SETI@Home may not work properly on newer Android (7 and 8) devices

疑难解答 : Unix/Linux : SETI@Home may not work properly on newer Android (7 and 8) devices
留言板合理

To post messages, you must log in.

前 · 1 · 2 · 3 · 4

作者消息
Profile Jord
志愿者测试人员
Avatar

发送消息
已加入:9 Jun 99
贴子:15175
积分:4,362,181
近期平均积分:3
Netherlands
消息 1961455 - 发表于:22 Oct 2018, 7:11:46 UTC - 回复消息 1961444.  

The ARMv7 is a 32-bit CPU. You won't get the v8.01 app on that device.

Not all Android devices manage to run all the Seti applications correctly. If you see continuous errors, your device is one that cannot be used here, no need to continue trying.

You can still try it with any other project out there.
ID: 1961455 · 举报违规帖子     回复 引用
Mike Ryan

发送消息
已加入:24 Jun 99
贴子:46
积分:24,363,752
近期平均积分:47
United States
消息 1961444 - 发表于:22 Oct 2018, 4:18:26 UTC

Just for my own morbid curiosity, is there any way (other than "just wait") to get BOINC to try the 64 bit v8.01 application aarch64-android-linux-gnu?

I have a couple of phones I've added to my computing mix. My very old LG Risio H343 has been happily crunching tasks under Android 6.0 with its Snapdragon 400 (32 bit 4x Cortex-A53) for a couple of weeks. So far seems to be averaging around 140 credits per day.

My newer phone is a Motorola e5 Cruise running Android 8.0 (Oreo) on a Snapdragon 425 (64 bit 4x Cortex-A53). So far ALL of the work units sent to this phone have failed. I believe around 300 work units have now been sent, but none tried have been the 8.01 application. From what this thread says it seems like that application will eventually be tried -- and perhaps it will run fine.

Any ideas? Or should I just shut this one down?
ID: 1961444 · 举报违规帖子     回复 引用
Profile Pierre A Renaud FCD @ team Carl Sagan
Avatar

发送消息
已加入:3 Apr 99
贴子:978
积分:9,101,544
近期平均积分:65
Canada
消息 1943596 - 发表于:9 Jul 2018, 21:07:30 UTC - 回复消息 1943317.  
最近的修改日期:9 Jul 2018, 21:17:19 UTC

I'd be glad to add it directly to their database, but to my knowledge it can only be done though the
non-intensive WUProp@home project (BOINC / Add Project / http://wuprop.boinc-af.org).

I personally only run it for a while for new computers/tablets/smartphones or when I attach a new project - in order to help them build their useful database.

WUProp@home is a non-intensive project that uses Internet-connected computers
to collect workunits properties of BOINC projects such as computation time, memory 
requirements, checkpointing interval or report limit. You can participate by downloading 
and running a free program on your computer.

You can add this:
ulefone Power_3 - SDK:25 ABI: arm64-v8a
OpenCL GPU Mali-G71 (4096MB) driver: 1.02
Tullio

Apr 3, 1999 - May 3, 2020
ID: 1943596 · 举报违规帖子     回复 引用
Profile tullio
志愿者测试人员

发送消息
已加入:9 Apr 04
贴子:8012
积分:2,930,782
近期平均积分:1
Italy
消息 1943317 - 发表于:8 Jul 2018, 17:49:23 UTC - 回复消息 1943248.  
最近的修改日期:8 Jul 2018, 17:58:00 UTC

You can add this:
ulefone Power_3 - SDK:25 ABI: arm64-v8a
OpenCL GPU Mali-G71 (4096MB) driver: 1.02
Tullio
ID: 1943317 · 举报违规帖子     回复 引用
Profile Pierre A Renaud FCD @ team Carl Sagan
Avatar

发送消息
已加入:3 Apr 99
贴子:978
积分:9,101,544
近期平均积分:65
Canada
消息 1943248 - 发表于:8 Jul 2018, 9:54:44 UTC
最近的修改日期:8 Jul 2018, 9:58:19 UTC

Because my 8 logical cores' LG Q6 (Android 7.1.1 PIE) keeps erroring out SETI@home WUs, it's now crunching World Community Grid data without problems. My much older 2 cores' Polaroid A4 (Android 4.4.2 non-PIE) didn't have issues with SETI@home WUs.

