LinuxLists.cc - [PATCH] mm: fix COW faults after mlock()

2018-09-24 13:09:34

Subject: [PATCH] mm: fix COW faults after mlock()

After mlock() on newly mmap()ed shared memory I observe page faults.

The problem is that populate_vma_page_range() doesn't set FOLL_WRITE
flag for writable shared memory in mlock() path, arguing that like:
/*
* We want to touch writable mappings with a write fault in order
* to break COW, except for shared mappings because these don't COW
* and we would not want to dirty them for nothing.
*/

But they are actually COWed. The most straightforward way to avoid it
is to set FOLL_WRITE flag for shared mappings as well as for private ones.

This is the partial revert of commit 5ecfda041e4b4 ("mlock: avoid
dirtying pages and triggering writeback"). So it re-enables dirtying.

The fix works for me (arm64, kernel v4.19-rc4 and v4.9), but after digging
into the code I still don't understand why we need to do copy-on-write on
shared memory. If comment above was correct when 5ecfda041e4b4 became
upstreamed (2011), shared mappings were not COWed back in 2011, but are
COWed now. If so, this is another issue to be fixed.

Signed-off-by: Yury Norov <[email protected]>
---
mm/gup.c | 5 ++---
1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/mm/gup.c b/mm/gup.c
index 1abc8b4afff6..1899e8bac06b 100644
--- a/mm/gup.c
+++ b/mm/gup.c
@@ -1202,10 +1202,9 @@ long populate_vma_page_range(struct vm_area_struct *vma,
gup_flags &= ~FOLL_POPULATE;
/*
* We want to touch writable mappings with a write fault in order
- * to break COW, except for shared mappings because these don't COW
- * and we would not want to dirty them for nothing.
+ * to break COW.
*/
- if ((vma->vm_flags & (VM_WRITE | VM_SHARED)) == VM_WRITE)
+ if (vma->vm_flags & VM_WRITE)
gup_flags |= FOLL_WRITE;

/*
--
2.17.1

2018-09-24 21:23:51

by Kirill A. Shutemov

[permalink] [raw]

Subject: Re: [PATCH] mm: fix COW faults after mlock()

On Mon, Sep 24, 2018 at 04:08:52PM +0300, Yury Norov wrote:
> After mlock() on newly mmap()ed shared memory I observe page faults.
>
> The problem is that populate_vma_page_range() doesn't set FOLL_WRITE
> flag for writable shared memory in mlock() path, arguing that like:
> /*
> * We want to touch writable mappings with a write fault in order
> * to break COW, except for shared mappings because these don't COW
> * and we would not want to dirty them for nothing.
> */
>
> But they are actually COWed. The most straightforward way to avoid it
> is to set FOLL_WRITE flag for shared mappings as well as for private ones.

Huh? How do shared mapping get CoWed?

In this context CoW means to create a private copy of the page for the
process. It only makes sense for private mappings as all pages in shared
mappings do not belong to the process.

Shared mappings will still get faults, but a bit later -- after the page
is written back to disc, the page get clear and write protected to catch
the next write access.

Noticeable exception is tmpfs/shmem. These pages do not belong to normal
write back process. But the code path is used for other filesystems as
well.

Therefore, NAK. You only create unneeded write back traffic.

--
Kirill A. Shutemov

2018-09-24 23:49:25

by Yury Norov

[permalink] [raw]

Subject: Re: [PATCH] mm: fix COW faults after mlock()

On Tue, Sep 25, 2018 at 12:22:47AM +0300, Kirill A. Shutemov wrote:
> External Email
>
> On Mon, Sep 24, 2018 at 04:08:52PM +0300, Yury Norov wrote:
> > After mlock() on newly mmap()ed shared memory I observe page faults.
> >
> > The problem is that populate_vma_page_range() doesn't set FOLL_WRITE
> > flag for writable shared memory in mlock() path, arguing that like:
> > /*
> > * We want to touch writable mappings with a write fault in order
> > * to break COW, except for shared mappings because these don't COW
> > * and we would not want to dirty them for nothing.
> > */
> >
> > But they are actually COWed. The most straightforward way to avoid it
> > is to set FOLL_WRITE flag for shared mappings as well as for private ones.
>
> Huh? How do shared mapping get CoWed?
>
> In this context CoW means to create a private copy of the page for the
> process. It only makes sense for private mappings as all pages in shared
> mappings do not belong to the process.
>
> Shared mappings will still get faults, but a bit later -- after the page
> is written back to disc, the page get clear and write protected to catch
> the next write access.
>
> Noticeable exception is tmpfs/shmem. These pages do not belong to normal
> write back process. But the code path is used for other filesystems as
> well.
>
> Therefore, NAK. You only create unneeded write back traffic.

Hi Kirill,

(My first reaction was exactly like yours indeed, but) on my real
system (Cavium OcteonTX2), and on my qemu simulation I can reproduce
the same behavior: just mlock()ed memory causes faults. That faults
happen because page is mapped to the process as read-only, while
underlying VMA is read-write. So faults get resolved well by just
setting write access to the page.

Maybe I use term COW wrongly here, but this is how faultin_page()
works, and it sets FOLL_COW bit before return (which is ignored
on upper level).

I realize that proper fix may be more complex, and if so I'll
thankfully take it and drop this patch from my tree, but this is
all that I have so far to address the problem.

The user code below is reproducer.

Thanks,
Yury

int i, ret, len = getpagesize() * 1000;
char tmpfile[] = "/tmp/my_tmp-XXXXXX";
int fd = mkstemp(tmpfile);

ret = ftruncate(fd, len);
if (ret) {
printf("Failed to ftruncate: %d\n", errno);
goto out;
}

ptr = mmap(NULL, len, PROT_READ | PROT_WRITE, MAP_SHARED, fd, 0);
if (ptr == MAP_FAILED) {
printf("Failed to mmap memory: %d\n", errno);
goto out;
}

ret = mlock(ptr, len);
if (ret) {
printf("Failed to mlock: %d\n", errno);
goto out;
}

printf("Touch...\n");

for (i = 0; i < len; i++)
ptr[i] = (char) i; /* Faults here. */

printf("\t... done\n");
out:
close(fd);
unlink(tmpfile);

2018-09-25 10:49:33

by Kirill A. Shutemov

[permalink] [raw]

Subject: Re: [PATCH] mm: fix COW faults after mlock()

On Tue, Sep 25, 2018 at 02:48:43AM +0300, Yury Norov wrote:
> On Tue, Sep 25, 2018 at 12:22:47AM +0300, Kirill A. Shutemov wrote:
> > External Email
> >
> > On Mon, Sep 24, 2018 at 04:08:52PM +0300, Yury Norov wrote:
> > > After mlock() on newly mmap()ed shared memory I observe page faults.
> > >
> > > The problem is that populate_vma_page_range() doesn't set FOLL_WRITE
> > > flag for writable shared memory in mlock() path, arguing that like:
> > > /*
> > > * We want to touch writable mappings with a write fault in order
> > > * to break COW, except for shared mappings because these don't COW
> > > * and we would not want to dirty them for nothing.
> > > */
> > >
> > > But they are actually COWed. The most straightforward way to avoid it
> > > is to set FOLL_WRITE flag for shared mappings as well as for private ones.
> >
> > Huh? How do shared mapping get CoWed?
> >
> > In this context CoW means to create a private copy of the page for the
> > process. It only makes sense for private mappings as all pages in shared
> > mappings do not belong to the process.
> >
> > Shared mappings will still get faults, but a bit later -- after the page
> > is written back to disc, the page get clear and write protected to catch
> > the next write access.
> >
> > Noticeable exception is tmpfs/shmem. These pages do not belong to normal
> > write back process. But the code path is used for other filesystems as
> > well.
> >
> > Therefore, NAK. You only create unneeded write back traffic.
>
> Hi Kirill,
>
> (My first reaction was exactly like yours indeed, but) on my real
> system (Cavium OcteonTX2), and on my qemu simulation I can reproduce
> the same behavior: just mlock()ed memory causes faults. That faults
> happen because page is mapped to the process as read-only, while
> underlying VMA is read-write. So faults get resolved well by just
> setting write access to the page.

mlock() doesn't guarntee that you'll never get a *minor* fault. Write back
or page migration will get these pages write-protected.

