lkml.org 
[lkml]   [2009]   [Apr]   [2]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
    /
    Date
    From
    SubjectRe: [tip:perfcounters/core] perf_counter: kerneltop: update to new ABI

    * Jaswinder Singh Rajput <jaswinder@kernel.org> wrote:

    > On Thu, 2009-04-02 at 20:11 +0200, Ingo Molnar wrote:
    > > * Jaswinder Singh Rajput <jaswinder@kernel.org> wrote:
    > >
    > > > On Thu, 2009-04-02 at 19:05 +0530, Jaswinder Singh Rajput wrote:
    > > > > On Thu, 2009-04-02 at 12:03 +0000, Peter Zijlstra wrote:
    > > > > > Commit-ID: 373b153677f26a263ef297d77a5e045a31f6486c
    > > > > > Gitweb: http://git.kernel.org/tip/373b153677f26a263ef297d77a5e045a31f6486c
    > > > > > Author: Peter Zijlstra <a.p.zijlstra@chello.nl>
    > > > > > AuthorDate: Thu, 2 Apr 2009 11:12:02 +0200
    > > > > > Committer: Ingo Molnar <mingo@elte.hu>
    > > > > > CommitDate: Thu, 2 Apr 2009 13:53:00 +0200
    > > > > >
    > > > > > perf_counter: kerneltop: update to new ABI
    > > > > >
    > > > > > Update to reflect the new record_type ABI changes.
    > > > > >
    > > > >
    > > > > perfstat is still having many issues:
    > > > >
    > > > > 1. 0:6: bus-cycles is not valid for AMD, so it fails
    > > > >
    > > > > 2. ./perfstat -e 0:1,0:2,0:3,0:4,0:5,0:6 ls
    > > > > can be replaced by ./perfstat -e 0:* ls
    > > > >
    > > > > 3. Similarly ./perfstat -e 1:1,1:2,1:3,1:4,1:5,1:6 ls
    > > > > can be replaced by ./perfstat -e 1:* ls
    > > > >
    > > > > 4. All events can be replaced by ./perfstat -e * ls
    > > > >
    > > >
    > > > 5. This command is invalid:
    > > > Sample output:
    > > >
    > > > $ ./perfstat -e 1 -e 3 -e 5 ls -lR /usr/include/ >/dev/null
    > >
    > > yes, this should be fixed in perfstat: '-e 1' should be accepted as
    > > '-e 0:1'. Patches welcome.
    > >
    >
    > if you treat '-e 1' as 'e 0:1' then how you will treat '-e 1:1' ?
    >
    > I think 'e 1:*' is better option to select all software events and
    > '-e 0:*' for all hardware events.

    That is not what i suggested though. I suggested '-e 1' to mean to
    default to the more common case: hw counters. I.e.:

    -e 1 == -e 0:1
    -e 2 == -e 0:2
    -e 3 == -e 0:3
    ...

    sw counters could be specified only via the longer form.

    Anyway, most people will use symbolic names anyway (-e cycles,
    etc.), so i suspect the issue is mostly irrelevant.

    Ingo


    \
     
     \ /
      Last update: 2009-04-02 20:31    [W:2.410 / U:0.292 seconds]
    ©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site