU event.
>>>> Since a Compat value can only match one identifier, when adding the
>>>> same event alias to PMUs with different identifiers, each identifier
>>>> needs to be defined once, which is not streamlined enough.
>>>>
>>>> So let "Compat" support using regular expression to match identifiers
>>>> for uncore PMU alias. For example, if the "Compat" value is set to
>>>> "43401|43c01", it would be able to match PMU identifiers such as "43401"
>>>> or "43c01", which correspond to CMN600_r0p0 or CMN700_r0p0.
>>>>
>>>> Signed-off-by: Jing Zhang <renyu.zj@xxxxxxxxxxxxxxxxx>
>>>> ---
>>>>  tools/perf/util/pmu.c | 23 +++++++++++++++++++++--
>>>>  tools/perf/util/pmu.h |  1 +
>>>>  2 files changed, 22 insertions(+), 2 deletions(-)
>>>>
>>>> diff --git a/tools/perf/util/pmu.c b/tools/perf/util/pmu.c
>>>> index e215985..7e2242f 100644
>>>> --- a/tools/perf/util/pmu.c
>>>> +++ b/tools/perf/util/pmu.c
>>>> @@ -28,6 +28,7 @@
>>>>  #include "strbuf.h"
>>>>  #include "fncache.h"
>>>>  #include "util/evsel_config.h"
>>>> +#include <regex.h>
>>>>
>>>>  struct perf_pmu perf_pmu__fake = {
>>>>         .name = "fake",
>>>> @@ -875,6 +876,24 @@ static bool pmu_uncore_alias_match(const char *pmu_name, const char *name)
>>>>         return res;
>>>>  }
>>>>
>>>> +bool pmu_uncore_identifier_match(const char *compat, const char *id)
>>>> +{
>>>> +       regex_t re;
>>>> +       regmatch_t pmatch[1];
>>>> +       int match;
>>>> +
>>>> +       if (regcomp(&re, compat, REG_EXTENDED) != 0) {
>>>> +               /* Warn unable to generate match particular string. */
>>>> +               pr_info("Invalid regular expression %s\n", compat);
>>>> +               return false;
>>>> +       }
>>>> +
>>>> +       match = !regexec(&re, id, 1, pmatch, 0);
>>>
>>> I wonder if we can make the regular expressions like
>>> "^(434|436|43c|43a)" more like "(434|436|43c|43a).*", so that we fully
>>> match the id string, by here doing:
>>>
>>> if (match) {
>>>   /* Ensure a full match. */
>>>   match = pmatch[0].rm_so == 0 && pmatch[0].rm_eo == strlen(id);
>>> }
>>>
>>
>> Ok, will do.
>>
>>
>>> I think longer term we can use jevents.py to generate a pmu-events.l,
>>> which would have a contents something like:
>>>
>>> (434|436|43c|43a).*  { return PMU_....;}
>>>
>>> That should make the matching faster but may add some restrictions
>>> onto the regular expression.
>>
>> Could you please describe the function of pmu-event.l in more detail? I may not fully understand it.
> 
> So for now there's no need for a pmu-event.l, I'm fine with the code
> as-is. The issue for using regular expressions is that we need to
> compile (regcomp) then use them (regexec), and in this new code the
> result of the parsing is discarded - perhaps we can save on some
> compiling with a 1 element cache, let's wait to see performance data
> saying it is an issue. If we were to compile the regular expressions
> at build time with flex then the runtime cost, any caching, etc. is
> unnecessary.
> 

I see, thanks for the explanation.

Thanks,
Jing


> Hope this makes sense. Thanks,
> Ian
> 
>> Thanks,
>> Jing
>>
>>>
>>> Thanks,
>>> Ian
>>>
>>>> +       regfree(&re);
>>>> +
>>>> +       return match;
>>>> +}
>>>> +
>>>>  static int pmu_add_cpu_aliases_map_callback(const struct pmu_event *pe,
>>>>                                         const struct pmu_events_table *table __maybe_unused,
>>>>                                         void *vdata)
>>>> @@ -915,8 +934,8 @@ static int pmu_add_sys_aliases_iter_fn(const struct pmu_event *pe,
>>>>         if (!pe->compat || !pe->pmu)
>>>>                 return 0;
>>>>
>>>> -       if (!strcmp(pmu->id, pe->compat) &&
>>>> -           pmu_uncore_alias_match(pe->pmu, pmu->name)) {
>>>> +       if (pmu_uncore_alias_match(pe->pmu, pmu->name) &&
>>>> +                       pmu_uncore_identifier_match(pe->compat, pmu->id)) {
>>>>                 perf_pmu__new_alias(pmu,
>>>>                                 pe->name,
>>>>                                 pe->desc,
>>>> diff --git a/tools/perf/util/pmu.h b/tools/perf/util/pmu.h
>>>> index bd5d804..fc155ce 100644
>>>> --- a/tools/perf/util/pmu.h
>>>> +++ b/tools/perf/util/pmu.h
>>>> @@ -240,6 +240,7 @@ void pmu_add_cpu_aliases_table(struct perf_pmu *pmu,
>>>>  char *perf_pmu__getcpuid(struct perf_pmu *pmu);
>>>>  const struct pmu_events_table *pmu_events_table__find(void);
>>>>  const struct pmu_metrics_table *pmu_metrics_table__find(void);
>>>> +bool pmu_uncore_identifier_match(const char *compat, const char *id);
>>>>
>>>>  int perf_pmu__convert_scale(const char *scale, char **end, double *sval);
>>>>
>>>> --
>>>> 1.8.3.1
>>>>



[Index of Archives]     [Kernel Newbies]     [Security]     [Netfilter]     [Bugtraq]     [Linux FS]     [Yosemite Forum]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux RAID]     [Samba]     [Video 4 Linux]     [Device Mapper]     [Linux Resources]

  Powered by Linux