eas-dev

eas-dev@lists.linaro.org

310 discussions

by Leo Yan

Hi all, This is second round profiling for EASv5 on Hikey board. I refined the energy model data according to Dietmar's suggestion (also appreciate for Dietmar's detailed review and guidance). Please help review and looking forward any comment and suggestion. * Overview This is the second round profiling for EASv5 patches on Hikey board. According to review comments from eas-dev mailing list, refined the energy model for Hikey, also developed several python/shell scripts for automatic analysis. So profiling data will be devided into three parts: - C-state profiling data - P-state profiling data - Scheduler performance profiling data * Hardware Environment - Platform: 96boards Hikey - SoC: Hi6220, 2 clusters, 4xCA53 CPUs in each cluster - CPU clock: 2 clusters with 8 CPUs have same clock source and support 208mHz/432mHz/729mHz/960mHz/1200mHz - Support CPU and cluster level low power mode * Software Environment - Kernel (4.2 + EAS RFCv5) [1] https://github.com/Leo-Yan/linux/tree/profile_easv5_hikey - ARM-TF [2] https://github.com/96boards/arm-trusted-firmware/tree/hikey - Enable CPUIdle with PSCI - Enable CPUFreq with cpufreq-dt driver - Profiling scritps: calc_idle_diff.py [3]: calculate C-state's difference for different configurations calc_pstate_time.py [4]: calculate P-state's difference for different configurations calc_sched_preformance.py [5]: calculate scheduler performance * Conventions Below are some conventions which used in below tables: CLS0: Cluster 0 CLS1: Cluster 1 WFI: CPU WFI state C2: CPU power down state M2: Cluster power down state DC: Duty cycle Configuration | Mailine EASv5 Enable CPUFreq CPUFreq | ENERGY_AWARE ondemand sched ------------- | ------- ----- ------------ -------- ------- Mainline (ndm) | Yes No No Yes No noEAS (ndm) | Yes Yes No Yes No EAS (ndm) | Yes Yes Yes Yes No EAS (sched) | Yes Yes Yes No Yes * Profiling: C-state & P-state The detailed profiling result have been uploaded to git-hub [6]; i manually adjust with more readable format for below statistics. - Case MP3: C-state DC Mainline (ndm) noEAS (ndm) EAS (ndm) EAS (sched) CLS0: WFI 1.17s -1.0s -1.1s -1.1s CLS0: C2 2.10s -1.3s -1.7s -1.9s CLS0: M2 25.01s +2.5s +3.3s -4.9s CLS1: WFI 445.06ms -179.3ms -384.1ms -445.1ms CLS1: C2 182.95ms +552.2ms -183.6ms -185.9ms CLS1: M2 19.24s -1.5s +3.7s +10.7s P-State Statistics ------------------------------------------------------------------------------------------------- | Cluster Level (ms) | ------ ------------------------------------------------------------------------------------------ | Item | 208MHz | 432MHz | 729MHz | 960MHz | 1.2GHz | Cycles | ------------------------------------------------------------------------------------------------- | MAINLINE | 1075.99 | 4803.60 | 105.06 | 37.31 | 5860.00 | 9443367.46 | | EAS DIS | -77.20%| +57.87%| +734.47%| +468.29%| -32.79%| -5.80%| | EAS NDM | -83.61%| -71.82%| +264.43%| +947.55%| +1.84%| -10.65%| | EAS SCHED | +795.42%| -100.00%| -100.00%| -100.00%| -98.56%| -77.71%| ------------------------------------------------------------------------------------------------- | CPU Level (ms) | ------ ------------------------------------------------------------------------------------------ | Item | 208MHz | 432MHz | 729MHz | 960MHz | 1.2GHz | Cycles | ------------------------------------------------------------------------------------------------- | MAINLINE | 1131.04 | 4814.53 | 116.39 | 29.15 | 5931.09 | 9545269.85 | | EAS DIS | -74.57%| +59.29%| +655.51%| +632.27%| -32.44%| -5.43%| | EAS NDM | -80.69%| -71.87%| +235.64%| +1215.77%| +0.69%| -11.48%| | EAS SCHED | +755.65%| -100.00%| -100.00%| -100.00%| -98.57%| -77.84%| ------------------------------------------------------------------------------------------------- - Case rt-app 6%: C-state DC Mainline (ndm) noEAS (ndm) EAS (ndm) EAS (sched) CLS0: WFI 7.87s +204.7ms +1.1s +286.0ms CLS0: C2 3.96s -272.9ms -4.0s -675.7ms CLS0: M2 9.29ms -83us -6.8ms -4.5ms CLS1: WFI 8.80s +50.1ms +260.2ms -1.6s CLS1: C2 1.16s +38.5ms -1.1s -1.1s CLS1: M2 73.76ms +51.4ms +49.7ms +127.9ms P-State Statistics ------------------------------------------------------------------------------------------------- | Cluster Level (ms) | ------ ------------------------------------------------------------------------------------------ | Item | 208MHz | 432MHz | 729MHz | 960MHz | 1.2GHz | Cycles | ------------------------------------------------------------------------------------------------- | MAINLINE | 16.58 | 18030.00 | 220.34 | 0.00 | 57.73 | 8022312.50 | | EAS DIS | -1.09%| +0.89%| -96.88%| +1972.00%| +88.79%| -0.08%| | EAS NDM | -0.42%| -99.69%| +9857.34%| +2738.00%| +87.08%| +101.66%| | EAS SCHED | +118.58%| -97.74%| +9394.42%| 0.00%| +310.84%| +95.94%| ------------------------------------------------------------------------------------------------- | CPU Level (ms) | ------ ------------------------------------------------------------------------------------------ | Item | 208MHz | 432MHz | 729MHz | 960MHz | 1.2GHz | Cycles | ------------------------------------------------------------------------------------------------- | MAINLINE | 54.72 | 40950.00 | 402.41 | 0.00 | 61.01 | 18068347.05 | | EAS DIS | +4.08%| +2.27%| -94.49%| +988.60%| +85.60%| +1.09%| | EAS NDM | -3.58%| -99.72%| +5357.03%| +1942.60%| +84.48%| -10.21%| | EAS SCHED | +38.23%| -98.75%| +5370.45%| 0.00%| +340.36%| -8.09%| ------------------------------------------------------------------------------------------------- - Case rt-app 13%: C-state DC Mainline (ndm) noEAS (ndm) EAS (ndm) EAS (sched) CLS0: WFI 8.79s -100.1ms -3.7s -4.3s CLS0: C2 2.76s -1.3s -2.8s -2.7s CLS0: M2 9.71ms -4.8ms -3.0ms -1.6ms CLS1: WFI 8.87s +19.0ms -3.9s -8.5s CLS1: C2 1.57s -918.8ms -1.5s -1.5s CLS1: M2 72.32ms +7.0ms +120.8ms +20.2s P-State Statistics ------------------------------------------------------------------------------------------------- | Cluster Level (ms) | ------ ------------------------------------------------------------------------------------------ | Item | 208MHz | 432MHz | 729MHz | 960MHz | 1.2GHz | Cycles | ------------------------------------------------------------------------------------------------- | MAINLINE | 17.03 | 2730.00 | 15310.00 | 0.00 | 59.01 | 12414704.24 | | EAS DIS | -2.52%| +382.78%| -52.91%| 0.00%| +2.14%| -11.19%| | EAS NDM | -0.29%| -99.27%| -98.28%| +2703000.00%| +4628.01%| +137.63%| | EAS SCHED | +3068.94%| -91.27%| -84.86%| +1078619.00%| +6612.15%| +37.03%| ------------------------------------------------------------------------------------------------- | CPU Level (ms) | ------ ------------------------------------------------------------------------------------------ | Item | 208MHz | 432MHz | 729MHz | 960MHz | 1.2GHz | Cycles | ------------------------------------------------------------------------------------------------- | MAINLINE | 55.77 | 8880.00 | 37850.00 | 0.00 | 66.17 | 31519814.16 | | EAS DIS | +0.22%| +378.04%| -52.76%| 0.00%| +5.32%| -0.16%| | EAS NDM | -0.02%| -99.13%| -99.02%| +2705652.00%| +4169.43%| -5.84%| | EAS SCHED | +937.83%| -91.40%| -85.80%| +1954704.00%| +9605.80%| -2.15%| ------------------------------------------------------------------------------------------------- - Case rt-app 19%: C-state DC Mainline (ndm) noEAS (ndm) EAS (ndm) EAS (sched) CLS0: WFI 9.33s -422.5ms -5.1s -5.2s CLS0: C2 312.17ms +29.7ms -301.8ms -303.5ms CLS0: M2 8.01ms +5us -3.7ms -5.5ms CLS1: WFI 8.50s -200.8ms -4.1s -7.8s CLS1: C2 675.16ms -194.5ms -659.0ms -563.9ms CLS1: M2 120.55ms +2.9ms -41.3ms +18.7s P-State Statistics ------------------------------------------------------------------------------------------------- | Cluster Level (ms) | ------ ------------------------------------------------------------------------------------------ | Item | 208MHz | 432MHz | 729MHz | 960MHz | 1.2GHz | Cycles | ------------------------------------------------------------------------------------------------- | MAINLINE | 16.57 | 3170.00 | 18020.00 | 19.87 | 110.11 | 14660673.76 | | EAS DIS | +1.51%| +57.41%| -6.66%| +802.47%| +4.98%| +0.49%| | EAS NDM | +4.59%| -100.00%| -85.46%| +143432.96%| +218.05%| +102.67%| | EAS SCHED | +517.32%| -95.68%| -80.73%| +57897.68%| +1565.83%| +8.29%| ------------------------------------------------------------------------------------------------- | CPU Level (ms) | ------ ------------------------------------------------------------------------------------------ | Item | 208MHz | 432MHz | 729MHz | 960MHz | 1.2GHz | Cycles | ------------------------------------------------------------------------------------------------- | MAINLINE | 54.95 | 12210.00 | 51480.00 | 9.84 | 120.12 | 42968660.00 | | EAS DIS | +4.55%| +55.61%| -5.48%| +5073.68%| +13.30%| +3.20%| | EAS NDM | -1.69%| -100.00%| -90.22%| +403660.16%| +310.75%| -1.29%| | EAS SCHED | +437.62%| -96.49%| -79.72%| +307381.50%| +3399.13%| -2.38%| ------------------------------------------------------------------------------------------------- - Case rt-app 25%: C-state DC Mainline (ndm) noEAS (ndm) EAS (ndm) EAS (sched) CLS0: WFI 9.37s +55.2ms -3.4s -4.6s CLS0: C2 13.79ms -3.4ms +23.4ms -2.4ms CLS0: M2 10.33ms -2.8ms -6.4ms -8.5ms CLS1: WFI 9.28s +109.2ms -3.3s -8.3s CLS1: C2 15.87ms -12.4ms +49.4ms +77.8ms CLS1: M2 73.39ms +51.7ms +99.5ms +17.2s P-State Statistics ------------------------------------------------------------------------------------------------- | Cluster Level (ms) | ------ ------------------------------------------------------------------------------------------ | Item | 208MHz | 432MHz | 729MHz | 960MHz | 1.2GHz | Cycles | ------------------------------------------------------------------------------------------------- | MAINLINE | 16.56 | 0.00 | 21330.00 | 10.37 | 57.41 | 15631861.68 | | EAS DIS | -1.75%| +1926.00%| -0.80%| +1.54%| +110.80%| -0.25%| | EAS NDM | +10.69%| 0.00%| -95.58%| +233651.21%| +5021.06%| +75.86%| | EAS SCHED | +293.18%| +10204.00%| -99.76%| +120440.02%| +8984.65%| +17.41%| ------------------------------------------------------------------------------------------------- | CPU Level (ms) | ------ ------------------------------------------------------------------------------------------ | Item | 208MHz | 432MHz | 729MHz | 960MHz | 1.2GHz | Cycles | ------------------------------------------------------------------------------------------------- | MAINLINE | 55.26 | 0.00 | 74930.00 | 29.43 | 58.58 | 54734011.68 | | EAS DIS | +0.11%| +6195.50%| +2.00%| -63.97%| +153.11%| +2.21%| | EAS NDM | -11.35%| 0.00%| -97.28%| +162964.90%| +9599.76%| -0.64%| | EAS SCHED | +90.25%| +40570.00%| -99.73%| +125517.36%| +24717.17%| -2.66%| ------------------------------------------------------------------------------------------------- - Case rt-app 31%: C-state DC Mainline (ndm) noEAS (ndm) EAS (ndm) EAS (sched) CLS0: WFI 8.39s +238.5ms +192.0ms -4.3s CLS0: C2 3.20ms -762us +1.6ms +82.3ms CLS0: M2 7.33ms -3.7ms -965us -2.8ms CLS1: WFI 8.39s +246.4ms +207.0ms -4.7s CLS1: C2 2.48ms -2.1ms -491us +458.0ms CLS1: M2 122.45ms +150.1ms -2.9ms +9.9s P-State Statistics ------------------------------------------------------------------------------------------------- | Cluster Level (ms) | ------ ------------------------------------------------------------------------------------------ | Item | 208MHz | 432MHz | 729MHz | 960MHz | 1.2GHz | Cycles | ------------------------------------------------------------------------------------------------- | MAINLINE | 16.80 | 0.00 | 23140.00 | 86.87 | 109.07 | 17086833.60 | | EAS DIS | -1.19%| 0.00%| -1.82%| -70.00%| +136.66%| -1.09%| | EAS NDM | -2.44%| 0.00%| -1.38%| -88.81%| +1.65%| -1.79%| | EAS SCHED | +3041.90%| 0.00%| -85.91%| +114.55%| +18860.30%| +60.83%| ------------------------------------------------------------------------------------------------- | CPU Level (ms) | ------ ------------------------------------------------------------------------------------------ | Item | 208MHz | 432MHz | 729MHz | 960MHz | 1.2GHz | Cycles | ------------------------------------------------------------------------------------------------- | MAINLINE | 56.27 | 0.00 | 90510.00 | 260.88 | 119.29 | 66387086.96 | | EAS DIS | -1.28%| 0.00%| -0.99%| -77.14%| +117.20%| -1.03%| | EAS NDM | -1.76%| 0.00%| -0.53%| -96.21%| +3.29%| -0.88%| | EAS SCHED | +909.24%| 0.00%| -89.73%| +101.70%| +41110.50%| +0.00%| ------------------------------------------------------------------------------------------------- - Case rt-app 38%: C-state DC Mainline (ndm) noEAS (ndm) EAS (ndm) EAS (sched) CLS0: WFI 7.50s -233.6ms -3.1s -4.1s CLS0: C2 2.04ms -1.0ms +20.4ms +33.3ms CLS0: M2 7.73ms -2.3ms -2.9ms -1.1ms CLS1: WFI 7.51s -232.5ms -2.9s -3.4s CLS1: C2 1.65ms +108us +26.0ms +1.5s CLS1: M2 70.35ms +48.7ms +4.5ms +7.3s P-State Statistics ------------------------------------------------------------------------------------------------- | Cluster Level (ms) | ------ ------------------------------------------------------------------------------------------ | Item | 208MHz | 432MHz | 729MHz | 960MHz | 1.2GHz | Cycles | ------------------------------------------------------------------------------------------------- | MAINLINE | 16.86 | 0.00 | 13610.00 | 11410.00 | 60.01 | 20950808.88 | | EAS DIS | -4.74%| 0.00%| +1.98%| +1.67%| +85.52%| +2.10%| | EAS NDM | -1.25%| 0.00%| -26.67%| +24.80%| +11348.09%| +39.34%| | EAS SCHED | +3510.50%| 0.00%| -57.83%| -99.39%| +33561.06%| +36.59%| ------------------------------------------------------------------------------------------------- | CPU Level (ms) | ------ ------------------------------------------------------------------------------------------ | Item | 208MHz | 432MHz | 729MHz | 960MHz | 1.2GHz | Cycles | ------------------------------------------------------------------------------------------------- | MAINLINE | 55.54 | 0.00 | 53470.00 | 44830.00 | 71.27 | 82113506.32 | | EAS DIS | -0.32%| 0.00%| +0.97%| +1.43%| +83.58%| +1.30%| | EAS NDM | -2.57%| 0.00%| -42.40%| -9.99%| +25170.10%| +0.85%| | EAS SCHED | +1070.72%| 0.00%| -66.71%| -99.53%| +77534.35%| -2.93%| ------------------------------------------------------------------------------------------------- - Case rt-app 44%: C-state DC Mainline (ndm) noEAS (ndm) EAS (ndm) EAS (sched) CLS0: WFI 5.58s +148.0ms -2.1s -2.0s CLS0: C2 3.00ms +462us +3.4ms +31.7ms CLS0: M2 8.36ms +1.4ms -4.4ms +84us CLS1: WFI 5.58s +161.4ms -685.8ms -148.4ms CLS1: C2 3.66ms -3.7ms +6.2ms +896.1ms CLS1: M2 66.94ms +8.2ms +60.1ms +2.8s P-State Statistics ------------------------------------------------------------------------------------------------- | Cluster Level (ms) | ------ ------------------------------------------------------------------------------------------ | Item | 208MHz | 432MHz | 729MHz | 960MHz | 1.2GHz | Cycles | ------------------------------------------------------------------------------------------------- | MAINLINE | 16.31 | 0.00 | 15990.00 | 12870.00 | 63.64 | 24091670.48 | | EAS DIS | -2.58%| 0.00%| -2.44%| +0.62%| +11.94%| -0.82%| | EAS NDM | +0.86%| 0.00%| -80.86%| +96.27%| +5258.27%| +26.91%| | EAS SCHED | +2343.72%| +5931.20%| -75.42%| -82.34%| +36213.64%| +36.51%| ------------------------------------------------------------------------------------------------- | CPU Level (ms) | ------ ------------------------------------------------------------------------------------------ | Item | 208MHz | 432MHz | 729MHz | 960MHz | 1.2GHz | Cycles | ------------------------------------------------------------------------------------------------- | MAINLINE | 54.46 | 0.00 | 63200.00 | 50700.00 | 84.69 | 94857755.68 | | EAS DIS | -3.43%| 0.00%| -3.34%| -0.10%| +23.86%| -1.65%| | EAS NDM | +0.28%| 0.00%| -81.79%| +51.87%| +10561.96%| -1.79%| | EAS SCHED | +707.29%| +5931.20%| -79.65%| -87.02%| +73887.48%| -4.07%| ------------------------------------------------------------------------------------------------- - Case rt-app 50%: C-state DC Mainline (ndm) noEAS (ndm) EAS (ndm) EAS (sched) CLS0: WFI 6.28s +113.9ms -197.9ms +301.9ms CLS0: C2 1.40ms +2.3ms +1.7ms +3.9ms CLS0: M2 8.25ms -5.1ms -7.6ms -5.2ms CLS1: WFI 6.29s +111.4ms -120.5ms +297.2ms CLS1: C2 2.47ms -2.5ms -2.2ms -2.5ms CLS1: M2 118.94ms -44.1ms -116.6ms +538.7ms P-State Statistics ------------------------------------------------------------------------------------------------- | Cluster Level (ms) | ------ ------------------------------------------------------------------------------------------ | Item | 208MHz | 432MHz | 729MHz | 960MHz | 1.2GHz | Cycles | ------------------------------------------------------------------------------------------------- | MAINLINE | 16.58 | 0.00 | 0.57 | 27440.00 | 113.26 | 26482179.82 | | EAS DIS | -4.34%| 0.00%| +6781.74%| -0.98%| -31.93%| -1.04%| | EAS NDM | +3.08%| 0.00%| +13155.65%| +0.91%| +21.72%| +1.23%| | EAS SCHED | +433.29%| 0.00%| -100.00%| -7.51%| +1694.80%| +1.29%| ------------------------------------------------------------------------------------------------- | CPU Level (ms) | ------ ------------------------------------------------------------------------------------------ | Item | 208MHz | 432MHz | 729MHz | 960MHz | 1.2GHz | Cycles | ------------------------------------------------------------------------------------------------- | MAINLINE | 55.02 | 0.00 | 10.64 | 108480.00 | 135.34 | 104322408.72 | | EAS DIS | -3.56%| 0.00%| +818.70%| -1.13%| -2.61%| -1.08%| | EAS NDM | -3.13%| 0.00%| +2034.87%| -1.62%| +160.34%| -1.22%| | EAS SCHED | +131.86%| 0.00%| -100.00%| -8.49%| +4018.46%| -2.21%| ------------------------------------------------------------------------------------------------- * Profiling: performance sysbench --test=cpu --num-threads=1 --max-time=30 run: Respone Time Mainline (ndm) noEAS (ndm) EAS (ndm) EAS (sched) min 1.67ms +00.00ms +00.00ms +00.00ms avg 1.68ms +00.00ms +00.00ms +00.00ms max 2.35ms +04.54ms +05.32ms +7.77ms approx 95% 1.70ms +00.00ms +00.00ms -00.02ms rt-app performance is calculate with below formula: task performance = slack/(c_period - c_run) * 1024 energy mainline (ndm) noeas (ndm) eas (ndm) eas (sched) prf prf prf prf rt-app 6% 665 663 630 606 rt-app 13% 696 634 474 348 rt-app 19% 665 663 630 606 rt-app 25% 650 627 391 243 rt-app 31% 587 599 596 396 rt-app 38% 574 562 403 212 rt-app 44% 458 487 399 177 rt-app 50% 478 590 572 616 * Summary - With EAS (sched), CPUIdle duty cycle is much better than other confugirations; If we review the cpu idle's duty cycle with P-state's statistics, we can easily to know if cpu frequency scaling is driven by scheduler, then it will introduce more aggressive policy to increase frequency. So CPU can finish tasks more quick and finally we can get better cpu idle's duty cycle. - With EAS (sched), it also introduce much higher for cluster level's cycle number; that can be explained by the scheduler will place tasks on single CPU rather than spread them, so CPU still need run for almost the same time, but the cluster level will run more time due the tasks is running sequential on less CPUs. - With EAS (sched), the CPUs are mainly running at lowest point (208MHz) or higher OPPs (960MHz or 1.2GHz); but for mainline kernel with ondemand governor, CPUs have many chances run at middle OPPs (432MHz or 729MHz). - For EAS (ndm), it enters cluster level's idle state for much less time if we compare with EAS (sched); After reviewed the detailed cpu level's idle data, EAS (ndm) will spread tasks into two clusters, but EAS (sched) will let only one cluster to run tasks as possible. So EAS (sched) can power off cluster 1 for most time, but EAS (ndm) will spread tasks to two CPUs, but these two CPUs are placed in two clusters separtely. - EAS (ndm) and EAS (sched) are much better for CPU level's idle duty cycle compare with noEAS (ndm). After enabled EAS_FEATURE, the CPU will run into low power mode for much more time than noEAS (ndm); though cluster level's idle duty cycle cannot demonstrate this. EAS (ndm) and EAS (sched) also will let CPU run at higher OPP than noEAS (ndm). - rt-app 6% case is special, EAS(ndm) and EAS(sched) will spread tasks into two clusters; so there have no improvement for cluster level's idle duty cycle. [1] https://github.com/Leo-Yan/linux/tree/profile_easv5_hikey [2] https://github.com/96boards/arm-trusted-firmware/tree/hikey [3] https://github.com/Leo-Yan/utility/blob/master/profile_eas/calc_idle_diff.py [4] https://github.com/Leo-Yan/utility/blob/master/profile_eas/calc_pstate_time… [5] https://github.com/Leo-Yan/utility/blob/master/profile_eas/calc_sched_prefo… [6] https://github.com/Leo-Yan/utility/tree/master/profile_eas/hikey_easv5_roun… Thanks, Leo Yan

