Commit message (Collapse) | Author | Age | Files | Lines | |
---|---|---|---|---|---|
* | Merge branch 'master' into FDK | Willem Jan Palenstijn | 2016-10-06 | 1 | -137/+184 |
|\ | |||||
| * | Fix inefficient block split logic | Willem Jan Palenstijn | 2016-06-20 | 1 | -5/+19 |
| | | |||||
| * | Improve volume block reduction | Willem Jan Palenstijn | 2016-06-20 | 1 | -132/+165 |
| | | | | | | | | | | The previous version would make the blocks too large due to inefficient computation of overlap. | ||||
* | | Use CompositeGeometryManager for FDK | Willem Jan Palenstijn | 2016-04-18 | 1 | -3/+47 |
|/ | | | | | Also fix a number of scaling/weighting issues in FDK, and switch to standard cone_bp with FDKWeighting for the BP step. | ||||
* | Fix build | Willem Jan Palenstijn | 2016-03-01 | 1 | -6/+6 |
| | |||||
* | Query max texture size instead of hardcoding it | Willem Jan Palenstijn | 2016-02-16 | 1 | -6/+6 |
| | |||||
* | Avoid (unlikely) integer overflow | Willem Jan Palenstijn | 2016-02-15 | 1 | -1/+5 |
| | |||||
* | Skip some unnecessary splitting | Willem Jan Palenstijn | 2016-02-15 | 1 | -5/+8 |
| | |||||
* | Also split volumes in X/Y directions to respect CUDA limits | Willem Jan Palenstijn | 2016-02-15 | 1 | -21/+240 |
| | |||||
* | Refactor FP and BP jobs creation in the composite geometry manager | Nicola Vigano | 2016-02-10 | 1 | -14/+25 |
| | |||||
* | Add multi-GPU support to CompositeGeometryManager | Willem Jan Palenstijn | 2016-01-19 | 1 | -114/+320 |
| | |||||
* | Fix whitespace | Willem Jan Palenstijn | 2016-01-19 | 1 | -21/+21 |
| | |||||
* | Fix projections parallel to XZ or YZ planes | Willem Jan Palenstijn | 2016-01-14 | 1 | -0/+12 |
| | | | | | The early-cutoff optimization in CompositeGeometryManager was failing to properly handle +/-Inf. | ||||
* | Add utility functions for creating FP/BP JobLists | Willem Jan Palenstijn | 2015-12-04 | 1 | -2/+111 |
| | |||||
* | Add CompositeGeometryManager | Willem Jan Palenstijn | 2015-12-04 | 1 | -0/+884 |
This handles FP and BP operations on multiple data objects at once, splitting them to fit in GPU memory where necessary. |