On 12/30/11 16:16, Shawn Walker wrote:
On 12/30/11 14:05, Brock Pytlik wrote:
On 12/30/11 13:19, Shawn Walker wrote:
On 12/29/11 19:45, Brock Pytlik wrote:
Webrev:
https://cr.opensolaris.org/action/browse/pkg/bpytlik/performance
Bugs:
19104 CliTestCase.pkgsend needs to display traceback when pkgsend
tracesback
19113 old dictionary in imageplan.__find_all_conflicts should be
seeded
using old_excludes
19119 generic.get_varcet_keys shouldn't use startswith
19120 Image shouldn't load the action dictionary twice during an
update
19121 if actions.offsets includes number of lines with the same key,
things can be faster
19122 pkgplans should hold onto their manifests a bit longer
19123 maintaining a cache of created fmris while checking for
conflicting actions makes things faster
19124 conflicting actions should use sets of strings instead of
PkgFmris
19125 pkg_solver should stop creating the same pfmris over and over
19126 manifests should track what's been excluded from them
19127 compiling re's in facets is faster than using fnmatch
I see that you provided lots of numbers for differences in update, but
did you check install for solaris-small-server and the like?
I didn't. Are you thinking of installing them into an empty image or
some other situation? If you describe the experiment you'd like to see,
I'm happy to run it.
Yes, the empty image case. Imagine an AI install being done, and try
to simulate that.
Ok, on my x86 machine, it takes 21 seconds to plan the update and uses
139M. With the changes that I sent out in the webrev, it takes about
17.5 seconds to plan the update and uses 111M. So it's 16% faster and
uses 20% less memory.
src/modules/client/image.py:
============================
lines 2454-2455: I'm uncertain about this change though those concerns
may be for reasons that are no longer valid. I would swear this was
discussed before and it was discovered that this intentionally didn't
take facets into account. That may have been before we changed
imageplanning to do conflict checking though and the like.
Well, all I can say is that all the tests pass with this change. If
there's other reasons not to have facets there, I'm not aware of the
history. I can revert that change, but it means making the checks in the
manifest code more complex (to allow more restrictive excludes to be
provided but not less restrictive ones).
Not asking for that, just racking my brain trying to figure out why
I'm uneasy about this. As I said, my feeling of unease may no longer
apply. But I know for certain I considered the same change at an
earlier point and after talking with Bart or Danek (don't remember
which) came to the conclusion that it was intentional and shouldn't be
changed.
But if the test suite passes (I was already assuming it did), then I
can't give any specific objection.
So many parts of the system should have had more comments :-(
...
After talking w/ Bart, it sounds like whatever reason might have applied
in the past doesn't apply now, so I think we're good w/ this change.
[snip]
-Shawn
Brock
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
