Hi, I want to do a parallel-for-loop (with OpenMP), iterating over the contents of a (concurrent) hash map.
In std::unordered_set one cannot write the following, since parallel-for needs a random-access iterator.
#pragma omp parallel for
for (auto m : current) { ... }
Instead, one can write the following, which is good enough:
#pragma omp parallel for
for (size_t b=0; b<current.bucket_count(); b++)
for (auto m=current.begin(b); m!=current.end(b); m++) { ... }
Here the outer parallel-loop iterates over all buckets (random access), while the inner, sequential loop iterates over all elements in a given bucket.
Problem: it seems I cannot do the same in libcuckoo. I can get a sequential iterator over a locked_table, but this wouldn't have random access.
I couldn't find a way to get access to and iterate over the elements in a particular given bucket b; (i.e., current.begin(b) doesn't work, even not if current is locked_table).
Is there a way to achieve this? Or is it easy to extend the libcuckoo API with this functionality?
Hi, I want to do a parallel-for-loop (with OpenMP), iterating over the contents of a (concurrent) hash map.
In
std::unordered_setone cannot write the following, since parallel-for needs a random-access iterator.Instead, one can write the following, which is good enough:
Here the outer parallel-loop iterates over all buckets (random access), while the inner, sequential loop iterates over all elements in a given bucket.
Problem: it seems I cannot do the same in libcuckoo. I can get a sequential iterator over a
locked_table, but this wouldn't have random access.I couldn't find a way to get access to and iterate over the elements in a particular given bucket
b; (i.e.,current.begin(b)doesn't work, even not ifcurrentislocked_table).Is there a way to achieve this? Or is it easy to extend the libcuckoo API with this functionality?