[concurrency-interest] ConcurrentHashMapV8 now supports Spliterator
aleksandar.prokopec at gmail.com
Wed Jul 4 12:40:43 EDT 2012
Indeed - this is true. The cost of maintaining the order during `map`,
`filter` and friends is high.
It's still good to hear that splitters will be included.
I was thinking of an application where you just do lookups in the
concurrent linked lists most of the time, with occasional inserts and
occasional parallel reduce methods (which do not construct a new linked
list, but just traverse the existing one to obtain the reduction).
On 7/4/12 1:29 PM, Doug Lea wrote:
> On 07/04/12 04:20, Aleksandar Prokopec wrote:
>> Wow, this is great news!
>> It may allow us to write a wrapper for chmv8 in Scala parallel
>> collections. I
>> wish concurrent skip lists had a similar thing.
> Yes, adding the new functionality of ConcurrentHashMap to
> ConcurrentSkipListMap is on the todo list. ConcurrentHashMap
> will almost always be a lot faster for parallelism though.
> Performing parallel operations on collections that maintain
> ordering (which is basically ignored for purposes of parallelism)
> is usually more costly than just using a hash table (or array)
> and then sorting (in parallel) later when the ordering is needed.
> Concurrency-interest mailing list
> Concurrency-interest at cs.oswego.edu
More information about the Concurrency-interest