For the curious, here's a WUprop list of Android devices that can crunch through SETI@home WUs.
Apr 3, 1999 - May 3, 2020
ID: 1943248 · 举报违规帖子     回复 引用
Profile tullio
志愿者测试人员

发送消息
已加入:9 Apr 04
贴子:8012
积分:2,930,782
近期平均积分:1
Italy
消息 1942893 - 发表于:6 Jul 2018, 16:47:24 UTC

801 tasks run on my Ulefone 3, 800 tasks fail. Einstein@home tasks run also and get me credits.
Tullio
ID: 1942893 · 举报违规帖子     回复 引用
Profile Jord
志愿者测试人员
Avatar

发送消息
已加入:9 Jun 99
贴子:15175
积分:4,362,181
近期平均积分:3
Netherlands
消息 1942179 - 发表于:1 Jul 2018, 6:31:25 UTC - 回复消息 1939957.  
最近的修改日期:1 Jul 2018, 6:54:46 UTC

Why not? You have to charge your phone, at least every night. During that charging it will be doing BOINC.
Here, look at my hosts. They're one tablet and two phones. (Leave the studio-ST1022 out for the moment, as its battery is broken and so it isn't crunching anymore).

Of these the HUAWEI VNS-L31 - SDK:24 is my day to day phone. It charges during the night, and sometimes during the day when I am not specifically using it. But mostly during my own sleep cycle, of about 7 hours.

The other Huawei is a fun story as well, it's my old smart phone that's now just crunching. It has been crunching since October 2017 and is still figuring out which of the 12 (!!) science applications is the fastest and most reliable to do science with. I wonder if it will ever figure this out.

The tablet and my other phone are Android 7 on a 64bit CPU, so they've done the run through faulty v8.01 tasks and are now almost only running v8.00 apps.And see that RAC of 611? Although you have to see it without the ST1022's RAC, so then that's a RAC of ~570 done on these guys at a cost of just 1 Kilowatt hour every 20 weeks or so. That includes the power used by the 230mm BitFenix fan that's cooling them and the 12V AC->Molex adapter that's powering the fan.
ID: 1942179 · 举报违规帖子     回复 引用
Profile tullio
志愿者测试人员

发送消息
已加入:9 Apr 04
贴子:8012
积分:2,930,782
近期平均积分:1
Italy
消息 1939957 - 发表于:17 Jun 2018, 17:22:01 UTC

I found that to run both SETI@home and Einstein@home on my Ulefone smart phone with an ARM64 CPU I have to connect the phone to a power outlet, so it is not likely that smart phone users will contribute much to BOINC.
Tullio
ID: 1939957 · 举报违规帖子     回复 引用
Profile tullio
志愿者测试人员

发送消息
已加入:9 Apr 04
贴子:8012
积分:2,930,782
近期平均积分:1
Italy
消息 1935370 - 发表于:12 May 2018, 16:05:31 UTC
最近的修改日期:12 May 2018, 16:06:36 UTC

The SETI RAC of my Ulefone smartphone using Android 7.1.1 on a ARM64 CPU has almost reached that of the Windows 10 PC with a four core A10-6700. The Computer page says it has an OpenCl GPU which is not however used so far. 801 tasks run well, 800 fail.
Tullio
ID: 1935370 · 举报违规帖子     回复 引用
Stonelaughter

发送消息
已加入:12 Mar 14
贴子:1
积分:184,234
近期平均积分:0
消息 1934800 - 发表于:10 May 2018, 9:40:00 UTC

Would this explain why "HTC Power To Give" won't allocate CPU to SETI@Home tasks? They either sit there "Ready" for ever, or they "error out".

