* hackbench score comparison between 5.10.75-rt47 and 5.14.14-rt21 [not found] <CGME20211103031549epcas2p34418a4e6218ca93da57a0c373691bd41@epcas2p3.samsung.com> @ 2021-11-03 3:15 ` Chanho Park 2021-11-03 9:13 ` Sebastian Andrzej Siewior 0 siblings, 1 reply; 8+ messages in thread From: Chanho Park @ 2021-11-03 3:15 UTC (permalink / raw) To: linux-rt-users; +Cc: bigeasy, 'Thomas Gleixner' Dear RT folks, I found an uncomprehended value of hackbench while I tested preempt rt patches on my ARM64(Cortex A76 x 8) target. So, I decided to check it on QEMU x86_64 KVM with yocto. I executed both images with below command. $ runqemu qemux86-64 kvm nographic qemuparams="-smp cores=4" I was able to get similar score values with my arm64 target. It was half than 5.10.75 kernel like below. Any idea about this? Actually, I'm not sure it could be a regression or not. <5.10.75-rt47> root@qemux86-64:~# hackbench -l 10000 Running in process mode with 10 groups using 40 file descriptors each (== 400 tasks) Each sender will pass 10000 messages of 100 bytes Time: 49.898 <5.14.14-rt21> root@qemux86-64:~# hackbench -l 10000 Running in process mode with 10 groups using 40 file descriptors each (== 400 tasks) Each sender will pass 10000 messages of 100 bytes Time: 96.973 Best Regards, Chanho Park ^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: hackbench score comparison between 5.10.75-rt47 and 5.14.14-rt21 2021-11-03 3:15 ` hackbench score comparison between 5.10.75-rt47 and 5.14.14-rt21 Chanho Park @ 2021-11-03 9:13 ` Sebastian Andrzej Siewior 2021-11-05 1:41 ` Chanho Park 0 siblings, 1 reply; 8+ messages in thread From: Sebastian Andrzej Siewior @ 2021-11-03 9:13 UTC (permalink / raw) To: Chanho Park; +Cc: linux-rt-users, 'Thomas Gleixner' On 2021-11-03 12:15:49 [+0900], Chanho Park wrote: > Dear RT folks, Hi, > I found an uncomprehended value of hackbench while I tested preempt rt > patches on my ARM64(Cortex A76 x 8) target. > So, I decided to check it on QEMU x86_64 KVM with yocto. I executed both > images with below command. > > $ runqemu qemux86-64 kvm nographic qemuparams="-smp cores=4" > > I was able to get similar score values with my arm64 target. It was half > than 5.10.75 kernel like below. > Any idea about this? Actually, I'm not sure it could be a regression or not. > > <5.10.75-rt47> > root@qemux86-64:~# hackbench -l 10000 > Running in process mode with 10 groups using 40 file descriptors each (== > 400 tasks) > Each sender will pass 10000 messages of 100 bytes > Time: 49.898 > > <5.14.14-rt21> > root@qemux86-64:~# hackbench -l 10000 > Running in process mode with 10 groups using 40 file descriptors each (== > 400 tasks) > Each sender will pass 10000 messages of 100 bytes > Time: 96.973 The 5.14 series has a different SLUB implementation. Could you please make sure that SLUB_CPU_PARTIAL is disabled? So v5.14-rc3-rt1 should be worse than v5.10. Then v5.14-rc3-rt2 introduced adaptive spinning which should improve the situation. However it is slightly worse than v5.10 but it should have improved. Could verify that? Also could double check this on hardware? I have no idea how well the adaptive spinning is working in KVM and this (hackbench) is a micro benchmark for the memory allocator/SLUB and any spin/guest preemption can have a visible outcome. While I saw worse numbers here (hackbench) I didn't observe it in a real-work workload like a kernel build for instance. > Best Regards, > Chanho Park Sebastian ^ permalink raw reply [flat|nested] 8+ messages in thread
