Re: [PATCH] perf tools: Fix cumul hit based sub-total

[Frederic Weisbecker <fweisbec@gmail.com>]

On Fri, Aug 07, 2009 at 10:12:58AM +0200, Peter Zijlstra wrote:
> On Fri, 2009-08-07 at 07:11 +0200, Frederic Weisbecker wrote:
> > The callchain fractal mode builds each new total hits in a new branch
> > of profiling by using the parent's hits of the current branch plus the
> > hits of the children.
> > 
> > This is wrong, the total hits of a branch should be made of the sum of
> > every children hits, we must ignore the parent hits in this scope.
> > 
> > This patch also fixes another mistakes with the hit counting.
> > 
> > Now the rates are corrects.
> > 
> > Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
> > Cc: Peter Zijlstra <peterz@infradead.org>
> > Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
> > Cc: Peter Zijlstra <peterz@infradead.org>
> > Cc: Mike Galbraith <efault@gmx.de>
> > ---
> 
> Ah will this fix these instances where you'd have things like
> 
>                 |                     page_fault
>                 |                     |
>                 |                     |--100.00%-- __strcmp_sse2
>                 |                     |          copy_dep_chain
>                 |                     |          __libc_start_main
>                 |                     |
>                 |                      --100.00%-- gettimeofday
>                 |                                __libc_start_main



I haven't met such huge rate's mistake yet :)
It was more about +-20% of imprecision.

I guess this patch should fix what you've met. If you still see that,
please tell me.

BTW, I will soon add a debug option to check the percentage correctness
and things like that...