Next: Functions
Up: Parallel file
Previous: Parallel file
A parallel file is initialized by a series of required and optional
parameters. The following are all required to initialize a parallel file.
- name
- The name of the sepfile.
- tag
- The tag used by the serial program (e.g. < , >, tag2=).
- usage
- How the file is to be used as input (INPUT) or output (OUTPUT).
- isect_dict
- A python dictionary object that links the job descriptor
with the section number it corresponds to.
In addition there several optional parameters available.
- type
- The distribution method for the file. Currently three options are
amiable:
- BLOCK,
- different elements of the distributed axis are grouped together;
- SPREAD,
- elements along the axis are distributing in a round robin fashion; or
- COPY,
- the file is shared among all of the sections. If input, the
file will be copied to all the nodes. If output, the sectioned files
were added together to form the final output. The file will also be
shared among all processes running on the node. In order to avoid
multiple processes overwriting themselves a simple file locking
mechanism is available (see the programming section).
- axis
- The axis the data is distributing along. Required for all distribution
methods except COPY.
- reuse_par
- A parameter string to include when the file is being reused.
This is only applicable in the case of COPY distribution type.
Next: Functions
Up: Parallel file
Previous: Parallel file
Stanford Exploration Project
10/23/2004