If the multi-threaded app in question can have its core usage controlled via a command-line parameter (most Windows MT apps can, using the --nthreads parameter), you should be able to use an app_config.xml to achieve this behaviour.

Client configuration - Application configuration

You would need to set both <avg_ncpus> (so BOINC can schedule other projects around the MT task properly), and <cmdline>--nthreads (to stop the MT app itself stepping outside its allotted boundary).

Thanks, Richard.

That appears to have done the trick.

Set those options, and restarted BOINC.

Now have one instance running (two more queued) with only 4 CPUs dedicated to each


If we ever meet, I owe you several drinks for your assistance over the last couple of weeks!!!

So far there´s still no auto update mechniasm within Boinc, are there any special reasons why this feature has not been impemented? I think it´s pretty common for a lot of applications these days.

I would like to have separate Minimum and Max additional buffer sizes for GPUs and CPUs. That is because for GPUGrid I need a small buffer (less than an hour usually) to earn bonus points for quick returns, whereas for CPU projects I would like a larger buffer for the usual reasons.

So far there´s still no auto update mechniasm within Boinc, are there any special reasons why this feature has not been impemented? I think it´s pretty common for a lot of applications these days.

That is intentional. Most institutions and corporations will not install software that auto update. It is impossible to test software against all possible system configuration. It is better for users to actively install updates so that they know what software may have caused problem.

Personally, the first thing I do when installing software is to disable auto update.

It is intentional, that's correct. You can actually find code for it in the source code of the client, it's just not finished and the developers won't finish it, because it opens the door for hackers to send malicious code onto people's computers, if they hack a BOINC server that can send updates.

I would like to have separate Minimum and Max additional buffer sizes for GPUs and CPUs. That is because for GPUGrid I need a small buffer (less than an hour usually) to earn bonus points for quick returns, whereas for CPU projects I would like a larger buffer for the usual reasons.

I agree. Unfortunately, this would require to also have a separate project resource share option for the GPU and CPU. This would require a complete rewrite of the manager work fetch algorithm as well as the server software while keeping things compatible with older versions.

I wouldn't expect the developers to address such a large task any time in near future.

---

Boinc V 7.4.36
Win7 i5 3.33G 4GB NVidia 470

It is intentional, that's correct. You can actually find code for it in the source code of the client, it's just not finished and the developers won't finish it, because it opens the door for hackers to send malicious code onto people's computers, if they hack a BOINC server that can send updates.

Thanks for your reply. I just thought to have a non default option or button in Boinc to check for updates and then a delta update of only changed files would be available. So maybe not auto update but in-app delta update from only one secure Boinc server (this one here maybe). I am just thinking of e.g. Firefox which also handles delta updates and is installed on millions of systems. Anayway, thanks for checking and maybe one day the devs will rethink their decision.

Support for Android x86 architecture:

https://software.intel.com/en-us/articles/whats-new-oclsdk-2014-r2

... and hopefully for crunching on Android GPU :)

Support for Android x86 architecture:

Support for x86 Android x86 architecture is already available:

http://boinc.berkeley.edu/download_all.php

Android
Android version 2.3 and later for ARM, MIPS, or x86
We recommend that, rather than downloading BOINC from here,
you get it from the Google Play Store or the Amazon app store (for Kindle Fire)

Claggy

I have overlooked this feature.
Then the next wish would be detection of OpenCL feature in BOINC / Android

there will always be a workaround, https://play.google.com/store/apps/details?id=com.xh.openclinfo

I have overlooked this feature.
Then the next wish would be detection of OpenCL feature in BOINC / Android

Support is already there:

Thu Jan 22 22:08:29 GMT 2015||(battery charge level 0.0% < threshold 90.0%
Thu Jan 22 22:08:29 GMT 2015||Suspending computation - battery low
Thu Jan 22 22:08:28 GMT 2015||Not using a proxy
Thu Jan 22 22:08:28 GMT 2015||(to change preferences, visit a project web site or select Preferences in the Manager)
Thu Jan 22 22:08:28 GMT 2015||don't use GPU while active
Thu Jan 22 22:08:28 GMT 2015||don't compute while active
Thu Jan 22 22:08:28 GMT 2015||max disk usage: 0.45GB
Thu Jan 22 22:08:28 GMT 2015||max memory usage when idle: 609.14MB
Thu Jan 22 22:08:28 GMT 2015||max memory usage when active: 380.71MB
Thu Jan 22 22:08:28 GMT 2015||Preferences:
Thu Jan 22 22:08:28 GMT 2015||Reading preferences override file
Thu Jan 22 22:08:28 GMT 2015|Albert@Home|URL http://albert.phys.uwm.edu/; Computer ID 9756; resource share 100
Thu Jan 22 22:08:28 GMT 2015|Einstein@Home|URL http://einstein.phys.uwm.edu/; Computer ID 9721755; resource share 100
Thu Jan 22 22:08:28 GMT 2015|SETI@home Beta Test|URL http://setiweb.ssl.berkeley.edu/beta/; Computer ID 69176; resource share 100
Thu Jan 22 22:08:28 GMT 2015|PrimeGrid|URL http://www.primegrid.com/; Computer ID 417716; resource share 100
Thu Jan 22 22:08:28 GMT 2015|Asteroids@home|URL http://asteroidsathome.net/boinc/; Computer ID 63394; resource share 100
Thu Jan 22 22:08:28 GMT 2015||Config: report completed tasks immediately
Thu Jan 22 22:08:28 GMT 2015||Local time is UTC +0 hours
Thu Jan 22 22:08:28 GMT 2015||Disk: 2.21 GB total, 522.13 MB free
Thu Jan 22 22:08:28 GMT 2015||Memory: 761.42 MB physical, 0 bytes virtual
Thu Jan 22 22:08:28 GMT 2015||OS: Android: 3.4.10-g2407d99
Thu Jan 22 22:08:28 GMT 2015||Processor features: swp half thumb fastmult vfp edsp neon vfpv3 tls vfpv4
Thu Jan 22 22:08:28 GMT 2015||Using 1 CPUs
Thu Jan 22 22:08:28 GMT 2015||Processor: 2 ARM ARMv7 Processor rev 0 (v7l)
Thu Jan 22 22:08:28 GMT 2015||Host name: android_59b3f67b
Thu Jan 22 22:08:28 GMT 2015||No usable GPUs found
Thu Jan 22 22:08:28 GMT 2015||OpenCL CPU: QUALCOMM Krait(TM) (OpenCL driver vendor: QUALCOMM, driver version build:Merge, device version OpenCL 1.1 Krait(TM))
Thu Jan 22 22:08:28 GMT 2015||Data directory: /data/data/edu.berkeley.boinc/client
Thu Jan 22 22:08:28 GMT 2015||Running as a daemon
Thu Jan 22 22:08:28 GMT 2015||Libraries: libcurl/7.39.0 OpenSSL/1.0.1j zlib/1.2.6.f-codeaurora-mods-v1
Thu Jan 22 22:08:28 GMT 2015||log flags: file_xfer, sched_ops, task
Thu Jan 22 22:08:28 GMT 2015||Starting BOINC client version 7.4.41 for arm-android-linux-gnu

For OpenCL GPU support to be even possible, Device vendors need to start supplying OpenCL support in the first place, While my 2012 HTC One S has CPU OpenCL support, neither my 2012 Nexus 7, nor my Samsung S6 mini have any OpenCL support:

2012 HTC One S:

Name: QUALCOMM Snapdragon(TM)
Vendor: QUALCOMM
Shared Library: /system/lib/libOpenCL.so
OpenCL Version: OpenCL 1.1 QUALCOMM build: AU_LINUX_ANDROID_JB_REL_RB4.04.01.02.16.028_msm8960_JB_REL_RB4.2_Merge_release_AU (Merge)
Build Date: 12/06/12 Thu
Local Branch:
Remote Branch: m/jb_rel_rb4.2
Reconstruct Branch: NOTHING
Profile: EMBEDDED_PROFILE
Extensions:

Name: QUALCOMM Krait(TM)
Type: CPU
Available: CL_TRUE
Vendor: QUALCOMM
Driver Version: build:Merge
OpenCL Version: OpenCL 1.1 Krait(TM)
OpenCL C Version: OpenCL C 1.1 Krait(TM)
Profile: EMBEDDED_PROFILE
Queue properties: CL_QUEUE_PROFILING_ENABLE CL_QUEUE_OUT_OF_ORDER_EXEC_MODE_ENABLE
Extensions: cl_khr_global_int32_base_atomics cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_atomics cl_khr_local_int32_extended_atomics cl_khr_byte_addressable_store cles_khr_int64 cl_khr_fp16

HOST_UNIFIED_MEMORY: CL_TRUE
ADDRESS_BITS: 32bits
BUILT_IN_KERNELS: not supported before OpenCL 1.2
COMPILER_AVAILABLE: CL_TRUE
SINGLE_FP_CONFIG: | CL_FP_INF_NAN | CL_FP_ROUND_TO_NEAREST | CL_FP_ROUND_TO_ZERO | CL_FP_ROUND_TO_INF | CL_FP_FMA
DOUBLE_FP_CONFIG:
HALF_FP_CONFIG: | CL_FP_INF_NAN | CL_FP_ROUND_TO_NEAREST | CL_FP_ROUND_TO_ZERO | CL_FP_ROUND_TO_INF
GLOBAL_MEM_CACHE_SIZE: 1048576
GLOBAL_MEM_CACHE_TYPE: CL_READ_WRITE_CACHE
GLOBAL_MEM_CACHELINE_SIZE: 64
GLOBAL_MEM_SIZE: 399204352
IMAGE_SUPPORT: CL_TRUE
IMAGE2D_MAX_HEIGHT: 8192
IMAGE2D_MAX_WIDTH: 8192
IMAGE3D_MAX_DEPTH: 2048
IMAGE3D_MAX_HEIGHT: 2048
IMAGE3D_MAX_WIDTH: 2048
IMAGE_MAX_BUFFER_SIZE: not supported before OpenCL 1.2
IMAGE_MAX_ARRAY_SIZE: not supported before OpenCL 1.2
LINKER_AVAILABLE: not supported before OpenCL 1.2
LOCAL_MEM_SIZE: 32768
LOCAL_MEM_TYPE: CL_GLOBAL
MAX_CLOCK_FREQUENCY: 1500
MAX_MEM_ALLOC_SIZE: 399204352
MAX_WORK_ITEM_SIZES: (1, 1, 1)