Making pages write protected is what we rely on for proper dirty
accounting: filesystems need to know when page gets dirty and allocate
resources for properly write back the page. Once page is written back to
storage the page gets write protected again to catch the next write access
to the page.

I guess we can situation a bit better for shmem/tmpfs: we can populate
such shared mappings with FOLL_WRITE. But this patch is not good for the
task.

--
Kirill A. Shutemov

2018-10-11 05:39:03

by Chen, Rong A

[permalink] [raw]

Subject: [LKP] [mm] dd12385915: vm-scalability.median 18.6% improvement

Greeting,

FYI, we

commit: url: in testcase: vm-scalability
on test with following parameters:

runtime: 300s
size: 1T
test: msync
cpufreq_governor: performance

test-description: test-url:

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

To reproduce:

git clone bin/lkp run job.yaml

================= compiler/cpufreq_ gcc-7/performan
commit:
dd52cb8790 dd12385915
dd52cb879063ca54 ---------------- fail:runs | 1:4 0:4 2:4 0:4 2:4 %stddev \ 277824 ± 4% 0.46 ± 6% 2817 ± 26% 6509 ± 28% 142.50 ± 5% 281.87 ± 6% 152.54 ± 6% 448928 ± 6% 2.093e+08 ± 6% 18807972 ± 5% 88890 ± 2% 100860 ± 2% 9.00 6.00 67870 ± 4% 14.04 ± 8% 29.68 ± 7% 42.68 ± 5% 13.58 ± 4% 294613 7559 294837 7559 817.75 ± 6% 454.00 ± 8% 13651918 ± 29% 1.931e+08 ± 6% 508755 ± 6% 16723882 ± 4% 21421 ± 3% 3.002e+08 ± 5% 312454 ± 4% 107432 ± 20% 18727 ± 8% 1073224 ± 2% 308786 ± 3% 1342759 ± 2% 172701 ± 16% 1170057 13762 85607 107046 37561 ± 4% 84144 ± 7% 85958 ± 6% 82818 ± 7% 209438 ± 7% 218172 ± 8% 199461 ± 6% 1399 ± 2% 15.87 ± 25% -14.25 11.61 ± 28% 127503 ± 14% 0.00 ± 94% 1879 ± 4% 57.42 ± 4% 1.11 ± 30% 508750 ± 6% 15.71 ± 7% 21419 ± 3% 1.36 ± 5% 312453 ± 4% 24.42 ± 6% 30.15 ± 8% 1.09 ± 5% 11.33 ± 9% 12.77 ± 3% 39798103 ± 5% 30.09 4.902e+11 ± 5% 3.636e+09 ± 8% 66.28 9.761e+09 ± 7% 1.11 2.298e+12 ± 6% 19198 ± 5% 3.858e+09 ± 15% 5.213e+11 ± 5% 0.12 ± 9% 3.446e+08 ± 10% 2.839e+11 ± 5% 74.39 3.687e+08 ± 8% 1.272e+08 ± 10% 2.077e+12 ± 5% 5646 ± 3% 0.90 47044927 ± 6% 47046188 ± 6% 22566 ± 39% 120471 ± 14% 7661 ± 28% 7682 ± 28% 2752 ± 5% 267103 ± 2% 78076 ± 3% 7811 ± 6% 43289 ± 16% 293783 3445 21416 9153 ± 2% 267072 ± 2% 43289 ± 16% 293694 87180 ± 3% 85522443 ± 6% 85522443 ± 6% 2958 ± 5% 51287873 ± 6% 28504604 ± 8% 57168735 ± 6% 41153300 ± 7% 85685675 ± 6% 6509 ± 28% 2976 41153300 ± 7% 477249 ± 6% 43943485 ± 5% 84942346 ± 6% 8102736 ± 7% 47558626 ± 6% 14421 2020457 ± 15% 17434531 ± 8% 19.94 ± 25% 12.24 ± 26% 12.22 ± 26% 12.21 ± 26% 11.34 ± 26% 16.17 ± 23% 9.39 ± 27% 7.59 ± 23% 7.53 ± 23% 6.69 ± 24% 6.55 ± 24% 14.30 ± 5% 14.31 ± 5% 15.40 ± 3% 2.67 ± 41% 4.15 ± 27% 2.46 ± 19% 3.60 ± 24% 9.38 ± 3% 3.54 ± 3% 0.31 ±100% 0.00 0.34 ±102% 0.59 ± 66% 0.00 0.00 2.14 ± 34% 1.14 ± 34% 1.24 ± 34% 1.19 ± 33% 1.57 ± 33% 1.60 ± 34% 1.90 ± 32% 0.00 0.00 0.00 0.00 3.25 ± 30% 0.00 0.00 0.00 6.10 ± 23% 6.60 ± 23% 3.01 ± 25% 0.00 0.00 11.83 ± 21% 11.85 ± 21% 12.52 ± 21% 12.52 ± 21% 12.52 ± 21% 19.94 ± 25% 17.94 ± 23% 17.89 ± 23% 17.91 ± 23% 9.39 ± 27% 14.54 ± 6% 14.52 ± 5% 15.55 ± 4% 16.04 ± 4% 7.32 ± 17% 5.75 ± 23% 5.38 ± 22% 5.38 ± 22% 4.18 ± 7% 9.57 1.76 ± 30% 2.91 ± 17% 0.79 ± 22% 0.65 ± 39% 1.12 ± 32% 1.27 ± 36% 3.68 ± 2% 0.75 ± 22% 0.30 ± 24% 1.26 ± 3% 0.29 ± 9% 1.31 ± 3% 0.20 ± 27% 0.20 ± 27% 0.20 ± 27% 0.83 ± 11% 0.59 ± 7% 0.58 ± 5% 0.21 ± 18% 0.13 ± 38% 0.16 ± 19% 0.09 ± 27% 0.00 0.08 ± 32% 0.01 ±173% 0.07 ± 59% 0.02 ±173% 0.04 ±101% 0.06 ±101% 0.09 ± 28% 0.11 ± 38% 0.07 ± 70% 0.00 0.14 ± 18% 0.11 ± 70% 0.10 ± 69% 0.10 ± 69% 0.10 ± 65% 0.19 ± 15% 0.35 ± 25% 0.48 ± 19% 0.28 ± 19% 0.07 ±101% 0.09 ± 43% 0.80 ± 9% 0.55 ± 17% 0.46 ± 20% 0.09 ±102% 0.10 ± 79% 0.42 ± 20% 0.49 ± 13% 0.00 0.00 1.46 ± 13% 0.43 ± 14% 0.37 ± 28% 0.15 ± 39% 0.65 ± 8% 0.45 ± 12% 0.39 ± 28% 0.39 ± 28% 0.73 ± 10% 0.07 ±119% 0.65 ± 14% 0.22 ± 37% 0.92 ± 6% 1.04 ± 8% 1.18 ± 8% 0.69 ± 11% 0.00 0.00 1.22 ± 16% 1.00 ± 28% 1.22 ± 15% 1.22 ± 25% 1.18 ± 26% 2.56 ± 24% 1.32 ± 27% 1.41 ± 26% 1.75 ± 26% 1.78 ± 27% 2.17 ± 24% 0.84 ± 36% 0.88 ± 35% 0.88 ± 35% 0.00 0.00 1.41 ± 25% 0.00 3.42 ± 23% 0.00 0.00 6.12 ± 23% 2.14 ± 22% 6.65 ± 24% 7.61 ± 24% 2.75 ± 26% 3.01 ± 25% 3.01 ± 25% 3.01 ± 25% 16.28 ± 22% 16.28 ± 22% 12.56 ± 21% 12.52 ± 21% 12.52 ± 21% 12.55 ± 21% 12.54 ± 21% 9.01 ± 26% 3.84 ± 24% 2.89 ± 9% 4.27 ± 3% 1.14 ± 16% 0.39 ± 11% 0.29 ± 9% 0.40 ± 20% 0.21 ± 20% 0.16 ± 19% 0.09 ± 28% 0.17 ± 3% 0.07 ± 14% 0.00 0.09 ± 27% 0.01 ±173% 0.02 ±173% 0.09 ± 36% 0.09 ± 28% 0.01 ±173% 0.21 ± 23% 0.13 ± 32% 0.16 ± 5% 0.15 ± 33% 0.18 ± 23% 0.10 ± 36% 0.07 ± 69% 0.05 ± 63% 0.13 ± 27% 0.23 ± 27% 0.44 ± 20% 0.17 ± 29% 0.00 1.43 ± 12% 0.16 ± 42% 0.33 ± 19% 0.53 ± 9% 0.37 ± 21% 0.20 ± 38% 0.05 ±112% 0.45 ± 12% 0.46 ± 15% 0.46 ± 33% 0.61 ± 14% 2.14 ± 21%

