First important thing the tree fails to compile.  It seems that you
forgot to commit arch/alpha/mt.hh (and maybe some other mt.hh files?)
Please fix this ASAP!

> I just pushed a bunch of stuff that gets the inorder cpu model working
> in the main tree and validated against a few shorter regressions
> (hello,twolf,vortex) that I added to keep it to some level of
> functionality.
I looked through many of the diffs and overall, I have to say great
work!   My only major question is, should the InorderCPU now be
defaulted to compile?  What ISAs does it work with now?  I see that
you made it compile happily with o3, was there any other build issues?

> Of some level of annoyance is that HG committed a couple of the
> "null"-merges I had to do with some of the recent changes that were in
> the tree. I called them null-merges because I really didnt do anything
> other than accept the changes (it affected no inorder-code). However,
> it caused to heads in my tree that had to be merged. I tried to commit
> each patch separately to avoid the committing of this code (e.g. hg
> push -r 6178) but everything just backfired on me. Yes, quite
> annoying.
merging branches is fine.  The history is necessary.  I'm not sure why
you though that committing the patches individually would fix this,
that's unrelated.

> I think everything is still working and the damage done is a few
> redundant changesets, but let me know if there is a need of a revert.
> I kept all the patches together in a separate directory just in case
> something happened unexpectedly (and it did!).
You may have a slight misconception of how mercurial works.  If you do
a push, it  will make the remote tree look exactly like the local
tree.

> Lastly, I got the inorder model to work for a few longer regressions
> (eon, gzip, perl) but I need to do a couple of performance
> optimizations to speed up the simulation. Shouldnt take too long but
> we'll see.
Awesome.

One final thing.  It's clear that you were using mercurial queues a
lot.  I'm glad to see that.  One note is that after you qfold, you
should update the commit message.  There are several options for that.
 use the -m option to qfold.do a "hg qrefresh -m", or edit the header
in the patch file itself (don't forget that you *must* qpop and qpush
the patch again to apply the change if you choose this route.)


  Nate
_______________________________________________
m5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/m5-dev

Reply via email to