I am running a HTC U11 with Android 8.0.
ID: 1934800 · 举报违规帖子     回复 引用
Profile tullio
志愿者测试人员

发送消息
已加入:9 Apr 04
贴子:8012
积分:2,930,782
近期平均积分:1
Italy
消息 1934025 - 发表于:6 May 2018, 18:46:17 UTC

On this ARM64 CPU time and Runtime are exactly equal.
Tullio
ID: 1934025 · 举报违规帖子     回复 引用
Profile tullio
志愿者测试人员

发送消息
已加入:9 Apr 04
贴子:8012
积分:2,930,782
近期平均积分:1
Italy
消息 1932615 - 发表于:29 Apr 2018, 8:27:11 UTC
最近的修改日期:30 Apr 2018, 5:07:04 UTC

I have a new Ulefone smartphone with ARM64 and Android 7.1.1. Some SETI@home 8.00 tasks fail on it while 8.01 work and complete. Einstein@home tasks work too but I have not completed one yet. The funny thing is that SETI sees a Mali G71 GPU as coprocessor, while Einstein does not. The SETI 801 tasks are crunched by aarch64-android-linux-gnu. The Einstein tasks are crunched by arm-android-linux-gnu.
Tullio
Einstein tasks complete and validate
ID: 1932615 · 举报违规帖子     回复 引用
Profile Mr. Kevvy Crowdfunding Project Donor*Special Project $250 donor
志愿者负责人
志愿者测试人员
Avatar

发送消息
已加入:15 May 99
贴子:3205
积分:1,114,826,392
近期平均积分:3,319
Canada
消息 1930920 - 发表于:19 Apr 2018, 17:57:06 UTC
最近的修改日期:19 Apr 2018, 18:08:09 UTC

As always, thank you Jord... Dr. Korpela had indicated that it was the neon client that was failing. I will update my OP indicating.

I have tasks (example) that seem to be completing now very slowly on aarch64-android-linux-gnu but the neon ones all errored out.
ID: 1930920 · 举报违规帖子     回复 引用
Profile Jord
志愿者测试人员
Avatar

发送消息
已加入:9 Jun 99
贴子:15175
积分:4,362,181
近期平均积分:3
Netherlands
消息 1930918 - 发表于:19 Apr 2018, 17:44:11 UTC - 回复消息 1930913.  
最近的修改日期:19 Apr 2018, 17:44:48 UTC

Do know though that the v8.01 application (aarch64-android-linux-gnu) works fine on Android 7 (and maybe 8). It is just the v8.00 arm64-vfpv4 and v8.00 arm64-neon that don't work. If you allow for these to run errors, eventually they will be on the 1 task per day limit for these applications and you should get mostly tasks for the v8.01 app which will finish fine. Like so:

<core_client_version>7.4.53</core_client_version>
<![CDATA[
<stderr_txt>
Unable to resolve function unwind_backtrace_signal_arch
Unable to resolve function acquire_my_map_info_list
Unable to resolve function release_my_map_info_list
Unable to resolve function get_backtrace_symbols
Unable to resolve function free_backtrace_symbols
Unable to resolve function format_backtrace_line
Unable to resolve function load_symbol_table
Unable to resolve function free_symbol_table
Unable to resolve function find_symbol
one or more symbols not found. stackdumps unavailable
setiathome_v8 8.00 Revision: 3553 aarch64-linux-android-g++ (GCC) 4.9 20140827 (prerelease)
libboinc: BOINC 7.7.0

Work Unit Info:
...............
WU true angle range is :  2.179504
Triplet: peak=7.971074, time=32.25, period=0.7381, d_freq=1420280761.72, chirp=0, fft_len=16 
Spike: peak=24.17157, time=90.6, d_freq=1420285295.82, chirp=-11.657, fft_len=64k
Spike: peak=24.38553, time=90.6, d_freq=1420285295.82, chirp=-11.672, fft_len=64k
Autocorr: peak=19.17839, time=33.55, delay=2.2229, d_freq=1420283599.19, chirp=11.804, fft_len=128k
Unable to resolve function unwind_backtrace_signal_arch
Unable to resolve function acquire_my_map_info_list
Unable to resolve function release_my_map_info_list
Unable to resolve function get_backtrace_symbols
Unable to resolve function free_backtrace_symbols
Unable to resolve function format_backtrace_line
Unable to resolve function load_symbol_table
Unable to resolve function free_symbol_table
Unable to resolve function find_symbol
one or more symbols not found. stackdumps unavailable
Restarted at 38.55 percent.
Triplet: peak=8.216975, time=77.97, period=0.6095, d_freq=1420284035.8, chirp=-20.632, fft_len=64 
Triplet: peak=9.033834, time=105.1, period=0.1442, d_freq=1420281559.56, chirp=-25.79, fft_len=256 
Triplet: peak=8.289671, time=105.1, period=0.1442, d_freq=1420281560.43, chirp=-49.001, fft_len=256 
Triplet: peak=8.519686, time=23.36, period=0.9011, d_freq=1420279556.98, chirp=-51.579, fft_len=64 
Triplet: peak=8.200883, time=105.1, period=0.1442, d_freq=1420281557.19, chirp=-74.79, fft_len=256 
Triplet: peak=8.559072, time=105.1, period=0.1442, d_freq=1420281558.06, chirp=-98.001, fft_len=256 

