linux-kernel - Re: [PATCH v6 00/12] perf tools: fix perf stat with large socket IDs

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <30c4426d-b314-98ff-e6d5-6d5152f316e8@huawei.com>
Date:   Fri, 4 Dec 2020 11:48:36 +0000
From:   John Garry <john.garry@...wei.com>
To:     Jiri Olsa <jolsa@...hat.com>, James Clark <james.clark@....com>
CC:     <linux-perf-users@...r.kernel.org>, <linux-kernel@...r.kernel.org>,
        <namhyung@...nel.org>,
        "linux-arm-kernel@...ts.infradead.org" 
        <linux-arm-kernel@...ts.infradead.org>,
        Linuxarm <linuxarm@...wei.com>
Subject: Re: [PATCH v6 00/12] perf tools: fix perf stat with large socket IDs

On 03/12/2020 15:39, Jiri Olsa wrote:

+

> On Thu, Nov 26, 2020 at 04:13:16PM +0200, James Clark wrote:
>> Changes since v5:
>>    * Fix test for cpu_map__get_die() by shifting id before testing.
>>    * Fix test for cpu_map__get_socket() by not using cpu_map__id_to_socket()
>>      which is only valid in CPU aggregation mode.
>>
>> James Clark (12):
>>    perf tools: Improve topology test
>>    perf tools: Use allocator for perf_cpu_map
>>    perf tools: Add new struct for cpu aggregation
>>    perf tools: Replace aggregation ID with a struct
>>    perf tools: add new map type for aggregation
>>    perf tools: drop in cpu_aggr_map struct
>>    perf tools: Start using cpu_aggr_id in map
>>    perf tools: Add separate node member
>>    perf tools: Add separate socket member
>>    perf tools: Add separate die member
>>    perf tools: Add separate core member
>>    perf tools: Add separate thread member
> 
> Acked-by: Jiri Olsa <jolsa@...hat.com>
> 

Tested-by: John Garry <john.garry@...wei.com>

I still think that vendors (like us) need to fix/improve their firmware 
tables so that we don't get silly big numbers for socket/package IDs, 
like S5418-D0, below:

$./perf stat -a --per-die

  Performance counter stats for 'system wide':

S36-D0   48   72,216.31 msec cpu-clock      #   47.933 CPUs utilized
S36-D0   48        174     context-switches #   0.002 K/sec
S36-D0   48         48     cpu-migrations   #   0.001 K/sec
S36-D0   48         0     page-faults    #   0.000 K/sec
S36-D0   48   7,991,698     cycles    #   0.000 GHz
S36-D0   48   4,750,040     instructions   #   0.59  insn per cycle
S36-D0    1   <not supported>     branches
S36-D0   48      32,928     branch-misses    #   0.00% of all branches
S5418-D0   48   72,189.54 msec cpu-clock     #   47.915 CPUs utilized
S5418-D0   48        176     context-switches  #   0.002 K/sec
S5418-D0   48         48     cpu-migrations   #   0.001 K/sec
S5418-D0   48         0     page-faults     #   0.000 K/sec
S5418-D0   48   5,677,218     cycles    #    0.000 GHz
S5418-D0   48   3,872,285     instructions   #  0.68  insn per cycle
S5418-D0    1   <not supported>     branches
S5418-D0   48      29,208     branch-misses   #  0.00% of all branches

       1.506615297 seconds time elapsed

but at least it works now. Thanks.

> 
>>
>>   tools/perf/builtin-stat.c      | 128 ++++++++++++------------
>>   tools/perf/tests/topology.c    |  64 ++++++++++--
>>   tools/perf/util/cpumap.c       | 171 ++++++++++++++++++++++-----------
>>   tools/perf/util/cpumap.h       |  55 ++++++-----
>>   tools/perf/util/stat-display.c | 102 ++++++++++++--------
>>   tools/perf/util/stat.c         |   2 +-
>>   tools/perf/util/stat.h         |   9 +-
>>   7 files changed, 337 insertions(+), 194 deletions(-)
>>
>> -- 
>> 2.28.0
>>
> 
> .
>