On Fri, 29 Oct 2010 14:03:10 -0700 Mark Brown broonie@opensource.wolfsonmicro.com wrote:
On Fri, Oct 29, 2010 at 03:02:22PM +0300, Jarkko Nikula wrote:
Functions dapm_seq_run and dapm_seq_run_coalesced are modified so that a new write is issued when the sequence extends to an another device and DAPM context of a widget instead of originating DAPM context is used when doing pop waits and register writes.
I'm having a hard time parsing this bit of the changelog, I'm afraid. What I'd expect is that the comparison operation used for coalescing would be extended to compare based on the CODEC/DAPM context too and the split per device should naturally fall out of that. I didn't notice the comparison being updated?
It was these diffs for dapm_seq_run:
/* Do we need to apply any queued changes? */ - if (sort[w->id] != cur_sort || w->reg != cur_reg) { + if (sort[w->id] != cur_sort || w->reg != cur_reg || + w->dapm != cur_dapm) { if (!list_empty(&pending)) dapm_seq_run_coalesced(dapm, &pending);
So if DAPM is different the dapm_seq_run_coalesced is run now and next iterations are done for another DAPM context.
Looking at this now I think I can leave out dapm_seq_run_coalesced changes by changing the line above to dapm_seq_run_coalesced(cur_dapm, &pending). I'll check it for the next version.
Otherwise this looks pretty good, these are stylistic things:
Good notes. I'll change them.