romcc progress.

Eric W. Biederman ebiederman at lnxi.com
Mon Jun 16 12:54:01 CEST 2003


Please don't do html mail...
And please CC the linuxBIOS list.  I admit it is not a real
substitute for documentation but with these kinds of discussions
happening in public the information is distributed more widely
and more people can answer question or take action based on
a conversation.

Jeff Pipkins <jpipkins at austin.rr.com> writes:

> This is a really cool concept, and a very useful tool.  <br>
> <br>
> It took me a while to figure out what romcc was from reading the romcc code
> and the LinuxBIOS code.  I think it would be worthwhile to put a copy of
> the email below in the romcc directory in the source tree, especially if
> you think it'll be a while before any formal documenation exists.  

That sounds like a good idea.  I will see about that.

> It would
> also be helpful to have a short comment in the C code that depends on romcc
> to say so.  Otherwise, it's a real head-twister to see a C program setting
> up mtrr's!  ;^)<br>

Well changing mtrrs is not romcc specific.  We do that even in
the freebios1 tree.  With romcc you can just run the C code earlier.
The real difference is that with romcc rdmsr wrmsr are done with
builtins whereas with gcc you need to do inline asm.  

I am in the processes of setting of the APIs in LinuxBIOS so
that most code will work with or without romcc.  You comment
reminds me that I still have to sync of the rdmsr/wrmsr APIs...

> <br>
> Nice piece of work, thanks in advance for the fun I'm going to have with
> it!<br>

Welcome. 

And BTW I have just committed a change that reduces the algorithmic
complexity of parts of the register allocator.  So the compiler now
should have a much more reasonable runtime.

Eric




More information about the coreboot mailing list