[U-Boot] [PATCH 1/8] POST/arm: adaptations needed for POST on ARM to work

[Marek Vasut]

On Thursday, August 18, 2011 04:31:09 PM Mike Frysinger wrote:
> On Thursday, August 18, 2011 06:07:06 Marek Vasut wrote:
> > On Tuesday, August 16, 2011 08:03:54 PM Mike Frysinger wrote:
> > > On Monday, August 15, 2011 05:09:42 Valentin Longchamp wrote:
> > > > On 08/14/2011 09:07 PM, Mike Frysinger wrote:
> > > > > On Wednesday, August 03, 2011 08:37:00 Valentin Longchamp wrote:
> > > > >> --- a/include/post.h
> > > > >> +++ b/include/post.h
> > > > >> 
> > > > >> +/*
> > > > >> + * some ARM implementations have to use gd->ram_size, since
> > > > >> POST_WORD is + * defined in RAM
> > > > >> + */
> > > > >> +DECLARE_GLOBAL_DATA_PTR;
> > > > > 
> > > > > i'm not sure about this.  no other header has been allowed to do
> > > > > this in the past, and i dont think we should start now.
> > > > 
> > > > OK. Then we should move the post_word_load and post_word_store
> > > > function to post/post.c. Would this be accepted ?
> > > 
> > > that would add overhead that most people dont need.  i guess the only
> > > other option would be to add a CONFIG_POST_EXTERNAL_WORD_FUNCS and then
> > > post.h would just define the two funcs as externs.  it'd be up to the
> > > board porters to define them however they want.
> > 
> > We don't want externs. Why would moving it into post.c introduce any
> > overhead ?
> 
> because the current code expands into a single memory read/write for many
> arches.  moving it into post.c already means making it into an extern and
> now people have to call an external function instead of inlining the
> memory access.

I don't think I follow you here ... why won't you be able to inline that stuff if 
it's in post.c ?

> -mike