Parallella Community

I think the dma descriptor 'stride' values are better defined as 'increment'. I got confused by the name until i rtfm more closely, and the 2d dma examples aren't much help since they don't describe how the descriptor values are formed.

Since there is a 1d-copy function in e-lib, a 2d-copy one might also be useful, although personally i'd prefer both to be 'async' (eg. specify the channel, but don't wait in completion).

e_dma_copy_2d(id, dst, dststride, src, srcstride, width, height)
... do stuff
e_dma_wait(id)

Where the strides here are proper strides.

notzed,

Thanks for the feedback! I am wondering if this is a matter of coming from different processor camps?
http://www.analog.com/static/imported-f ... 01_pgr.pdf

The TS used modifier and stride interchangeably, but 'increment' would work as well.

We struggled with defining a DMA function library because there are are so many parameters that it seemed almost like programming the registers directly..

We'll think again about the sync/async. We figured that advanced users would just set up their own descriptors but maybe we should opt for a bigger set of functions for the DMA?

Appreciate your inputs!

Andreas