lkml.org 
[lkml]   [2014]   [Nov]   [5]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
SubjectRe: [PATCH 1/1] perf tools: perf diff for different binaries
Date
Hi Kan,

On Tue, 4 Nov 2014 17:07:43 +0000, Kan Liang wrote:
> Hi Namhyung,
>
>> > tchain_edit [.] f1
>> > 0.14% 3.913444 tchain_edit [.] f2
>> > 99.82% 1.005478 tchain_edit [.] f3
>>
>> Hmm.. I think it should be a default behavior for perf diff, otherwise -s
>> symbol is almost meaningless IMHO.
>
> I think we need both instruction level and function level diff.
> For debugging scaling issue, I think we need to do deeper analysis for some
> cache or lock issue. The function level is too high granularity.
>
> The new option can be used to debug scaling regression issue.
> If the binary/kernel is updated, it doesn't make sense to compare the
> symbol address, since it should be changed. So comparing the function
> should be more useful.
>
>
>> What about setting the
>> sort_sym.se_collapse in data_process() so that hists__match() can use
>> symbol names?
>
> Yes, we can set it if we only do function level diff. But I'd like to keep
> both. So I defined two sort keys.
> "symbol" means "symbol address executed at the time of sample "
> "symbol_name" means "name of function executed at the time of sample"

Hmm.. I don't think the symbol sort key provides the instruction level
diff that you want. If it finds a symbol it just use the start address
of the symbol, not the exact address of the sample. Am I missing
something?

Thanks,
Namhyung


\
 
 \ /
  Last update: 2014-11-05 08:04    [W:0.090 / U:2.068 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site