Nvidia Tegrastats-max

You are using an out of date browser. It may not display this or other websites correctly.
You should upgrade or use an culling browser.

[HOWTO] Monitor/Adapt arrangement resources with tegrastats (CPU load, speed, mem, etc.)

  • Thread starter Da_G
  • Beginning date
Da_G

Da_G

Inactive Senior RD / Moderator Emeritus
  • #1
Hi again :)

The Atrix 4G includes a nice trivial command-line app called "tegrastats" (root required)

The app has 4 functions:
"tegrastats -max" - sets cpu min and max frequencies to 1ghz (locks cpu speed @ max)
"tegrastats -default" - sets cpu min and max frequencies to default (seems to range from 300MHz to 1000MHz and accept many more steppings available than listed in /sys/devices/system/cpu/cpu0/cpufreq/scaling_available_frequencies)
"tegrastats [delay in ms]" - outputs a line per second to adb logcat detailing the current system load per cpu, overall, and the % load relative to current CPU speed. If [delay in ms] is specified that value will be used for the delay betwixt logcat updates rather than ane second.
"tegrastats -debug" - same equally higher up with more than detailed info

[size=+1]What does the logcat output mean?[/size]

RAM X/Y (lfb Z) - X is the amount of RAM in apply in megabytes. Y is the total corporeality of RAM available for applications. Z is the size of the largest gratis block.
Carveout X/Y (lfb Z) - ("carveout" refers to video memory, or GPU memory.) Ten is the amount of carveout retention in employ, in megabytes. Y is the total corporeality available. Z is the size of the largest free block.
GART Ten/Y (lfb Z) - ("GART" refers to system retention that has been mapped to exist visible to the GPU. It is typically allocated when carveout allocations fail.) X is the amount of GART memory in use, in megabytes Y is the total amount available. Z is the size of the largest free block.
IRAM X/Y (lfb Z) - ("IRAM" refers to dedicated memory attached to the AVP processor.) X is the corporeality of IRAM memory in apply, in kilobytes. Y is the total amount bachelor. Z is the size of the largest free block.
cpu [X%,Y%]@Z - X is the percentage of CPU0 being used relative to the current running frequency. Z. Y is the percent of CPU1 being used (or 'off' if second cadre is turned off.) Z is the CPU frequency in MHz (the frequency volition dynamically get up or downward depending on what workload the CPU is seeing.)
avp 10%@Y - ("avp" is the audio/video processor, which is not visible to either the Bone or applications. However, some forms of video decode/encode make heavy use of it.) X is the percent of the AVP that is being used (aka CPU load), relative to the current running frequency Y. Y is the AVP frequency in MHz (the frequency will dynamically become up or down depending on what workload the AVP is seeing)
vde Ten%@Y ("vde" is the video decode engine.) 10 is the percent of the VDE that is existence used (aka CPU load), relative to the current running frequency Y. Y is the VDE frequency in MHz.
emc X%@Y ("emc" is the external retentivity controller, which all sysmem/carveout/GART memory accesses go through) X is the per centum of the EMC memory bandwidth that is existence used, relative to the current running frequency Y. Y is the EMC frequency in MHz.

I've noticed that changing CPU limits with SetCPU or /sys/devices/system/cpu/cpu0/cpufreq/scaling_min_freq & scaling_max_freq seem to conflict with the built in Dynamic Frequency Scaler, sometimes it honors the request, other times overrides it (per the tegrastats output)

For more info about tegrastats etc.: http://programmer.download.nvidia.com/tegra/docs/linux_for_tegra_quickstart_20100903_10_8_2.pdf

Below are several sample output lines from "tegrastats -debug" from adb:

                                    03-16 19:11:10.669 10014 10014 E TegraStats: total0 101  idle0 96 | total1 146  idle1 146 03-16 19:xi:x.669 10014 10014 E TegraStats: DFS CPU load six, OS CPU load 2 03-16 xix:xi:10.669 10014 10014 E TegraStats: RAM 555/818MB (lfb 1x2MB) Carveout 48/192MB (lfb 38MB) GART 11204/32768KB (lfb 8MB) GART_Free(21564KB) GART_Pins(80KB) GART_Non-Pins(11124KB, lnpb(1056KB)) IRAM 72/256kB (lfb 184kB) | DFS(%@MHz): cpu [4%,0%]@501 avp 2%@120 vde 0%@120 emc 0%@108 03-16 nineteen:11:eleven.670 10014 10014 E TegraStats: total0 100  idle0 88 | total1 149  idle1 149 03-xvi 19:11:11.670 10014 10014 E TegraStats: DFS CPU load ix, OS CPU load 4 03-16 19:11:11.670 10014 10014 East TegraStats: RAM 555/818MB (lfb 1x2MB) Carveout 48/192MB (lfb 38MB) GART 11204/32768KB (lfb 8MB) GART_Free(21564KB) GART_Pins(80KB) GART_Non-Pins(11124KB, lnpb(1056KB)) IRAM 72/256kB (lfb 184kB) | DFS(%@MHz): cpu [12%,0%]@501 avp two%@120 vde 0%@120 emc 0%@108 03-xvi 19:eleven:12.679 10014 10014 Due east TegraStats: total0 100  idle0 90 | total1 105  idle1 104 03-xvi nineteen:11:12.679 10014 10014 E TegraStats: DFS CPU load xiii, Bone CPU load 5 03-16 19:11:12.679 10014 10014 E TegraStats: RAM 555/818MB (lfb 1x2MB) Carveout 48/192MB (lfb 38MB) GART 11204/32768KB (lfb 8MB) GART_Free(21564KB) GART_Pins(80KB) GART_Non-Pins(11124KB, lnpb(1056KB)) IRAM 72/256kB (lfb 184kB) | DFS(%@MHz): cpu [10%,0%]@501 avp 2%@120 vde 0%@120 emc 0%@108 03-16 xix:xi:13.679 10014 10014 East TegraStats: total0 101  idle0 99 | total1 49  idle1 49 03-xvi xix:eleven:13.679 10014 10014 E TegraStats: DFS CPU load eight, Os CPU load i 03-16 19:11:13.679 10014 10014 E TegraStats: RAM 555/818MB (lfb 1x2MB) Carveout 48/192MB (lfb 38MB) GART 11204/32768KB (lfb 8MB) GART_Free(21564KB) GART_Pins(80KB) GART_Non-Pins(11124KB, lnpb(1056KB)) IRAM 72/256kB (lfb 184kB) | DFS(%@MHz): cpu [one%,0%]@501 avp two%@120 vde 0%@120 emc 0%@108 03-sixteen xix:11:fourteen.679 10014 10014 E TegraStats: total0 100  idle0 99 | total1 68  idle1 68 03-16 19:eleven:14.689 10014 10014 East TegraStats: DFS CPU load 5, Os CPU load 0 03-sixteen 19:11:14.689 10014 10014 Due east TegraStats: RAM 555/818MB (lfb 1x2MB) Carveout 48/192MB (lfb 38MB) GART 11204/32768KB (lfb 8MB) GART_Free(21564KB) GART_Pins(80KB) GART_Non-Pins(11124KB, lnpb(1056KB)) IRAM 72/256kB (lfb 184kB) | DFS(%@MHz): cpu [1%,0%]@501 avp 2%@120 vde 0%@120 emc 0%@108 03-16 19:11:15.691 10014 10014 E TegraStats: total0 101  idle0 99 | total1 93  idle1 93 03-xvi nineteen:xi:xv.691 10014 10014 E TegraStats: DFS CPU load 4, OS CPU load one 03-sixteen xix:eleven:15.691 10014 10014 E TegraStats: RAM 555/818MB (lfb 1x2MB) Carveout 48/192MB (lfb 38MB) GART 11204/32768KB (lfb 8MB) GART_Free(21564KB) GART_Pins(80KB) GART_Non-Pins(11124KB, lnpb(1056KB)) IRAM 72/256kB (lfb 184kB) | DFS(%@MHz): cpu [1%,0%]@501 avp 2%@120 vde 0%@120 emc 0%@108 03-16 19:11:xvi.692 10014 10014 E TegraStats: total0 100  idle0 98 | total1 195  idle1 193 03-xvi 19:11:16.692 10014 10014 E TegraStats: DFS CPU load 5, OS CPU load ane 03-16 19:11:16.692 10014 10014 East TegraStats: RAM 555/818MB (lfb 1x2MB) Carveout 48/192MB (lfb 38MB) GART 11204/32768KB (lfb 8MB) GART_Free(21564KB) GART_Pins(80KB) GART_Non-Pins(11124KB, lnpb(1056KB)) IRAM 72/256kB (lfb 184kB) | DFS(%@MHz): cpu [2%,ane%]@501 avp 2%@120 vde 0%@120 emc 0%@108                                  
Last edited:
ahmarchi
  • #two
G can we get this app at the market? or is it a sideload apk
Da_G

