| Message ID | 20180913211923.7696-1-timofey.titovets@synesis.ru (mailing list archive) |
|---|---|
| Headers | show |
| Series | KSM replace hash algo with xxhash | expand |
On 9/13/18 5:19 PM, Timofey Titovets wrote: > From: Timofey Titovets <nefelim4ag@gmail.com> > > Currently used jhash are slow enough and replace it allow as to make KSM > less cpu hungry. > > About speed (in kernel): > ksm: crc32c hash() 12081 MB/s > ksm: xxh64 hash() 8770 MB/s > ksm: xxh32 hash() 4529 MB/s > ksm: jhash2 hash() 1569 MB/s > > By sioh Lee tests (copy from other mail): > Test platform: openstack cloud platform (NEWTON version) > Experiment node: openstack based cloud compute node (CPU: xeon E5-2620 v3, memory 64gb) > VM: (2 VCPU, RAM 4GB, DISK 20GB) * 4 > Linux kernel: 4.14 (latest version) > KSM setup - sleep_millisecs: 200ms, pages_to_scan: 200 > > Experiment process > Firstly, we turn off KSM and launch 4 VMs. > Then we turn on the KSM and measure the checksum computation time until full_scans become two. > > The experimental results (the experimental value is the average of the measured values) > crc32c_intel: 1084.10ns > crc32c (no hardware acceleration): 7012.51ns > xxhash32: 2227.75ns > xxhash64: 1413.16ns > jhash2: 5128.30ns > > In summary, the result shows that crc32c_intel has advantages over all > of the hash function used in the experiment. (decreased by 84.54% compared to crc32c, > 78.86% compared to jhash2, 51.33% xxhash32, 23.28% compared to xxhash64) > the results are similar to those of Timofey. > > But, > use only xxhash for now, because for using crc32c, > cryptoapi must be initialized first - that require some > tricky solution to work good in all situations. > > So: > - Fisrt patch implement compile time pickup of fastest implementation of xxhash > for target platform. > - Second replace jhash2 with xxhash > > Thanks. > > CC: Andrea Arcangeli <aarcange@redhat.com> > CC: linux-mm@kvack.org > CC: kvm@vger.kernel.org > CC: leesioh <solee@os.korea.ac.kr> > > Timofey Titovets (2): > xxHash: create arch dependent 32/64-bit xxhash() > ksm: replace jhash2 with xxhash > > include/linux/xxhash.h | 23 +++++++++++++ > mm/Kconfig | 2 ++ > mm/ksm.c | 93 +++++++++++++++++++++++++++++++++++++++++++++++--- > 3 files changed, 114 insertions(+), 4 deletions(-) This is wrong stat. ksm.c should not have any new lines at all.
пт, 14 сент. 2018 г. в 0:26, Pasha Tatashin <Pavel.Tatashin@microsoft.com>: > > > > On 9/13/18 5:19 PM, Timofey Titovets wrote: > > From: Timofey Titovets <nefelim4ag@gmail.com> > > > > Currently used jhash are slow enough and replace it allow as to make KSM > > less cpu hungry. > > > > About speed (in kernel): > > ksm: crc32c hash() 12081 MB/s > > ksm: xxh64 hash() 8770 MB/s > > ksm: xxh32 hash() 4529 MB/s > > ksm: jhash2 hash() 1569 MB/s > > > > By sioh Lee tests (copy from other mail): > > Test platform: openstack cloud platform (NEWTON version) > > Experiment node: openstack based cloud compute node (CPU: xeon E5-2620 v3, memory 64gb) > > VM: (2 VCPU, RAM 4GB, DISK 20GB) * 4 > > Linux kernel: 4.14 (latest version) > > KSM setup - sleep_millisecs: 200ms, pages_to_scan: 200 > > > > Experiment process > > Firstly, we turn off KSM and launch 4 VMs. > > Then we turn on the KSM and measure the checksum computation time until full_scans become two. > > > > The experimental results (the experimental value is the average of the measured values) > > crc32c_intel: 1084.10ns > > crc32c (no hardware acceleration): 