Yes. There's issue213 for the somewhat worrying memory usage of the 64-bit
configuration; other than that, I don't think more needs to be done at the
moment. Thanks for noticing, and another one closed. :-)

This bug can probably be closed, can't it?

"yY_greedy" should rather be "yY_eager", of course.

> Could you please add those configurations to 
> new-scripts/downward_configs.py?

Can you do it? Here's the necessary hints:

 * A* with blind heuristic
   => already there ("blind")
 * A* with m&s with 200000 states
   => could be called "oa200000" for consistency with the existing "oa50000".
      Add "num_states=200000" to the mas() options. If that doesn't work,
      consult Gabi's excellent docs. ;-)
 * Lazy greedy search with h^cea and its preferred operators
   => already there ("yY")
 * Eager greedy search with h^cea and its preferred operators
   => "yY_greedy"; should be like yY with "lazy" => "eager", I hope.

4 GB
> Just running a bunch of small tests locally without the scripts would also be
> OK. It's not like we need a whole lot of data.
I just like the scripts too much ;)

#! /usr/bin/env python
from issue102 import build_makefile_experiment

def compare_architectures():
     replace = [('32bit', []), ('64bit', [('-m32', '-m64')])]
     build_makefile_experiment(replace)

compare_architectures()

Could you please add those configurations to 
new-scripts/downward_configs.py?

That would be a good idea, yes.

On Mon, Nov 15, 2010 at 10:00 PM, Jendrik <downward.issues@googlemail.com>wrote:

>
> Jendrik <jendrik.seipp@mars.uni-freiburg.de> added the comment:
>
> I guess I will have to disable the memory limit for the tests, won't I?
>
> _______________________________________________________
> Fast Downward issue tracker <downward.issues@gmail.com>
> <http://issues.fast-downward.org/issue147>
> _______________________________________________________
>

-- 

--------------------------------------------------------------
"Adventure is just bad planning."
    Roald Amundsen
    Norwegian Arctic & Antarctic explorer
    (1872 - 1928)
--------------------------------------------------------------

Just increase the memory limit to whatever makes sense without causing the
machine to swap (physical RAM minus 512 MB should work fine). How much physical
RAM do you have available?

Just running a bunch of small tests locally without the scripts would also be
OK. It's not like we need a whole lot of data. Interesting configurations:

 * A* with blind heuristic on a set of problems of scaling size (domain doesn't
matter).
 * A* with m&s with 200000 states on logistics00
 * Lazy greedy search with h^cea and its preferred operators on the large
Satellite tasks.
 * Eager greedy search with h^cea and its preferred operators on the large
Satellite tasks.

I guess I will have to disable the memory limit for the tests, won't I?

To start - change -m32 to -m64 in the Makefile, and check if the planner still 
compiles and runs.
If it works, then we should compare peak memory (as well as run time) between the 
32 and 64 bit versions.

My desktop runs 64 bit Ubuntu. What do I have to do to test it?

I don't have access to a 64-bit system either. Gabi, Jendrik?

As it turns out - we don't have a 64 bit OS installed anywhere.
I can install a 64 bit version on my laptop, but that will take some time.
Can someone else try this?

IPC-2011 will have 6 GB of memory, so we should support a 64 bit version of Fast 
Downward.

From malte:
 It would be good to do some preparation of the code though. In
particular, I'd like to add some static assertions to verify the size of
certain key data structures. This would be even more relevant once we
support -m64 in addition to -m32.

For now, I will try to compile it with -m64 and see what happens.