11 years, 1 month

[RFC] sched: Central, scheduler-driven, power-perfomance control

by Patrick Bellasi

Hi eas-dev, as pre-announced by Morten on that ML, last week we posted on LKML an RFC introducing SchedTune, a proposal for a central tuning knob. The current posting addresses DVFS biasing only. There will be a follow up posting to deal with task placement via integration with EAS as the upstream discussions move along. The full cover letter can be found here: http://www.kernelhub.org/?msg=819910&p=2 For your convenience the patch set is available here: git://www.linux-arm.com/linux-power eas/stune/rfcv1 Cheers Patrick -- #include <best/regards.h> Patrick Bellasi

11 years, 1 month

EAS Profiling On Hikey (1st Round)

by Leo Yan

Hi all, Below are my trying for profiling EAS; please help review and welcome any suggestion or question. * Purpose This is the first round profiling for EASv5 patches on Hikey board; With profiling EASv5 patches on Hikey board, can get below info and feedback for EAS's developement: - Created the profiling enviornment for ARM64 - Collected manifestation after applied EASv5 patches on SoC with two CA53 clusters - I cannot measure hardware power cosumption, so currently _ONLY_ check CPU duty cycle for comparasion scheduler behavior * Hardware Environment - Platform: 96boards Hikey - SoC: Hi6220, 2 clusters, 4xCA53 CPUs in each cluster - CPU clock: 2 clusters with 8 CPUs have coupled clock source and support 208mHz/432mHz/729mHz/960mHz/1200mHz - Support CPU and cluster level low power mode * Software Environment - Kernel: 4.2rc4 + EAS RFCv5 - ARM-TF: [1] - Enabled CPUIdle with PSCI - Enabled CPUFreq with cpufreq-dt driver * Profiling Data CLS0: Cluster 0 CLS1: Cluster 1 CPU_PD: CPU power down state CLS_PW: Cluster power down state - Case Sysbench: sysbench --test=cpu --cpu-max-prime=20000 run Respone Time Mainline (ndm) noEAS (ndm) EAS (ndm) EAS (sched) min 4.19ms +00.00ms +00.00ms +00.00ms avg 4.21ms +00.00ms +00.00ms -00.01ms max 6.86ms +00.09ms +00.04ms +13.61ms approx 95% 4.23ms +00.00ms +00.00ms -00.02ms Idle Dutycycle Mainline (ndm) noEAS (ndm) EAS (ndm) EAS (sched) CLS0: WFI 000.20ms +000.35ms +000.34ms +000.89ms CLS0: CPU_PD 001.58ms +000.76ms +000.88ms +001.40ms CLS0: CLS_PD 001.82ms -000.41ms -000.12ms +001.10ms CLS1: WFI n.a. n.a. +2.9s +000.07ms CLS1: CPU_PD n.a. n.a. +001.30ms +6.7s CLS1: CLS_PD 42.11s +003.8ms -2.9s -6.8s - Case MP3: ./idlestat --trace -f mp3_trace.log -t 30 -p -c -w -o mp3_report.log -- rt-app ./doc/examples/mp3-long.json Idle Dutycycle Mainline (ndm) noEAS (ndm) EAS (ndm) EAS (sched) CLS0: WFI 067.31ms -022.10ms -022.20ms +019.70ms CLS0: CPU_PD 887.74ms +919.80ms -292.40ms -316.90ms CLS0: CLS_PD 17.08s -444.80ms +895.30ms +5.3s CLS1: WFI 000.59ms +002.30ms +000.28ms +196.70ms CLS1: CPU_PD n.a. +000.26ms n.a. +004.00ms CLS1: CLS_PD 28.80s +189.10ms -269.40ms -10.6s - Case rt-app 6%: ./idlestat --trace -f trace.log -t 30 -p -c -w -o report.log -- rt-app ./doc/examples/rt-app-6.json Idle Dutycycle Mainline (ndm) noEAS (ndm) EAS (ndm) EAS (sched) CLS0: WFI 7.82s +037.40ms -7.1s -7.3s CLS0: CPU_PD 4.26s -154.20ms -3.5s -4.2s CLS0: CLS_PD 005.76ms n.a. +17.6s +18.9s CLS1: WFI 6.46s +2.0s +2.4s -155.90ms CLS1: CPU_PD 6.01s -1.9s -5.3s -6.0s CLS1: CLS_PD 123.76ms -118.50ms -121.90ms +1.2s - Case rt-app 13%: ./idlestat --trace -f trace.log -t 30 -p -c -w -o report.log -- rt-app ./doc/examples/rt-app-13.json Idle Dutycycle Mainline (ndm) noEAS (ndm) EAS (ndm) EAS (sched) CLS0: WFI 9.26s -304.70ms -8.8s -4.2s CLS0: CPU_PD 1.11s -695.20ms -275.2ms -1.1s CLS0: CLS_PD 003.49ms -001.30ms +18.3s -613us CLS1: WFI 8.32s -8.3s -3.4s -8.3s CLS1: CPU_PD 2.65s -2.6s -2.6s -2.6s CLS1: CLS_PD 123.07ms +19.9s -121.6ms +20.3s - Case rt-app 19%: ./idlestat --trace -f trace.log -t 30 -p -c -w -o report.log -- rt-app ./doc/examples/rt-app-19.json Idle Dutycycle Mainline (ndm) noEAS (ndm) EAS (ndm) EAS (sched) CLS0: WFI 8.91s -256.70ms -473.4ms -2.7s CLS0: CPU_PD 428.42ms +000.86ms -425.5ms -356.20ms CLS0: CLS_PD 002.28ms +000.45ms +1.1ms n.a. CLS1: WFI 6.20s +224.80ms -6.2s -5.2s CLS1: CPU_PD 4.02s -193.20ms -4.0s -4.0s CLS1: CLS_PD 073.93ms +1.3s +20.0s +039.80ms - Case rt-app 25%: ./idlestat --trace -f trace.log -t 30 -p -c -w -o report.log -- rt-app ./doc/examples/rt-app-25.json Idle Dutycycle Mainline (ndm) noEAS (ndm) EAS (ndm) EAS (sched) CLS0: WFI 9.60s -170.90ms -3.5s -3.4s CLS0: CPU_PD 025.02ms -018.90ms -023.20ms -023.50ms CLS0: CLS_PD 004.43ms -001.80ms -003.50ms +004.40ms CLS1: WFI 9.72s -1.1s -9.6s -9.2s CLS1: CPU_PD 239.13ms +1.7s -237.90ms -152.00ms CLS1: CLS_PD 075.45ms +001.00ms +19.9s +20.1s - Case rt-app 31%: ./idlestat --trace -f trace.log -t 30 -p -c -w -o report.log -- rt-app ./doc/examples/rt-app-31.json Idle Dutycycle Mainline (ndm) noEAS (ndm) EAS (ndm) EAS (sched) CLS0: WFI 8.54s +108.10ms -5.6s -189.20ms CLS0: CPU_PD 001.97ms +001.70ms -001.40ms +005.50ms CLS0: CLS_PD 003.24ms -002.30ms +003.00ms +001.90ms CLS1: WFI 8.56s +108.60ms -8.6s -260.00ms CLS1: CPU_PD n.a. +001.70ms n.a. +199.60ms CLS1: CLS_PD 189.15ms -000.28ms +19.9s +449.60ms - Case rt-app 38%: ./idlestat --trace -f trace.log -t 30 -p -c -w -o report.log -- rt-app ./doc/examples/rt-app-38.json Idle Dutycycle Mainline (ndm) noEAS (ndm) EAS (ndm) EAS (sched) CLS0: WFI 8.77s +695.00ms -7.0s -4.9s CLS0: CPU_PD 000.96ms +001.50ms +001.70ms -000.01ms CLS0: CLS_PD 003.06ms -000.59ms -002.50ms +002.30ms CLS1: WFI 8.79s +913.70ms -8.8s -8.8s CLS1: CPU_PD 001.71ms +151.30ms -001.70ms -000.45ms CLS1: CLS_PD 123.32ms -120.60ms +19.6s +20.4s - Case rt-app 44%: ./idlestat --trace -f trace.log -t 30 -p -c -w -o report.log -- rt-app ./doc/examples/rt-app-44.json Idle Dutycycle Mainline (ndm) noEAS (ndm) EAS (ndm) EAS (sched) CLS0: WFI 5.76s +101.20ms -4.6s -4.5s CLS0: CPU_PD 003.07ms -000.18ms -001.70ms -001.80ms CLS0: CLS_PD 002.02ms -001.20ms +000.23ms +001.30ms CLS1: WFI 6.01s +108.60ms -5.9s -5.9s CLS1: CPU_PD 001.14ms +001.20ms +000.07ms +000.65ms CLS1: CLS_PD 190.75ms -115.50ms +19.6s +19.6s - Case rt-app 50%: ./idlestat --trace -f trace.log -t 30 -p -c -w -o report.log -- rt-app ./doc/examples/rt-app-50.json Idle Dutycycle Mainline (ndm) noEAS (ndm) EAS (ndm) EAS (sched) CLS0: WFI 6.62s -267.20ms -6.6s -6.6s CLS0: CPU_PD n.a. +001.80ms +001.50ms +000.72ms CLS0: CLS_PD 001.47ms +001.30ms -000.33ms +005.10ms CLS1: WFI 6.89s -348.20ms -6.9s -6.9s CLS1: CPU_PD 001.63ms -001.50ms -001.60ms -001.60ms CLS1: CLS_PD 001.87ms +123.50ms +19.9s +20.2s * Summary - For two same clusters case, EAS (ndm) has the best performance than other configurations. We can get most benefit for CPU/cluster's duty cycle with EAS (ndm); The reason is for almost all cases, EAS (ndm) increases much time for cluster's power down, than means the scheduler have optimized load balance within one cluster rather than spread tasks to two clusters. - EAS (sched) is not consistent for all cases, for case MP3 even worse than mainline; for cases rt-app 6%/13%/25%/38%/44%/50% it has almost same behavior with EAS (ndm), but it's not stable for cases rt-app 19%/31%. - EAS (sched) introduces very high latency for sysbench, so the max response time it will take 20.47ms, which is much higher than other three configs. - EAS (sched) and EAS (ndm) will impact idle states selection in CPUIdle, and it increases the possibility for CPU level's state rather than cluster level's state; this result is coming from sysbench case. [1] https://github.com/Leo-Yan/arm-trusted-firmware/tree/hikey_enable_low_power… Thanks, Leo Yan

