Benchmarks?

[dendisuhubdy]

Do you have numbers for benchmarks between other implementations for your MPMC queue?

[erez-strauss]

Hi Dendi, Good morning. Just saw your email, I'll send you multiple benchmark results tonight and add them to the web site. There are few points to notice: 1. Fully searielizable 2. Can work between processes 3. Provides access to sequence number of push/pop 4. Fixed size for speed, can be determined as template parameter or constructor parameter. 5. Limitations: small data types can fit up to 12 bytes, enough for pointer and index. Best regards, Erez +1 781 439 7937

…

On Mon, Nov 25, 2019, 07:15 Dendi Suhubdy ***@***.***> wrote: Do you have numbers for benchmarks between other implementations for your MPMC queue? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#1?email_source=notifications&email_token=ABH5IVWKC2XMIBAVOH4GJDDQVO6UPA5CNFSM4JRIADO2YY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4H3ZL4NA>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABH5IVSBUTAUBCCDGOV5OLTQVO6UPANCNFSM4JRIADOQ> .

[erez-strauss]

Hi Dendi, I used the benchmarks in the site: https://github.com/max0x7ba/atomic_queue It seems the my queue perform well, please see bellow. Regards, Erez

…

---- Running throughput benchmarks (higher is better) ---- boost::lockfree::spsc_queue, 1,s: 94,879,418 msg/sec boost::lockfree::spsc_queue, 1,i: 97,811,250 msg/sec boost::lockfree::queue, 1,s: 7,075,138 msg/sec boost::lockfree::queue, 1,i: 7,109,413 msg/sec boost::lockfree::queue, 2,s: 6,048,648 msg/sec boost::lockfree::queue, 2,i: 6,107,601 msg/sec boost::lockfree::queue, 3,s: 5,788,840 msg/sec boost::lockfree::queue, 3,i: 5,870,491 msg/sec es::lockfree::mpmc_queue<f,f>, 1,s: 41,906,117 msg/sec es::lockfree::mpmc_queue<f,f>, 1,i: 49,227,116 msg/sec es::lockfree::mpmc_queue<f,f>, 1,s: 50,219,724 msg/sec es::lockfree::mpmc_queue<f,f>, 1,i: 50,338,088 msg/sec es::lockfree::mpmc_queue<f,f>, 2,s: 21,006,724 msg/sec es::lockfree::mpmc_queue<f,f>, 2,i: 21,652,522 msg/sec es::lockfree::mpmc_queue<f,f>, 3,s: 13,689,788 msg/sec es::lockfree::mpmc_queue<f,f>, 3,i: 14,222,712 msg/sec es::lockfree::mpmc_queue<f,t>, 1,s: 42,043,597 msg/sec es::lockfree::mpmc_queue<f,t>, 1,i: 43,247,748 msg/sec es::lockfree::mpmc_queue<f,t>, 1,s: 43,408,077 msg/sec es::lockfree::mpmc_queue<f,t>, 1,i: 43,371,852 msg/sec es::lockfree::mpmc_queue<f,t>, 2,s: 10,491,380 msg/sec es::lockfree::mpmc_queue<f,t>, 2,i: 10,452,377 msg/sec es::lockfree::mpmc_queue<f,t>, 3,s: 8,374,590 msg/sec es::lockfree::mpmc_queue<f,t>, 3,i: 8,636,852 msg/sec es::lockfree::mpmc_queue<t,f>, 1,s: 43,342,007 msg/sec es::lockfree::mpmc_queue<t,f>, 1,i: 40,730,969 msg/sec es::lockfree::mpmc_queue<t,f>, 1,s: 42,999,163 msg/sec es::lockfree::mpmc_queue<t,f>, 1,i: 42,331,340 msg/sec es::lockfree::mpmc_queue<t,f>, 2,s: 13,216,232 msg/sec es::lockfree::mpmc_queue<t,f>, 2,i: 12,940,071 msg/sec es::lockfree::mpmc_queue<t,f>, 3,s: 8,918,911 msg/sec es::lockfree::mpmc_queue<t,f>, 3,i: 9,063,264 msg/sec es::lockfree::mpmc_queue<t,t>, 1,s: 41,980,251 msg/sec es::lockfree::mpmc_queue<t,t>, 1,i: 42,459,030 msg/sec es::lockfree::mpmc_queue<t,t>, 1,s: 42,229,862 msg/sec es::lockfree::mpmc_queue<t,t>, 1,i: 33,724,725 msg/sec es::lockfree::mpmc_queue<t,t>, 2,s: 11,530,558 msg/sec es::lockfree::mpmc_queue<t,t>, 2,i: 12,865,424 msg/sec es::lockfree::mpmc_queue<t,t>, 3,s: 9,680,444 msg/sec es::lockfree::mpmc_queue<t,t>, 3,i: 10,088,071 msg/sec pthread_spinlock, 1,s: 17,019,862 msg/sec pthread_spinlock, 1,i: 