On Sat, Jan 28, 2012 at 8:58 AM, Ali Saidi <[email protected]> wrote:

>
> On Jan 28, 2012, at 3:00 AM, Gabe Black wrote:
>
> >
> >>> My discussion of the general compile-time mechanism is just to
> emphasize
> >>> that the maintenance of this one aspect of the current SE/FS
> difference is
> >>> a stopgap replacement for this alternate ideal, not a perpetuation of
> the
> >>> now-meaningless distinction that Gabe has worked so hard to eradicate,
> so
> >>> that he doesn't accuse me of bitterly clinging to the past...
> >> Ok, but I'd really hate to see this delay Gabe long enough that we
> >> don't ever get it into the tree.  Gabe, do you think it'd be easy to
> >> cook up what Steve is talking about?  (Or does it already work in your
> >> tree?)
> >>
> >
> > I'll try adding an option that leaves out the devices since I assume
> > that's where the extra time is coming from. I'm thinking if you set
> > NO_DEVICES on the scons command line it will leave the devices out of
> > the build. That won't be set by default on anything since the same build
> > is used for both SE and FS and it doesn't make sense as a default, but
> > if you're using it for SE style stuff and build time is affected enough
> > for you to find out about that option, then it'll be available. There
> > may be complications so no promises. It may also not close recover the
> > build time, but I expect it will.
> I think that adding these sorts of things is going to create more
> confusion that it's worth. We're going to have people who compile with
> NO_DEVICE and don't understand why they can't run full-system code (or the
> random error we're going to produce about not being able to create a
> simobject) and we're going to have a group of people who don't know
> NO_DEVICE exists and so they never use it. A 30% compile time increase the
> first time you compile "SE" mode doesn't seem bad. The device models
> shouldn't need re-compile, so any further development for SE activities
> shouldn't effect them. This simples something that confuses a reasonable
> fraction of users (What is SE and FS?) and speeds up compiles tremendously
> for those of us who compile multiple binaries frequently.


If, as Gabe suggests, we maintain some of the SE conditional compilation
flags as NO_DEVICES, but don't expose that flag directly by including it in
any of the "official" named builds, then I expect that only people who know
what they're doing will ever use it.  So I don't see it adding to the
confusion.

Also, as I mentioned before, the argument about "it's only the first time
you compile" only really cuts it if scons magically becomes both bug-free
and omniscient.

Steve
_______________________________________________
gem5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/gem5-dev

Reply via email to