Opened 16 years ago
Closed 16 years ago
#254 closed enhancement (done)
Support real value types in NetworkSimplex
| Reported by: | Peter Kovacs | Owned by: | Peter Kovacs |
|---|---|---|---|
| Priority: | major | Milestone: | LEMON 1.1 release |
| Component: | core | Version: | hg main |
| Keywords: | Cc: | ||
| Revision id: |
Description
Attachments (1)
Change History (7)
Changed 16 years ago by
| Attachment: | ns-numerical-6ac5d9ae1d3d.patch added |
|---|
comment:1 Changed 16 years ago by
comment:2 Changed 16 years ago by
| Milestone: | LEMON 1.2 release → LEMON 1.1 release |
|---|
comment:3 Changed 16 years ago by
| Status: | new → assigned |
|---|
comment:4 Changed 16 years ago by
| Summary: | Support real value types in NetworkSimplex → Support real value types in NetworkSimplex, PHASE I. |
|---|
A follow-up ticket #261 has been created for the remaining issues in the floating point support.
comment:5 Changed 16 years ago by
| Summary: | Support real value types in NetworkSimplex, PHASE I. → Support real value types in NetworkSimplex |
|---|
This ticket is about floating point types, #261 is about floating point data.
comment:6 Changed 16 years ago by
| Resolution: | → done |
|---|---|
| Status: | assigned → closed |
Note: See
TracTickets for help on using
tickets.
The attached patch [6ac5d9ae1d3d] supports real value types, but expects integer data (it doesn't use Tolerance, so double values could cause problems). A possible feature could be to support real data in the future. The changeset is on the top of [9ad8d2122b50], see #234.
Moreover it contains an important numerical stability fix. Using a feature of the XTI data structure unlucky tree updates could cause overflow problems related to the node potentials. This feature is not notably faster (at most about 1 percent faster) according to our benchmark tests on various inputs, since it could be used rarely (especially on large networks). On the other hand if it was used more, it would surely cause numerical problems. So it seemed better to remove it.