Two parts of this pull request:
1) getReference -> getObjectDependencies: This fixes the currently broken distillation Noise code, which is not correctly handled by the scheduler
2) re-organisation of the perambulator module to batch sources for batch deflation.
Code is tested to produce the same results (up to solver precision) on my laptop. It also compiles on Tursa. Nelson is testing the batch-deflation speed-up and correctness on Tursa currently.
With CG solver this produces bit identical perambulators. We also get good agreement using mixed precision solver on tursa and tesseract. ready to merge.
Two parts of this pull request: 1) getReference -> getObjectDependencies: This fixes the currently broken distillation Noise code, which is not correctly handled by the scheduler 2) re-organisation of the perambulator module to batch sources for batch deflation.
Code is tested to produce the same results (up to solver precision) on my laptop. It also compiles on Tursa. Nelson is testing the batch-deflation speed-up and correctness on Tursa currently.