// @author: Bert Maher <bertrand@fb.com>
-#include <thread>
+#include <folly/ProducerConsumerQueue.h>
+
+#include <cstdio>
#include <iostream>
-#include <stdio.h>
-#include <pthread.h>
+#include <thread>
+
+#include <glog/logging.h>
#include <folly/Benchmark.h>
-#include <folly/ProducerConsumerQueue.h>
#include <folly/portability/GFlags.h>
+#include <folly/portability/PThread.h>
#include <folly/stats/Histogram.h>
#include <folly/stats/Histogram-defs.h>
-#include <glog/logging.h>
namespace {
typedef unsigned long LatencyType;
typedef ProducerConsumerQueue<LatencyType> LatencyQueueType;
-template<class QueueType>
+template <class QueueType>
struct ThroughputTest {
explicit ThroughputTest(size_t size, int iters, int cpu0, int cpu1)
: queue_(size),
QueueType queue_;
std::atomic<bool> done_;
- int iters_;
+ const int iters_;
int cpu0_;
int cpu1_;
};
-template<class QueueType>
+template <class QueueType>
struct LatencyTest {
explicit LatencyTest(size_t size, int iters, int cpu0, int cpu1)
: queue_(size),
QueueType queue_;
std::atomic<bool> done_;
int time_cost_;
- int iters_;
+ const int iters_;
int cpu0_;
int cpu1_;
Histogram<int> hist_;
BENCHMARK_PARAM(BM_ProducerConsumerAffinity, 1048574);
BENCHMARK_PARAM(BM_ProducerConsumerLatency, 1048574);
-}
+} // namespace
int main(int argc, char** argv) {
google::InitGoogleLogging(argv[0]);
#if 0
/*
-Benchmark on Intel(R) Xeon(R) CPU E5-2660 0 @ 2.20GHz
-Latency histogram:
- log(nsec)
- min max count
- 6 7 5124
- 7 8 4799
- 8 9 49
- 9 10 2
- 10 11 1
- 11 12 5
- 12 13 3
- 13 14 9
- 14 15 8
+Benchmark
+
+$ lscpu
+Architecture: x86_64
+CPU op-mode(s): 32-bit, 64-bit
+Byte Order: Little Endian
+CPU(s): 24
+On-line CPU(s) list: 0-23
+Thread(s) per core: 1
+Core(s) per socket: 1
+Socket(s): 24
+NUMA node(s): 1
+Vendor ID: GenuineIntel
+CPU family: 6
+Model: 60
+Model name: Intel Core Processor (Haswell, no TSX)
+Stepping: 1
+CPU MHz: 2494.244
+BogoMIPS: 4988.48
+Hypervisor vendor: KVM
+Virtualization type: full
+L1d cache: 32K
+L1i cache: 32K
+L2 cache: 4096K
+NUMA node0 CPU(s): 0-23
+
+$ ../buck-out/gen/folly/test/producer_consumer_queue_benchmark
+5 6 1 5
+6 7 1893 11358
+7 8 39671 277697
+8 9 34921 279368
+9 10 17799 160191
+10 11 3685 36850
+11 12 1075 11825
+12 13 456 5472
+13 14 422 5486
+14 15 64 896
+15 16 7 105
+16 17 3 48
+17 18 3 51
============================================================================
folly/test/ProducerConsumerQueueBenchmark.cpp relative time/iter iters/s
============================================================================
----------------------------------------------------------------------------
-BM_ProducerConsumer(1048574) 7.52ns 132.90M
-BM_ProducerConsumerAffinity(1048574) 8.28ns 120.75M
-BM_ProducerConsumerLatency(1048574) 10.00s 99.98m
+BM_ProducerConsumer(1048574) 5.82ns 171.75M
+BM_ProducerConsumerAffinity(1048574) 7.36ns 135.83M
+BM_ProducerConsumerLatency(1048574) 1.67min 9.99m
============================================================================
*/
#endif