Re: Live range splitting in new allocator

[Russ Allbery <rra at stanford dot edu>]

Richard Kenner <kenner@vlsi1.ultra.nyu.edu> writes:

> You misunderstand.  The issue isn't giving up optimization.  That's
> totally orthogonal.

I have certainly heard the exact rational that I gave, namely the ability
of the end user to debug the installed binaries, as the reason given for
autoconf's default usage of "-g -O" and the makefile standard to avoid
stripping binaries when installing them.  The former has definitely
resulted in optimization with debugging exercised more than would be the
case otherwise.

Please bear in mind that I was commenting specifically on Linus's comment
about people expecting -O2 to work with -g, not on the more general
question of whether -g should change what optimizations should be used.  I
agree with the policy that you state on the latter point.

-- 
Russ Allbery (rra@stanford.edu)             <http://www.eyrie.org/~eagle/>