On 10/07/2020 01:39 PM, Mike Leach wrote:
Hi,
On Wed, 7 Oct 2020 at 13:29, Suzuki K Poulose suzuki.poulose@arm.com wrote:
On 10/07/2020 12:21 PM, Mike Leach wrote:
Hi Suzuki,
On Wed, 7 Oct 2020 at 12:13, Suzuki K Poulose suzuki.poulose@arm.com wrote:
On 10/07/2020 12:07 PM, Mike Leach wrote:
Hi Suzuki,
On Wed, 7 Oct 2020 at 10:36, Suzuki K Poulose suzuki.poulose@arm.com wrote:
Hi Mike
On 08/27/2020 03:34 PM, Mike Leach wrote: > Adds handling for the cs_etm/cs_config directory that contains loaded > configurations that may be selected by the perf command line using the > @name syntax. > > Adds flag to mark the hash ID found as a configuration name rather > than a sink name. > > Signed-off-by: Mike Leach mike.leach@linaro.org
Do we need this to be controlled by "@" ?
Can we make it work like :
cat /sys/bus/event_source/devices/cs_etm/format/cs_sys_cfg config2:36
cat /sys/bus/event_source/devices/cs_etm/events/autofdo cs_sys_cfg=1,sinkid=<hash-for-autodo>
Not sure how this works on the command line?
Are you suggesting the replacing the easy to use: perf -e cs_etm/@autofdo/ ... with:
Sorry I was not clear in my comments above
perf -e cs_etm/cs_sys_cfg=1,sinkid=<some value i've got to look up and copy>/ ... or are you suggesting that the two syntax variants should be equivalent - as per the @sink / sinkid=<sink_hash> are?
No. This will become :
perf record -e cs_etm/autofdo/
Perf will be able to expand "autofdo" to "cs_sys_cfg=1,sinkid=<autofdo-hash>" based on the fact that
cat /sys/bus/event_source/devices/cs_etm/events/autofdo cs_sys_cfg=1,sinkid=<hash-for-autodo>
and set the config2 automatically. So that you don't need special handling in perf tool.
OK - didn't realise it would do that - but it does sound better! Not seeing an events directory in my cs_etm on the DB410 - presumably I have to create that?
Correct, we don't have the event group yet, mainly because our configs were 1bit fields, except for the sinkid. We could add that as well.
e.g,
cat /sys/..event_source/.../cs_etm/events/tmc_etr0 sinkid=<hash-for-tmc_etr0>
Cheers Suzuki
OK - thanks. I'll stick to what is immediately relevant for this set now. As this is a new command option / use case, not using the @ is fine.
Sure, np.
Not sure that changing the existing command line we have been using for years for sinks will be equally well received!
Btw, that doesn't mean we have to stop supporting the existing way /@tmc_etr0/. Both are covered by different code in the perf tool.
Cheers Suzuki