Automated benchmarks of boost::concurrent_flat_map and boost::concurrent_node_map against oneapi::tbb::concurrent_hash_map and gtl::parallel_flat_hash_map.
parallel_load: T threads concurrently perform N operations update, successful lookup and unsuccessful lookup, randomly chosen with probabilities 10%, 45% and 45%, respectively, on a concurrent map of (int,int) pairs. The keys used by all operations are also random, where update and successful lookup follow a Zipf distribution over [1, N/10] with skew exponent s, and unsuccessful lookup follows a Zipf distribution with the same skew s over [1 + N/10, 2N/10] (so, not overlapping with the former interval).
boost::concurrent_flat_mapandboost::concurrent_node_mapare exercised using both regular and bulk visitation: in the latter case, lookup keys are buffered in a local array and then processed at once each time the buffer reachesbulk_visit_size.
gcc-x64: GCC 12, x64clang-x64: Clang 15, x64vs-x64: Visual Studio 2022, x64clang-arm64: Clang 12, ARM64gcc-x86: GCC 12, x86clang-x86: Clang 15, x86vs-x86: Visual Studio 2022, x86
 |
 |
 |
|---|---|---|
| 500k updates, 4.5M lookups skew=0.01 |
500k updates, 4.5M lookups skew=0.5 |
500k updates, 4.5M lookups skew=0.99 |
 |
 |
 |
|---|---|---|
| 5M updates, 45M lookups skew=0.01 |
5M updates, 45M lookups skew=0.5 |
5M updates, 45M lookups skew=0.99 |
 |
 |
 |
|---|---|---|
| 500k updates, 4.5M lookups skew=0.01 |
500k updates, 4.5M lookups skew=0.5 |
500k updates, 4.5M lookups skew=0.99 |
 |
 |
 |
|---|---|---|
| 5M updates, 45M lookups skew=0.01 |
5M updates, 45M lookups skew=0.5 |
5M updates, 45M lookups skew=0.99 |
 |
 |
 |
|---|---|---|
| 500k updates, 4.5M lookups skew=0.01 |
500k updates, 4.5M lookups skew=0.5 |
500k updates, 4.5M lookups skew=0.99 |
 |
 |
 |
|---|---|---|
| 5M updates, 45M lookups skew=0.01 |
5M updates, 45M lookups skew=0.5 |
5M updates, 45M lookups skew=0.99 |
 |
 |
 |
|---|---|---|
| 500k updates, 4.5M lookups skew=0.01 |
500k updates, 4.5M lookups skew=0.5 |
500k updates, 4.5M lookups skew=0.99 |
 |
 |
 |
|---|---|---|
| 5M updates, 45M lookups skew=0.01 |
5M updates, 45M lookups skew=0.5 |
5M updates, 45M lookups skew=0.99 |
 |
 |
 |
|---|---|---|
| 500k updates, 4.5M lookups skew=0.01 |
500k updates, 4.5M lookups skew=0.5 |
500k updates, 4.5M lookups skew=0.99 |
 |
 |
 |
|---|---|---|
| 5M updates, 45M lookups skew=0.01 |
5M updates, 45M lookups skew=0.5 |
5M updates, 45M lookups skew=0.99 |
 |
 |
 |
|---|---|---|
| 500k updates, 4.5M lookups skew=0.01 |
500k updates, 4.5M lookups skew=0.5 |
500k updates, 4.5M lookups skew=0.99 |
 |
 |
 |
|---|---|---|
| 5M updates, 45M lookups skew=0.01 |
5M updates, 45M lookups skew=0.5 |
5M updates, 45M lookups skew=0.99 |
 |
 |
 |
|---|---|---|
| 500k updates, 4.5M lookups skew=0.01 |
500k updates, 4.5M lookups skew=0.5 |
500k updates, 4.5M lookups skew=0.99 |
 |
 |
 |
|---|---|---|
| 5M updates, 45M lookups skew=0.01 |
5M updates, 45M lookups skew=0.5 |
5M updates, 45M lookups skew=0.99 |