Da_G

Inactive Senior RD / Moderator Emeritus
  • #iii
It's included with the stock Atrix ROM.

To become information technology running, go adb going (either via installing the android sdk or grabbing aroot or another ane click tool which includes adb.exe)
Open two command prompts.
in 1,


which will whorl down the listing of various messages in the arrangement log, and continue scrolling in real time
at present you lot'll get the output in the other command prompt :)
  • #4
da_g! my quondam fuze dev :). is at that place anyway to get overclocking by manipulating these values?
ahmarchi
  • #5
Chiliad check this out... i did some enquiry and i constitute a pdf well-nigh the nvidia tegra.

Audio Book

Control

To modify the master audio book, you lot can employ the
ALSA amixer

utility, like:

amixer set Primary 300

That sets themaster volume to 300 (range of 0
-

ahmarchi
  • #6
ive read the pdf, talks nearly the tegratstat control and has ALOT of other absurd lilliputian info..... if u desire it i could send it to you lot to look at
Da_G

Da_G

Inactive Senior RD / Moderator Emeritus
  • #7
@tassadar898:

hi :) maybe? lets observe out!

@ahjdmarchi:

yeah, had information technology linked in the first post when i was composing it, deleted in ane of the edits, readded it just now :)

ahmarchi
  • #8
nm to anynyone who wnats to do it... AKA noobs like me heres some instrucitions ( in depth and so Thou dosent take to waste product fourth dimension)

open beginning CMD
go to your androidsdk/tool
run adb logcat

leave that one open up its will give a live reading and keep updating very fast

now open another cmd
go to your androidsdk/tools again
type adb vanquish
su
tegrastats -max ( for example)

now quick await at you other CMD window, it will say inverse max frequency etc... there done. thats how you use the tegratsatsts comman

NOW G ane terminal question for you..... is this a permanent mod, or does it have to be run every fourth dimension the telephone is rebooted/restarted???

Concluding edited:
Da_G

Da_G

Inactive Senior RD / Moderator Emeritus
  • #nine
@ahjdmarchi:

Information technology'southward only valid for the current session (reboot or poweroff negates information technology)
You could stick information technology into an init script if you desire it to occur every bootup.

ahmarchi
  • #x
anywhere that explains how to make an init script.....ive read though that PDF file... man theres some in depth stuff...
jug6ernaut
  • #eleven
Squeamish find, should exist able to some good stuff with this...
  • #12
Could someone post the atrix4g tegrastats file. I'd like to see if its unlike from the one on my gtab.. FWIW.. on the Xoom, its not included! At the very least, run an md5sum..

3c5fa74a48d4e58e336addf5541ef441
and the i included in the latest cm7 nightly for the gtab
26672690e5bec623bb401995509886de

they are the same size.. 9656 bytes.. yet they differ.. presumably not in function..

Nvidia is horrificly ambiguous when information technology comes to their approach to utilities and implementation (dvfs).. nvflash is a nice treat (apparently not on Xoom, gtab works cracking) - its good for when you get those ultra informative messages from the bootloader -- Magic Value mismatch.. "Its magical".. simply yeah.. tegrastats considering they must not like cpufreq. also much command for the finish user. :)

what board is atrix based on? harmony or ventana?

KeRmiT80
  • #13
i run the stats while doing the setcpu stress test, tegrastats -max

both cores were running at 100%@501 mhz

shouldn't run at ane ghz each core?

why in fix cpu i get only 501mhz and my cpu benchmarks scores are half than before?