280 +-+-------- | + + 260 +-+ + + .+. 240 +-+ + +. | + 220 +-+ | 200 +-+ | 180 +-+ 160 +-+ | O O O 140 O-+ O | 120 +-+-------- 2600 +-+------- | 2400 +-+ 2200 +-+ +.+.. | .. +..+. 2000 +-++ | 1800 +-+ | 1600 +-+ 1400 +-+ | O 1200 O-+O O O | 1000 +-+------- 900 +-+-------- 850 +-+ |.. .+.. 800 +-+ .+ +..+..+. 750 +-++. 700 +-+ 650 +-+ | 600 +-+ 550 +-+ 500 +-+ 450 +-+ O O O O O 400 +-+ 350 +-+-------- 8000 +-+------- | : 7000 +-+ : | : : |: : 6000 +-+ : | :.+.. 5000 +-+ + | 4000 +-+ | | 3000 +-+ O O O O O 2000 +-+------- 3e+08 | 2.5e+08 +-+ | | 2e+08 +-+ | 1.5e+08 +-+ | 1e+08 +-+ O O O O O | 5e+07 +-+ | 1.6e+09 | 1.5e+09 +-+ O 1.4e+09 +-+ | 1.3e+09 +-+ | 1.2e+09 +-+ | 1.1e+09 +-+ |.. .+..+.. 1e+09 +-++ | 9e+08 6e+08 |.. +..+.. | + 5.5e+08 +-++ | | 5e+08 +-+ | 4.5e+08 +-+ | O O O O | 4e+08 O-+ | | 3.5e+08 [*] bisect-good sample
[O] bisect-bad sample