16,972,283 msg/sec pthread_spinlock, 2,s: 13,216,552 msg/sec pthread_spinlock, 2,i: 15,108,251 msg/sec pthread_spinlock, 3,s: 8,572,469 msg/sec pthread_spinlock, 3,i: 8,111,074 msg/sec moodycamel::ReaderWriterQueue, 1,s: 524,768,600 msg/sec moodycamel::ReaderWriterQueue, 1,i: 525,209,331 msg/sec moodycamel::ConcurrentQueue, 1,s: 19,944,087 msg/sec moodycamel::ConcurrentQueue, 1,i: 17,440,607 msg/sec moodycamel::ConcurrentQueue, 2,s: 9,402,469 msg/sec moodycamel::ConcurrentQueue, 2,i: 9,313,832 msg/sec moodycamel::ConcurrentQueue, 3,s: 9,907,164 msg/sec moodycamel::ConcurrentQueue, 3,i: 9,885,286 msg/sec tbb::spin_mutex, 1,s: 43,699,952 msg/sec tbb::spin_mutex, 1,i: 45,849,644 msg/sec tbb::spin_mutex, 2,s: 31,424,220 msg/sec tbb::spin_mutex, 2,i: 31,655,436 msg/sec tbb::spin_mutex, 3,s: 20,302,226 msg/sec tbb::spin_mutex, 3,i: 20,300,173 msg/sec tbb::speculative_spin_mutex, 1,s: 1,158,635 msg/sec tbb::speculative_spin_mutex, 1,i: 1,640,915 msg/sec tbb::speculative_spin_mutex, 2,s: 26,297,905 msg/sec tbb::speculative_spin_mutex, 2,i: 26,999,266 msg/sec tbb::speculative_spin_mutex, 3,s: 19,215,439 msg/sec tbb::speculative_spin_mutex, 3,i: 17,928,715 msg/sec tbb::concurrent_bounded_queue, 1,s: 9,623,481 msg/sec tbb::concurrent_bounded_queue, 1,i: 9,829,076 msg/sec tbb::concurrent_bounded_queue, 2,s: 12,395,821 msg/sec tbb::concurrent_bounded_queue, 2,i: 12,357,749 msg/sec tbb::concurrent_bounded_queue, 3,s: 9,992,534 msg/sec tbb::concurrent_bounded_queue, 3,i: 9,729,767 msg/sec AtomicQueue, 1,s: 45,046,351 msg/sec AtomicQueue, 1,i: 45,631,194 msg/sec AtomicQueue, 2,s: 9,660,551 msg/sec AtomicQueue, 2,i: 9,411,904 msg/sec AtomicQueue, 3,s: 7,199,061 msg/sec AtomicQueue, 3,i: 7,217,304 msg/sec AtomicQueueB, 1,s: 38,497,614 msg/sec AtomicQueueB, 1,i: 38,168,571 msg/sec AtomicQueueB, 2,s: 9,038,393 msg/sec AtomicQueueB, 2,i: 8,800,752 msg/sec AtomicQueueB, 3,s: 7,142,461 msg/sec AtomicQueueB, 3,i: 6,970,981 msg/sec OptimistAtomicQueue, 1,s: 528,869,800 msg/sec OptimistAtomicQueue, 1,i: 551,472,984 msg/sec OptimistAtomicQueue, 2,s: 28,805,222 msg/sec OptimistAtomicQueue, 2,i: 29,170,093 msg/sec OptimistAtomicQueue, 3,s: 30,603,599 msg/sec OptimistAtomicQueue, 3,i: 30,253,940 msg/sec OptimistAtomicQueueB, 1,s: 39,230,915 msg/sec OptimistAtomicQueueB, 1,i: 41,735,853 msg/sec OptimistAtomicQueueB, 2,s: 15,855,172 msg/sec OptimistAtomicQueueB, 2,i: 16,442,681 msg/sec OptimistAtomicQueueB, 3,s: 17,087,497 msg/sec OptimistAtomicQueueB, 3,i: 17,179,175 msg/sec AtomicQueue2, 1,s: 46,846,907 msg/sec AtomicQueue2, 1,i: 44,904,326 msg/sec AtomicQueue2, 2,s: 9,330,049 msg/sec AtomicQueue2, 2,i: 9,253,054 msg/sec AtomicQueue2, 3,s: 7,381,856 msg/sec AtomicQueue2, 3,i: 7,301,297 msg/sec AtomicQueueB2, 1,s: 34,568,239 msg/sec AtomicQueueB2, 1,i: 34,640,445 msg/sec AtomicQueueB2, 2,s: 8,209,669 msg/sec AtomicQueueB2, 2,i: 8,092,231 msg/sec AtomicQueueB2, 3,s: 6,799,506 msg/sec AtomicQueueB2, 3,i: 6,415,026 msg/sec OptimistAtomicQueue2, 1,s: 463,135,274 msg/sec OptimistAtomicQueue2, 1,i: 379,537,674 msg/sec OptimistAtomicQueue2, 2,s: 21,497,154 msg/sec OptimistAtomicQueue2, 2,i: 19,282,535 msg/sec OptimistAtomicQueue2, 3,s: 24,191,905 msg/sec OptimistAtomicQueue2, 3,i: 23,956,296 msg/sec OptimistAtomicQueueB2, 1,s: 28,988,530 msg/sec OptimistAtomicQueueB2, 1,i: 28,985,125 msg/sec OptimistAtomicQueueB2, 2,s: 11,995,527 msg/sec OptimistAtomicQueueB2, 2,i: 11,812,101 msg/sec OptimistAtomicQueueB2, 3,s: 15,137,048 msg/sec OptimistAtomicQueueB2, 3,i: 14,708,241 msg/sec

