Re: question about understanding/exploring agents

2009-01-16 Thread Timothy Pratley
> So I guess two mysteries solved with one patch! Actually I spoke too soon... the mt2.clj is still giving the 'unexpected' behavior... I am still mystified. But going to bed now :) --~--~-~--~~~---~--~~ You received this message because you are subscribed to the

Re: question about understanding/exploring agents

2009-01-16 Thread Timothy Pratley
Sorry for the quad-post, but I just noticed: http://groups.google.com/group/clojure/browse_thread/thread/6ac273446b4a7183 And actually that is what made the big difference in performance (I'd done a svn update for something unrelated). So I guess two mysteries solved with one patch! Boris I sugges

Re: question about understanding/exploring agents

2009-01-16 Thread Timothy Pratley
On a hunch I rearranged such that the agents are not recreated for each run, but instead are reused from a pre-established set: http://groups.google.com/group/clojure/web/mt2.2.clj Strangely (from my understanding of agents) this has a significant impact on the result!!! threads: 1 "Elapsed tim

Re: question about understanding/exploring agents

2009-01-16 Thread Timothy Pratley
> goes up by only %15 for 3 threads... I would expect it to go up by %50 Actually I would expect it to go up by %100 t1t2 t1 --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Clojure" group. To post to this grou

Re: question about understanding/exploring agents

2009-01-16 Thread Timothy Pratley
> There are two pools of threads servicing agent actions. The send pool   > is fixed in size and based on the number of available cores. The send- > off pool is variable in size and grows as needed to accommodate the   > largest number of simultaneous pending send-off calls that your   > program p

Re: question about understanding/exploring agents

2009-01-15 Thread Stephen C. Gilardi
On Jan 15, 2009, at 9:26 AM, bOR_ wrote: I think I know when to use one and when to use the other, but the extra clarification doesn't hurt. However, what happens if you get it wrong? use send where you should have used send-off, and visa versa? I would like to know what they do differently.

Re: question about understanding/exploring agents

2009-01-15 Thread bOR_
I think I know when to use one and when to use the other, but the extra clarification doesn't hurt. However, what happens if you get it wrong? use send where you should have used send-off, and visa versa? I would like to know what they do differently. On 15 jan, 13:12, "Stephen C. Gilardi" wrote

Re: question about understanding/exploring agents

2009-01-15 Thread Stephen C. Gilardi
On Jan 15, 2009, at 2:57 AM, bOR_ wrote: That is, if I understand blocking correctly. Currently assuming that blocking only happens when two things would like to write the same ref? Blocking in this case refers to this definition: http://en.wikipedia.org/wiki/Blocking_(computing) You should

Re: question about understanding/exploring agents

2009-01-14 Thread bOR_
Hi Timothy, that is indeed what I was surprised about. I have been reading up on agents (using the forum and the website), and one thing that I noticed is that rich is very brief on when to use send vs send-off. I tried figuring out what the difference is by sourcecode, but it just seems to call

Re: question about understanding/exploring agents

2009-01-14 Thread Timothy Pratley
Hi Chris > What exactly are you trying to measure? I think what Boris is expecting is that for 4 CPUs, running 1,2,3,4 equal work threads will take the same amount of time. This is true when he calls loopfib, but not true when he calls loopmult: threads: 1 "Elapsed time: 205.458949 msecs" "Ela

Re: question about understanding/exploring agents

2009-01-14 Thread Chouser
On Wed, Jan 14, 2009 at 6:00 AM, bOR_ wrote: > > However, on this 4core machine, the fib behaves as I would expect when > I scale up the number of threads, while the multiplication barely > seems to benefit from the 4 cores. The only difference I see between > the two functions is that the fib is

question about understanding/exploring agents

2009-01-14 Thread bOR_
Hi all. I am trying to figure out what the effect of the agent-function is on the efficiency of concurrency. Here is something I do not really understand. I've a fibonacci function and a simple multiplication, both are wrapped in their respective dotimes 100k loop. However, on this 4core machine