On 28 Sep 1998, David Maslen wrote:
> > 2.1.122 seems rock-solid when compiled with egcs-1.1b, -O6 and -mcpu=i686.
>
> So to summarise using this egcs combination to compile a dev kernel on
> SMP will give me a faster/better kernel?
egcs (mostly) is reputed to generate better code than gcc. It doesn't
pass all of the test-suite, but doesn't seem to break the kernel. (I
doubt gcc passes all of the tests either, mind you.)
> Or have we just decided it will probably work, and might not work on a
> stable kernel?
It does not work on 2.0 series kernels. There are patches which seem
to catch the breakages, but I wouldn't advise touching 2.0 with gcc >
2.7.2.x. I forget what the official word from Linus at al is on 2.1
at the moment, but I think that it's along the lines of:
"If egcs/ggc 2.8 breaks things by optimising then we'll fix
it. We _won't_ work around code generation bugs."
But don't quote me.
Matthew.
