Hi All,

I wanted to understand the reason behind defining LOG_LEVEL_* macros as multiple of 10's, also same goes with LOG_MARKER_* macros. If we set the LOG macros to 1,2,3 and so on, we can avoid the mod and division operation when printing the msgs to console.

Please let me your thoughts on this.

Link to this question on Phabricator - https://developer.trustedfirmware.org/T624

Thanks Sheetal

Hi Sheetal I didn't quite get reason for the optimization from multiples of 10 to consecutive numbers. The mod operation you refer to are assertions which will be compiled out when DEBUG=0 or when ENABLE_ASSSERTIONS=0. Other than that there is a single divide by 10 operation within the default implementation of plat_log_get_prefix().

If the logging utility is considered performance critical, removing the single divide by 10 will not result in any gain. Usually the bottleneck for logs are UART baudrates and the CPU spends most of time spinning in a wait loop for UART to be free. Hence I don't see any need for change for log levels macro if the intent is performance optimization.

Best Regards Soby Mathew

-- TF-A mailing list TF-A@lists.trustedfirmware.org https://lists.trustedfirmware.org/mailman/listinfo/tf-a