[uClibc]Re: setjmp/longjmp and others

Miles Bader miles at lsi.nec.co.jp
Tue Jan 8 04:25:57 UTC 2002


Erik Andersen <andersen at codepoet.org> writes:
> > Note that adding '-falign-functions=0' to the compile flags for
> > uClibc can decrease it's size by several percent.
> 
> I have that now for x86.  It wasn't certain that it was
> safe as a general case optimization.  Are you certain that
> would be safe for ARM, sparc, etc?

`-falign-functions' seems to be a gcc 3.x thing; gcc 2.95 only supports the
machine-dependent option `-malign-functions' (on selected processors).
Even if this option were safe, I suspect that many people compiling uClibc,
especially those using non-mainstream processors, are not using gcc 3.x yet.

I presume that gcc will maintain minimal instruction alignment, even with
`-falign-functions=0', for processors that need it, but the documentation
doesn't seem to say so explicitly.

-Miles
-- 
Quidquid latine dictum sit, altum viditur.





More information about the uClibc mailing list