* RE: hackbench score comparison between 5.10.75-rt47 and 5.14.14-rt21 2021-11-03 9:13 ` Sebastian Andrzej Siewior @ 2021-11-05 1:41 ` Chanho Park 2021-11-12 14:00 ` 'Sebastian Andrzej Siewior' 0 siblings, 1 reply; 8+ messages in thread From: Chanho Park @ 2021-11-05 1:41 UTC (permalink / raw) To: 'Sebastian Andrzej Siewior' Cc: linux-rt-users, 'Thomas Gleixner' Hi, > -----Original Message----- > From: Sebastian Andrzej Siewior <bigeasy@linutronix.de> > Sent: Wednesday, November 3, 2021 6:14 PM > To: Chanho Park <chanho61.park@samsung.com> > Cc: linux-rt-users@vger.kernel.org; 'Thomas Gleixner' <tglx@linutronix.de> > Subject: Re: hackbench score comparison between 5.10.75-rt47 and 5.14.14- > rt21 > > On 2021-11-03 12:15:49 [+0900], Chanho Park wrote: > > Dear RT folks, > Hi, > > > I found an uncomprehended value of hackbench while I tested preempt rt > > patches on my ARM64(Cortex A76 x 8) target. > > So, I decided to check it on QEMU x86_64 KVM with yocto. I executed > > both images with below command. > > > > $ runqemu qemux86-64 kvm nographic qemuparams="-smp cores=4" > > > > I was able to get similar score values with my arm64 target. It was > > half than 5.10.75 kernel like below. > > Any idea about this? Actually, I'm not sure it could be a regression or > not. > > > > <5.10.75-rt47> > > root@qemux86-64:~# hackbench -l 10000 > > Running in process mode with 10 groups using 40 file descriptors each > > (== > > 400 tasks) > > Each sender will pass 10000 messages of 100 bytes > > Time: 49.898 > > > > <5.14.14-rt21> > > root@qemux86-64:~# hackbench -l 10000 > > Running in process mode with 10 groups using 40 file descriptors each > > (== > > 400 tasks) > > Each sender will pass 10000 messages of 100 bytes > > Time: 96.973 > > The 5.14 series has a different SLUB implementation. Could you please make > sure that SLUB_CPU_PARTIAL is disabled? > So v5.14-rc3-rt1 should be worse than v5.10. Then v5.14-rc3-rt2 introduced > adaptive spinning which should improve the situation. However it is > slightly worse than v5.10 but it should have improved. > Could verify that? > > Also could double check this on hardware? I have no idea how well the > adaptive spinning is working in KVM and this (hackbench) is a micro > benchmark for the memory allocator/SLUB and any spin/guest preemption can > have a visible outcome. > While I saw worse numbers here (hackbench) I didn't observe it in a real- > work workload like a kernel build for instance. I checked the same test on my aarch64 target. I'll do more realistic benchmark such as compile bench. <5.10.73-rt54 aarch64> root@euto-v9-sadk:~# hackbench -l 10000 Time: 24.994 <5.15.0-rt17 aarch64 w/o CONFIG_SLUB_CPU_PARTIAL> root@euto-v9-sadk:~# hackbench -l 10000 Time: 31.372 <5.15.0-rt17 aarch64 w/ CONFIG_SLUB_CPU_PARTIAL> root@euto-v9-sadk:~# hackbench -l 10000 Time: 35.269 Best Regards, Chanho Park ^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: hackbench score comparison between 5.10.75-rt47 and 5.14.14-rt21 2021-11-05 1:41 ` Chanho Park @ 2021-11-12 14:00 ` 'Sebastian Andrzej Siewior' 2021-11-16 9:37 ` Vlastimil Babka 0 siblings, 1 