Greeting, FYI, we noticed a 57.1% improvement of vm-scalability.throughput due to commit: commit: 4b922b23ce8026a6cdd79ecd57aaa515d8144f2a ("mm/swap: Split swap cache into 64MB trunks") git://bee.sh.intel.com/git/yhuang/linux.git swap_optimize_v4 in testcase: vm-scalability on test machine: 72 threads Intel(R) Xeon(R) CPU E5-2699 v3 @ 2.30GHz with 128G memory with following parameters: thp_enabled: never thp_defrag: never nr_task: 16 disk: 1pmem test: swap-w-seq unit_size: 96G size: 96G cpufreq_governor: performance test-description: The motivation behind this suite is to exercise functions and regions of the mm/ of the Linux kernel which are of interest to us. test-url: https://git.kernel.org/cgit/linux/kernel/git/wfg/vm-scalability.git/ In addition to that, the commit also has significant impact on the following tests: +------------------+-----------------------------------------------------------------------+ | testcase: change | vm-scalability: vm-scalability.throughput 57.6% improvement | | test machine | 72 threads Intel(R) Xeon(R) CPU E5-2699 v3 @ 2.30GHz with 128G memory | | test parameters | cpufreq_governor=performance | | | disk=1pmem | | | nr_task=32 | | | size=96G | | | test=swap-w-seq | | | thp_defrag=never | | | thp_enabled=never | | | unit_size=96G | +------------------+-----------------------------------------------------------------------+ Details are as below: --------------------------------------------------------------------------------------------------> To reproduce: git clone git://git.kernel.org/pub/scm/linux/kernel/git/wfg/lkp-tests.git cd lkp-tests bin/lkp install job.yaml # job file is attached in this email bin/lkp run job.yaml testcase/path_params/tbox_group/run: vm-scalability/never-never-16-1pmem-swap-w-seq-96G-96G-performance/lkp-hsw-ep4 91526a72b886b53e 4b922b23ce8026a6cdd79ecd57 ---------------- -------------------------- %stddev change %stddev \ | \ 2977834 57% 4677757 vm-scalability.throughput 2501424 ± 3% -5% 2368700 ± 5% vm-scalability.time.maximum_resident_set_size 95.11 -27% 69.01 vm-scalability.time.elapsed_time 95.11 -27% 69.01 vm-scalability.time.elapsed_time.max 1427 -28% 1028 vm-scalability.time.system_time 27580435 19% 32703586 interrupts.CAL:Function_call_interrupts 55.27 9% 60.14 turbostat.RAMWatt 176 5% 185 turbostat.PkgWatt 7320 ± 83% -6e+03 822 ± 57% latency_stats.sum.pipe_read.__vfs_read.vfs_read.SyS_read.entry_SYSCALL_64_fastpath 526 ± 27% 77% 928 ± 17% vmstat.swap.si 1430761 40% 1999866 vmstat.swap.so 360122 49% 535772 vmstat.system.in 4624 ± 10% 61% 7446 vmstat.system.cs 4.084e+08 ± 4% 69% 6.886e+08 perf-stat.node-stores 7.14e+08 ± 24% 60% 1.141e+09 ± 4% perf-stat.node-loads 3.818e+09 ± 6% 22% 4.653e+09 ± 4% perf-stat.cache-misses 1.402e+10 ± 10% 25% 1.746e+10 ± 3% perf-stat.cache-references 443182 ± 11% 19% 525174 ± 4% perf-stat.context-switches 0.41 17% 0.48 perf-stat.ipc 38.30 ± 3% 6% 40.66 perf-stat.iTLB-load-miss-rate% 1.837e+09 ± 6% -11% 1.634e+09 ± 5% perf-stat.node-load-misses 40411593 -15% 34404698 perf-stat.iTLB-load-misses 72.33 ± 6% -19% 58.85 ± 3% perf-stat.node-load-miss-rate% 4582 ± 5% -11% 4079 ± 5% perf-stat.cpu-migrations 1.041e+09 ± 3% -8% 9.572e+08 perf-stat.branch-misses 65214946 ± 5% -23% 50246132 ± 3% perf-stat.iTLB-loads 2.208e+12 -9% 2.014e+12 ± 3% perf-stat.instructions 5.646e+11 -10% 5.091e+11 ± 3% perf-stat.branch-instructions 3.016e+08 ± 4% -13% 2.63e+08 ± 6% perf-stat.dTLB-store-misses 0.19 ± 7% -18% 0.15 ± 4% perf-stat.dTLB-store-miss-rate% 56.96 -23% 43.72 perf-stat.node-store-miss-rate% 5.421e+12 ± 3% -22% 4.229e+12 ± 4% perf-stat.cpu-cycles 5.85e+08 ± 9% -25% 4.409e+08 ± 18% perf-stat.dTLB-load-misses perf-stat.iTLB-load-misses 4.3e+07 ++----------------------------------------------------------------+ 4.2e+07 *+.. | | . ..*... .*... ..*... | 4.1e+07 ++ *. . .. . ..*....*.. *....*.... ..*... | 4e+07 ++ *. *. *.. *....* | | 3.9e+07 ++ | 3.8e+07 ++ | 3.7e+07 ++ | | | 3.6e+07 ++ | 3.5e+07 ++ O O | O O | 3.4e+07 ++ O | 3.3e+07 ++----------------------------------------------------------------+ perf-stat.node-stores 7.5e+08 ++----------------------------------------------------------------+ O O | 7e+08 ++ | 6.5e+08 ++ O O O | | | 6e+08 ++ | | | 5.5e+08 ++ | | ..