Hi, I need a queue structure for passing bytes for a uart driver. I was using the freertos queue implementation but a co-worker insist on writing our own queue struct. Is there any benefit of rewriting our own?
It might be possible to write a slightly more efficient queue knowing that one end is alway an interrupt routine, and maybe some others knowable special cases (like it is always 1 byte at a time being sent, if that is how you are doing it). I am not sure if the improvements are worth the extra work and risk of bugs.
While the FreeRTOS queue works well and is easy to use it does have some drawbacks. It’s oriented toward one character at a time rather than block transfers. I did go the separate FIFO driver route in order to minimize interrupt overhead. I read and write blocks of data through a FIFO, also making use of DMA to reduce USART TX interrupts by as much as 100 to 1 (sending large Modbus messages).
If your data rate is low use the FreeRTOS queue, but if you have a high data rate your co-worker is right…roll your own buffer management. USART interrupt overhead can be significant.
It’s oriented toward one character at a time rather than block transfers
A queue can be created to hold blocks of any size, although each block in a queue has to be the same size. To queue really large blocks you would queue a pointer to the block instead of the block itself.
The FreeRTOS queues are designed for their purpose.