---- Running ping-pong benchmarks (lower is better) ---- boost::lockfree::spsc_queue: 0.000000178 sec/round-trip boost::lockfree::queue: 0.000000404 sec/round-trip es::lockfree::mpmc_queue<f,f>: 0.000000184 sec/round-trip es::lockfree::mpmc_queue<f,t>: 0.000000185 sec/round-trip es::lockfree::mpmc_queue<t,f>: 0.000000229 sec/round-trip es::lockfree::mpmc_queue<t,t>: 0.000000223 sec/round-trip pthread_spinlock: 0.000000402 sec/round-trip moodycamel::ReaderWriterQueue: 0.000000170 sec/round-trip moodycamel::ConcurrentQueue: 0.000000314 sec/round-trip tbb::spin_mutex: 0.000000374 sec/round-trip tbb::speculative_spin_mutex: 0.000051170 sec/round-trip tbb::concurrent_bounded_queue: 0.000000382 sec/round-trip AtomicQueue: 0.000000170 sec/round-trip AtomicQueueB: 0.000000184 sec/round-trip OptimistAtomicQueue: 0.000000183 sec/round-trip OptimistAtomicQueueB: 0.000000211 sec/round-trip AtomicQueue2: 0.000000187 sec/round-trip AtomicQueueB2: 0.000000208 sec/round-trip OptimistAtomicQueue2: 0.000000225 sec/round-trip OptimistAtomicQueueB2: 0.000000231 sec/round-trip

On Mon, Nov 25, 2019 at 8:09 AM Erez Strauss ***@***.***> wrote: Hi Dendi, Good morning. Just saw your email, I'll send you multiple benchmark results tonight and add them to the web site. There are few points to notice: 1. Fully searielizable 2. Can work between processes 3. Provides access to sequence number of push/pop 4. Fixed size for speed, can be determined as template parameter or constructor parameter. 5. Limitations: small data types can fit up to 12 bytes, enough for pointer and index. Best regards, Erez +1 781 439 7937 On Mon, Nov 25, 2019, 07:15 Dendi Suhubdy ***@***.***> wrote: > Do you have numbers for benchmarks between other implementations for your > MPMC queue? > > — > You are receiving this because you are subscribed to this thread. > Reply to this email directly, view it on GitHub > <#1?email_source=notifications&email_token=ABH5IVWKC2XMIBAVOH4GJDDQVO6UPA5CNFSM4JRIADO2YY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4H3ZL4NA>, > or unsubscribe > <https://github.com/notifications/unsubscribe-auth/ABH5IVSBUTAUBCCDGOV5OLTQVO6UPANCNFSM4JRIADOQ> > . >