Re: new syntax for perf event
From: Jiri Olsa
Date: Thu Jan 05 2012 - 04:17:25 EST
hi, any feedback?
thanks,
jirka
On Wed, Dec 21, 2011 at 05:16:13PM +0100, Jiri Olsa wrote:
> On Tue, Dec 20, 2011 at 12:39:21PM +0100, Peter Zijlstra wrote:
> > On Tue, 2011-12-20 at 12:30 +0100, Peter Zijlstra wrote:
> > > event config-0:7
> > > umask config-8:15
> > > usr config-16
> > > os config-17
> > > edge config-18
> > > any config-21
> > > inv config-23
> > > cmask config-24:31
> > >
> > > nhm-dmnd_data_rd config1-0
> > > nhm-dmnd_rfo config1-1
> > > nhm-dmnd_ifetch config1-2
> > > ...
> > >
> > > (the real syntax will likely be a little more complex in order to be
> > > able to express the various other architectures their raw input format).
> >
> > for instance AMD would need:
> >
> > event config-0:7,32-35
> > guest config-40
> > host config-41
> >
> > (assuming you read that as a mask from LSB to MSB)
> >
> > And I still haven't actually read how the P4 works so I can't help out
> > there.
>
>
> ok first try ;) description + grammar attached
>
> - event has a general format:
> pmu/event_def[:modifier] [ ',' pmu/event_def[:modifier] ] ...
>
> where event_def depends on the chosen pmu:
> 'cpu' 'tracepoint' 'breakpoint'
>
>
> - to keep some of the old interface, symbolic and cache events stays,
> since they are recognizable by keywords (event_symbol, event_cache)
>
>
> - raw events are defined with the help of event keywords like:
> cpu/raw,config=1,config1=0,config2=0
> cpu/hw,config=1,config1=0
>
> * first symbol after '/' defines the type: raw/hw/cache
> which match PERF_TYPE_RAW/PERF_TYPE_HARDWARE/PERF_TYPE_HW_CACHE
>
> * the rest consists assignments for 'config/config1/config1' keywords
> that match with the perf_event_attr:config/config1/config values
> (possibly extended once we have sysfs events representation)
>
>
> - tracepoint events are defined like:
> tracepoint/system:tracepoint
>
> where tracepoint could contain '*?'
>
>
> - breakpoint events are defined like:
> breakpoint/addr:type
>
>
> - each event could be followed by modifier definition
>
>
> - I think pmu names should match registered pmu names,
> but we could make some shortcuts so we dont need to type 'tracepoint/',
> but just 'tp/' or smth..
>
>
> - as for tracepoints it looks like the old format does not clash with the new one
> and we could add shortcut (omit the 'tracepoint/') like:
> event: event_tracepoint
>
>
> - not sure I should include 'software' pmu, as it seems to covered by
> event_symbol events
>
>
> thanks for comments,
> jirka
>
>
>
> grammar
> -------
> events:
> event_mod ',' event_mod | event_mod
>
> event_mod:
> event | event ':' modifier
>
> event:
> event_symbol |
> event_cache |
> 'cpu' '/' event_cpu |
> 'tracepoint' '/' event_tracepoint
> 'breakpoint' '/' event_breakpoint
>
> event_symbol:
> cpu-cycles|cycles
> stalled-cycles-frontend|idle-cycles-frontend
> stalled-cycles-backend|idle-cycles-backend
> instructions
> cache-references
> cache-misses
> branch-instructions|branches
> branch-misses
> bus-cycles
> cpu-clock
> task-clock
> page-faults|faults
> minor-faults
> major-faults
> context-switches|cs
> cpu-migrations|migrations
> alignment-faults
> emulation-faults
>
> event_cache:
> cache_type
> cache_type '-' cache_result_op
> cache_type '-' cache_result_op '-' cache_result_op
>
> cache_type:
> L1-dcache|l1-d|l1d|L1-data
> L1-icache|l1-i|l1i|L1-instruction
> LLC|L2
> dTLB|d-tlb|Data-TLB
> iTLB|i-tlb|Instruction-TLB
> branch|branches|bpu|btb|bpc
> node
>
> cache_result_op:
> load|loads|read
> store|stores|write
> prefetch|prefetches
> speculative-read|speculative-load
> refs|Reference|ops|access
> misses|miss
>
> event_cpu:
> 'raw' ',' event_cpu_def
> 'hw' ',' event_cpu_def
> 'cache' ',' event_cpu_def
>
> event_cpu_def:
> event_cpu_ass ',' event_cpu_ass |
> event_cpu_ass
>
> event_cpu_ass:
> event_cpu_elem '=' value
>
> event_cpu_elem:
> 'config' | 'config1' | 'config2'
>
> event_tracepoint:
> system ':' tracepoint
>
> event_breakpoint:
> addr ':' type
>
> modifier:
> [ukhp]{1,4}
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/