reply; 8+ messages in thread From: 'Sebastian Andrzej Siewior' @ 2021-11-12 14:00 UTC (permalink / raw) To: Chanho Park; +Cc: linux-rt-users, 'Thomas Gleixner' On 2021-11-05 10:41:40 [+0900], Chanho Park wrote: > Hi, Hi, > I checked the same test on my aarch64 target. I'll do more realistic > benchmark such as compile bench. > > <5.10.73-rt54 aarch64> > root@euto-v9-sadk:~# hackbench -l 10000 > Time: 24.994 > > <5.15.0-rt17 aarch64 w/o CONFIG_SLUB_CPU_PARTIAL> > root@euto-v9-sadk:~# hackbench -l 10000 > Time: 31.372 > > <5.15.0-rt17 aarch64 w/ CONFIG_SLUB_CPU_PARTIAL> > root@euto-v9-sadk:~# hackbench -l 10000 > Time: 35.269 The last two are equal, is that on purporse? Any feedback on v5.14-rc3-rt1 and v5.14-rc3-rt2? v5.13-rt1 should be identical to you 5.10 numbers. > Best Regards, > Chanho Park Sebastian ^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: hackbench score comparison between 5.10.75-rt47 and 5.14.14-rt21 2021-11-12 14:00 ` 'Sebastian Andrzej Siewior' @ 2021-11-16 9:37 ` Vlastimil Babka 2021-11-16 9:42 ` 'Sebastian Andrzej Siewior' 0 siblings, 1 reply; 8+ messages in thread From: Vlastimil Babka @ 2021-11-16 9:37 UTC (permalink / raw) To: 'Sebastian Andrzej Siewior', Chanho Park Cc: linux-rt-users, 'Thomas Gleixner' On 11/12/21 15:00, 'Sebastian Andrzej Siewior' wrote: > On 2021-11-05 10:41:40 [+0900], Chanho Park wrote: >> Hi, > Hi, > >> I checked the same test on my aarch64 target. I'll do more realistic >> benchmark such as compile bench. >> >> <5.10.73-rt54 aarch64> >> root@euto-v9-sadk:~# hackbench -l 10000 >> Time: 24.994 >> >> <5.15.0-rt17 aarch64 w/o CONFIG_SLUB_CPU_PARTIAL> >> root@euto-v9-sadk:~# hackbench -l 10000 >> Time: 31.372 >> >> <5.15.0-rt17 aarch64 w/ CONFIG_SLUB_CPU_PARTIAL> >> root@euto-v9-sadk:~# hackbench -l 10000 >> Time: 35.269 > > The last two are equal, is that on purporse? I see the first of last two is without SLUB_CPU_PARTIAL and the second with SLUB_CPU_PARTIAL, so not equal? > Any feedback on v5.14-rc3-rt1 and v5.14-rc3-rt2? v5.13-rt1 should be > identical to you 5.10 numbers. > >> Best Regards, >> Chanho Park > > Sebastian > ^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: hackbench score comparison between 5.10.75-rt47 and 5.14.14-rt21 2021-11-16 9:37 ` Vlastimil Babka @ 2021-11-16 9:42 ` 'Sebastian Andrzej Siewior' 2021-11-16 13:42 ` Chanho Park 0 siblings, 1 reply; 8+ messages in thread From: 'Sebastian Andrzej Siewior' @ 2021-11-16 9:42 UTC (permalink / raw) To: Vlastimil Babka; +Cc: Chanho Park, linux-rt-users, 'Thomas Gleixner' On 2021-11-16 10:37:22 [+0100], Vlastimil Babka wrote: > On 11/12/21 15:00, 'Sebastian Andrzej Siewior' wrote: > > On 2021-11-05 10:41:40 [+0900], Chanho Park wrote: > >> Hi, > > Hi, > > > >> I checked the same test on my aarch64 target. I'll do more realistic > >> benchmark such as compile bench. > >> > >> <5.10.73-rt54 aarch64> > >> root@euto-v9-sadk:~# hackbench -l 10000 > >> Time: 24.994 > >> > >> <5.15.0-rt17 aarch64 w/o CONFIG_SLUB_CPU_PARTIAL> > >> root@euto-v9-sadk:~# hackbench -l 10000 > >> Time: 31.372 > >> > >> <5.15.0-rt17 aarch64 w/ CONFIG_SLUB_CPU_PARTIAL> > >> root@euto-v9-sadk:~# hackbench -l 10000 > >> Time: 35.269 > > > > The last two are equal, is that on purporse? > > I see the first of last two is without SLUB_CPU_PARTIAL and the second with > SLUB_CPU_PARTIAL, so not equal? indeed. I didn't notice the missing o. Thanks!. > > Any feedback on v5.14-rc3-rt1 and v5.14-rc3-rt2? v5.13-rt1 should be > > identical to you 5.10 numbers. > > > >> Best Regards, > >> Chanho Park Sebastian ^ permalink raw reply [flat|nested] 8+ messages in thread