*.... | 5e+08 *+...*. *.... ..*...*....*....*...*....*... | 4.5e+08 ++ *.. . | | *....*... | 4e+08 ++ *....| | * 3.5e+08 ++----------------------------------------------------------------+ perf-stat.node-store-miss-rate_ 62 *+---------------------------------------------------------------------+ 60 ++ *....*....*....*....*....*.....*....*....*.... | | *.. .* 58 ++ .. .*.... .. | 56 ++ .. *. | | *. | 54 ++ | 52 ++ | 50 ++ | | | 48 ++ | 46 ++ | | O | 44 ++ O O O | 42 O+---------------------------------------------------------------------+ perf-stat.ipc 0.5 ++-------------------------------------------------------------------+ | | 0.48 O+ O O O | | | 0.46 ++ O | | | 0.44 ++ | | | 0.42 ++ | | ..*....*.... ..* 0.4 ++ .*.... ..*... ..*.. *.. | | .. *.... ..*....*....*.. *....*.. | 0.38 *+ *.. | | | 0.36 ++-------------------------------------------------------------------+ vm-scalability.time.system_time 1600 ++-------------------------------------------------------------------+ *... | 1500 ++ . ..*.... ..*....*...*.... | | *.. *....*....*.. *....*....*....*....*... | 1400 ++ .| | * 1300 ++ | | | 1200 ++ | | | 1100 ++ | O O O O | 1000 ++ O | | | 900 ++-------------------------------------------------------------------+ vm-scalability.time.elapsed_time 105 ++--------------------------------------------------------------------+ | | 100 *+... ..*.... ..*.... | 95 ++ *.. *....*....*....*....*.. *....*....*....*....*....| | * 90 ++ | | | 85 ++ | | | 80 ++ | 75 ++ | | | 70 ++ O O | O O O | 65 ++--------------------------------------------------------------------+ vm-scalability.time.elapsed_time.max 105 ++--------------------------------------------------------------------+ | | 100 *+... ..*.... ..*.... | 95 ++ *.. *....*....*....*....*.. *....*....*....*....*....| | * 90 ++ | | | 85 ++ | | | 80 ++ | 75 ++ | | | 70 ++ O O | O O O | 65 ++--------------------------------------------------------------------+ turbostat.RAMWatt 61 ++---------------------------------------------------------------------+ | O | 60 O+ O O O | | | 59 ++ | | | 58 ++ | | | 57 ++ | | *..... *.. | 56 ++ *.... ..*... .. *. .. .. | | .. *....*.. . .. .. *.... .. | 55 ++. * . .. * *....*....* * .. | 54 ++---------------------------------------*-----------------------------+ vm-scalability.throughput 5e+06 ++----------------------------------------------------------------+ | O | O O O O | 4.5e+06 ++ | | | | | 4e+06 ++ | | | 3.5e+06 ++ | | | | | 3e+06 ++ ..*.... ..*....* | ..*...*....*....*....*...*....*....*...*....*.. *. | *.. | 2.5e+06 ++----------------------------------------------------------------+ vmstat.swap.so 2.1e+06 ++---O------------------------------------------------------------+ | | 2e+06 O+ O O | | O | 1.9e+06 ++ | 1.8e+06 ++ | | | 1.7e+06 ++ | | | 1.6e+06 ++ | 1.5e+06 ++ | | ..*.... ..* 1.4e+06 ++ ..*...*....*.... ..*...*....*.... ..*.. *....*...*.. | *.. *.. *. | 1.3e+06 ++----------------------------------------------------------------+ vmstat.system.in 560000 ++---O-------------------------------------------------------------+ 540000 O+ | | O O | 520000 ++ O | 500000 ++ | 480000 ++ | 460000 ++ | | | 440000 ++ | 420000 ++ | 400000 ++ ..*.... ..*.. | 380000 *+...*....*...*.... ..*.. *...*....*.. .. | | *.. | 360000 ++ *...*....*....* 340000 ++-----------------------------------------------------------------+ vmstat.system.cs 8000 ++-------------------------------------------------------------------+ | O O | 7500 O+ O O | 7000 ++ | | | 6500 ++ | 6000 ++ | | | 5500 ++ ..*... | 5000 ++ *.... *.. . | | . *.. *....*....*.. . *... | 4500 ++.. .. .. . .. . | 4000 ++ .. *.... . *....* * *....* * | 3500 ++-------------------------------------------------------------------+ [*] bisect-good sample [O] bisect-bad sample Disclaimer: Results have been estimated based on internal Intel analysis and are provided for informational purposes only. Any difference in system hardware or software design or configuration may affect actual performance. Thanks, Xiaolong