-
Bug
-
Resolution: Fixed
-
High
-
PDK
-
PDK-10528
-
PROCESSOR_SDK_08.00.00
-
PROCESSOR_SDK_08.01.00
-
j721e-evm
The Freertos (nonos) implementation of the cache maintenance operations is not optimal. In it, we are looping through the ‘size’ and maintaining one cache line at a time. In SYSBIOS, this was only done for small buffer sizes, but for larger buffer sizes there is a HW optimized method where you program the HW with the starting address and full buffer size, and then wait for the HW to finish. Please refer to SYSBIOS implementation and copy that. They choose to do cache line looping option when the buffer size is below some threshold, and the one shot operation and wait when the buffer is above that threshold.