* RE: hackbench score comparison between 5.10.75-rt47 and 5.14.14-rt21 2021-11-16 9:42 ` 'Sebastian Andrzej Siewior' @ 2021-11-16 13:42 ` Chanho Park 2021-11-19 11:12 ` 'Sebastian Andrzej Siewior' 0 siblings, 1 reply; 8+ messages in thread From: Chanho Park @ 2021-11-16 13:42 UTC (permalink / raw) To: 'Sebastian Andrzej Siewior', 'Vlastimil Babka' Cc: linux-rt-users, 'Thomas Gleixner' Hi, > -----Original Message----- > From: 'Sebastian Andrzej Siewior' <bigeasy@linutronix.de> > Sent: Tuesday, November 16, 2021 6:43 PM > To: Vlastimil Babka <vbabka@suse.cz> > Cc: Chanho Park <chanho61.park@samsung.com>; linux-rt- > users@vger.kernel.org; 'Thomas Gleixner' <tglx@linutronix.de> > Subject: Re: hackbench score comparison between 5.10.75-rt47 and 5.14.14- > rt21 > > On 2021-11-16 10:37:22 [+0100], Vlastimil Babka wrote: > > On 11/12/21 15:00, 'Sebastian Andrzej Siewior' wrote: > > > On 2021-11-05 10:41:40 [+0900], Chanho Park wrote: > > >> Hi, > > > Hi, > > > > > >> I checked the same test on my aarch64 target. I'll do more > > >> realistic benchmark such as compile bench. > > >> > > >> <5.10.73-rt54 aarch64> > > >> root@euto-v9-sadk:~# hackbench -l 10000 > > >> Time: 24.994 > > >> > > >> <5.15.0-rt17 aarch64 w/o CONFIG_SLUB_CPU_PARTIAL> > > >> root@euto-v9-sadk:~# hackbench -l 10000 > > >> Time: 31.372 > > >> > > >> <5.15.0-rt17 aarch64 w/ CONFIG_SLUB_CPU_PARTIAL> > > >> root@euto-v9-sadk:~# hackbench -l 10000 > > >> Time: 35.269 > > > > > > The last two are equal, is that on purporse? > > > > I see the first of last two is without SLUB_CPU_PARTIAL and the second > > with SLUB_CPU_PARTIAL, so not equal? > > indeed. I didn't notice the missing o. Thanks!. Sorry for this confusion. > > > > Any feedback on v5.14-rc3-rt1 and v5.14-rc3-rt2? v5.13-rt1 should be > > > identical to you 5.10 numbers. I got below value on v5.14-rc3-rt1. <5.14.0-rc3-rt1 aarch64> Time: 35.948 Unfortunately, I've got below panic on v5.13-rt1. I stacked same patches for my board both v5.13-rt1 and v514-rc3-rt1. [ 0.112294] Freeing unused kernel memory: 5440K [ 0.131064] Run /sbin/init as init process [ 0.137371] ------------[ cut here ]------------ [ 0.137373] WARNING: CPU: 7 PID: 1 at arch/arm64/kernel/fpsimd.c:286 task_fpsimd_load+0x98/0xc0 [ 0.137381] Modules linked in: [ 0.137385] CPU: 7 PID: 1 Comm: init Not tainted 5.13.0-rt1-00032-gdee4765ed8ff #317 [ 0.137388] Hardware name: Samsung ExynosAuto v9 SADK board (DT) [ 0.137389] pstate: 40400009 (nZcv daif +PAN -UAO -TCO BTYPE=--) [ 0.137392] pc : task_fpsimd_load+0x98/0xc0 [ 0.137394] lr : task_fpsimd_load+0x18/0xc0 [ 0.137396] sp : ffff80001004bd60 [ 0.137397] x29: ffff80001004bd60 x28: ffff000800098000 x27: 0000000000000000 [ 0.137400] x26: 0000000000000000 x25: 0000000000000000 x24: 0000000045585401 [ 0.137402] x23: 0000000000000000 x22: ffff0008000986d8 x21: ffff000800098000 [ 0.137405] x20: ffff80001004beb0 x19: 0000000000000008 x18: 0000000000000014 [ 0.137408] x17: 00000000ccf04c01 x16: 0000000027fe4405 x15: 00000000083da15a [ 0.137411] x14: 0000000000000000 x13: 