Eventually it probably will make sense to combine these files, but I’m not 
convinced that we’ve reach that point yet.  For instance, the only common 
options between the two files are the benchmark option and now the kernel and 
script options.  All other options supported by fs.py are not supported by 
Ruby.  I think there is some nice cleanliness achieved by keeping these files 
separate and I fear a lot of nasty nested conditional statements will result by 
combining them.  Remember these files are the minimal baseline implementations 
and typically users build complexity on top of them.

I think the se.py and ruby_se.py combination made more sense because there was 
more code shared between the two and less difference between what Classic and 
Ruby supported.  For instance, there is no equivalent FSConfig.py file that the 
se files shared thus there was a lot more common functionality that was pushed 
up to those top-level se files.  Also se mode doesn’t support multiple systems 
connected together via Ethernet, which Classic supports but Ruby does not.

Brad



From: Nilay Vaish [mailto:[email protected]]
Sent: Tuesday, August 09, 2011 7:35 PM
To: Nilay Vaish; Beckmann, Brad; Default
Subject: Re: Review Request: Config: Add ruby_fs.py's functionality to fs.py

This is an automatically generated e-mail. To reply, visit: 
http://reviews.m5sim.org/r/818/



On August 9th, 2011, 9:07 a.m., Brad Beckmann wrote:

I just have a couple minor edits to make to the comments.  However, I wonder 
what is the real benefit for combining the two fs files.  I don't believe there 
is a lot of duplicate code between the two.  Most of the added options in fs.py 
are not supported/related to Ruby.  Also with your change, the main body of 
fs.py is now a large if/else statement with the if being the functionality from 
ruby_fs.py and the else being the fs.py functionality.  Overall, I feel this is 
an unnecessary change that will negatively impact those of us who have other 
internal patches that modify these files.  Why do it?

There are a couple of reasons. We should not have different set of options for

ruby_fs.py and fs.py. Since we have already combined se.py and ruby_se.py, we

should do the same for fs scripts as well. The main body of fs.py being a large

if/else is either because I have not taken out all those things that are common,

and that we still lack in the integration that Ruby and M5 should achieve. As

far as internal patches are concerned, unless the patches can be updated, they

should be used with the version with which they were created. Going ahead, we

are only going to add more features, sooner or later these features will step

on the code used in the patches.


- Nilay


On August 6th, 2011, 12:43 a.m., Nilay Vaish wrote:
Review request for Default.
By Nilay Vaish.

Updated 2011-08-06 00:43:50

Description

Config: Add ruby_fs.py's functionality to fs.py

This patch combines fs.py and ruby_fs.py and removes ruby_fs.py. Using

ruby with fs.py would require mentioning --ruby on the command line. This

same as se.py.


Diffs

 *   configs/example/fs.py (7a9a7f2a3d46)
 *   configs/example/ruby_fs.py (7a9a7f2a3d46)

View Diff<http://reviews.m5sim.org/r/818/diff/>


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

Reply via email to