11 years, 1 month

[RFCv5] Energy cost model for energy-aware scheduling

by Morten Rasmussen

Hi eas-dev, In case you don't follow LKML, we have posted RFCv5 of the EAS patch set for review and comments. The main addition in this posting is scheduler-driven DVFS along with a number of smaller fixes and improvements. The full cover letter can be found here: https://lkml.org/lkml/2015/7/7/754 For your convenience the patch set is available here: http://www.linux-arm.org/git?p=linux-power.git branch: energy_model_rfc_v5 Additional patches introducing a central tuning knob that influences both task placement and DVFS will be posted later for discussion. Thanks, Morten

11 years, 2 months

Re: [Eas-dev] [RESEND PATCH v2] doc: measure the efficiency of cpufreq governors

by Amit Kucheria

(This might be of interest to folks on eas-dev, adding them to cc) On Wed, Jun 24, 2015 at 11:53 AM, Vincent Guittot <vincent.guittot(a)linaro.org> wrote: > On 24 June 2015 at 03:41, <pang.xunlei(a)zte.com.cn> wrote: >> Hi Vincent, >> >> Vincent Guittot <vincent.guittot(a)linaro.org> wrote 2015-06-23 PM 09:43:55: >>> Re: [RESEND PATCH v2] doc: measure the efficiency of cpufreq governors >>> >>> Hi Xunlei, >>> >>> I have run the bench on a quad A15 with sched-dvfs (but without eas >>> patches unlike you) >>> #sudo ./test.sh 4 100 1000 >>> Frequency domain CPU0~CPU3, run 100ms, sleep 1000ms: >>> powersave efficiency: 0% >>> performance efficiency: 100% >>> conservative efficiency: 18% >>> ondemand efficiency: 48% >>> cfs efficiency: 24% >>> >>> #sudo ./test.sh 4 200 1000 >>> Frequency domain CPU0~CPU3, run 200ms, sleep 1000ms: >>> powersave efficiency: 0% >>> performance efficiency: 100% >>> conservative efficiency: 40% >>> ondemand efficiency: 68% >>> cfs efficiency: 30% >>> >>> $ sudo ./test.sh 4 50 1000 >>> Frequency domain CPU0~CPU3, run 50ms, sleep 1000ms: >>> powersave efficiency: 0% >>> performance efficiency: 100% >>> conservative efficiency: 0% >>> ondemand efficiency: 25% >>> cfs efficiency: 19% >>> >>> As an example, here is the result when ondemand parameter are tuned >>> for the platform >>> sudo ./test.sh 4 100 1000 >>> Frequency domain CPU0~CPU3, run 100ms, sleep 1000ms: >>> powersave efficiency: 0% >>> performance efficiency: 100% >>> conservative efficiency: 19% >>> ondemand efficiency: 94% >>> cfs efficiency: 23% >>> >>> Beside these results, i have seen variation in the results that >>> confirm the interest of having more statistics like, min, man stdev >> >> I guess the hardware environment has something to do with this, on my >> platform, there're 11 available freuencies in total: 1200Mhz~2200Mhz, >> the step size is 100Mhz. > > Yes for sure, it was just to give some figures with a different platform. > Regarding the variation, i have seen these variations for the same > platform with the same SW. Nevertheless, this variations are somewhat > normal if we consider the default sampling rate of 164ms for some > governor compared to a run duration of 100ms > > Regards, > Vincent > >> >> Also it may get different results when running with different workload >> loops. >> >> -Xunlei >> >>> >>> Regards, >>> Vincent >>> >>> >>> >>> On 18 June 2015 at 13:35, <pang.xunlei(a)zte.com.cn> wrote: >>> > Just tested on my Intel EAS test environment(implemented x86 frequency >>> > invariant hook). >>> > With EAS disabled and sched-dvfs enabled. >>> > >>> > #./test.sh 3 100 1000 >>> > Frequency domain CPU0~CPU2, run 100ms, sleep 1000ms: >>> > powersave efficiency: 0% >>> > performance efficiency: 100% >>> > conservative efficiency: 92% >>> > ondemand efficiency: 97% >>> > cfs efficiency: 79% >>> > >>> > #./test.sh 3 200 1000 >>> > Frequency domain CPU0~CPU2, run 200ms, sleep 1000ms: >>> > powersave efficiency: 0% >>> > performance efficiency: 100% >>> > conservative efficiency: 97% >>> > ondemand efficiency: 99% >>> > cfs efficiency: 89% >>> > >>> > #./test.sh 3 50 1000 >>> > Frequency domain CPU0~CPU2, run 50ms, sleep 1000ms: >>> > powersave efficiency: 0% >>> > performance efficiency: 100% >>> > conservative efficiency: 93% >>> > ondemand efficiency: 96% >>> > cfs efficiency: 58% >>> > >>> > #./test.sh 3 1000 100 >>> > Frequency domain CPU0~CPU2, run 1000ms, sleep 100ms: >>> > powersave efficiency: 0% >>> > performance efficiency: 100% >>> > conservative efficiency: 99% >>> > ondemand efficiency: 99% >>> > cfs efficiency: 97% >>> > >>> > Seems sched-dvfs is computing inefficient at low cpu usage(implies >> power >>> > efficient), >>> > but computing efficient at high cpu usage. >>> > >>> > -Xunlei >>> > >>> > Xunlei Pang <xlpang(a)126.com> wrote 2015-06-18 PM 05:06:07: >>> >> [RESEND PATCH v2] doc: measure the efficiency of cpufreq governors >>> >> >>> >> From: Xunlei Pang <pang.xunlei(a)linaro.org> >>> >> >>> >> DVFS adds a latency in the execution of task because of the time to >>> >> decide to move at max freq. We need to measure this latency and check >>> >> that the governor stays in an acceptable range. >>> >> >>> >> When workgen runs a json file, a log file is created for each thread. >>> >> This log file records the number of loop that has been executed and >>> >> the duration for executing these loops (per phase). We can use these >>> >> figures to evaluate to latency that is added by a cpufreq governor >>> >> and its "performance efficiency". >>> >> >>> >> We use the run+sleep pattern to do the measurement, for the run time >> per >>> >> loop, the performance governor should run the expected duration as >> the >>> >> CPU stays a max freq. At the opposite, the powersave governor will >> give >>> >> use the longest duration (as it stays at lowest OPP). Other governor >>> > will >>> >> be somewhere between the 2 previous duration as they will use several >>> > OPP >>> >> and will go back to max frequency after a defined duration which >> depends >>> >> on its monitoring period. >>> >> >>> >> The formula: >>> >> >>> >> duration of powersave gov - duration of the gov >>> >> -------------------------------------------------------- x 100% >>> >> duration of powersave gov - duration of performance gov >>> >> >>> >> will give the efficiency of the governor. 