On 01.03.2008 03:06, ron minnich wrote:
On Fri, Feb 29, 2008 at 5:58 PM, Carl-Daniel Hailfinger wrote:
- Maybe the name of mainboard_pre_payload() is unfortunate and we
should have pre_payload() which calls cpu_pre_payload(), chipset_pre_payload() and mainboard_pre_payload() in a row. That way, we don't have to put any GeodeLX ROM cache disabling below mainboard() and keep that stuff local to the southbridge code
My concern here is we're going to walk our way back to the v1 style functions that end up mostly empty (e.g. mainboard_pre_payload(){}) on most targets. Mainboard_pre_payload is a hack but I have no good ideas on workarounds. Let's think this out some more. We'll get it.
Actually, these empty functions should be zero cost in the binary with gcc, but they clutter the source code. Hmm. I'd like something like "overridable functions" similar to weak functions, but where gcc is able to strip out the call if no non-empty function exists.
The chipsets are always changing our ideas on how to set up the software :-)
Indeed.
Regards, Carl-Daniel