stats: Method stats source

Information
Submitter:	Andreas Hansson
Repository:	gem5
Branch:	default
Bugs:
Depends On:
Reviewers
Groups:	Default
People:

Description

Changeset 10193:ae6a0ea78f4f
---------------------------
stats: Method stats source

This source for stats binds an object and a method / function from the object
to a stats object. This allows pulling out stats from object methods without
needing to go through a global, or static shim.

Syntax is somewhat unpleasant, but the templates and method pointer type
specification were quite tricky.  Interface is very clean though; and similar
to .functor

Testing Done

This is cool, but it'd be better to just change FunctorProxy to be more generic.  If FunctorProxy held a std::function<Counter()> (instead of the T*), then anything that took no arguments and returned a stats::Counter could be used.  This would allow pretty much everything to be used: functions, classes, lambdas, and anything in <functional>.   For the specific case you're trying to solve for, you could simply use a lambda or std::mem_fn and std::bind to wrap objects.

Check out: http://en.cppreference.com/w/cpp/utility/functional

Ali Saidi April 23, 2014, 11:42 a.m. (April 23, 2014, 11:42 a.m.)

I learn something new every day... but unfortunately, i don't think gcc 4.4 supports std::bind

Nathan Binkert April 23, 2014, 12:42 p.m. (April 23, 2014, 12:42 p.m.)

Wow, is 4.4 the minimum version?  You guys should move forward.  Pretty easy to get newer compilers than that on almost any system.

You could use boost::bind for ancient versions.  std::bind was basically lifted from boost.

Stephan Diestelhorst April 23, 2014, 1:27 p.m. (April 23, 2014, 1:27 p.m.)

Good point, Nate.  I actually had a look at these (not only to figure out the template craziness) as well, but came to the same conclusion as Ali.  And I am not too sure about boost, I get a mixed (at best) vibe about it around here ;-)

Nathan Binkert April 23, 2014, 1:30 p.m. (April 23, 2014, 1:30 p.m.)

Perhaps you don't want to use boost as a general library, but using it as a way to effectively get backports of modern C++11 features seems like a win.  My experience is that C++11 has been a huge boon to productivity and correctness, you guys should do whatever you can to start using as many of the features as possible.

Andreas Hansson April 23, 2014, 3:19 p.m. (April 23, 2014, 3:19 p.m.)

I would think the best route is to adopt a more recent version of gcc as the minimum, if we really want to make this happen. With gcc 4.9 released a few days ago perhaps this is the right time.

Andreas Hansson May 2, 2014, 1:38 a.m. (May 2, 2014, 1:38 a.m.)

I'm working on the assumption that this patch is good to go as is.

As a separate thread I'll send out a mail concerning bumping gcc to 4.6.

This is cool, but it'd be better to just change FunctorProxy to be more generic.  If FunctorProxy held a std::function<Counter()> (instead of the T*), then anything that took no arguments and returned a stats::Counter could be used.  This would allow pretty much everything to be used: functions, classes, lambdas, and anything in <functional>.   For the specific case you're trying to solve for, you could simply use a lambda or std::mem_fn and std::bind to wrap objects.

Check out: http://en.cppreference.com/w/cpp/utility/functional

You have a pending review.

Review Board 2.0.15

This change has been marked as submitted.

Screenshots

Files

Status: Closed (submitted)