100% means as efficient as >>> >> the perf governor and 0% means as efficient as the powersave >> governor. >>> >> >>> >> This patch offers json files and shell scripts to do the measurement, >>> >> >>> >> Usage: ./test.sh <cpus> <runtime> <sleeptime> >>> >> cpus: number of cpus in the CPU0's frequency domain >>> >> runtime: running time in ms per loop of the workload pattern >>> >> sleeptime: sleeping time in ms per loop of the workload pattern >>> >> >>> >> Example: >>> >> "./test.sh 4 100 1000" means >>> >> CPU0~CPU3 sharing frequency, "100ms run + 1000ms sleep" workload >>> > pattern. >>> >> >>> >> test result on my machine: >>> >> ~#./test.sh 4 100 1000 >>> >> Frequency domain CPU0~CPU3, run 100ms, sleep 1000ms: >>> >> powersave efficiency: 0% >>> >> performance efficiency: 100% >>> >> conservative efficiency: 28% >>> >> ondemand efficiency: 95% >>> >> >>> >> NOTE: Make sure there are "sed", "cut", "grep", "rt-app", etc tools >> on >>> >> your test machine, and run the script under root privilege. >>> >> >>> >> Signed-off-by: Xunlei Pang <pang.xunlei(a)linaro.org> >>> >> --- >>> >> doc/examples/cpufreq_governor_efficiency/README | 54 >> ++++++++++++++ >>> >> .../cpufreq_governor_efficiency/calibration.json | 27 +++++++ >>> >> .../cpufreq_governor_efficiency/calibration.sh | 11 +++ >>> >> doc/examples/cpufreq_governor_efficiency/dvfs.json | 27 +++++++ >>> >> doc/examples/cpufreq_governor_efficiency/dvfs.sh | 38 ++++++++++ >>> >> doc/examples/cpufreq_governor_efficiency/test.sh | 82 +++++++++++ >>> >> +++++++++++ >>> >> 6 files changed, 239 insertions(+) >>> >> create mode 100644 doc/examples/cpufreq_governor_efficiency/README >>> >> create mode 100644 >>> > doc/examples/cpufreq_governor_efficiency/calibration.json >>> >> create mode 100755 >>> > doc/examples/cpufreq_governor_efficiency/calibration.sh >>> >> create mode 100644 >> doc/examples/cpufreq_governor_efficiency/dvfs.json >>> >> create mode 100755 doc/examples/cpufreq_governor_efficiency/dvfs.sh >>> >> create mode 100755 doc/examples/cpufreq_governor_efficiency/test.sh >>> >> >>> >> diff --git a/doc/examples/cpufreq_governor_efficiency/README b/doc/ >>> >> examples/cpufreq_governor_efficiency/README >>> >> new file mode 100644 >>> >> index 0000000..cc8efe1 >>> >> --- /dev/null >>> >> +++ b/doc/examples/cpufreq_governor_efficiency/README >>> >> @@ -0,0 +1,54 @@ >>> >> +Measure the efficiency of cpufreq governors using rt-app >>> >> + >>> >> +BACKGROUND: >>> >> + DVFS adds a latency in the execution of task because of the time >> to >>> >> + decide to move at max freq. We need to measure this latency and >>> > check >>> >> + that the governor stays in an acceptable range. >>> >> + >>> >> + When workgen runs a json file, a log file is created for each >>> > thread. >>> >> + This log file records the number of loop that has been executed >> and >>> >> + the duration for executing these loops (per phase). We can use >>> > these >>> >> + figures to evaluate to latency that is added by a cpufreq >> governor >>> >> + and its "performance efficiency". >>> >> + >>> >> + We use the run+sleep pattern to do the measurement, for the run >>> > time per >>> >> + loop, the performance governor should run the expected duration >> as >>> > the >>> >> + CPU stays a max freq. At the opposite, the powersave governor >> will >>> > give >>> >> + use the longest duration (as it stays at lowest OPP). Other >>> > governor will >>> >> + be somewhere between the 2 previous duration as they will use >>> > several OPP >>> >> + and will go back to max frequency after a defined duration which >>> > depends >>> >> + on its monitoring period. >>> >> + >>> >> + The formula: >>> >> + >>> >> + duration of powersave gov - duration of the gov >>> >> + -------------------------------------------------------- x 100% >>> >> + duration of powersave gov - duration of performance gov >>> >> + >>> >> + will give the efficiency of the governor. 100% means as >> efficient >>> > as >>> >> + the perf governor and 0% means as efficient as the powersave >>> > governor. >>> >> + >>> >> + This test offers json files and shell scripts to do the >>> > measurement, >>> >> + >>> >> +USAGE: >>> >> + ./test.sh <cpus> <runtime> <sleeptime> >>> >> + cpus: number of cpus in the CPU0's frequency domain >>> >> + runtime: running time in ms per loop of the workload pattern >>> >> + sleeptime: sleeping time in ms per loop of the workload pattern >>> >> + >>> >> +Example: >>> >> + "./test.sh 4 100 1000" means >>> >> + CPU0~CPU3 sharing frequency, "100ms run + 1000ms sleep" workload >>> > pattern. >>> >> + >>> >> + test result on an Intel machine: >>> >> + ~#./test.sh 4 100 1000 >>> >> + Frequency domain CPU0~CPU3, run 100ms, sleep 1000ms: >>> >> + powersave efficiency: 0% >>> >> + performance efficiency: 100% >>> >> + conservative efficiency: 28% >>> >> + ondemand efficiency: 95% >>> >> + >>> >> +NOTE: >>> >> + Make sure there are "sed", "cut", "grep", "rt-app", etc tools >>> >> on your test >>> >> + machine, and run the script under root privilege. >>> >> + >>> >> diff --git a/doc/examples/cpufreq_governor_efficiency/ >>> >> calibration.json >>> > b/doc/examples/cpufreq_governor_efficiency/calibration.json >>> >> new file mode 100644 >>> >> index 0000000..4377990 >>> >> --- /dev/null >>> >> +++ b/doc/examples/cpufreq_governor_efficiency/calibration.json >>> >> @@ -0,0 +1,27 @@ >>> >> +{ >>> >> + "tasks" : { >>> >> + "thread" : { >>> >> + "instance" : 1, >>> >> + "cpus" : [0], >>> >> + "loop" : 1, >>> >> + "phases" : { >>> >> + "run" : { >>> >> + "loop" : 1, >>> >> + "run" : 200000, >>> >> + }, >>> >> + "sleep" : { >>> >> + "loop" : 1, >>> >> + "sleep" : 200000, >>> >> + } >>> >> + } >>> >> + } >>> >> + }, >>> >> + "global" : { >>> >> + "default_policy" : "SCHED_FIFO", >>> >> + "calibration" : "CPU0", >>> >> + "lock_pages" : true, >>> >> + "ftrace" : true, >>> >> + "logdir" : "./", >>> >> + } >>> >> +} >>> >> + >>> >> diff --git a/doc/examples/cpufreq_governor_efficiency/calibration.sh >>> >> b/doc/examples/cpufreq_governor_efficiency/calibration.sh >>> >> new file mode 100755 >>> >> index 0000000..d10e644 >>> >> --- /dev/null >>> >> +++ b/doc/examples/cpufreq_governor_efficiency/calibration.sh >>> >> @@ -0,0 +1,11 @@ >>> >> +#!