                                  RT_Non-Pins(48KB, lnpb(48KB)) IRAM 124/256kB (lfb 132kB) | DFS(%@MHz): cpu [100%,100%]@501 avp 2%@240 vde 0%@240 emc nine%@300 06-21 xv:51:42.402 12552 12552 Due east TegraStats: RAM 365/818MB (lfb 14x4MB) Carveout 25/192MB (lfb 119MB) GART 48/32768KB (lfb 32MB) GART_Free(32720KB) GART_Pins(0KB) GART_Non-Pins(48KB, lnpb(48KB)) IRAM 124/256kB (lfb 132kB) | DFS(%@MHz): cpu [100%,100%]@501 avp 2%@240 vde 0%@240 emc 9%@300 06-21 15:51:42.512  1419  2430 I         : odm_disp_esd_thread is running(96) state=one 06-21 15:51:43.402 12552 12552 East TegraStats: RAM 365/818MB (lfb 14x4MB) Carveout 25/192MB (lfb 119MB) GART 48/32768KB (lfb 32MB) GART_Free(32720KB) GART_Pins(0KB) GART_Non-Pins(48KB, lnpb(48KB)) IRAM 124/256kB (lfb 132kB) | DFS(%@MHz): cpu [100%,100%]@501 avp 2%@240 vde 0%@240 emc viii%@300 06-21 15:51:44.402 12552 12552 E TegraStats: RAM 365/818MB (lfb 14x4MB) Carveout 25/192MB (lfb 119MB) GART 48/32768KB (lfb 32MB) GART_Free(32720KB) GART_Pins(0KB) GART_Non-Pins(48KB, lnpb(48KB)) IRAM 124/256kB (lfb 132kB) | DFS(%@MHz): cpu [100%,100%]@501 avp 2%@240 vde 0%@240 emc 9%@300 06-21 15:51:45.412 12552 12552 Due east TegraStats: RAM 365/818MB (lfb 14x4MB) Carveout 25/192MB (lfb 119MB) GART 48/32768KB (lfb 32MB) GART_Free(32720KB) GART_Pins(0KB) GART_Non-Pins(48KB, lnpb(48KB)) IRAM 124/256kB (lfb 132kB) | DFS(%@MHz): cpu [100%,100%]@501 avp ii%@240 vde 0%@240 emc 9%@300 06-21 15:51:45.882  5171  5171 D dalvikvm: GC_EXPLICIT freed 2692 objects / 188120 bytes in 128ms 06-21 fifteen:51:46.412 12552 12552 Eastward TegraStats: RAM 365/818MB (lfb 14x4MB) Carveout 25/192MB (lfb 119MB) GART 48/32768KB (lfb 32MB) GART_Free(32720KB) GART_Pins(0KB) GART_Non-Pins(48KB, lnpb(48KB)) IRAM 124/256kB (lfb 132kB) | DFS(%@MHz): cpu [100%,100%]@501 avp 2%@240 vde 0%@240 emc 9%@300 06-21 15:51:47.422 12552 12552 E TegraStats: RAM 365/818MB (lfb 14x4MB) Carveout 25/192MB (lfb 119MB) GART 48/32768KB (lfb 32MB) GART_Free(32720KB) GART_Pins(0KB) GART_Non-Pins(48KB, lnpb(48KB)) IRAM 124/256kB (lfb 132kB) | DFS(%@MHz): cpu [100%,100%]@501 avp two%@240 vde 0%@240 emc ix%@300 06-21 15:51:48.422 12552 12552 E TegraStats: RAM 365/818MB (lfb 14x4MB) Carveout 25/192MB (lfb 119MB) GART 48/32768KB (lfb 32MB) GART_Free(32720KB) GART_Pins(0KB) GART_Non-Pins(48KB, lnpb(48KB)) IRAM 124/256kB (lfb 132kB) | DFS(%@MHz): cpu [100%,100%]@501 avp 2%@240 vde 0%@240 emc nine%@300                                
  • #14
Was the phone hot?

cd /sys/devices/system/cpu/cpu0/
cat scaling_max_freq cpuinfo_cur_freq cpuinfo_max_freq scaling_cur_freq

it could be in thermal throttling fashion. Yous should find out what temp sensor is in the device. Depending on the ref board, it volition have a diff sensor. Sensor on my TF101 is quite easy to admission via sysfs.. gtab not so much..

cd /sys/
detect | grep temp

possibly the in a higher place might plough something up.

