Re: [tree-ssa] Patch: add -fuse-global-var

[Dale Johannesen <dalej at apple dot com>]

On Jan 19, 2004, at 2:37 PM, law@redhat.com wrote:
> In message <1074551138.25137.159.camel@frodo.toronto.redhat.com>, 
> Diego Novillo
>  writes:
> > On Mon, 2004-01-19 at 17:12, Dale Johannesen wrote:
> > > This adds a new flag, -fuse-global-var, to control whether the
> > > .GLOBAL_VAR mechanism
> > > is used or not.  The default is unchanged.  Two reasons:
> > > - enable best code gen when the user is willing to accept slow
> > > compilation (e.g. benchmarking)
> > > - permit construction of small testcases for bugs dependent on
> > > .GLOBAL_VAR being used
> > > Bootstrapped and tested on Darwin.
> >
> > I don't think this is a good idea.  Since .global_var is obscure 
> > enough
> > for users, I think we might as well add a --param switch that controls
> > the heuristic used to trigger the creation of .global_var.
> Yea, let's make this a --param.  Cleaning up this kind of thing is 
> actually
> one of the things we have to do to integrate with the mainline anyway.

Well, all right.  It's documented in either case, and seems to me 
--param is
more obscure for users.  Another mysterious big number, as opposed to
"on / off / let the compiler choose".  Why do you prefer --param?