Some Eclipse Foundation services are deprecated, or will be soon. Please ensure you've read this important communication.
Bug 301893 - Should sort input to SAT4J
Summary: Should sort input to SAT4J
Status: CLOSED FIXED
Alias: None
Product: Equinox
Classification: Eclipse Project
Component: p2 (show other bugs)
Version: 3.6   Edit
Hardware: PC Windows XP
: P3 normal (vote)
Target Milestone: 3.6 M6   Edit
Assignee: Daniel Le Berre CLA
QA Contact:
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2010-02-04 16:55 EST by DJ Houghton CLA
Modified: 2010-02-24 09:34 EST (History)
5 users (show)

See Also:

Attachments
Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description DJ Houghton CLA 2010-02-04 16:55:04 EST 
We have been investigating problems where either an optimal solution was not found or was not found within specified conflict limits. It is difficult to reproduce these problems on separate machines because the input order depends on which VM the person is running, etc.

We should consider sorting the input to SAT4J to get consistent solutions across VMs.
Comment 1 Daniel Le Berre CLA 2010-02-23 10:41:12 EST 
The code to sort the IU is already inside the projector code, but only activated in debug mode.

In p2cudf, I activated that code to have reproducible results and did not noticed a decrease of performance, even if we are treating more than 50K IUs.

I suggest that we always sort the IUs to facilitate future debugging and maintenance of p2.

Waiting for a go before doing a commit.
Comment 2 John Arthorne CLA 2010-02-23 12:20:12 EST 
I now this inconsistency has caused problems a few times in the past so being more consistent here would help. I noticed when this was last visited in bug 267504 there was some performance concerns with it, but it sounds like you no longer see this hit.
Comment 3 Daniel Le Berre CLA 2010-02-23 17:42:21 EST 
I committed the code on HEAD.