Ram initialization and small c.

Eric W. Biederman ebiederman at lnxi.com
Tue Feb 25 11:32:00 CET 2003


"Ronald G. Minnich" <rminnich at lanl.gov> writes:

> On 25 Feb 2003, Jeremy Jackson wrote:
> 
> > Below is a trivial C program with a function call and a few loops.  When
> > compiled with gcc using the stated flags, it does not use any stack. 
> > The assembler output is below.
> 
> neat!
> 
> I wonder if we'll be able to work with the gcc community somehow to get 
> what Eric needs with a standard gcc.

I suspect a careful gcc port of gcc to x86-noram can handle it, and
long term that is probably where the effort needs to go.

Short term I suspect there would be a lot of distraction with just
the mechanics of gcc.  If someone knows gcc better than me feel free.

For the first proof of concept pass I am going to tackle something
simple and stand alone.  When that works we will have something
to use for the short term, and something we can point at and
say hey it works.  What will it take to get gcc to do something
similar.

One of the very nice things about C is that practically every
optimization can be done by hand.  So a C compiler really only
needs to do register allocation, instruction selection, and
instruction scheduling.  Which means you really don't need a compiler
that is especially smart.

Anyway we will see in a little bit what makes the most sense.

Eric




More information about the coreboot mailing list