Greeting,

FYI, we noticed a -77.9% regression of pft.faults_per_sec_per_cpu due to commit:

commit: e1f8a89c4f33f5b32cfb047f07fdf6d38cda854a ("drm: introduce sync objects (v4)") git://people.freedesktop.org/~airlied/linux.git drm-syncobj-sem

in testcase: pft on test machine: qemu-system-x86_64 -enable-kvm -cpu Penryn -smp 4 -m 2G with following parameters:

iterations: 20x

test-description: Pft is the page fault test micro benchmark. test-url: https://github.com/gormanm/pft

Details are as below: -------------------------------------------------------------------------------------------------->

To reproduce:

git clone https://github.com/01org/lkp-tests.git cd lkp-tests bin/lkp qemu -k <bzImage> job-script # job-script is attached in this email

testcase/path_params/tbox_group/run: pft/20x/vm-vp-2G

v4.12-rc2 e1f8a89c4f33f5b32cfb047f07 ---------------- -------------------------- %stddev %change %stddev \ | \ 878620 ± 15% -77.9% 194209 ± 12% pft.faults_per_sec_per_cpu 7.16 ± 23% +257.2% 25.58 ± 12% pft.time.elapsed_time 7.16 ± 23% +257.2% 25.58 ± 12% pft.time.elapsed_time.max 1223 ± 10% +199.9% 3667 ± 15% pft.time.involuntary_context_switches 61828 ± 2% +10106.0% 6310202 ± 0% pft.time.minor_page_faults 145.40 ± 3% +20.6% 175.33 ± 5% pft.time.percent_of_cpu_this_job_got 9.37 ± 22% +345.2% 41.72 ± 15% pft.time.system_time 5.04 ± 11% -45.6% 2.74 ± 9% mpstat.cpu.usr% 131734 ± 12% +20.3% 158487 ± 5% softirqs.TIMER 1335 ± 11% +105.1% 2738 ± 19% vmstat.system.cs 5390 ± 4% -27.6% 3904 ± 38% slabinfo.kmalloc-96.active_objs 5420 ± 3% -21.5% 4252 ± 27% slabinfo.kmalloc-96.num_objs 68.30 ±209% +13218.2% 9096 ± 84% latency_stats.max.pipe_read.__vfs_read.vfs_read.SyS_read.entry_SYSCALL_64_fastpath 588552 ± 32% -74.2% 151873 ± 37% latency_stats.sum.ep_poll.SyS_epoll_wait.do_syscall_64.return_from_SYSCALL_64 74.20 ±203% +12326.8% 9220 ± 82% latency_stats.sum.pipe_read.__vfs_read.vfs_read.SyS_read.entry_SYSCALL_64_fastpath 160550 ± 14% -100.0% 0.00 ± -1% meminfo.AnonHugePages 204792 ± 0% -100.0% 0.00 ± -1% meminfo.CmaFree 204800 ± 0% -100.0% 0.00 ± -1% meminfo.CmaTotal 3041 ± 0% +13.4% 3449 ± 0% meminfo.KernelStack 51198 ± 0% -100.0% 0.00 ± -1% proc-vmstat.nr_free_cma 3033 ± 1% +13.6% 3446 ± 0% proc-vmstat.nr_kernel_stack 73394 ± 9% +8258.9% 6134982 ± 2% proc-vmstat.numa_hit 73394 ± 9% +8258.9% 6134982 ± 2% proc-vmstat.numa_local 67006 ± 9% +9043.7% 6126912 ± 2% proc-vmstat.pgfault 11161 ± 7% -100.0% 0.00 ± -1% proc-vmstat.thp_deferred_split_page 11217 ± 7% -100.0% 0.00 ± -1% proc-vmstat.thp_fault_alloc 7.16 ± 23% +257.2% 25.58 ± 12% time.elapsed_time 7.16 ± 23% +257.2% 25.58 ± 12% time.elapsed_time.max 1223 ± 10% +199.9% 3667 ± 15% time.involuntary_context_switches 61828 ± 2% +10106.0% 6310202 ± 0% time.minor_page_faults

pft.faults_per_sec_per_cpu

1e+06 ++--------------------------------------*--------------------------+ 900000 ++ *.. .. | | * : .. | 800000 ++ : : : ...* | 700000 ++ ...* : : : *... | *... : : : : | 600000 ++ : : : : | 500000 ++ : : : : | 400000 ++ : : : : | | : : : : | 300000 ++ : : : : | 200000 O+ O : O : O : O : O O O | : : : : O O | 100000 ++ : : : : | 0 ++-----------*-------------*-------------------O-------------------+

[*] 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