Disclaimer:
Results for informational design or

Thanks,
Rong Chen
noticed a 18.6% improvement of vm-scalability.median due to commit:
dd12385915f4f83b738467e599b053c33dffbd48 ("[PATCH] mm: fix COW faults after mlock()")
href="https://github.com/0day-ci/linux/commits/Yury-Norov/mm-fix-COW-faults-after-mlock/20180925-174527">https://github.com/0day-ci/linux/commits/Yury-Norov/mm-fix-COW-faults-after-mlock/20180925-174527
href="https://github.com/thesofproject/linux">https://github.com/thesofproject/linux master
machine: 80 threads Skylake with 64G memory
The motivation behind this suite is to exercise functions and regions of the mm/ of the Linux kernel which are of interest to us.
href="https://git.kernel.org/cgit/linux/kernel/git/wfg/vm-scalability.git/">https://git.kernel.org/cgit/linux/kernel/git/wfg/vm-scalability.git/
--------------------------------------------------------------------------------->
href="https://github.com/intel/lkp-tests.git">https://github.com/intel/lkp-tests.git
bin/lkp install job.yaml # job file is attached in this email
========================================================================
governor/kconfig/rootfs/runtime/size/tbox_group/test/testcase/ucode:
ce/x86_64-rhel-7.2/debian-x86_64-2018-04-03.cgz/300s/1T/lkp-ivb-d02/msync/vm-scalability/0x20
(" platform-drivers-x86 for v4.17-4")
("mm: fix COW faults after mlock()")
dd12385915f4f83b738467e599
--------------------------
%reproduction fail:runs
| |
-25% :4 dmesg.WARNING:at#for_ip_error_entry/0x
18% 1:4 perf-profile.calltrace.cycles-pp.error_entry
-18% 2:4 perf-profile.children.cycles-pp.error_entry
0% 0:4 perf-profile.children.cycles-pp.schedule_timeout
-17% 1:4 perf-profile.self.cycles-pp.error_entry
%change %stddev
| \
+18.6% 329586 ± 6% vm-scalability.median
-46.9% 0.24 ± 27% vm-scalability.median_stddev
+5.7e+06% 1.608e+08 ± 6% vm-scalability.time.file_system_inputs
-73.7% 1714 ± 3% vm-scalability.time.major_page_faults
-40.0% 85.50 ± 4% vm-scalability.time.percent_of_cpu_this_job_got
-35.2% 182.76 ± 5% vm-scalability.time.system_time
-46.3% 81.85 ± 5% vm-scalability.time.user_time
-31.1% 309105 ± 8% vm-scalability.time.voluntary_context_switches
-44.9% 1.154e+08 ± 5% vm-scalability.workload
-27.3% 13664118 ± 6% interrupts.CAL:Function_call_interrupts
-10.4% 79618 ± 2% softirqs.RCU
+31.5% 132623 ± 2% softirqs.SCHED
+2.8e+06% 256051 ± 5% vmstat.io.bi
-50.0% 3.00 vmstat.memory.buff
-25.0% 50893 ± 5% vmstat.system.in
+2.4 16.47 ± 7% mpstat.cpu.idle%
+13.5 43.20 ± 2% mpstat.cpu.iowait%
-10.1 32.62 ± 5% mpstat.cpu.sys%
-5.9 7.70 ± 4% mpstat.cpu.usr%
+88.5% 555281 slabinfo.buffer_head.active_objs
+88.5% 14248 slabinfo.buffer_head.active_slabs
+88.5% 555706 slabinfo.buffer_head.num_objs
+88.5% 14248 slabinfo.buffer_head.num_slabs
-9.7% 738.25 ± 7% slabinfo.proc_inode_cache.num_objs
+24.4% 565.00 ± 18% slabinfo.skbuff_head_cache.active_objs
+56.9% 21419299 ± 8% cpuidle.C1.time
+51.1% 2.919e+08 ± 2% cpuidle.C1E.time
+33.4% 678785 ± 2% cpuidle.C1E.usage
+73.1% 28943328 cpuidle.C3.time
+78.8% 38298 cpuidle.C3.usage
+29.2% 3.879e+08 ± 4% cpuidle.C6.time
+28.8% 402526 ± 4% cpuidle.C6.usage
+61.4% 173441 ± 12% cpuidle.POLL.time
-25.9% 13876 ± 6% cpuidle.POLL.usage
+17.7% 1263701 meminfo.Active(file)
+18.0% 364452 ± 2% meminfo.Dirty
-20.9% 1061562 ± 3% meminfo.Inactive
-39.3% 104801 ± 3% meminfo.Inactive(anon)
-18.2% 956761 ± 3% meminfo.Inactive(file)
-20.3% 10968 ± 2% meminfo.PageTables
+29.8% 111114 meminfo.SReclaimable
+23.8% 132534 meminfo.Slab
+23.5% 46379 ± 4% meminfo.Writeback
-26.0% 62266 ± 5% sched_debug.cfs_rq:/.exec_clock.avg
-24.6% 64822 ± 5% sched_debug.cfs_rq:/.exec_clock.max
-26.6% 60788 ± 5% sched_debug.cfs_rq:/.exec_clock.min
-41.9% 121723 ± 4% sched_debug.cfs_rq:/.min_vruntime.avg
-40.0% 130873 ± 3% sched_debug.cfs_rq:/.min_vruntime.max
-43.1% 113416 ± 7% sched_debug.cfs_rq:/.min_vruntime.min
-16.2% 1172 ± 11% sched_debug.cpu.curr->pid.stddev
+104.2% 32.42 ± 35% sched_debug.cpu.nr_uninterruptible.max
+91.2% -27.25 sched_debug.cpu.nr_uninterruptible.min
+98.0% 22.99 ± 28% sched_debug.cpu.nr_uninterruptible.stddev
+26.0% 160648 ± 4% sched_debug.cpu.sched_goidle.min
-91.3% 0.00 ±158% sched_debug.rt_rq:/.rt_time.stddev
-27.7% 1358 ± 5% turbostat.Avg_MHz
-15.7 41.70 ± 4% turbostat.Busy%
+0.6 1.71 ± 7% turbostat.C1%
+33.4% 678785 ± 2% turbostat.C1E
+7.6 23.32 ± 2% turbostat.C1E%
+78.8% 38298 turbostat.C3
+1.0 2.31 turbostat.C3%
+28.8% 402527 ± 4% turbostat.C6
+6.6 31.00 ± 5% turbostat.C6%
+39.3% 42.01 ± 3% turbostat.CPU%c1
+60.0% 1.75 ± 6% turbostat.CPU%c3
+28.3% 14.54 ± 6% turbostat.CPU%c6
-18.3% 10.43 ± 3% turbostat.CorWatt
-25.5% 29652069 ± 6% turbostat.IRQ
-8.0% 27.69 turbostat.PkgWatt
-41.5% 2.869e+11 ± 28% perf-stat.branch-instructions
-32.3% 2.461e+09 ± 28% perf-stat.branch-misses
-7.3 58.95 perf-stat.cache-miss-rate%
-26.2% 7.2e+09 ± 28% perf-stat.cache-misses
+5.9% 1.17 perf-stat.cpi
-35.6% 1.48e+12 ± 27% perf-stat.cpu-cycles
+14.7% 22028 ± 5% perf-stat.cpu-migrations
-42.9% 2.205e+09 ± 16% perf-stat.dTLB-load-misses
-36.1% 3.332e+11 ± 28% perf-stat.dTLB-loads
-0.0 0.09 ± 18% perf-stat.dTLB-store-miss-rate%
-52.5% 1.638e+08 ± 17% perf-stat.dTLB-store-misses
-35.2% 1.839e+11 ± 28% perf-stat.dTLB-stores
+7.5 81.93 perf-stat.iTLB-load-miss-rate%
-45.1% 2.024e+08 ± 27% perf-stat.iTLB-load-misses
-65.7% 43583231 ± 21% perf-stat.iTLB-loads
-39.0% 1.266e+12 ± 28% perf-stat.instructions
+10.3% 6228 ± 4% perf-stat.instructions-per-iTLB-miss
-5.5% 0.85 perf-stat.ipc
-44.5% 26119814 ± 5% perf-stat.minor-faults
-44.5% 26120941 ± 5% perf-stat.page-faults
-78.5% 4857 ± 4% proc-vmstat.allocstall_movable
-86.8% 15866 ± 5% proc-vmstat.allocstall_normal
-78.7% 1631 ± 11% proc-vmstat.compact_fail
-78.4% 1659 ± 11% proc-vmstat.compact_stall
-37.9% 1710 ± 4% proc-vmstat.kswapd_low_wmark_hit_quickly
+17.7% 314329 ± 2% proc-vmstat.nr_active_file
+17.3% 91559 proc-vmstat.nr_dirty
-21.7% 6118 ± 28% proc-vmstat.nr_free_cma
-39.4% 26248 ± 3% proc-vmstat.nr_inactive_anon
-17.7% 241780 ± 4% proc-vmstat.nr_inactive_file
-20.5% 2740 proc-vmstat.nr_page_table_pages
+29.8% 27796 proc-vmstat.nr_slab_reclaimable
+26.1% 11542 ± 2% proc-vmstat.nr_writeback
+17.7% 314320 ± 2% proc-vmstat.nr_zone_active_file
