29 Dec
2007
29 Dec
'07
2:56 a.m.
Tim Docker wrote:
I found it worthwhile to try and visualise what's going on here. Let's say I have 4 calculations that I want to run in parallel. The first doesn't need a request; the second needs to make a single request (A1); the third needs to make two requests where the second (B2) depends on the result of the first (B1), etc. The resulting parallel operations will be done in 3 batches, looking like:
This sounds similar to "futures", where a request for a parallel computation returns immediately, but the "value" returned is just a placeholder for the result, which will be filled in when it becomes available. Paul.