This patchset consists of refactoring to allow the decoder to be created in advance when the AUX records are iterated over. The AUX record flags are used to communicate whether the data is formatted or not which is the reason this refactoring is required.
These changes result in some simplifications, removal of early exit conditions etc.
A change was also made to --dump-raw-trace code to allow the formatted/unformatted status to persist and for the decoder to not be continually deleted and recreated.
The changes apply on top of the previous patchset "[PATCH v7 0/2] perf cs-etm: Split Coresight decode by aux records".
Changes since v1: * Change 'decoders_per_cpu' variable name to 'decoders' and add a comment * Add a warning that piped mode is best effort, suggested by Suzuki
James Clark (6): perf cs-etm: Refactor initialisation of kernel start address perf cs-etm: Split setup and timestamp search functions perf cs-etm: Only setup queues when they are modified perf cs-etm: Suppress printing when resetting decoder perf cs-etm: Use existing decoder instead of resetting it perf cs-etm: Pass unformatted flag to decoder
.../perf/util/cs-etm-decoder/cs-etm-decoder.c | 14 +- tools/perf/util/cs-etm.c | 185 +++++++++--------- 2 files changed, 97 insertions(+), 102 deletions(-)