-39.4% 26251 ± 3% proc-vmstat.nr_zone_inactive_anon
-17.7% 241712 ± 4% proc-vmstat.nr_zone_inactive_file
+18.2% 103074 proc-vmstat.nr_zone_write_pending
-43.5% 48347251 ± 5% proc-vmstat.numa_hit
-43.5% 48347251 ± 5% proc-vmstat.numa_local
-35.3% 1913 ± 3% proc-vmstat.pageoutrun
-43.2% 29135174 ± 5% proc-vmstat.pgactivate
-18.2% 23322955 ± 5% proc-vmstat.pgalloc_dma32
-56.0% 25151136 ± 8% proc-vmstat.pgalloc_normal
-46.5% 22026100 ± 6% proc-vmstat.pgdeactivate
-43.4% 48485032 ± 5% proc-vmstat.pgfree
-73.7% 1714 ± 3% proc-vmstat.pgmajfault
+2.7e+06% 80419921 ± 6% proc-vmstat.pgpgin
-46.5% 22026099 ± 6% proc-vmstat.pgrefill
+72.3% 822527 ± 11% proc-vmstat.pgrotated
-52.5% 20878181 ± 5% proc-vmstat.pgscan_direct
-41.7% 49530834 ± 5% proc-vmstat.pgscan_kswapd
-85.2% 1196177 ± 5% proc-vmstat.pgsteal_direct
-39.2% 28921712 ± 6% proc-vmstat.pgsteal_kswapd
-13.1% 12538 ± 3% proc-vmstat.slabs_scanned
-66.2% 682455 ± 8% proc-vmstat.workingset_activate
-44.0% 9754788 ± 5% proc-vmstat.workingset_refault
-17.5 2.42 ±110% perf-profile.calltrace.cycles-pp.do_access
-10.7 1.50 ±109% perf-profile.calltrace.cycles-pp.page_fault.do_access
-10.7 1.50 ±109% perf-profile.calltrace.cycles-pp.do_page_fault.page_fault.do_access
-10.7 1.50 ±109% perf-profile.calltrace.cycles-pp.__do_page_fault.do_page_fault.page_fault.do_access
-10.0 1.37 ±109% perf-profile.calltrace.cycles-pp.handle_mm_fault.__do_page_fault.do_page_fault.page_fault.do_access
-8.6 7.58 ± 25% perf-profile.calltrace.cycles-pp.__handle_mm_fault.handle_mm_fault.__do_page_fault.do_page_fault.page_fault
-8.1 1.29 ±110% perf-profile.calltrace.cycles-pp.do_rw_once
-4.5 3.05 ± 30% perf-profile.calltrace.cycles-pp.do_page_mkwrite.__handle_mm_fault.handle_mm_fault.__do_page_fault.do_page_fault
-4.5 3.00 ± 30% perf-profile.calltrace.cycles-pp.__xfs_filemap_fault.do_page_mkwrite.__handle_mm_fault.handle_mm_fault.__do_page_fault
-3.3 3.44 ± 23% perf-profile.calltrace.cycles-pp.__do_fault.__handle_mm_fault.handle_mm_fault.__do_page_fault.do_page_fault
-3.2 3.38 ± 23% perf-profile.calltrace.cycles-pp.__xfs_filemap_fault.__do_fault.__handle_mm_fault.handle_mm_fault.__do_page_fault
-3.0 11.27 ± 7% perf-profile.calltrace.cycles-pp.__do_page_cache_readahead.ondemand_readahead.filemap_fault.__xfs_filemap_fault.__do_fault
-3.0 11.29 ± 8% perf-profile.calltrace.cycles-pp.ondemand_readahead.filemap_fault.__xfs_filemap_fault.__do_fault.__handle_mm_fault
-2.7 12.73 ± 6% perf-profile.calltrace.cycles-pp.filemap_fault.__xfs_filemap_fault.__do_fault.__handle_mm_fault.handle_mm_fault
-2.1 0.56 ±116% perf-profile.calltrace.cycles-pp.__alloc_pages_slowpath.__alloc_pages_nodemask.__do_page_cache_readahead.ondemand_readahead.filemap_fault
-2.1 2.10 ± 12% perf-profile.calltrace.cycles-pp.__alloc_pages_nodemask.__do_page_cache_readahead.ondemand_readahead.filemap_fault.__xfs_filemap_fault
-1.6 0.89 ± 68% perf-profile.calltrace.cycles-pp.shrink_page_list.shrink_inactive_list.shrink_node_memcg.shrink_node.do_try_to_free_pages
-1.5 2.06 ± 40% perf-profile.calltrace.cycles-pp.shrink_inactive_list.shrink_node_memcg.shrink_node.do_try_to_free_pages.try_to_free_pages
-1.4 7.95 ± 14% perf-profile.calltrace.cycles-pp.mpage_readpages.__do_page_cache_readahead.ondemand_readahead.filemap_fault.__xfs_filemap_fault
-0.6 2.95 ± 15% perf-profile.calltrace.cycles-pp.add_to_page_cache_lru.mpage_readpages.__do_page_cache_readahead.ondemand_readahead.filemap_fault
+0.5 0.81 ± 17% perf-profile.calltrace.cycles-pp.xfs_bmap_add_extent_hole_delay.xfs_bmapi_reserve_delalloc.xfs_file_iomap_begin.iomap_apply.iomap_page_mkwrite
+0.6 0.63 ± 22% perf-profile.calltrace.cycles-pp.follow_page_pte.__get_user_pages.populate_vma_page_range.__mm_populate.vm_mmap_pgoff
+0.8 1.10 ± 40% perf-profile.calltrace.cycles-pp.__set_page_dirty.mark_buffer_dirty.__block_commit_write.block_commit_write.iomap_page_mkwrite_actor
+1.0 1.59 ± 26% perf-profile.calltrace.cycles-pp.mark_buffer_dirty.__block_commit_write.block_commit_write.iomap_page_mkwrite_actor.iomap_apply
+1.1 1.06 ± 36% perf-profile.calltrace.cycles-pp.alloc_set_pte.finish_fault.__handle_mm_fault.handle_mm_fault.__get_user_pages
+1.1 1.11 ± 34% perf-profile.calltrace.cycles-pp.finish_fault.__handle_mm_fault.handle_mm_fault.__get_user_pages.populate_vma_page_range
+1.2 3.35 ± 19% perf-profile.calltrace.cycles-pp.xfs_file_iomap_begin.iomap_apply.iomap_page_mkwrite.__xfs_filemap_fault.do_page_mkwrite
+1.2 2.35 ± 47% perf-profile.calltrace.cycles-pp.kmem_cache_alloc.alloc_buffer_head.alloc_page_buffers.create_empty_buffers.create_page_buffers
+1.2 2.46 ± 45% perf-profile.calltrace.cycles-pp.alloc_page_buffers.create_empty_buffers.create_page_buffers.__block_write_begin_int.iomap_page_mkwrite_actor
+1.2 2.43 ± 46% perf-profile.calltrace.cycles-pp.alloc_buffer_head.alloc_page_buffers.create_empty_buffers.create_page_buffers.__block_write_begin_int
+1.4 2.98 ± 35% perf-profile.calltrace.cycles-pp.create_empty_buffers.create_page_buffers.__block_write_begin_int.iomap_page_mkwrite_actor.iomap_apply
+1.4 3.04 ± 33% perf-profile.calltrace.cycles-pp.create_page_buffers.__block_write_begin_int.iomap_page_mkwrite_actor.iomap_apply.iomap_page_mkwrite
+2.0 3.86 ± 38% perf-profile.calltrace.cycles-pp.__block_write_begin_int.iomap_page_mkwrite_actor.iomap_apply.iomap_page_mkwrite.__xfs_filemap_fault
+2.1 2.09 ± 50% perf-profile.calltrace.cycles-pp.memcpy_erms.memcpy_to_page._copy_to_iter.copy_page_to_iter.shmem_file_read_iter
+2.1 2.11 ± 50% perf-profile.calltrace.cycles-pp.memcpy_to_page._copy_to_iter.copy_page_to_iter.shmem_file_read_iter.do_iter_readv_writev
+2.2 2.19 ± 50% perf-profile.calltrace.cycles-pp._copy_to_iter.copy_page_to_iter.shmem_file_read_iter.do_iter_readv_writev.do_iter_read
+2.2 2.22 ± 50% perf-profile.calltrace.cycles-pp.copy_page_to_iter.shmem_file_read_iter.do_iter_readv_writev.do_iter_read.loop_queue_work
+2.9 6.13 ± 26% perf-profile.calltrace.cycles-pp.iomap_page_mkwrite_actor.iomap_apply.iomap_page_mkwrite.__xfs_filemap_fault.do_page_mkwrite
+3.0 3.00 ± 51% perf-profile.calltrace.cycles-pp.shmem_file_read_iter.do_iter_readv_writev.do_iter_read.loop_queue_work.kthread_worker_fn
