520b9852d7
android: bug fix for number cpu core detection. issue #1300 .
...
previous attempt to only read sysconf for number of cores is insufficient, due to a bug in the Android system. Bug got fixed in April 2012 but is still present on many devices.
workaround: read /sys/devices/system/cpu and parse number of cores from there. sysconf uses /proc/stat/ which gets adapted to changes in cpu throttling. Read both sources and compare higher results.
2013-07-17 14:01:39 +02:00
4d0109e3fb
Client: add test harness for file copy (make sure we can handle > 2GB)
2013-07-16 21:40:13 -07:00
9ff8554ee4
client: when launching the GPU detection child process, use boinc_fopen(), more checks for error returns.
2013-07-16 01:29:00 -07:00
8321ef0430
client: Use get_exit_status to wait until the GPU detection process completes and log any error code.
2013-07-15 19:42:21 -04:00
88135a7000
client: wait as long as necessary for the GPU detection child process to exit: eliminate the 3-second limit.
2013-07-15 14:43:29 -07:00
338487fd13
client: don't write name twice in job log
2013-07-14 22:56:37 -07:00
8949c04a77
client: add exit status and result name to job log entries
2013-07-14 13:19:22 -07:00
abfd522840
client: if system clock goes backward, show details in event log
2013-07-14 11:54:20 -07:00
f654cf947e
client: parse <client_new_version_text> in cc_config.xml
2013-07-13 16:45:15 -07:00
ba6fbc8071
client: if suspend for battery reasons, show parameters in event log
...
E.g., if suspend because battery temperature,
show current temperature and temperature limit.
Could do this for other things too, e.g. exclusive apps.
2013-07-13 16:31:25 -07:00
bcaf537228
android: bug fix for changing p_ncpus in host_info. issue #1300 .
...
previous attempt to acquire full wake lock during setup routine has not fixed this bug.
changed p_ncpus detection from sysconf(_SC_NPRROCESSORS_ONLN) to _SC_NPROCESSORS_CONF in client's get_host_info for Android devices.
2013-07-13 13:46:29 +02:00
1d590434ea
client: Fix a race condition on system reboot when using a child process for detecting GPUs, add debugging support to child process.
2013-07-12 18:35:47 -07:00
07017027d5
client: Clearing the devnums_pci_slot_sort vector after the first Nvidia OpenCL device is detected causes a SIGSEGV for any subsequent Nvidia OpenCL device detections.
2013-07-11 00:52:43 -04:00
27d611c505
client: compile fix for non-Windows builds.
2013-07-10 00:05:27 -07:00
4774296d1b
client: API test harness tweak
2013-07-09 17:44:03 -07:00
78c295c5de
client: compile fix
2013-07-09 16:39:04 -07:00
faae3aac3a
client: tweaks for API test harness - parse cc_config.xml
2013-07-09 16:21:33 -07:00
b52d98b640
client: change per-project runnable job limit to a flat 1000
2013-07-09 13:52:50 -07:00
782a11e22f
client: don't fetch work if project has > max(2000, ncpus*100) runnable jobs
2013-07-09 11:17:56 -07:00
a8485f3d3f
Various: Fix some compile warnings; from Gianfranco
2013-07-09 10:34:32 -07:00
0feedfe811
client: Fixes for launching a child process for detecting GPUs.
...
We must call parse_cmdline() to set the working directory before read_config_file().
2013-07-09 07:37:01 -07:00
3c3318a867
client: Fixes for launching a child process for detecting GPUs.
...
Allow relative path to the client executable.
Bug fix: last item in arg list for run_program() should be NULL, not empty string.
2013-07-09 05:55:53 -07:00
4d47e2f170
client: don't request work from a project w/ > 1000 runnable jobs
...
Because of O(N^2) algorithms, the client becomes CPU-intensive
when there are lots of jobs.
This limit could be somewhat lower.
2013-07-07 13:13:57 -07:00
4cb34a123a
client: don't apply CPU throttling to apps that use < .5 CPUs (like GPU, NCI)
2013-07-05 20:48:35 -07:00
93c970f87c
client: compile fix for Win
2013-07-05 14:18:45 -07:00
519a0bcbef
API: add test harness for the runtime system
...
- If you run the client with --run_test_app,
runs "test_app" in the current directory and interacts with it
(and does nothing else).
It can suspend/resume it with arbitrary timing;
this is controlled in run_test_app() (app_start.cpp).
- example app: add --critical_section option.
This lets you test the runtime system for apps that do
most of their work in a critical section (like GPU apps).
- Add some logging messages (conditioned by DEBUG_BOINC_API)
to the runtime system.
