Greetings everyone
Using beaglebone white i have done some experiments trying to use edma to sample Gpio1 port from its datain register.
I input an 1Mhz square to gpmc_ad6 pin and trigger the transfer/transfers via xdma_event_intr2 pin.
Two settings take place
Setting number one:
A-sync transfer
acnt=2
bcnt=1000
ccnt=1
an 30 Mhz square connected to xdma_event_intr2 pin triggering the transfers(Bcnt*Ccnt).
the result is that from the plotted data it seems that i get ~13MSPS = ~13Samples in one period=~26Bytes
Setting number two:
A_B-sync transfer
acnt=2
bcnt=1000
ccnt=1
an 30 Mhz square connected to xdma_event_intr2 pin now triggering only a single transfer(Ccnt).
the result is that from the plotted data it seems that i get ~20MSPS = ~20Samples in one period=~40Bytes
From what i understand the faster A_B-sync makes sence because is just a much more efficient way to move data around , i also understand that setting acnt to 2000 and bcnt to 1 would probably be even more efficient but since i want my source to be a 16bit -register this is no use.
Can someone verify these results and tell me if these are the max data rates available or if there is something to be done to make this senario work faster?
thanks a lot.
Edit: does anyone know what need to be done to be able to dma_alloc_coherent more than 2MB?