Best spike: peak=24.38553, time=90.6, d_freq=1420285295.82, chirp=-11.672, fft_len=64k
Best autocorr: peak=19.17839, time=33.56, delay=2.2229, d_freq=1420283599.19, chirp=11.804, fft_len=128k
Best gaussian: peak=0, mean=0, ChiSq=0, time=-2.124e+11, d_freq=0,
	score=-12, null_hyp=0, chirp=0, fft_len=0 
Best pulse: peak=0.5680848, time=84.9, period=0.01807, d_freq=1420284106.26, score=0.9061, chirp=82.527, fft_len=16 
Best triplet: peak=9.033834, time=105.1, period=0.1442, d_freq=1420281559.56, chirp=-25.79, fft_len=256 

Flopcounter: 16219663652779.679688

Spike count:    2
Autocorr count: 1
Pulse count:    0
Triplet count:  7
Gaussian count: 0
03:07:14 (15539): called boinc_finish(0)

</stderr_txt>
]]>

ID: 1930918 · 举报违规帖子     回复 引用
Profile Mr. Kevvy Crowdfunding Project Donor*Special Project $250 donor
志愿者负责人
志愿者测试人员
Avatar

发送消息
已加入:15 May 99
贴子:3205
积分:1,114,826,392
近期平均积分:3,319
Canada
消息 1930913 - 发表于:19 Apr 2018, 17:06:10 UTC
最近的修改日期:1 Aug 2019, 17:21:24 UTC

(This has been well-known here for a while, at least a year, but I think it needs a pinned thread for those who aren't forum regulars.)

The SETI@Home client may not work on some newer Android smartphones, tablets, etc. Work units may crash and show as "Error while computing" in the device's Tasks listing. Here is an example task from my own Samsung Galaxy S7. The error shows:

process got signal 6
</message>
<stderr_txt>
CANNOT LINK EXECUTABLE "../../projects/setiathome.berkeley.edu/setiathome_8.00_arm-android-linux-gnu__armv7-neon": /data/data/edu.berkeley.boinc/client/projects/setiathome.berkeley.edu/setiathome_8.00_arm-android-linux-gnu__armv7-neon: has text relocations


The technical reason behind this is that the SETI@Home Android client is 32-bit and has never been ported to Android's 64-bit CLANG C++ suite. Some devices have a 32-bit subsystem to run these older 32-bit apps. (much like desktop OSes do so you can still use your older programs in Windows 10 64-bit, for example) but some do not and the client will not launch. Even if the 32-client does work, tasks may complete slowly compared to a native 64-bit client. (Thanks to Dr. Korpela for confirming this.)

If you have a newer Android device and it doesn't seem to be doing any work, you can confirm this issue by clicking your display name at the top-right of this page, then the View link beside Computers on this account, then find the device and click Tasks beside it. Under the Status column look for Error while computing.

If this is noted, go back to Computers on this account and click Details. Beside Operating system will be the Android version. If it shows (Android 7.0) or higher (ie 8 or 9) then this is the issue and SETI@Home may not work properly on this device until new client is released. After enough work units have errored out, some may start to complete slowly as they are based on a different Android platform (see Ageless' post after this one.)

In the interim, if it doesn't work at all or is too slow, there are other BOINC projects with updated clients which are compatible with Android 7+. Unfortunately there is no ETA for an updated Android SETI@Home client at this time.
ID: 1930913 · 举报违规帖子     回复 引用
前 · 1 · 2 · 3 · 4

疑难解答 : Unix/Linux : SETI@Home may not work properly on newer Android (7 and 8) devices


 
©2020 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.