Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

ANN bench fix latency measurement overhead #2084

Merged
merged 5 commits into from
Jan 24, 2024
+8 −10
Merged

ANN bench fix latency measurement overhead #2084

Changes from 4 commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
508c57e
Decrease latency measurement overhead
tfeher Jan 9, 2024
6efd970
Fix timer start, use kAvgIteratins for latency
tfeher Jan 23, 2024
ec33ce6
Update cpp/bench/ann/src/common/benchmark.hpp
achirkin Jan 24, 2024
1b3c8ae
Merge branch 'branch-24.02' into ann_bench_timers
achirkin Jan 24, 2024
6241439
Fix copyright year
tfeher Jan 24, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
16 changes: 7 additions & 9 deletions cpp/bench/ann/src/common/benchmark.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -287,11 +287,11 @@ void bench_search(::benchmark::State& state,
std::make_shared<buf<std::size_t>>(current_algo_props->query_memory_type, k * query_set_size);

cuda_timer gpu_timer;
auto start = std::chrono::high_resolution_clock::now();
{
nvtx_case nvtx{state.name()};

auto algo = dynamic_cast<ANN<T>*>(current_algo.get())->copy();
auto algo = dynamic_cast<ANN<T>*>(current_algo.get())->copy();
auto start = std::chrono::high_resolution_clock::now();
for (auto _ : state) {
[[maybe_unused]] auto ntx_lap = nvtx.lap();
[[maybe_unused]] auto gpu_lap = gpu_timer.lap();
Expand All @@ -314,17 +314,15 @@ void bench_search(::benchmark::State& state,

queries_processed += n_queries;
}
auto end = std::chrono::high_resolution_clock::now();
auto duration = std::chrono::duration_cast<std::chrono::duration<double>>(end - start).count();
if (state.thread_index() == 0) { state.counters.insert({{"end_to_end", duration}}); }
state.counters.insert({"Latency", {duration, benchmark::Counter::kAvgIterations}});
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I have also applied Artem's suggestion to store latency values with benchmark::Counter::kAvgIterations marker.

Earlier we manually divided by number of iterations and let gbench average over threads using kAvgThreads. Since iterations are counted as total iterations performed by all threads, using kAviIterations leads to the same results (without manual divisions by iterations).

}
auto end = std::chrono::high_resolution_clock::now();
auto duration = std::chrono::duration_cast<std::chrono::duration<double>>(end - start).count();
if (state.thread_index() == 0) { state.counters.insert({{"end_to_end", duration}}); }
state.counters.insert(
{"Latency", {duration / double(state.iterations()), benchmark::Counter::kAvgThreads}});

state.SetItemsProcessed(queries_processed);
if (cudart.found()) {
double gpu_time_per_iteration = gpu_timer.total_time() / (double)state.iterations();
state.counters.insert({"GPU", {gpu_time_per_iteration, benchmark::Counter::kAvgThreads}});
state.counters.insert({"GPU", {gpu_timer.total_time(), benchmark::Counter::kAvgIterations}});
}

// This will be the total number of queries across all threads
Expand Down
Loading