[Shootout-list] Threads and Threads-Flow Proposal

[Bengt Kleberg]

Isaac Gouy wrote:
...deleted
> *Someone* needs to compare the relative performance between languages
> at 3000 threads and 300 threads and 30 threads - if the relative
> performance doesn't change then the "improvement" is neutral. If the
> relative performance changes then we need to understand why, so we can
> decide if it's 3000 or 300 or 30 that really do what we need.

are you saying that you know that the current value is ''the right 
one'', that gives the correct relative performance?
and that it is therefore up to the person that suggests a change to 
prove that a change is not harmful?


for my now removed precursors to these tests (ie, they tested the same 
mechanisms (threads and message passing), but in another way) i worked 
on this problem. my solution was to measure at different numbers of 
threads. unfortunatly this was impossible, given the shootouts very 
narrow focus.


sometimes it seems to me as if the very narrow focus creates problems, 
instead of solving them.


bengt