+3.1 3.06 ± 51% perf-profile.calltrace.cycles-pp.do_iter_readv_writev.do_iter_read.loop_queue_work.kthread_worker_fn.kthread
+3.3 3.26 ± 52% perf-profile.calltrace.cycles-pp.do_iter_read.loop_queue_work.kthread_worker_fn.kthread.ret_from_fork
+4.1 10.16 ± 21% perf-profile.calltrace.cycles-pp.iomap_apply.iomap_page_mkwrite.__xfs_filemap_fault.do_page_mkwrite.__handle_mm_fault
+4.4 11.01 ± 19% perf-profile.calltrace.cycles-pp.iomap_page_mkwrite.__xfs_filemap_fault.do_page_mkwrite.__handle_mm_fault.handle_mm_fault
+5.9 8.94 ± 38% perf-profile.calltrace.cycles-pp.secondary_startup_64
+9.5 9.53 ± 27% perf-profile.calltrace.cycles-pp.__xfs_filemap_fault.do_page_mkwrite.__handle_mm_fault.handle_mm_fault.__get_user_pages
+9.6 9.59 ± 27% perf-profile.calltrace.cycles-pp.do_page_mkwrite.__handle_mm_fault.handle_mm_fault.__get_user_pages.populate_vma_page_range
+10.1 21.96 ± 18% perf-profile.calltrace.cycles-pp.__handle_mm_fault.handle_mm_fault.__get_user_pages.populate_vma_page_range.__mm_populate
+10.4 22.23 ± 18% perf-profile.calltrace.cycles-pp.handle_mm_fault.__get_user_pages.populate_vma_page_range.__mm_populate.vm_mmap_pgoff
+11.2 23.77 ± 17% perf-profile.calltrace.cycles-pp.__get_user_pages.populate_vma_page_range.__mm_populate.vm_mmap_pgoff.ksys_mmap_pgoff
+11.2 23.77 ± 17% perf-profile.calltrace.cycles-pp.__mm_populate.vm_mmap_pgoff.ksys_mmap_pgoff.do_syscall_64.entry_SYSCALL_64_after_hwframe
+11.2 23.77 ± 17% perf-profile.calltrace.cycles-pp.populate_vma_page_range.__mm_populate.vm_mmap_pgoff.ksys_mmap_pgoff.do_syscall_64
-17.5 2.42 ±110% perf-profile.children.cycles-pp.do_access
-9.2 8.71 ± 25% perf-profile.children.cycles-pp.__do_page_fault
-9.2 8.69 ± 25% perf-profile.children.cycles-pp.do_page_fault
-9.2 8.73 ± 25% perf-profile.children.cycles-pp.page_fault
-8.1 1.29 ±110% perf-profile.children.cycles-pp.do_rw_once
-3.3 11.29 ± 8% perf-profile.children.cycles-pp.__do_page_cache_readahead
-3.2 11.29 ± 7% perf-profile.children.cycles-pp.ondemand_readahead
-2.8 12.78 ± 7% perf-profile.children.cycles-pp.filemap_fault
-2.6 13.46 ± 6% perf-profile.children.cycles-pp.__do_fault
-1.6 5.70 ± 14% perf-profile.children.cycles-pp.__alloc_pages_nodemask
-1.6 4.13 ± 17% perf-profile.children.cycles-pp.__alloc_pages_slowpath
-1.5 3.85 ± 16% perf-profile.children.cycles-pp.try_to_free_pages
-1.5 3.85 ± 16% perf-profile.children.cycles-pp.do_try_to_free_pages
-1.4 2.76 ± 29% perf-profile.children.cycles-pp.page_vma_mapped_walk
-1.4 8.19 ± 12% perf-profile.children.cycles-pp.mpage_readpages
-1.3 0.48 ± 69% perf-profile.children.cycles-pp.pte_alloc_one
-1.1 1.77 ± 35% perf-profile.children.cycles-pp.page_referenced_one
-0.7 0.07 ±113% perf-profile.children.cycles-pp.filemap_map_pages
-0.6 0.08 ± 70% perf-profile.children.cycles-pp.shrink_slab
-0.5 0.66 ± 29% perf-profile.children.cycles-pp.free_pcppages_bulk
-0.5 0.82 ± 38% perf-profile.children.cycles-pp.free_unref_page_list
-0.4 3.24 ± 9% perf-profile.children.cycles-pp.add_to_page_cache_lru
-0.3 0.46 ± 16% perf-profile.children.cycles-pp.__radix_tree_replace
-0.1 0.15 ± 53% perf-profile.children.cycles-pp.find_vma
-0.1 1.12 ± 6% perf-profile.children.cycles-pp.pagevec_lru_move_fn
-0.1 0.15 ± 23% perf-profile.children.cycles-pp.replace_slot
-0.1 1.17 ± 8% perf-profile.children.cycles-pp.__lru_cache_add
-0.1 0.06 ± 65% perf-profile.children.cycles-pp.drain_local_pages_wq
-0.1 0.06 ± 65% perf-profile.children.cycles-pp.drain_pages
-0.1 0.06 ± 65% perf-profile.children.cycles-pp.drain_pages_zone
-0.1 0.69 ± 16% perf-profile.children.cycles-pp.page_cache_tree_insert
-0.1 0.46 ± 20% perf-profile.children.cycles-pp.unmap_page_range
-0.1 0.46 ± 19% perf-profile.children.cycles-pp.unmap_vmas
-0.1 0.10 ± 50% perf-profile.children.cycles-pp.page_mapped
-0.1 0.04 ± 58% perf-profile.children.cycles-pp.super_cache_count
-0.1 0.09 ± 27% perf-profile.children.cycles-pp.ptep_clear_flush_young
-0.1 0.03 ±100% perf-profile.children.cycles-pp.down_read_trylock
+0.1 0.06 ± 11% perf-profile.children.cycles-pp.find_next_and_bit
+0.1 0.14 ± 34% perf-profile.children.cycles-pp.task_tick_fair
+0.1 0.08 ± 24% perf-profile.children.cycles-pp.cpumask_next_and
+0.1 0.13 ± 30% perf-profile.children.cycles-pp.pmd_devmap_trans_unstable
+0.1 0.09 ± 21% perf-profile.children.cycles-pp.xfs_fsb_to_db
+0.1 0.14 ± 18% perf-profile.children.cycles-pp.kthread_queue_work
+0.1 0.15 ± 11% perf-profile.children.cycles-pp.loop_queue_rq
+0.1 0.19 ± 27% perf-profile.children.cycles-pp.memset_erms
+0.1 0.22 ± 15% perf-profile.children.cycles-pp.current_kernel_time64
+0.1 0.19 ± 7% perf-profile.children.cycles-pp.blk_mq_dispatch_rq_list
+0.1 0.12 ± 65% perf-profile.children.cycles-pp.touch_atime
+0.1 0.27 ± 31% perf-profile.children.cycles-pp.xfs_file_iomap_end
+0.1 0.23 ± 8% perf-profile.children.cycles-pp.blk_mq_run_hw_queue
+0.1 0.23 ± 6% perf-profile.children.cycles-pp.__blk_mq_run_hw_queue
+0.1 0.23 ± 6% perf-profile.children.cycles-pp.blk_mq_sched_dispatch_requests
+0.1 0.23 ± 7% perf-profile.children.cycles-pp.blk_mq_do_dispatch_sched
+0.1 0.34 ± 18% perf-profile.children.cycles-pp.current_time
+0.2 0.51 ± 14% perf-profile.children.cycles-pp.try_to_wake_up
+0.2 0.63 ± 5% perf-profile.children.cycles-pp.set_page_dirty
+0.2 0.44 ± 24% perf-profile.children.cycles-pp.xfs_vn_update_time
+0.2 0.23 ± 7% perf-profile.children.cycles-pp.__blk_mq_delay_run_hw_queue
+0.2 0.25 ± 46% perf-profile.children.cycles-pp.update_load_avg
+0.2 0.98 ± 6% perf-profile.children.cycles-pp._raw_spin_lock_irqsave
+0.2 0.73 ± 3% perf-profile.children.cycles-pp.radix_tree_tag_clear
+0.2 0.65 ± 5% perf-profile.children.cycles-pp.xfs_iunlock
+0.2 0.28 ± 15% perf-profile.children.cycles-pp.blk_mq_flush_plug_list
+0.2 0.30 ± 11% perf-profile.children.cycles-pp.blk_flush_plug_list
+0.2 0.61 ± 19% perf-profile.children.cycles-pp.tick_sched_timer
+0.2 0.69 ± 16% perf-profile.children.cycles-pp._raw_spin_unlock_irqrestore
+0.2 0.20 ± 19% perf-profile.children.cycles-pp.blk_finish_plug
+0.2 0.23 ± 61% perf-profile.children.cycles-pp.mpage_end_io
+0.2 1.68 ± 4% perf-profile.children.cycles-pp.__radix_tree_lookup
+0.3 0.70 ± 18% perf-profile.children.cycles-pp.follow_page_pte
+0.3 0.64 ± 18% perf-profile.children.cycles-pp.blk_mq_make_request