- boinc_finish() waits for the timer thread to write final messages;
make sure it doesn't do anything else
(like suspend the worker thread) during this period
2013-07-04 16:00:10 -07:00
46733f2747
Merge branch 'master' of ssh://boinc.berkeley.edu/boinc-v2
2013-07-03 11:13:54 -07:00
4b82bac42b
client: fix compiler warning.
2013-07-03 02:34:43 -07:00
39af029598
client: mostly revert dddf586, which could lead to way overcommitted CPU
2013-07-03 00:56:01 -07:00
9978c40bcb
client: fix bug where <no_project_notices> from account manager didn't work
2013-07-02 20:53:50 -07:00
42720dab7c
client: Add the various OS compatibility shreds to the assembly manifest so we can properly detect Windows 8.1 or better.
2013-07-01 19:01:04 -04:00
e8d09695a5
client: minor housekeeping cleanup when correlating CUDA GPU reporting order with apple's OpenCL.
2013-06-30 03:02:45 -07:00
2d1a3f8589
client: use a child process for detecting GPUs on all platforms, not just on Macs.
2013-06-29 03:30:13 -07:00
a6e276b93e
client: apparently CUDA does not always report GPUs in order of ascending PCI slot number, but Apple's OpenCL does. We must take this into account win correlating NVIDIA OpenCL GPUs with CUDA GPUs.
...
Since NVIDIA provides drivers for both OpenCL and CUDA on Windows and Linux, I am assuming (until shown otherwise) that they will report GPUs in the same order.
2013-06-28 02:37:57 -07:00
631e236b08
client: tweaks to code for detecting GPUs via a child process.
...
Added safety features requested by Rom Walton:
* Change COPROC_ATI::get_available_ram and COPROC_NVIDIA::get_available_ram to static routines to prevent calling them without first loading CAL or CUDA libraries.
* Add tests for NULL library calls in these routines.
* Add comments warning about need to call from a separate child process on dual-GPU laptops, proper library initialization, etc.
2013-06-27 02:36:20 -07:00
3614870952
client: don't request work from NCI project if "no new work" set
2013-06-26 20:36:44 -07:00
f03c4a0393
client: tweaks to code for detecting GPUs via a child process.
2013-06-26 05:42:15 -07:00
4d74c5abbd
client: tweaks to code for detecting GPUs via a child process and change sprintf calls to safer snprintf.
2013-06-26 05:00:25 -07:00
737ab61bce
client: tweaks to code for detecting GPUs via a child process and change sprintf calls to safer snprintf.
2013-06-26 02:24:36 -07:00
5d8ee12063
client: fix bug in commit c8e1b51: matching of ATI/AMD CAL and OpenCL GPU enumeration for (most) cases where CAL supports some older GPUs but OpenCL does not.
2013-06-25 20:49:50 -07:00
e0eb88058a
client: bug fix for detecting GPUs via a child process.
2013-06-25 05:03:26 -07:00
152a1969e5
client: fix compile break on Windows
2013-06-25 04:50:33 -07:00
e2b2370e9d
client: optionally detect GPUs via a child process, for dual_GPU laptops.
...
Some dual-GPU laptops (e.g., Macbook Pro) don't power down the more powerful GPU until all applications which used them exit. To save battery life, the client launches a second instance of the client as a child process to detect and get info about the GPUs.
The child process writes the info to a temp file which our main client then reads.
This option is enabled at compile time by defining USE_CHILD_PROCESS_TO_DETECT_GPUS as non-zero in gpu_detect.cpp
2013-06-25 04:31:34 -07:00
e11319d5b4
client(Android): removed logcat output from client_msgs, not necessary since client log is shown in GUI.
2013-06-24 12:14:41 +02:00
37eb54f12f
client (Android): get "user active" from GUI; don't run if user active
...
On Android, get the "user active" state from the GUI
(as DEVICE_STATUS::user_active).
Never compute if user active (ignore preferences).
2013-06-22 09:58:39 -07:00
727f220766
client (Android): remove apps from mem when suspending and on batteries
...
If we're on batteries and suspending for any reason other than CPU throttling,
remove apps from memory.
This reduces battery drain in some cases.
2013-06-22 09:35:16 -07:00
47562339e3
client: if idle instance, fetch work even if uploading
2013-06-21 19:57:25 -07:00
dbbe474b77
client, Android: always use mmap instead of shmget for client/app comm
2013-06-21 15:41:48 -07:00
57a6d3d17a
client (Android): make max battery temperature a preference
...
Note: internal change only; there's no GUI for this yet
2013-06-20 21:47:34 -07:00
cfd3ba9d13
client: AMD GPU ID tweaks, from [P3D] Crashtest
2013-06-20 12:19:36 -07:00
Joachim Fritzsch