At Tue, 2 Jun 2015 12:07:32 +0100, Mark Brown wrote:
On Tue, Jun 02, 2015 at 12:53:00PM +0200, Takashi Iwai wrote:
Mark Brown wrote:
Ugh, and the functions are macros so can't be used directly. I'd still expect to see these ops be defined in some central place and reused.
Maybe we can lift up some time later, but it's nothing more than an optimization, so in a low priority for now.
It's not like it's hard to do and given that my concerns around this series mostly center around code reuse and abstraction I'd rather not have big warning signs type issues sitting there obscuring other things.
Well, the page allocator ops of snd-hda-intel have workarounds for some chips that need the cache type changes of allocated pages. Tegra and SKL don't need such mess.
Of course, we may split ops again to several pieces, but it doesn't look good, either. Or, a bit more code refactoring is needed; use the standard allocator but apply the post-process and pre-process for allocation and release of pages to fiddle with the memory caches on demand.
Takashi