Hi Alejandro,
The behaviour I’m seeing is different performance results in read operations when working in a standalone single-core configuration,
Where the driver is a single task running on M4 only, and is based on your svf522revb_m4 example that comes with the Vybrid MQX 4.1 s/w,
With only QSPI1 enabled and active,
Where the results for reading a 64 bytes block were fixed on 47 microseconds,
And our application configuration, which is a dual core (with A5 active also) where the QSPI1 driver is implemented as the highest priority pre-emptive task,
And the QSPI0 is used for the NOR-FLASH (where the code is located) which is being accessed by A5.
In the application configuration (dual core, dual QSPI i/f’s) I’ve got inconsistent results for the read operation in the range of 360-590 microsec.,
Which suggest that the read operation, although implemented as a highest priority pre-emptive task, was interrupted by another event/s.
I’ve checked in the QSPI example drivers that come with the Vybrid MQX 4.1 s/w, and saw that in all of them,
The option of both i/f’s (QSPI0 & QSPI1) active at the same time is disabled by s/w ( please see BSP_QUADSPI_MEMORY_CHANNEL flag),
And wondered whether that could be the reason.
Looking forward to your answer,
Thanks,
Tali