
28 Nov
2007
28 Nov
'07
7:29 p.m.
Brad Clow:
When I (deeply) force the worker thread's results to be strict, I observe both cores working, but the execution time (elapsed) slower.
I can only speculate, but since you emphasise deep forcing, I wonder how deep is the structure returned from the worker thread? Could it be deep enough to hurt CPU cache performance (compared to an unforced version where production and consumption are interleaved)?