[Shootout-list] ring of processes

Bengt Kleberg bengt.kleberg@ericsson.com
Mon, 04 Oct 2004 17:19:34 +0200


Brent Fulgham wrote:
> On 2004-10-04 05:48:12 -0700 Bengt Kleberg <bengt.kleberg@ericsson.com> 
> wrote:
> 

deleted...

>> i suggest we use a ratio of 128 messages to each process, until 
>> another language show us that we need to change it upwards.
>> in the test we create n number of process and send 128*n messages.
> 
> 
> Fine.

let me be even more specific here (i made a mistake in my code as 
perhaps somebody already have seen). since we send messages between all 
processes we only need to initiate sending of 128 messages from the 
master/controller process. (with the high/low prio that means 64 of each)
128*n messages are then sent in total between all the processes.

...deleted

> 
> What about sending randomly within the ring?  These would be like 
> 'spokes' on the
> ring -- a send from P1 to P100 and P5 to P72, for example.

i know we are talking about pseudo random, but i still think it would be 
better with a deterministic sending algorithm.

if we had both ways around the ring, would it, in a haskell program, 
still be silly to look at the sender?


bengt