2012 Nexus 7:

OpenCL platform not found.
OpenCL shared libraries we tried to load:
/system/vendor/lib/libOpenCL.so
/system/lib/libOpenCL.so
/system/vendor/lib/libPVROCL.so
/system/vendor/lib/egl/libGLES_mali.so
/system/vendor/lib/libllvm-a3xx.so

2014 Samsung S6 mini:

OpenCL platform not found.
OpenCL shared libraries we tried to load:
/system/vendor/lib/libOpenCL.so
/system/lib/libOpenCL.so
/system/vendor/lib/libPVROCL.so
/system/vendor/lib/egl/libGLES_mali.so
/system/vendor/lib/libllvm-a3xx.so

Claggy

What I would definitely like to see in BOINC is the feature I already requested: When double-clicking on a WU, the WUs' property window should pop up.

This would make checking or searching for a specific WU much more omfortable.

Moreover, why not implement a up-down-feature (like arrows) within the properties window, so one could cycle through all WUs?

Ah yes, please, could you make the snooze feature adjustable (both GPU and BOINC total), so one could set the amount of time BOINC and/or the GPU would snooze? This is in the "the list of past requests" but has not been implemented yet.


Thanks a lot for considering this.


Many greetings!

I also endorse an adjustable SNOOZE. When I am into gaming, I have to completely shut down boinc, and often I forget to reactivate it, sometimes days at a time. I wish I could just snooze it for a few hours. Thanx

Why not use the exclusive applications setup then run from cc_config.xml?
With that you can tell BOINC to suspend when a defined program is detected in memory, while when you exit that, BOINC will then resume running the science applications. It's possible to run multiple exclusions, and defined specifically for CPU and GPU.

e.g.
&lt;exclusive_app&gt;gtav.exe&lt;/exclusive_app&gt; will suspend all of BOINC (CPU and GPU) when it detects you started GTA V.
&lt;exclusive_gpu_app&gt;gtav.exe&lt;/exclusive_gpu_app&gt; will suspend only GPU tasks, but continue to run CPU tasks.

It's possible to add exclusions for the CPU through BOINC Manager (Tools->Computing preferences->exclusive applications), or when using (testing) BOINC 7.5.0 add the exclusions for GPU through that menu as well (but then in Options->Exclusive Applications).

Hi,

I would like to propose an extended functionality to the exclusive applications feature in BOINC. I often find myself gaming and while gaming I may need to free up 1 or 2 cores of the CPU. It would be very nice if it would be possible when BOINC detects that a defined exclusive application becomes active, it could load a specific processor usage profile for that excl. app.

I suppose for others defining network and disk and memory usage for an excl. app. could be useful as well. Perhaps this is easier to implement with custom computer roles (home idle, home gaming, etc.).

Regards, Robert

CPU scheduler

I would like to request to implement that I could set my boinc client to run at 60% between 07:00 to 17:00 and 100% between 17:00 to 07:00.

That way boinc won't interfere with my dayly work when i need the computer to be very responsive.
At the moment I have to do this manually. Unfortunatley i forget to put it to 100% sometimes and the project miss out on the other 40%.

Is this possible with current builds or maybe in future builds?

It would be nice if there was a possibility to upload partially completed task - I repeatedly lost quite a lot of CPU time being not able to meet task deadline :-( As retired person I can't allow PC run round the clock any more :-( Moreover, I prefer to use tablet prior PC for reading mail, browsing topics of my interest on web etc... thus I power PC only sparsely nowadays ;-) It's critical when I have to leave home for longer time (for holiday, e.g.) and there is no possibility to finish computing in time.

Similarly, I'd like to be able to download tasks of some size only (for the same reason).

jp

I would like to see an option to limit the GPU CU based on percentage, like it is done for CPUs for supporting GPUs, see Asynchronous Shading from http://www.anandtech.com/show/9124/amd-dives-deep-on-asynchronous-shading and my post from 23 Oct 2012 :P

I recently reinstalled Boinc Manager with the VirtualBox when updating to 7.6.9 64bit. I noticed that it installed an old, outdated version of VirtualBox (4.3.12). File hippo update manager tells me that the current version of Virtual Box is 5.0.4. Version 4.3.12 is more than a year old and may be insecure. Is there any compelling reason to stay with old version or should I update to the latest version as my update manager suggests?

---