/bin/sh >>> >> + >>> >> +set -e >>> >> + >>> >> +echo performance > >>> > /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor >>> >> + >>> >> +sleep 1 >>> >> + >>> >> +pLoad=$(rt-app calibration.json 2>&1 |grep pLoad |sed 's/.*= >>> > $.*$ns.*/\1/') >>> >> +sed 's/"calibration" : .*,/"calibration" : '$pLoad',/' -i dvfs.json >>> >> + >>> >> diff --git a/doc/examples/cpufreq_governor_efficiency/dvfs.json b/ >>> >> doc/examples/cpufreq_governor_efficiency/dvfs.json >>> >> new file mode 100644 >>> >> index 0000000..b413156 >>> >> --- /dev/null >>> >> +++ b/doc/examples/cpufreq_governor_efficiency/dvfs.json >>> >> @@ -0,0 +1,27 @@ >>> >> +{ >>> >> + "tasks" : { >>> >> + "thread" : { >>> >> + "instance" : 1, >>> >> + "cpus" : [0], >>> >> + "loop" : 5, >>> >> + "phases" : { >>> >> + "running" : { >>> >> + "loop" : 1, >>> >> + "run" : 100000, >>> >> + }, >>> >> + "sleeping" : { >>> >> + "loop" : 1, >>> >> + "sleep" : 1000000, >>> >> + } >>> >> + } >>> >> + } >>> >> + }, >>> >> + "global" : { >>> >> + "default_policy" : "SCHED_OTHER", >>> >> + "calibration" : 90, >>> >> + "lock_pages" : true, >>> >> + "ftrace" : true, >>> >> + "logdir" : "./", >>> >> + } >>> >> +} >>> >> + >>> >> diff --git a/doc/examples/cpufreq_governor_efficiency/dvfs.sh b/doc/ >>> >> examples/cpufreq_governor_efficiency/dvfs.sh >>> >> new file mode 100755 >>> >> index 0000000..8591fc7 >>> >> --- /dev/null >>> >> +++ b/doc/examples/cpufreq_governor_efficiency/dvfs.sh >>> >> @@ -0,0 +1,38 @@ >>> >> +#!/bin/sh >>> >> + >>> >> +#echo $1 $2 $3 >>> >> +set -e >>> >> + >>> >> +if [ $1 ] && [ $2 ] ; then >>> >> + for i in $(seq 0 1 $(expr $2 - 1)); do >>> >> + echo $1 > >> /sys/devices/system/cpu/cpu$i/cpufreq/scaling_governor >>> >> + #cat /sys/devices/system/cpu/cpu$i/cpufreq/scaling_governor >>> >> + done >>> >> + >>> >> + sleep 3 >>> >> +fi >>> >> + >>> >> +if [ $3 ] ; then >>> >> + sed 's/"run" : .*,/"run" : '$3',/' -i dvfs.json >>> >> +fi >>> >> + >>> >> +if [ $4 ] ; then >>> >> + sed 's/"sleep" : .*,/"sleep" : '$4',/' -i dvfs.json >>> >> +fi >>> >> + >>> >> +#cat dvfs.json >>> >> + >>> >> +rt-app dvfs.json 2> /dev/null >>> >> + >>> >> +if [ $1 ] ; then >>> >> + mv -f rt-app-thread-0.log rt-app_$1_run$3us_sleep$4us.log >>> >> + >>> >> + sum=0 >>> >> + for i in $(cat rt-app_$1_run$3us_sleep$4us.log | sed 'n;d' | sed >>> >> '1d' |cut -f 3); do >>> >> + sum=$(expr $sum + $i) >>> >> + done >>> >> + sum=$(expr $sum / 5) >>> >> + echo $sum >>> >> + rm -f rt-app_$1_run$3us_sleep$4us.log >>> >> +fi >>> >> + >>> >> diff --git a/doc/examples/cpufreq_governor_efficiency/test.sh b/doc/ >>> >> examples/cpufreq_governor_efficiency/test.sh >>> >> new file mode 100755 >>> >> index 0000000..d72fc6a >>> >> --- /dev/null >>> >> +++ b/doc/examples/cpufreq_governor_efficiency/test.sh >>> >> @@ -0,0 +1,82 @@ >>> >> +#!/bin/sh >>> >> + >>> >> +set -e >>> >> + >>> >> +set_calibration() { >>> >> + calibration.sh >>> >> +} >>> >> + >>> >> +test_efficiency() { >>> >> + >>> >> + FILENAME="results_$RANDOM$$.txt" >>> >> + >>> >> + if [ -e /sys/devices/system/cpu/cpu0/cpufreq/ >>> >> scaling_available_governors ]; then >>> >> + for i in $(cat /sys/devices/system/cpu/cpu0/cpufreq/ >>> >> scaling_available_governors); do >>> >> + export gov_$i=$(echo $i) >>> >> + done >>> >> + else >>> >> + echo "cpufreq is not available!" >>> >> + exit >>> >> + fi >>> >> + >>> >> + if [ ! $gov_performance ] ; then >>> >> + echo "Can't find performance governor!" >>> >> + exit >>> >> + fi >>> >> + >>> >> + if [ ! $gov_powersave ] ; then >>> >> + echo "Can't find powersave governor!" >>> >> + exit >>> >> + fi >>> >> + >>> >> + # Get powersave data >>> >> + dvfs.sh powersave $1 $2 $3 > $FILENAME >>> >> + powersave=$(cat $FILENAME |sed -n '1p') >>> >> + >>> >> + # Get performance data >>> >> + dvfs.sh performance $1 $2 $3 > $FILENAME >>> >> + performance=$(cat $FILENAME |sed -n '1p') >>> >> + >>> >> + if [ $performance -ge $powersave ] ; then >>> >> + echo "Error! Probably not input all the cpus in the same >>> >> frequency domain" >>> >> + exit >>> >> + fi >>> >> + >>> >> + denominator=$(expr $powersave - $performance) >>> >> + echo "powersave efficiency: 0%" >>> >> + echo "performance efficiency: 100%" >>> >> + >>> >> + # Calcuate other governors data >>> >> + for gov_next in $gov_conservative $gov_ondemand $gov_cfs; do >>> >> + if [ "$gov_next" != "" ] ; then >>> >> + dvfs.sh $gov_next $1 $2 $3 > $FILENAME >>> >> + data=$(cat $FILENAME |sed -n '1p'); >>> >> + numerator=$(expr $powersave - $data) >>> >> + numerator=$(expr $numerator \* 100) >>> >> + if [ $numerator -lt 0 ] ; then >>> >> + let numerator=0 >>> >> + fi >>> >> + data=$(expr $numerator / $denominator) >>> >> + echo "$gov_next efficiency: $data%" >>> >> + fi >>> >> + done >>> >> + >>> >> + rm -f $FILENAME >>> >> +} >>> >> + >>> >> +if [ $# -lt 3 ]; then >>> >> + echo "Usage: ./test.sh <cpus> <runtime> <sleeptime>" >>> >> + echo "cpus: number of cpus in the CPU0's frequency domain" >>> >> + echo "runtime: running time in ms per loop of the workload >> pattern" >>> >> + echo "sleeptime: sleeping time in ms per loop of the workload >>> > pattern" >>> >> + echo -e "\nExample: \n\"./test.sh 4 100 1000\" means\nCPU0~CPU3 >>> >> sharing frequency, \"100ms run + 1000ms sleep\" workload pattern.\n" >>> >> + exit >>> >> +fi >>> >> + >>> >> +echo "Frequency domain CPU0~CPU$(expr $1 - 1), run $2ms, sleep >> $3ms:" >>> >> + >>> >> +sleep 1 >>> >> +PATH=$PATH:. >>> >> +set_calibration >>> >> +test_efficiency $1 $(expr $2 \* 1000) $(expr $3 \* 1000) >>> >> + >>> >> -- >>> >> 1.9.1 >>> >>