0000ffffd305cbf8 x12: 000000000000000f [ 0.137414] x11: 0000ffffd305cfed x10: 000000000000001f x9 : 0000000000000000 [ 0.137416] x8 : ffff80001004c000 x7 : 0000000000000000 x6 : 0000000000000001 [ 0.137418] x5 : 0000000080080007 x4 : 0000000000000000 x3 : 0000000000000001 [ 0.137421] x2 : 0000000000000000 x1 : 0000000000000000 x0 : 0000000000000000 [ 0.137424] Call trace: [ 0.137425] task_fpsimd_load+0x98/0xc0 [ 0.137427] fpsimd_restore_current_state+0x78/0xac [ 0.137430] do_notify_resume+0x10c/0x1300 [ 0.137434] work_pending+0xc/0x5f8 [ 0.137436] ---[ end trace 0000000000000002 ]--- Best Regards, Chanho Park ^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: hackbench score comparison between 5.10.75-rt47 and 5.14.14-rt21 2021-11-16 13:42 ` Chanho Park @ 2021-11-19 11:12 ` 'Sebastian Andrzej Siewior' 0 siblings, 0 replies; 8+ messages in thread From: 'Sebastian Andrzej Siewior' @ 2021-11-19 11:12 UTC (permalink / raw) To: Chanho Park Cc: 'Vlastimil Babka', linux-rt-users, 'Thomas Gleixner' On 2021-11-16 22:42:01 [+0900], Chanho Park wrote: > Hi, Hi, > > On 2021-11-16 10:37:22 [+0100], Vlastimil Babka wrote: > > > On 11/12/21 15:00, 'Sebastian Andrzej Siewior' wrote: > > > > On 2021-11-05 10:41:40 [+0900], Chanho Park wrote: > > > >> Hi, > > > > Hi, > > > > > > > >> I checked the same test on my aarch64 target. I'll do more > > > >> realistic benchmark such as compile bench. > > > >> > > > >> <5.10.73-rt54 aarch64> > > > >> root@euto-v9-sadk:~# hackbench -l 10000 > > > >> Time: 24.994 > > > >> > > > >> <5.15.0-rt17 aarch64 w/o CONFIG_SLUB_CPU_PARTIAL> > > > >> root@euto-v9-sadk:~# hackbench -l 10000 > > > >> Time: 31.372 > > > >> > > > >> <5.15.0-rt17 aarch64 w/ CONFIG_SLUB_CPU_PARTIAL> > > > >> root@euto-v9-sadk:~# hackbench -l 10000 > > > >> Time: 35.269 > > > > > > > > The last two are equal, is that on purporse? > > > > > > I see the first of last two is without SLUB_CPU_PARTIAL and the second > > > with SLUB_CPU_PARTIAL, so not equal? > > > > indeed. I didn't notice the missing o. Thanks!. > > Sorry for this confusion. > > > > > > > Any feedback on v5.14-rc3-rt1 and v5.14-rc3-rt2? v5.13-rt1 should be > > > > identical to you 5.10 numbers. > > I got below value on v5.14-rc3-rt1. > > <5.14.0-rc3-rt1 aarch64> > Time: 35.948 So this matches your v5.15 numbers with CONFIG_SLUB_CPU_PARTIAL. In v5.14-rc3-rt2 adaptive spinning was introduced. Based on these numbers it makes no difference for you. > Unfortunately, I've got below panic on v5.13-rt1. I stacked same patches for my board both v5.13-rt1 and v514-rc3-rt1. Yes, good. > Best Regards, > Chanho Park Sebastian ^ permalink raw reply [flat|nested] 8+ messages in thread
end of thread, other threads:[~2021-11-19 11:12 UTC | newest]
Thread overview: 8+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
[not found] <CGME20211103031549epcas2p34418a4e6218ca93da57a0c373691bd41@epcas2p3.samsung.com>
2021-11-03 3:15 ` hackbench score comparison between 5.10.75-rt47 and 5.14.14-rt21 Chanho Park
2021-11-03 9:13 ` Sebastian Andrzej Siewior
2021-11-05 1:41 ` Chanho Park
2021-11-12 14:00 ` 'Sebastian Andrzej Siewior'
2021-11-16 9:37 ` Vlastimil Babka
2021-11-16 9:42 ` 'Sebastian Andrzej Siewior'
2021-11-16 13:42 ` Chanho Park
2021-11-19 11:12 ` 'Sebastian Andrzej Siewior'
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox