Hello Simon,
On Fri, 13 Nov 2015 19:12:13 -0700, Simon Glass sjg@chromium.org wrote:
IMO, keeping it simple does not play well with making DM and malloc available pre-SDRAM init :) --- but don't miscontrue my meaning: I'm fine with DM and malloc pre-SDRAM on platforms which have enough IRAM / SRAM / lockable cache / whatever for it. My point is just that sooner or later, someone will start wanting not to do again after SDRAM init what they've already done before it, because it avoids double HW init issues, becaue it saves time, etc. And that's going to be way sooner than later, IMO.
Yes that is quite possibility true. I'm happy to review any patches if it helps. My comment about it possibly not being worthwhile solving is just that. Things have a way of changing, so what looks unnecessary now might look essential tomorrow. In a way, the simple malloc() fits into that category.
And so evolve the uses of DM, of the DT, etc.
Alright, so I will post my RFC and read your comments with much interest!
Regards, Simon
Amicalement,