Sjors-
I have put together 2 separate data sets that had been refined to a 3D map and shiny ptcls were made, in order to do
a final 3D refinement that may provide a more isotropic map. Again, I took each half data set with its shiny.star file
and did a refinement on the large cluster and it worked fine, so there are no bad files, and the two half datasets
are matched up in box size, pixel size etc, were ctf corrected in the same way.
I made a merged star file with new groups.
The first time I run the combined data set it gives a core dump and a Segmentation fault but no real information
on why, even though I have at least 2x the memory of the half data set refinements. In fact in the last run
I had about 64 Gb RAM per MPI process and it crashed at the same point.
The job always gets through the step of estimating the noise and then starts on Expectation iteration 1 and writes the
itr_000 files and then crashes. I have tried so many different permutations and can still find nothing wrong
with any of the data.
If I restart I get the error below that I sent in before, but in looking at the code I have no idea what file cannot
be read / does not exist.
Especially since both half data sets refine just fine, so no missing data files.
No idea how to proceed to get past this problem and get a final map.
The combined data set has 76K ptcls in 320 boxes and the half data sets ran fine with 32 Gb of RAM per MPI process
available, so is this still a memory issue?
cat r9-6ct0.err
Cannot read file It does not exist
File: ./src/image.h line: 1040
--------------------------------------------------------------------------
MPI_ABORT was invoked on rank 2 in communicator MPI_COMM_WORLD
with errorcode 1.
C Akey
|