Creating a Sub-view of a non - RA (hasSlicing) range.

[Dmitry Olshansky]

On 30-Jun-12 18:25, Andrei Alexandrescu wrote:
> On 6/30/12 9:15 AM, Dmitry Olshansky wrote:
>> On 30-Jun-12 15:35, Tobias Pankrath wrote:
>>>> Say I have a forward range (for example, an SList[]). I would like to
>>>> create a new range containing only the first 5 elements of that old
>>>> range. How can I do that?
>>>>
>>>
>>> std.algorithm.take.
>>>
>>> But I would generally avoid SList.
>>
>> Indeed. I'd be hard pressed to devise realistic use case where it
>> performs better.
>
> Singly-linked lists are frequently used with lock-free algorithms.

As post involved terms like "generally" I pointed out that indeed List 
should not be used "generally" :) Their value in certain scenarios is 
remarkable, or rather of linked storage such as in free lists, and (like 
you mentioned) certain lock-free algorithms.

Let's not forget however that what people seek is lock-free queues and 
other high-level abstractions. S-lists just happen to be a means to an 
end that works on current hardware.

I totally expect a new wave of simpler lock-free algorithms with things 
like Intel's TSX and something analogous that AMD will have.
(others would either copy or redesign similar things as it's an obvious 
need for the future multicores)

There a lot of links but this gives a nice overview and is fairly new:
http://arstechnica.com/business/2012/02/transactional-memory-going-mainstream-with-intel-haswell/
> Generally it's simplistic to compare slists with arrays as to "which
> performs better" because they have different primitives.

Agreed. Though it's not at all uncommon to mix things up in awful ways. 
*cough* Java *cough*


-- 
Dmitry Olshansky