previously, performance measurement based on platform clock, source
clock is 38.4mhz, for better align with CPC calculation, change
clock to dsp clock, this will be done through enable config:
CONFIG_TIMING_FUNCTIONS, and this config only be used with perf_cnt
header file, so this change only have impact on performance build.
Signed-off-by: Baofeng Tian <baofeng.tian@intel.com>
Heap memory profiling is based on runtime zephyr API.
It will print out each memory allocation with allocated bytes
and free bytes, display as below:
zephyr: heap allocatd: 22c0 free: 152a90 max allocated: 1000
Signed-off-by: Baofeng Tian <baofeng.tian@intel.com>
PERFORMANCE_COUNTERS_RUN_AVERAGE is not an independent macro,
it depend on: PERFORMANCE_COUNTERS, so it is duplicated,
remove this macro and use one macro to control this module,
also reduced unnecessary profiling log print.
Signed-off-by: Baofeng Tian <baofeng.tian@intel.com>
After add this file, with -o app/perf_overlay.conf in build command
line, profiling functionality will be enabled and mtrace/sof log will
have performance data for further analysis.
Signed-off-by: Baofeng Tian <baofeng.tian@intel.com>
Adrian Bonislawski