+0.3 0.42 ± 83% perf-profile.children.cycles-pp.menu_select
+0.3 0.94 ± 20% perf-profile.children.cycles-pp.file_update_time
+0.3 0.76 ± 28% perf-profile.children.cycles-pp.radix_tree_tag_set
+0.3 0.70 ± 17% perf-profile.children.cycles-pp.submit_bio
+0.3 0.70 ± 17% perf-profile.children.cycles-pp.generic_make_request
+0.4 1.11 ± 8% perf-profile.children.cycles-pp.radix_tree_lookup_slot
+0.4 0.46 ± 67% perf-profile.children.cycles-pp.__pte_alloc
+0.4 1.06 ± 36% perf-profile.children.cycles-pp.__hrtimer_run_queues
+0.4 0.64 ± 12% perf-profile.children.cycles-pp.follow_page_mask
+0.5 1.46 ± 21% perf-profile.children.cycles-pp.alloc_set_pte
+0.5 1.59 ± 19% perf-profile.children.cycles-pp.xfs_vm_set_page_dirty
+0.7 1.83 ± 10% perf-profile.children.cycles-pp.find_get_entry
+0.7 1.42 ± 19% perf-profile.children.cycles-pp.finish_fault
+0.8 0.77 ± 70% perf-profile.children.cycles-pp.mempool_alloc
+0.8 0.77 ± 71% perf-profile.children.cycles-pp.bio_alloc_bioset
+0.8 2.06 ± 50% perf-profile.children.cycles-pp.smp_apic_timer_interrupt
+0.9 1.86 ± 20% perf-profile.children.cycles-pp.mark_buffer_dirty
+0.9 2.09 ± 50% perf-profile.children.cycles-pp.apic_timer_interrupt
+1.0 2.23 ± 16% perf-profile.children.cycles-pp.__block_commit_write
+1.0 2.22 ± 16% perf-profile.children.cycles-pp.block_commit_write
+1.1 3.69 ± 16% perf-profile.children.cycles-pp.xfs_file_iomap_begin
+1.4 2.77 ± 41% perf-profile.children.cycles-pp.alloc_buffer_head
+1.5 2.87 ± 41% perf-profile.children.cycles-pp.alloc_page_buffers
+1.5 3.29 ± 36% perf-profile.children.cycles-pp.create_empty_buffers
+1.6 3.34 ± 34% perf-profile.children.cycles-pp.create_page_buffers
+1.8 4.00 ± 35% perf-profile.children.cycles-pp.__block_write_begin_int
+1.9 2.71 ± 31% perf-profile.children.cycles-pp.new_slab
+1.9 2.76 ± 30% perf-profile.children.cycles-pp.___slab_alloc
+1.9 2.76 ± 30% perf-profile.children.cycles-pp.__slab_alloc
+2.1 2.11 ± 50% perf-profile.children.cycles-pp.memcpy_to_page
+2.2 2.20 ± 50% perf-profile.children.cycles-pp._copy_to_iter
+2.2 3.61 ± 26% perf-profile.children.cycles-pp.kmem_cache_alloc
+2.3 2.25 ± 50% perf-profile.children.cycles-pp.copy_page_to_iter
+3.0 6.38 ± 26% perf-profile.children.cycles-pp.iomap_page_mkwrite_actor
+3.0 3.01 ± 51% perf-profile.children.cycles-pp.shmem_file_read_iter
+3.3 3.30 ± 53% perf-profile.children.cycles-pp.do_iter_read
+4.2 10.27 ± 21% perf-profile.children.cycles-pp.iomap_apply
+4.3 6.46 ± 22% perf-profile.children.cycles-pp.intel_idle
+4.4 11.08 ± 19% perf-profile.children.cycles-pp.iomap_page_mkwrite
+5.0 12.65 ± 17% perf-profile.children.cycles-pp.do_page_mkwrite
+5.5 8.29 ± 35% perf-profile.children.cycles-pp.cpuidle_enter_state
+5.9 8.94 ± 38% perf-profile.children.cycles-pp.secondary_startup_64
+5.9 8.94 ± 38% perf-profile.children.cycles-pp.cpu_startup_entry
+6.0 8.97 ± 38% perf-profile.children.cycles-pp.do_idle
+10.9 27.21 ± 15% perf-profile.children.cycles-pp.do_syscall_64
+11.0 27.24 ± 15% perf-profile.children.cycles-pp.entry_SYSCALL_64_after_hwframe
+11.2 23.80 ± 17% perf-profile.children.cycles-pp.vm_mmap_pgoff
+11.2 23.77 ± 17% perf-profile.children.cycles-pp.__mm_populate
+11.2 23.77 ± 17% perf-profile.children.cycles-pp.populate_vma_page_range
+11.3 23.80 ± 17% perf-profile.children.cycles-pp.ksys_mmap_pgoff
+11.3 23.81 ± 17% perf-profile.children.cycles-pp.__get_user_pages
-7.8 1.25 ±110% perf-profile.self.cycles-pp.do_rw_once
-3.4 0.42 ±108% perf-profile.self.cycles-pp.do_access
-1.0 1.88 ± 30% perf-profile.self.cycles-pp.page_vma_mapped_walk
-0.9 3.35 ± 18% perf-profile.self.cycles-pp.do_mpage_readpage
-0.4 0.70 ± 41% perf-profile.self.cycles-pp.page_referenced_one
-0.1 0.24 ± 28% perf-profile.self.cycles-pp.xfs_get_blocks
-0.1 0.15 ± 23% perf-profile.self.cycles-pp.replace_slot
-0.1 0.26 ± 33% perf-profile.self.cycles-pp.page_referenced
-0.1 0.10 ± 50% perf-profile.self.cycles-pp.page_mapped
-0.1 0.09 ± 27% perf-profile.self.cycles-pp.ptep_clear_flush_young
-0.1 0.03 ±100% perf-profile.self.cycles-pp.down_read_trylock
-0.0 0.14 ± 12% perf-profile.self.cycles-pp.mem_cgroup_commit_charge
+0.0 0.12 ± 21% perf-profile.self.cycles-pp.kmem_cache_free
+0.1 0.06 ± 11% perf-profile.self.cycles-pp.find_next_and_bit
+0.1 0.15 ± 19% perf-profile.self.cycles-pp.iomap_page_mkwrite_actor
+0.1 0.08 ± 23% perf-profile.self.cycles-pp.radix_tree_lookup_slot
+0.1 0.09 ± 21% perf-profile.self.cycles-pp.xfs_fsb_to_db
+0.1 0.16 ± 14% perf-profile.self.cycles-pp.__mark_inode_dirty
+0.1 0.16 ± 31% perf-profile.self.cycles-pp.memset_erms
+0.1 0.08 ± 48% perf-profile.self.cycles-pp.memcpy_from_page
+0.1 0.28 ± 12% perf-profile.self.cycles-pp.lock_page_memcg
+0.1 0.22 ± 30% perf-profile.self.cycles-pp.mark_buffer_dirty
+0.1 0.25 ± 12% perf-profile.self.cycles-pp.__block_commit_write
+0.1 0.25 ± 32% perf-profile.self.cycles-pp.xfs_add_to_ioend
+0.1 0.28 ± 15% perf-profile.self.cycles-pp.set_page_dirty
+0.1 0.22 ± 15% perf-profile.self.cycles-pp.current_kernel_time64
+0.1 0.19 ± 47% perf-profile.self.cycles-pp.security_file_permission
+0.1 0.18 ± 71% perf-profile.self.cycles-pp.update_load_avg
+0.1 0.27 ± 24% perf-profile.self.cycles-pp.__xfs_filemap_fault
+0.1 0.38 ± 26% perf-profile.self.cycles-pp.xfs_ilock
+0.1 0.59 ± 5% perf-profile.self.cycles-pp.radix_tree_tag_clear
+0.2 0.34 ± 15% perf-profile.self.cycles-pp.follow_page_pte
+0.2 0.18 ± 43% perf-profile.self.cycles-pp.shmem_file_read_iter
+0.2 1.62 ± 3% perf-profile.self.cycles-pp.__radix_tree_lookup
+0.2 0.36 ± 33% perf-profile.self.cycles-pp.xfs_bmapi_reserve_delalloc
+0.2 0.53 ± 29% perf-profile.self.cycles-pp.kmem_cache_alloc
+0.2 0.76 ± 10% perf-profile.self.cycles-pp._raw_spin_lock_irqsave
+0.2 0.61 ± 23% perf-profile.self.cycles-pp.__block_write_begin_int
+0.3 0.48 ± 13% perf-profile.self.cycles-pp.follow_page_mask
+0.3 0.34 ± 51% perf-profile.self.cycles-pp.__get_user_pages
+0.3 0.76 ± 28% perf-profile.self.cycles-pp.radix_tree_tag_set
+0.3 0.79 ± 11% perf-profile.self.cycles-pp.find_get_entry
+0.4 0.82 ± 18% perf-profile.self.cycles-pp.xfs_file_iomap_begin
+0.4 0.98 ± 16% perf-profile.self.cycles-pp.__handle_mm_fault
+4.3 6.46 ± 22% perf-profile.self.cycles-pp.intel_idle

