A structured tracefile consists of a number of files, which can be interpreted as one conceptual data set. See the approximate structure in the figure below. Depending on the organization of actual files, the following component files will be written:

<trace>

is the tracefile name, which is determined automatically or set in the

LOGFILE-NAME

configuration option.

Additionally, five data files may be created for the given trace. These files are Summary Data files. They have common name

<trace>.stf.sum.<suffix>

(where suffix is one of

fnc

,

cop

,

msg

,

cnt

, or

rep

) and formally are not a part of the trace. You can use these files as additional input for Intel® Trace Analyzer. For details of Summary Data usage, see

Intel® Trace Analyzer User and Reference Guide

.

The records for routine entry/exit and counters are contained in the process files. The anchor files are used by Intel® Trace Analyzer to fast-forward within the record files; they can be deleted, but that may result in slower operation of Intel® Trace Analyzer.

Make sure that you use different names for traces from different runs; otherwise you will experience difficulties in identifying which process files belong to an index file, and which ones are left over from a previous run. To catch all component files, use the

stftool

with the

--remove

option to delete a STF file, or put the files into single-file STF format for transmission or archival with the

stftool --convert

option.

The number of actual process files will depend on the setting of the

STF-USE-HW-STRUCTURE

and

STF-PROCS-PER-FILE

configuration options described below.