Messages in this thread | | | Date | Mon, 16 Apr 2012 14:16:28 +0200 | From | Jiri Olsa <> | Subject | Re: [RFCv2 0/8] perf tool: Add new event group management |
| |
On Sun, Apr 15, 2012 at 05:16:55PM +0200, Peter Zijlstra wrote: > On Wed, 2012-04-04 at 23:16 +0200, Jiri Olsa wrote: > > - added 'group=...' syntax to group events (patches 1-5). > > > - added 'group:1=...' syntax to specify which event to sample (leader) > > I'm still sad about not actually using {} for grouping. Esp. since you > use ';' which needs escaping anyway:
yea, I realized we need ending separator after all ;) so right we could use {} then..
so smth like:
'{cycles,faults}' - anonymous group
'group1{cycles,faults} - group with name 'group1'
'group1:1{cycles,faults} - group with name 'group1' and modifier '1'
'group1{cycles,faults},{instructions,cache-misses},faults' '{cycles,faults},group1:2{instructions,cache-misses},faults' ls - group/event combinations
group: group_spec '{' events '}' group_spec: name ':' mod | empty
also, any thoughts about the displaying question? ;)
> I'm looking on how to present this data in perf and it seems we need > to reset all siblings once we read/store them (in kernel) to the > leader sample. > > My current thinking is to store siblings' sum values for each > hists entry of the sample (perf report count unit) .. and display > them in similar way we display callchains: for each hists entry > display the sum value for each sibling. > > Could you provide more of your world examples? Your expectations about > presenting this.. > > Maybe we want to make the reset optional, and do some do some other > math with siblings' values..?
Not sure it could affect the syntax, but it'd be better to have whole picture
thanks, jirka
| |