Other Parts Discussed in Thread: SYSBIOS
Hello,
There were questions about the work of HSR with the implementation of PRU-ICSS.
We evaluated the HSR demo project (hsr_prp_app) from Sysbios_ind_sdk, where the same code can be used on the AM3 and AM4 platform.
We modified the project to automatically send data with a minimum delay and increase the length of the "package" from 64 to 1500 bytes every 1.5-2 seconds.
(The line of lengths is 64, 96, 128, 192, 256, 512, 768, 1024, 1280, 1500 bytes)
We got an incomprehensible effect. On short packets (64-256 data bytes) on the AM437x Industrial EVM rev 1.4A are transmitted significantly slower than on AM335x ICE Board Rev 2.1A (see the Appendix). With a length of 512 - 1500 bytes speed begin to equalize and approximately reach the maximum possible network bandwidth for "packets" up to 1500 bytes length.
If you collect a chain of two cards with a generator on the AM335x ICE via the AM437x EVM on a PC, the data is lost and processed, limited to the performance of the AM437x.
If the generator made on the AM437x data go through the AM335x ICE in the PC lossless.
We think that the even on AM335x ICE Board. 5-10 Mbps below the available network bandwidth.
How this behavior can be explain?