vm-scalability.time.user_time

-----------------------------------------------------------+
.+..+.+.. .+.. .+.. + +.. |
+. +..+.+. +.. + +.. .+ |
+ +. |
|
|
|
|
|
|
|
O O O O |
O O O O O O O O O O O O O O
O O O |
-----------------------------------------------------------+

vm-scalability.time.system_time

-----------------------------------------------------------+
+.. .+. .+ .+.. |
.. +. +..+..+. + .+. +.+..+..+ |
.+.+ +. |
|
|
|
|
|
|
|
O |
O O O O |
O O O O O O O O O O O O O O O O
-----------------------------------------------------------+

vm-scalability.time.percent_of_cpu_this_job_got

-----------------------------------------------------------+
.+.. .+.. .+. .+.. .+.+.. |
.+ +. +. +. +..+. +..+..+ |
|
|
|
|
|
|
|
|
O |
O O O O O O O O |
O O O O O O O O O O O O
-----------------------------------------------------------+

vm-scalability.time.major_page_faults

-----------------------------------------------------------+
|
|
|
|
+.. .+. .+..+. + +.. |
+ +. +..+. +..+.. .. + .. |
+..+.. + + + + |
+.+ |
|
|
O O |
O O O O O O O
O O O O O O O O O O O |
-------------------------------O---------------------------+

vm-scalability.time.file_system_inputs

+-+---------------------------------------------------------------+
|
O O |
|
O O O O
O O |
|
|
|
|
O O O O O O O O O |
O O O O |
|
|
0 +-+---------------------------------------------------------------+

vm-scalability.time.file_system_outputs

+-+O-O--O--O------O-----O-O-----O----O-------O--------------------+
|
|
O O O O O O O |
|
|
O O O O O O O O
|
|
|
|
.+.+..+.. .+.. .+.. .+.. |
+.+. + +..+.+. +.+. +.+..+.+ |
|
+-+---------------------------------------------------------------+

vm-scalability.workload

+-+---------------------------------------------------------------+
+.+..+.. +.. +.. +.. |
.. + .. .. |
+.+ + +..+.+ +.+ +.+..+.+ |
|
|
|
|
|
O O O O O O |
|
O O O O O O O |
|
O O O O O O O O
+-+---------------------------------------------------------------+

have been estimated based on internal Intel analysis and are provided
purposes only. Any difference in system hardware or software
configuration may affect actual performance.

Attachments:

(No filename) (46.81 kB)
config-4.17.0-rc7-00044-gdd12385 (167.16 kB)
job-script (7.31 kB)
job.yaml (4.95 kB)
reproduce (157.71 kB)
Download all attachments