[Bug 2080973] [NEW] [FFe] Update lpmd to version 0.0.6
Public bug reported: == Description == As part of the Kobuk engagement, Intel has requested us to update lpmd to v0.0.5. As v0.0.6 is already out, I am filing a FFe for that version. https://bugs.launchpad.net/ubuntu/+source/intel-lpmd/+bug/2076286 == Changelog == The upstream intel-lpmd repository available on Github (https://github.com/intel/intel-lpmd) doesn't provide a proper CHANGELOG or NEWS file, so I attached the git log v0.0.3..v0.0.6 output in this bug. Besides internal changes and code cleanups, the major improvements seemed to happen in v0.0.4, which has the following changes: 1. Enhance HFI monitor so that it can handle back-to-back HFI LPM hints. Previously it only handles first HFI LPM hints and ignore the subsequential HFI LPM hints until it exits Low Power mode. 2. Enhance HFI monitor to handle HFI hints for banned CPUs. 3. Introduce support for multiple Low Power states. With this, intel_lpmd can define multiple Low Power states that have different requirements for EPP/EPB/ITMT setting, IRQ migration, and task migration. And enter different Low Power state based on different utilization threshold. 4. Introduce support for work load type hint. Intel firmware is able to predict workload type and expose this hint to userspace. Add support to define Low Power states and switch between them based on workload type hint. 5. Allow change EPP during Low Power modes transition. 6. Minor fixes and cleanups. == Testing impact == The v0.0.6 package was built on my PPA and the build logs (attached) were collected from it https://launchpad.net/~phckopper/+archive/ubuntu/dev/+sourcepub/16441578/+listing-archive-extra Installation logs on Noble are also attached. Functional testing was performed by Intel. As no other package depends on it, this should not cause any chain effects. ** Affects: intel-lpmd (Ubuntu) Importance: Undecided Status: New -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/2080973 Title: [FFe] Update lpmd to version 0.0.6 To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/intel-lpmd/+bug/2080973/+subscriptions -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 2080973] Re: [FFe] Update lpmd to version 0.0.6
** Attachment added: "Build log" https://bugs.launchpad.net/ubuntu/+source/intel-lpmd/+bug/2080973/+attachment/5818661/+files/buildlog_ubuntu-oracular-amd64.intel-lpmd_0.0.6-0ubuntu1_BUILDING.txt.gz.txt -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/2080973 Title: [FFe] Update lpmd to version 0.0.6 To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/intel-lpmd/+bug/2080973/+subscriptions -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 2080973] Re: [FFe] Update lpmd to version 0.0.6
Installation logs: sudo apt install ./intel-lpmd_0.0.6-0ubuntu1_amd64.deb [sudo] password for phckopper: Reading package lists... Done Building dependency tree... Done Reading state information... Done Note, selecting 'intel-lpmd' instead of './intel-lpmd_0.0.6-0ubuntu1_amd64.deb' The following NEW packages will be installed: intel-lpmd 0 upgraded, 1 newly installed, 0 to remove and 31 not upgraded. Need to get 0 B/48.3 kB of archives. After this operation, 154 kB of additional disk space will be used. Get:1 /home/phckopper/Downloads/intel-lpmd_0.0.6-0ubuntu1_amd64.deb intel-lpmd amd64 0.0.6-0ubuntu1 [48.3 kB] Selecting previously unselected package intel-lpmd. (Reading database ... 430074 files and directories currently installed.) Preparing to unpack .../intel-lpmd_0.0.6-0ubuntu1_amd64.deb ... Unpacking intel-lpmd (0.0.6-0ubuntu1) ... Setting up intel-lpmd (0.0.6-0ubuntu1) ... Created symlink /etc/systemd/system/org.freedesktop.intel_lpmd.service → /usr/lib/systemd/system/intel_lpmd.service. Created symlink /etc/systemd/system/multi-user.target.wants/intel_lpmd.service → /usr/lib/systemd/system/intel_lpmd.service. Processing triggers for dbus (1.14.10-4ubuntu4.1) ... Processing triggers for man-db (2.12.0-4build2) ... -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/2080973 Title: [FFe] Update lpmd to version 0.0.6 To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/intel-lpmd/+bug/2080973/+subscriptions -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 2080973] Re: [FFe] Update lpmd to version 0.0.6
** Description changed: == Description == - As part of the Kobuk engagement, Intel has requested us to update lpmd to v0.0.5. As v0.0.6 + Intel has requested us to update lpmd to v0.0.5. As v0.0.6 is already out, I am filing a FFe for that version. https://bugs.launchpad.net/ubuntu/+source/intel-lpmd/+bug/2076286 == Changelog == The upstream intel-lpmd repository available on Github (https://github.com/intel/intel-lpmd) doesn't provide a proper CHANGELOG or NEWS file, so I attached the git log v0.0.3..v0.0.6 output in this bug. Besides internal changes and code cleanups, the major improvements seemed to happen in v0.0.4, which has the following changes: - 1. Enhance HFI monitor so that it can handle back-to-back HFI LPM -hints. Previously it only handles first HFI LPM hints and ignore -the subsequential HFI LPM hints until it exits Low Power mode. - - 2. Enhance HFI monitor to handle HFI hints for banned CPUs. - - 3. Introduce support for multiple Low Power states. -With this, intel_lpmd can define multiple Low Power states that -have different requirements for EPP/EPB/ITMT setting, IRQ migration, -and