diamondlightsource / savu Goto Github PK

the astra import statement seems to hang when running in mpi

line 213 in utils - need to remove squeeze from result (wrt SWTP single frame sinogram centre of mass/alignment correction). It needs the extra dimension in the result even though only 1 (i.e. single slice)

There are a series of processes, which need to be updated to the new Plugin runner test style

There are quite a few things which could do with a refactor.

1. hard coded strings should be dealt with as statics
2. there are some places where there are multiple locations in the NXtomo where things can be found, they should all be tested, as some files don't contain everything they need.

Currently it is failing as the system has changed, Mark needs to understand more about how to use the experiment

After using pandas, a text log will be fine to process.

In GpuPlugin the run_process(...) function looks in "processes" for "GPU". This is currently fixed to be "CPU0" so it doesn't run the plugin (I hope that's a little clearer...). I don't know what your plans are for developing this?

So grepping about we are definatly getting some threads wrong. there are also some odd timing issues. A little annoying as its a transitory problem, which makes me think more barriers is needed

Should just e a case of testing the downsample plugin with the new setup process.

Some example data from Steve would be good here

want to return new centre of rotation value not just the frame from sinogram alignment

Should make it possible to make savu run in windows, and also make the Travis build quicker.

i.e in savu.plugins there should probably be
savu.plugins.filters
savu.plugins.reconstruction
and any others which are required.

There are links to local data in the test, causing it to fail, Disabled for now.

create a new CGLS plugin which uses the multithreaded code options.

should extent AllCpusPlugin and make use of
self.availble_cpus

Would be nice to be able to say after pip installing savu

import savu
savu.full_test()

and it run all the tests properly

So point out something like setting the PYTHONPATH and then using python tomo_recon.py ...

the dark and flatfield plugin seems to work incorrectly over MPI. if I were to guess the first part of the flatfield is applied to all sections of the projections, there is a tell tale issue with the top left of each part which is the timestamp on the images.

To sub-sample the data

Should be able to state that either all datasets should be used in a plugin or a plugin should be applied to all datasets without specifying them individually

There is an import error here, looks like it should be a simple fix

Should allow you to load, save, add and remove items, and modify parameters.

There seems to be an issue when generating the initial size of the hdf5 file associated with the paganin step, not sure whether this is not passing the right information though, or why this is happening.

So this only runs in one thread on CPU0, but tells the process how many threads it can use.