Re: Generic Embedded List Nodes

[Warren <ve3wwg@gmail.com>]

On Wednesday, 22 June 2016 20:37:08 UTC-4, Randy Brukardt  wrote:
...
> The bounded containers are designed for the purpose of low-overhead 
> operations; in particular, they don't do allocation/deallocation of memory 
> for individual objects. You ought to check out whether those are high enough 
> performance for your purposes before reinventing the wheel... (especially 
> using the check suppression implemented in the latest GNAT versions).

I just posted (to Niklas) what will likely be the solution- a large array indexed by socket number. That is prolly the cleanest way I can do this cheap, at the expense of a little bit of memory.

> (Most programmers, myself included, are terrible at determining what matters 
> for performance of a particular application. The only way to be sure that 
> something is too slow is to try it...)
> 
>                            Randy.

Agreed. In my day job (where we try lots of things), I do benchmarks on the C++ servers I write and maintain.  Experience shows that it takes very little to slow down the transaction rate and there are often surprises.  

Here though, in Ada, I'm just playing around with a version of it as a hobby horse. It would be nice if I could someday compare the Ada version against the C++ version. So I'm just chipping away at this.

Warren