It looks like we added a python option to run_hw_trace.py to set limit to the number of kernels traced when mlperf was added. This option was overwriting any set DYNAMIC_KERNEL_END environment variable by default and setting it to 0. This PR allows the environment variable to be overwritten only if the -l option is specified in the script (and does not overwrite by default).
It looks like we added a python option to
run_hw_trace.pyto set limit to the number of kernels traced when mlperf was added. This option was overwriting any set DYNAMIC_KERNEL_END environment variable by default and setting it to 0. This PR allows the environment variable to be overwritten only if the-loption is specified in the script (and does not overwrite by default).