task migration. And enter different Low Power state based on -different utilization threshold. - - 4. Introduce support for work load type hint. -Intel firmware is able to predict workload type and expose this hint -to userspace. Add support to define Low Power states and switch -between them based on workload type hint. - - 5. Allow change EPP during Low Power modes transition. - - 6. Minor fixes and cleanups. + 1. Enhance HFI monitor so that it can handle back-to-back HFI LPM + hints. Previously it only handles first HFI LPM hints and ignore + the subsequential HFI LPM hints until it exits Low Power mode. + + 2. Enhance HFI monitor to handle HFI hints for banned CPUs. + + 3. Introduce support for multiple Low Power states. + With this, intel_lpmd can define multiple Low Power states that + have different requirements for EPP/EPB/ITMT setting, IRQ migration, + and task migration. And enter different Low Power state based on + different utilization threshold. + + 4. Introduce support for work load type hint. + Intel firmware is able to predict workload type and expose this hint + to userspace. Add support to define Low Power states and switch + between them based on workload type hint. + + 5. Allow change EPP during Low Power modes transition. + + 6. Minor fixes and cleanups. == Testing impact == The v0.0.6 package was built on my PPA and the build logs (attached) were collected from it https://launchpad.net/~phckopper/+archive/ubuntu/dev/+sourcepub/16441578/+listing-archive-extra Installation logs on Noble are also attached. Functional testing was performed by Intel. As no other package depends on it, this should not cause any chain effects. ** Description changed: == Description == - Intel has requested us to update lpmd to v0.0.5. As v0.0.6 - is already out, I am filing a FFe for that version. + Intel has requested us to update lpmd to v0.0.5. As v0.0.6 is already out, I + am filing a FFe for that version. https://bugs.launchpad.net/ubuntu/+source/intel-lpmd/+bug/2076286 == Changelog == The upstream intel-lpmd repository available on Github (https://github.com/intel/intel-lpmd) doesn't provide a proper CHANGELOG or NEWS file, so I attached the git log v0.0.3..v0.0.6 output in this bug. Besides internal changes and code cleanups, the major improvements seemed to happen in v0.0.4, which has the following changes: 1. Enhance HFI monitor so that it can handle back-to-back HFI LPM hints. Previously it only handles first HFI LPM hints and ignore the subsequential HFI LPM hints until it exits Low Power mode. 2. Enhance HFI monitor to handle HFI hints for banned CPUs. 3. Introduce support for multiple Low Power states. With this, intel_lpmd can define multiple Low Power states that have different requirements for EPP/EPB/ITMT setting, IRQ migration, and task migration
[Bug 2075533] [NEW] Fix performance regression caused by bad default EPP value on Emerald Rapids
Public bug reported: On Intel Emerald Rapids machines (fifth generation Xeon Scalable server processors), we ship the Energy Performance Preference (EPP) default for balance_performance (the default setting) as 128. However, during an internal investigation together with Intel, we have determined that 32 is a more suitable value. This leads to significant improvements in both performance and energy: POV-Ray: 32% faster | 12% less energy OpenSSL: 12% faster | energy within 1% Build Linux Kernel: 29% faster | 18% less energy Therefore, we should move our default EPP for balance_performance to 32. This is currently handled in the intel_pstate driver and should be upstreamed soon. However, as there are significant performance and energy gains, I propose we patch this as soon as possible. ** Affects: linux (Ubuntu) Importance: Undecided Status: New -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/2075533 Title: Fix performance regression caused by bad default EPP value on Emerald Rapids To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2075533/+subscriptions -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 2077470] [NEW] [SRU][Noble] Bad EPP defaults cause performance regressions on select Intel CPUs
Public bug reported: [ Impact ] * Intel currently provides an Energy Performance Preference (EPP) tunable via its intel_pstate to control power management. It ranges from 0 (performance) to 255 (powersaving). * When the governor is set to powersave, which is the default, the EPP value of balance_performance controls how the CPU frequency scaling occurs. However, on some platforms (Emerald Rapids, Meteor Lake, Arrow Lake) it is set to a very conservative value out of the box, resulting in degraded performance by default. * These EPP tuning values work as a black-box and vary per-generation, so having the latest values available is important to keep systems running at maximum achievable performance. Both Intel and Canonical perform this kinds of tests and supply values. [ Test Plan ] * Perform a benchmark on a stock Ubuntu 24.04 image * On Emerald Rapids: * Change the EPP to 32 by executing echo 32 | sudo tee /sys/devices/system/cpu/cpu*/cpufreq/energy_performance_preference * Observe the performance difference. In our internal testing, we found the following results when setting EPP to 32: POV-Ray:32% faster | 12% less energy OpenSSL:12% faster | energy within 1% Build Linux Kernel: 29% faster | 18% less energy * On Meteor Lake and Arrow Lake: * Change the EPP to 64 by executing echo 64 | sudo tee /sys/devices/system/cpu/cpu*/cpufreq/energy_performance_preference * Observe the performance difference. Intel found out the following for Meteor Lake: > Speedometer 2.1 > score: +19% > Perf/watt: +5.25% > > Webxprt 4 score > score: +12% > Perf/watt: +6.12% > > 3DMark Wildlife extreme unlimited score > score: +3.2% > Perf/watt: +11.5% > > Geekbench6 MT > score: +2.14% > Perf/watt: +0.32% > > Also update balance_power EPP default to 179. With this change: > Video Playback power is reduced by 52% > Team video conference power is reduced by 35% * The same outcome should be obtained from Arrow Lake as per Intel. [ Where problems could occur ] * The regression risks are very low. The CPU should perform in a more efficient zone, resulting in overall power savings, despite the more aggressive scaling. This was verified using a Yokogawa WT310 power meter and an Emerald Rapids reference platform. * Tests on Meteor Lake and Arrow Lake were performed by Intel. They were not independently verified by Canonical. [ Other Info ] * Patch for Emerald Rapids: https://lore.kernel.org/lkml/Zqu6zjVMoiXwROBI@capivara/ (merged in 6.11-rc3 https://lore.kernel.org/lkml/cajz5v0hq3o7e3ygvfkbzjkjbx7pddyd9qc+re4efyzjs12o...@mail.gmail.com/) * Patch for Meteor Lake: https://lore.kernel.org/lkml/cajz5v0i8rnws9ogxxqvocg4rhhn7x3elcrupogrs1ip8hns...@mail.gmail.com/ * Patch for Arrow Lake: https://lore.kernel.org/lkml/CAJZ5v0g+R6JK=T9EjmweRPKmcorw93JmPM4u-Z=pywv6kh7...@mail.gmail.com/ (both merged in 6.11-rc1 https://lore.kernel.org/lkml/cajz5v0h7ohr7ipj6ubgtlez238jzktshh93cye_gdx5j+zb...@mail.gmail.com/) ** Affects: linux (Ubuntu) Importance: Undecided Status: New -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/2077470 Title: [SRU][Noble] Bad EPP defaults cause performance regressions on select Intel CPUs To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2077470/+subscriptions -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 2075533] Re: Fix performance regression caused by bad default EPP value on Emerald Rapids
*** This bug is a duplicate of bug 2077470 *** https://bugs.launchpad.net/bugs/2077470 ** This bug has been marked a duplicate of bug 2077470 [SRU][Noble] Bad EPP defaults cause performance regressions on select Intel CPUs -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/2075533 Title: Fix performance regression caused by bad default EPP value on Emerald Rapids To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2075533/+subscriptions -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 2077470] Re: [SRU][Noble] Bad EPP defaults cause performance regressions on select Intel CPUs
** Changed in: linux (Ubuntu Noble) Status: New => In Progress -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/2077470 Title: [SRU][Noble] Bad EPP defaults cause performance regressions on select Intel CPUs To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2077470/+subscriptions -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 2080973] Re: [FFe] Update lpmd to version 0.0.6
As per Intel "We discussed needing this version in 24.10 with Stephane [Verdy] and he agreed this is closest to what we will need to support LNL. Additional patches will need to be backported to 0.0.6 later. But at the very least we should get 0.0.6 as baseline for 24.10.". As additional context for the package FFe: intel-lpmd is the low-power mode daemon that is used on newer hybrid architectures (Meteor Lake onwards) where LP-cores are present (and not only the E-cores and P-cores of Alder Lake etc). lpmd manages the transition to low-power mode via dbus and can provide significant energy savings on mobile platforms. On v0.0.4 onwards, several changes were introduced enhancing this energy-saving capabilities. For instance, enhancing HFI support allows the daemon to handle scheduling hints from the CPU. By being able to enter multiple power states, it can exploit the sleep depth-latency trade-off more effectively and reduce power without affecting user experience. Finally, handling work load type hints allows the daemon to selectively enter low-power mode with increased accuracy. Due to the changes outlined above, I believe that this package should be accepted for a FFe as it provides significantly better handling of low- power modes on recent Intel mobile processors. As the processors themselves are very recent, it is expected we might require more frequent exceptions while the tooling around the LP-cores stabilize. Finally, I reiterate that the package has no reverse dependencies apt-cache rdepends intel-lpmd intel-lpmd Reverse Depends: And it has been thoroughly validated by Intel. Furthermore, as this is exclusively an user-space tool, any regressions should only cause performance/power degradation. -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/2080973 Title: [FFe] Update lpmd to version 0.0.6 To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/intel-lpmd/+bug/2080973/+subscriptions -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 2080973] Re: [FFe] Update lpmd to version 0.0.6
** Tags added: pe-sponsoring-request -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/2080973 Title: [FFe] Update lpmd to version 0.0.6 To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/intel-lpmd/+bug/2080973/+subscriptions -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