KeRmiT80
  • #15
i searched in cpufreq dir
                                    # cd /sys/devices/organisation/cpu/cpu0/cpufreq cd /sys/devices/system/cpu/cpu0/cpufreq # cat scaling_max_freq cpuinfo_cur_freq cpuinfo_max_freq scaling_cur_freq true cat scaling_max_freq cpuinfo_cur_freq cpuinfo_max_freq scaling_cur_freq million 300867 1000000 scaling_cur_freq: No such file or directory # cd /sys cd /sys # find | grep temp find | grep temp ./devices/platform/tegra_tmon/temp1_input ./devices/platform/tegra_tmon/temp2_input ./devices/platform/cpcap_battery/power_supply/bombardment/temp ./devices/w1 omnibus master/w1_master_attempts ./module/yaffs/parameters/yaffs_wr_attempts                                  

temp1 is 4200 temp2 is 3600

phone is cold... but i recall you are right

http://nv-tegra.nvidia.com/gitweb/?p=linux-2.6.git;a=log;h=dcb48435e04dba5aafe87cd2ebaa1e0f09bf90cb

                                    [ARM/tegra] RM: Updated thermal throttling algorithm.    (A) Interval (... 90C) - no throttling, free running organisation  (B) Interval (90C, 115C) - active throttling of CPU frequency with  - slope: -100 MHz/2 sec  [B]- depression limit: l% of CPU frequency maximum[/B]  (C) Above 115C - h/west shutdown.                                  
how practise i clean upward the TT?

edit: i managed to lower temp below 31C, 88F (even so 501 mhz max)

Last edited:
  • #xvi
quick question,
I have tried aura rom and honeycomb rom, aura heated my phone up pretty bad so i switched to honeycomb rom, went into setcpu, it was at 1.3gigs, turned it downward to max of 1gigs... withal heated upwardly pretty bad. (about 30-32 hours of normal utilise on stock two.two.two, united kingdom of great britain and northern ireland tmobile's vs nearly maximum viii hours on both of these roms....)
perchance the cpu was withal clocked at 1.3 i dunnoo... now i am scared to flash another rom ... any ideas how to bank check whats going on???
btw the process "10" consumes the battery about l% under normal use, dont know what tht is, Diviance told me it was some native linux process, wasnt able to impale it, it doesnt show nether any task managers...
Diviance thought my download was corrupt, retried everything, same event....
IUH1991
  • #17
quick question,
I have tried aura rom and honeycomb rom, aureola heated my phone up pretty bad and then i switched to honeycomb rom, went into setcpu, it was at 1.3gigs, turned it down to max of 1gigs... withal heated up pretty bad. (nigh 30-32 hours of normal use on stock two.2.2, uk tmobile'south vs nigh maximum 8 hours on both of these roms....)
maybe the cpu was notwithstanding clocked at 1.three i dunnoo... now i am scared to flash another rom ... any ideas how to check whats going on???
btw the process "X" consumes the battery about fifty% under normal use, dont know what tht is, Diviance told me information technology was some native linux process, wasnt able to kill it, it doesnt show under any task managers...
Diviance thought my download was corrupt, retried everything, aforementioned result....

Just install fauxs stock kernel. your phones overheating because it's overclocked.

Sent from my MB860 using XDA App

  • #18
will try tomorrow morn... mee zzzZZZZzzzz
thx though
BravoMotorola
  • #xix
Ok simply wondering because I saw this thread dorsum up, but what the hell always happened to Da_G? Did he dice or something? Honestly where did he become?
  • #20
I'm more than interested in the pdf file linked, its gone. Anyone have a copy?

Similar threads

wedgess

whitehilooppat49.blogspot.com

Source: https://forum.xda-developers.com/t/howto-monitor-adjust-system-resources-with-tegrastats-cpu-load-speed-mem-etc.996397/

0 Response to "Nvidia Tegrastats-max"

Post a Comment

Iklan Atas Artikel

Iklan Tengah Artikel 1

Iklan Tengah Artikel 2

Iklan Bawah Artikel