Ticket #1217 (closed: fixed)
SANS: Mulitple reductions after sample size change produce different results
Reported by: | Martyn Gigg | Owned by: | Martyn Gigg |
---|---|---|---|
Priority: | blocker | Milestone: | Iteration 23 |
Component: | Keywords: | ||
Cc: | Blocked By: | ||
Blocking: | Tester: |
Description
Change History
comment:1 Changed 10 years ago by Martyn Gigg
- Status changed from new to closed
- Resolution set to fixed
comment:2 Changed 10 years ago by Martyn Gigg
- Status changed from closed to reopened
- Resolution fixed deleted
comment:3 Changed 10 years ago by Martyn Gigg
- Status changed from reopened to testing
- Resolution set to fixed
comment:4 Changed 10 years ago by Nick Draper
- Priority changed from major to blocker
- Status changed from testing to reopened
- Resolution fixed deleted
Multiple reductions still appear to alter the reduced value, although the alterations only appear as low Q range and are 10-5 out on a single run.
This gets worse with subsequent runs
Tested on SVN4799
comment:5 Changed 10 years ago by Martyn Gigg
(In [4833]) Added a couple of parallel protected regions to Q1D as it was giving different answers on multiple runs. Note that it still gives a difference of about 10-14 between multiple runs but I believe this is due to limitations of OpenMP and summing double precision numbers in different orders. Re #1217
Note: See
TracTickets for help on using
tickets.
(In [4346]) Fix SANS interface to specify the correct sample geometry upon consecutive reductions. Fixes #1217