HP Caliper User Guide Release 5.5 (5900-2351, August 2012)

Database: /home/sujoys/db3
Measurement scope: per-process
Sampling Specification
Sampling event: CPU_CYCLES
Sampling period: 500000 events
Sampling period variation: 25000 (5.00% of sampling period)
Sampling counter privilege: user (user-space sampling)
Data granularity: 16 bytes
Data sampled: IP
--------------------------------------------------------------------------
If the data collected in two or more of the databases is similar data (for example, an fprof
collection), then the report has this data summed together. If the data collected is different (for
example, an fprof collection and a dcache collection), then the report will contain both sets of
data.
For example, the Function Summary for a merged fprof and dcache report will have separate
columns to specify all of the following:
Sampled Dcache
IP Dcache Latency
Samples Misses Cycles
If you do not specify a database, HP Caliper uses the database created from your latest run. (Of
course, there is no merging of data, but you do not get an error.)
By default, all processes with the same basename will have their data merged together, regardless
of where they might reside in the process hierarchy of the input databases. To change this, use
the --group-by module option, which will group data by load module with the same basename.
HP Caliper supports merge reports for all measurements except the ones below. For these
measurements, the data is not merged across collection runs. Instead, the reports are appended
one after another:
cgprof (HP-UX only)
cpu (HP-UX only)
pmu_trace
scgprof
You cannot merge databases to a single database. You can only merge databases to a report.
Using the caliper diff Command to Difference Data Collected in Two Databases
Use caliper diff to create a report that differences the data collected in two databases. In the
report, the contributing collection runs are appended one after another.
The syntax for this command is:
caliper diff [report_options] database2 database1
HP Caliper will produce a report that shows the difference in data collected between matching
processes in the two databases, but only for processes with the same measurement type. Any
process that does not match a process in the other database will be ignored. See Example 3.
118 Controlling the Content of Reports