cf.COLLAPSE_PARALLEL_MODE¶
-
cf.
COLLAPSE_PARALLEL_MODE
(*arg)[source]¶ Which mode to use when collapse is run in parallel. There are three possible modes:
- This attempts to maximise parallelism, possibly at the expense of extra communication. This is the default mode.
- This minimises communication, possibly at the expense of the degree of parallelism. If collapse is running slower than you would expect, you can try changing to mode 1 to see if this improves performance. This is only likely to work if the output of collapse will be a sizeable array, not a single point.
- This is here for debugging purposes, but we would expect this to maximise communication possibly at the expense of parallelism. The use of this mode is, therefore, not recommended.
Parameters: - arg:
int
, optional The new value (0, 1 or 2).
Returns: int
The value prior to the change, or the current value if no new value was specified.
Examples:
>>> cf.COLLAPSE_PARALLEL_MODE() 0 >>> cf.COLLAPSE_PARALLEL_MODE(1) 0 >>> cf.COLLAPSE_PARALLEL_MODE() 1