7012.51ns > > xxhash32: 2227.75ns > > xxhash64: 1413.16ns > > jhash2: 5128.30ns > > > > In summary, the result shows that crc32c_intel has advantages over all > > of the hash function used in the experiment. (decreased by 84.54% compared to crc32c, > > 78.86% compared to jhash2, 51.33% xxhash32, 23.28% compared to xxhash64) > > the results are similar to those of Timofey. > > > > But, > > use only xxhash for now, because for using crc32c, > > cryptoapi must be initialized first - that require some > > tricky solution to work good in all situations. > > > > So: > > - Fisrt patch implement compile time pickup of fastest implementation of xxhash > > for target platform. > > - Second replace jhash2 with xxhash > > > > Thanks. > > > > CC: Andrea Arcangeli <aarcange@redhat.com> > > CC: linux-mm@kvack.org > > CC: kvm@vger.kernel.org > > CC: leesioh <solee@os.korea.ac.kr> > > > > Timofey Titovets (2): > > xxHash: create arch dependent 32/64-bit xxhash() > > ksm: replace jhash2 with xxhash > > > > include/linux/xxhash.h | 23 +++++++++++++ > > mm/Kconfig | 2 ++ > > mm/ksm.c | 93 +++++++++++++++++++++++++++++++++++++++++++++++--- > > 3 files changed, 114 insertions(+), 4 deletions(-) > > This is wrong stat. ksm.c should not have any new lines at all. Sorry, just copy-paste error when i rework patchset. Must be: include/linux/xxhash.h | 23 +++++++++++++++++++++++ mm/Kconfig | 1 + mm/ksm.c | 4 ++-- And i leave some useless new lines in second patch, i can drop them byself and resend if that needed. Thanks.
From: Timofey Titovets <nefelim4ag@gmail.com> Currently used jhash are slow enough and replace it allow as to make KSM less cpu hungry. About speed (in kernel): ksm: crc32c hash() 12081 MB/s ksm: xxh64 hash() 8770 MB/s ksm: xxh32 hash() 4529 MB/s ksm: jhash2 hash() 1569 MB/s By sioh Lee tests (copy from other mail): Test platform: openstack cloud platform (NEWTON version) Experiment node: openstack based cloud compute node (CPU: xeon E5-2620 v3, memory 64gb) VM: (2 VCPU, RAM 4GB, DISK 20GB) * 4 Linux kernel: 4.14 (latest version) KSM setup - sleep_millisecs: 200ms, pages_to_scan: 200 Experiment process Firstly, we turn off KSM and launch 4 VMs. Then we turn on the KSM and measure the checksum computation time until full_scans become two. The experimental results (the experimental value is the average of the measured values) crc32c_intel: 1084.10ns crc32c (no hardware acceleration): 7012.51ns xxhash32: 2227.75ns xxhash64: 1413.16ns jhash2: 5128.30ns In summary, the result shows that crc32c_intel has advantages over all of the hash function used in the experiment. (decreased by 84.54% compared to crc32c, 78.86% compared to jhash2, 51.33% xxhash32, 23.28% compared to xxhash64) the results are similar to those of Timofey. But, use only xxhash for now, because for using crc32c, cryptoapi must be initialized first - that require some tricky solution to work good in all situations. So: - Fisrt patch implement compile time pickup of fastest implementation of xxhash for target platform. - Second replace jhash2 with xxhash Thanks. CC: Andrea Arcangeli <aarcange@redhat.com> CC: linux-mm@kvack.org CC: kvm@vger.kernel.org CC: leesioh <solee@os.korea.ac.kr> Timofey Titovets (2): xxHash: create arch dependent 32/64-bit xxhash() ksm: replace jhash2 with xxhash include/linux/xxhash.h | 23 +++++++++++++ mm/Kconfig | 2 ++ mm/ksm.c | 93 +++++++++++++++++++++++++++++++++++++++++++++++--- 3 files changed, 114 insertions(+), 4 deletions(-)