Axiom Computer Algebra System - Patches: patch #2074, Bug #4733 (rounding of negative...
You are not allowed to post comments on this tracker with your current authentication level.
patch #2074: Bug #4733 (rounding of negative numbers)
Submitter: | Bill Page <billpage1> | ||
Submitted: | Wed 15 Oct 2003 04:07:07 AM UTC | ||
Category: | None | Priority: | 5 - Normal |
Status: | Done | Privacy: | Public |
Assigned to: | billpage1 | Open/Closed: | Open |
Sat 18 Oct 2003 03:35:10 PM UTC, comment #1: |
Dylan Thurston <dpt> |
Wed 15 Oct 2003 04:07:07 AM UTC, original submission:
Just adding Dylan's name, since he did most of the analysis work ... |
Bill Page <billpage1> |
Attached Files
Depends on the following items: None found
Items that depend on this one: None found
CC list is empty
There are 0 votes so far. Votes easily highlight which items people would like to see resolved in priority, independently of the priority of the item set by tracker managers.
Follow 3 latest changes.
Date | Changed by | Updated Field | Previous Value | => | Replaced by |
---|---|---|---|---|---|
2005-01-21 | billpage1 | Status | None | Done | |
2004-12-16 | billpage1 | Assigned to | None | billpage1 | |
2003-10-15 | billpage1 | Patch Code | Modified - New Version |
Powered by Savane 3.13-758e.
Corresponding source code
I think this patch could be improved: the definition of shift2 could be modified to not do multiplications (and so presumably be faster).
I also wonder how much the original authors worried about rounding behaviour in some of these calls.
In the function 'normalize', for instance, the old semantics gave 'round towards 0' behaviour; using the new semantics of shift, it would be easy to give 'round towards -infinity' behaviour. But both are wrong: the right method for normalizing floating point numbers, to avoid accumulation of errors, is 'round towards even'.