11 years, 3 months

eas-test branch ci failed

by Alex Shi

Hi Fathi, I just updated the eas-test branch on https://git.linaro.org/kernel/eas-backports.git But the dashboard said deploy linaro kernel failed: https://validation.linaro.org/dashboard/streams/public/team/linaro/eas/bund… And I didn't find the eas on https://ci.linaro.org/view/kernel-ci/ or lsk-ci. Guess I miss sth, could you lead me to right ci or testing url? Thanks&Regards! Alex

11 years, 4 months

Updates to cpu capacity consolidation patches in the backports tree

by Amit Kucheria

Hi Alex, In 4.1-rc1, several patches (see 36ee28e4 onwards) related to cpu capacity consolidation were merged. It would be a good idea to refresh the eas-backport tree so that these patches are cherry-picked directly from mainline into stable/sched-upstream branch and their equivalent versions in stable/sched-core are removed. Regards, Amit

11 years, 4 months

[RFC internal v2 0/4] scheduler-based cpu frequency scaling

by Michael Turquette

This series implements an event-driven cpufreq governor that scales cpu frequency as a function of cfs runqueue utilization. The intent of this RFC is to get some discussion going about how the scheduler can become the policy engine for selecting cpu frequency, what limitations exist and what design do we want to take to get to a solution. This series depends on having frequency-invariant representations for load. This requires Vincent's recently merged cpu capacity rework patches, as well as a new patch from Morten included here. Morten's patch will likely make an appearance in his energy aware scheduling v4 series. Thanks to Juri Lelli <juri.lelli(a)arm.com> for contributing to the development of the governor. A git branch with these patches can be pulled from here: https://git.linaro.org/people/mike.turquette/linux.git sched-freq Smoke testing has been done on an OMAP4 Pandaboard and an Exynos 5800 Chromebook2. Extensive benchmarking and regression testing has not yet been done. Before sinking too much time into extensive testing I'd like to get feedback on the general design. Michael Turquette (3): sched: sched feature for cpu frequency selection sched: export get_cpu_usage & capacity_orig_of sched: cpufreq_sched_cfs: PELT-based cpu frequency scaling Morten Rasmussen (1): arm: Frequency invariant scheduler load-tracking support arch/arm/include/asm/topology.h | 7 + arch/arm/kernel/smp.c | 53 ++++++- arch/arm/kernel/topology.c | 17 +++ drivers/cpufreq/Kconfig | 24 +++ include/linux/cpufreq.h | 3 + kernel/sched/Makefile | 1 + kernel/sched/cpufreq_sched_cfs.c | 314 +++++++++++++++++++++++++++++++++++++++ kernel/sched/fair.c | 20 ++- kernel/sched/features.h | 6 + kernel/sched/sched.h | 9 ++ 10 files changed, 450 insertions(+), 4 deletions(-) create mode 100644 kernel/sched/cpufreq_sched_cfs.c -- 1.9.1

11 years, 5 months

[PATCH 0/4] scheduler-based cpu frequency scaling

by Michael Turquette

This series implements an event-driven cpufreq governor that scales cpu frequency as a function of cfs runqueue utilization. The intent of this RFC is to get some discussion going about how the scheduler can become the policy engine for selecting cpu frequency, what limitations exist and what design do we want to take to get to a solution. This work is a different take on the patches I posted in November: http://lkml.kernel.org/r/<1413958051-7103-1-git-send-email-mturquette(a)linaro.org> This series depends on having frequency-invariant representations for load. This requires Vincent's recently merged cpu capacity rework patches, as well as a new patch from Morten included here. Morten's patch will likely make an appearance in his energy aware scheduling v4 series. Thanks to Juri Lelli <juri.lelli(a)arm.com> for contributing to the development of the governor. A git branch with these patches can be pulled from here: https://git.linaro.org/people/mike.turquette/linux.git sched-freq Smoke testing has been done on an OMAP4 Pandaboard and an Exynos 5800 Chromebook2. Extensive benchmarking and regression testing has not yet been done. Before sinking too much time into extensive testing I'd like to get feedback on the general design. Michael Turquette (3): sched: sched feature for cpu frequency selection sched: export get_cpu_usage & capacity_orig_of sched: cpufreq_sched_cfs: PELT-based cpu frequency scaling Morten Rasmussen (1): arm: Frequency invariant scheduler load-tracking support arch/arm/include/asm/topology.h | 7 + arch/arm/kernel/smp.c | 53 ++++++- arch/arm/kernel/topology.c | 17 +++ drivers/cpufreq/Kconfig | 24 ++++ include/linux/cpufreq.h | 3 + kernel/sched/Makefile | 1 + kernel/sched/cpufreq_cfs.c | 311 ++++++++++++++++++++++++++++++++++++++++ kernel/sched/fair.c | 48 +++---- kernel/sched/features.h | 6 + kernel/sched/sched.h | 39 +++++ 10 files changed, 475 insertions(+), 34 deletions(-) create mode 100644 kernel/sched/cpufreq_cfs.c -- 1.9.1

11 years, 5 months

EAS on Juno LSK conflicting patches

by Juri Lelli

Hi Alex, Robin asked me to share with you the patches that I found having conflicts on the Juno LSK integration branch. It would be helpful to know your feelings about them (are they fundamental? can they be dropped? etc.). The aim here is to understand if we can eventually come up with a sched-upstream branch that is more easily applicable on top of the Juno integration branch: http://git.linaro.org/landing-teams/working/arm/kernel.git integration-lsk-3.10-juno-android I'll try to give you a list of patches that I had to remove or revert from sched-upstream when trying to build a branch that supports Juno and has EASv3 on top (so that we could test it on Android). I'll also add comments about patches that I'm more familiar with. Please be aware that I had to come up with this in quite a hurry, as we wanted to show EAS results coming from Juno at the last Connect. This means that the rationale for deciding to keep or drop a patch in the branch was merely "drop everything that seems to have non trivial conflicts". At the end I was fairly happy with the behaviours, but this doesn't necessarily mean that we still not need some of what I dropped. 080bdd1 exit.c: unexport __set_special_pids() cdba026 ptrace: revert "Prepare to fix racy accesses on task breakpoints" 73bcb2e mm: remove free_area_cache 802d8c2 ptrace/x86: revert "hw_breakpoints: Fix racy access to ptrace breakpoints" 6472001 ptrace/arm: revert "hw_breakpoints: Fix racy access to ptrace breakpoints" f0a709a sched: Add new scheduler syscalls to support an extended scheduling parameters ABI 631db36 sched/core: Fix htmldocs warnings 4f89001 sched: Preserve the nice level over sched_setscheduler() and sched_setparam() calls c9d415b sched: Fix up scheduler syscall LTP fails 0525ffc sched: Fix up attr::sched_priority warning b62aed3 sched: Move SCHED_RESET_ON_FORK into attr::sched_flags 7d214b1 sched: Fix __sched_setscheduler() nice test 9db9c75 sched: Fix information leak in sys_sched_getattr() 10d4f89 sched: Add 'flags' argument to sched_{set,get}attr() syscalls as this preceding set comes with the (or are subsequent fixes of) the SCHED_DEADLINE patchset, I think we could just ignore them for now 02fee98 ARM: introduce common set_auxcr/get_auxcr functions 6a4f9e7 sched: Implement task_nice() as static inline function this preceding patch is maybe not a crucial fix? 6261e04 sched: Adjust p->sched_reset_on_fork when nothing else changes 33e8006 sched: Replace hardcoding of -20 and 19 with MIN_NICE and MAX_NICE this two preceding patches depend on 6a4f9e7 15bfe13 arm64: use common reboot infrastructure 52cefcb arm64: Fix definition of arm_pm_restart to match the declaration 8c7b4f7 arm64: kernel: add CPU idle call e1bf4b0 time: Change the return type of clockevents_notify() to integer ff5e6b8 tracing: Do not do anything special with tracepoint_string when tracing is disabled 2e576ce tracepoint: add generic tracepoint definitions for IPI tracing 5174675 ARM: 7872/1: Support arch_irq_work_raise() via self IPIs 128647d ARM: SMP: basic IPI triggered completion support eccd85f ARM: add IPI tracepoints 5be433d arm64: enable generic clockevent broadcast e947689 arm64: Support arch_irq_work_raise() via self IPIs 5716acb ARM64: add IPI tracepoints I then had to revert the following patches. db9cfbf sched/idle: Optimize try-to-wake-up IPI 50299a0 sched/idle: Avoid spurious wakeup IPIs 7f219d1 sched, trace: Add a tracepoint for IPI-less remote wakeups 95ec016 sched/idle: Simplify wake_up_idle_cpu() And finally apply the attached patches to make the branch compile. Please don't hesitate to ask for more information, as I understand that all this might be quite confusing :). Best, - Juri

11 years, 5 months

Jump to page:

2026

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

eas-dev