Re: I have thought more about my scalable FIFO MPMC priority queue

Re: I have thought more about my scalable FIFO MPMC priority queue

[optikos]

On Thursday, June 27, 2013 11:46:56 AM UTC-5, aminer wrote:
> there is a requirement,
> it's mandatory to start the same number of consumer threads than the 
> number of producer threads than the number of queues and cores that you 
> will pass to the constructor.

In the school of thought of making lemonade out of the lemons of this seemingly never-ending monologue, we could utilize Anime's monologue about Anime's library design in Pascal to be a running Ada-centric constructive critique of what a quite-different library would look like in Ada.

Although Amine will never read this reply to Anime's never-ending monologue regarding Pascal, for posterity if anyone ever designs anything like Anime's library for Ada, any design that imposes any requirements on the quantity consumer threads other than "There shall sometimes be at least one consumer thread." and/or that imposes any requirements on the quantity of producer threads other than "There shall sometimes be at least one producer thread." is deeply flawed.  The effective "pool" of producer threads should be of any nonzero size (and even permitted to be transiently zero-sized in some applications).  The effective "pool" of consumer threads should be of any nonzero size (and even permitted to be transiently zero-sized in some applications).  The effective "pool" of producer threads might or might not be disjoint from the effective "pool" of consumer threads.  These effective "pools" (from the scheduler's shorter-time-window dynamic-behavior perspective) might or might in fact not be Ada pools or Ada subpools (from a longer-timeline resource-allocation perspective).