Ticket #4986 (closed: wontfix)
Evaluate the speed of finding detector for a given Q position
Reported by: | Janik Zikovsky | Owned by: | Janik Zikovsky |
---|---|---|---|
Priority: | critical | Milestone: | Release 2.1 |
Component: | Mantid | Keywords: | |
Cc: | saviciat@… | Blocked By: | |
Blocking: | Tester: | Andrei Savici |
Description
We want to see how much time it would take to locate the detector ID that contributed to a given Q point, as compared to storing the Zeroes of every detector id and every bin and the memory that uses.
Change History
comment:3 Changed 9 years ago by Janik Zikovsky
- Status changed from accepted to verify
- Resolution set to wontfix
Hi Nick,
You asked me to consider whether it would be possible to calculate the contributing pixels/bins to an inelastic MD workspace on the fly, rather than doing the calculation at the beginning and carrying one MDEvent for each bin of each pixel (possibly with 0 weight).
After some discussion with Andrei I've come to the conclusion that it is very unlikely to be possible to calculate this on the fly. The main problem is that a nice regular shape in Q-space will be distorted when converted to detector/bin space. Even converting each vertex would not be sufficient because the shapes will be curved. It will be very tricky to get it right, and likely computationally intensive; so I don't think it is worth pursuing.
The approach of converting from detector/bin to Q is much more straightforward, at the cost of added memory. If only a small region of Q-space is of interest, the algorithm can reject pixels falling outside the range, saving memory.
Cheers,
Janik Zikovsky Scientific Data Analysis Group, Spallation Neutron Source