Adding data to a hashmap in parallel

Question

I've got an instance of an std::unordered_map serving as a database of some sort. (It's basically just a regular hashmap).
I have 4 threads retrieving long strings from a network (I'm using the c++11 std::threads) and within each of those threads, i tokenize the strings they receive. The tokenization itself is paralellized as well using the new Microsoft AMP library (so the process of tokenization is done by the GPU threads).

I would like each of the 4 "main" threads to add the tokens as keys into the hashmap as the GPU threads finish tokenizing the strings (and then repeat the process).

How would i go about synchronizing all of this so that i can safely add new keys into my database from each of the threads ?

Have a std::mutex at the same scope as your std::unordered_map<> and have each thread lock the mutex before touching the hash table. Honestly, though, if you're concerned about throughput, you'd probably be better off with a proper concurrent container rather than a manually synchronized one. — ildjarn
– ildjarn, Commented Apr 8, 2012 at 17:52
It's probably more efficient to have one hashmap per thread and then merge them pairwise concurrently as you finish up. — James Youngman
– James Youngman, Commented Apr 8, 2012 at 17:57

Rick · Accepted Answer · 2012-04-08 17:55:00Z

2

use concurrent_unordered_map instead of unordered_map in concurrent_unordered_map.h

answered Apr 8, 2012 at 17:55

Rick

3,35319 silver badges17 bronze badges

Sign up to request clarification or add additional context in comments.

Collectives™ on Stack Overflow

Adding data to a hashmap in parallel

1 Answer 1

Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related