cupti_query strange output with Tesla T4

I compiled and ran this sample from cuda-10/extras/CUPTI/sample/cupti_query on linux with three different cards: K80, V100 and T4

Why is this error happening with running on T4?

cupti_query-tesla-v100.txt (2.36 KB)
cupti_query-tesla-k80.txt (9.69 KB)
cupti_query-tesla-T4.txt (198 Bytes)

CUPTI sample cupti_query is based on the Event and Metric APIs which are not supported on the Turing and higher architectures. Event and metrics APIs are being replaced by new Profiling APIs which are supported for Volta and higher architectures. Details can be found at https://docs.nvidia.com/cupti/Cupti/r_main.html#r_profiling_migration