I’m trying to use DMA to transfer data from memory to a UART. I’m using lots of CMSIS code. When DMA didn’t seem to work, I stepped through the library code. In the CMSIS function GPDMA_Setup(…), about half way through a number of values are bit-OR’d together to set the pDMAch->DMACCControl word value. No matter what value I use for the transfer size, those bits appear zeroed when examining the Control word value. I have altered the code, including ORing the values into a temporary variable (which works correctly, the transfer size is in that) then setting DMACCControl=temp_val. This sets all the bits of DMACCControl excepting the transfer size!
I’ve tried two different boards (with different chips) and tried different DMA channels - same results. I’ve scoured the NXP manual looking for some kind of protection for these bits and found none - besides, the code is much like the UART-DMA example included with the CMSIS code - which couldn’t be wrong, could it?? The most recent errata sheet that I had from NXP lists no DMA problems.
Might there be something systemically wrong with the chips (that NXP hasn’t revealed, yet) or a very weird problem with the OpenOCD/Olimex-ARM-USB-OCD debugger? These seem unlikely too.
I’m stumped at this point. If anyone has a clue, I’d appreciate it!
-f