Hello all,
I've got an application where I'd like to be able to use the PL330 DMA to do 2D
mem-to-mem transfers (i.e. I have two source buffers and I want to interleave
them into a destination buffer one 32-bit word at a time). I'll need to do this
using two separate transfers, since I may not have both buffers available
simultaneously.
I think that this should be possible by adding the appropriate ADDH to the
load/store loop in _ldst_memtomem(), after having set my source and destination
burst size to 4 and length to 1:
static inline int _ldst_memtomem(unsigned dry_run, u8 buf[],
const struct _xfer_spec *pxs, int cyc,
u16 ld_stride, u16 st_stride )
{
...
while (cyc--) {
off += _emit_LD(dry_run, &buf[off], ALWAYS);
off += _emit_ST(dry_run, &buf[off], ALWAYS);
if ( ld_stride > 4 )
off += _emit_ADDH(dry_run, &buf[off], SRC, ld_stride-4);
if ( st_stride > 4 )
off += _emit_ADDH(dry_run, &buf[off], DST, st_stride-4);
...
}
I'd then make the above available through a "device_prep_interleaved_dma"-style
interface.
Depending on the src_inc and dst_inc values in the dma_interleaved_template, it
might be possible to increase the burst length/size of either the source or
destination as well, which would require changing the code above, but I'm just
thinking of a simple implementation right now.
Is anyone aware of any other issues I should be concerned with when trying to
implement 2D transfers with the pl330 DMA?
Thanks.
-Jeremy Trimble
--
To unsubscribe from this list: send the line "unsubscribe dmaengine" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
