Have tried in
"BOINC/Data"
"BOINC/Data/Projects"
"BOINC/Data/Projects/Einstein"
And exited and restarted BOINC each time.
TIA
dunx
Which file is that? and who are you asking? as your message was posted to the thread, and not any message in particular.
All the files for the app_info.xml and the app_info.xml itself need to go into the project folder of the project that the app is for, ie BOINC/Data/Projects/Einstein
Don't forget to change the 1.000000 tags in Claggy's example to the appropriate fraction (i.e. 0.5 for 2 tasks at once, 0.333 for 3 tasks, etc.) or the app_info.xml file won't be of any benefit.
I have a question concerning the value
220200960.000000
in the file app_info.xml.
What does this entry mean?
Do I have to specify the total memory of the video card or the memory which should be used for each GPU task or the minimum requirements for a GPU task?
i second this question about gpu_ram.
playing with it give me nothing useful
Hi,
let me guess: 220200960.000000 informs BOINC how much GPU RAM a WU needs. Setting it to low and starting several WUs per GPU might crash some WUs due to missing RAM, so do not mess around with this value.
some WUs require 230 mb, some 400 mb, so which value have we to set?
No WU requires 400 mb RAM, this was a misunderstanding. When running several WUs in parallel, the log would indeed sometimes indicate a memory consumption of 400+ MB. This is because the debugging code that prints out this number assumed only one task running at a time. Basically it records the free mem when the task starts and during the task after memory has been allocated, as there is no other way to query the CUDA driver for the memory that THIS process is consuming. If more than one app starts at the same time, the log info will therefore be misleading, and the new versions have a log messages that explicitly states "assuming one task at a time".
RE: Will this work with my
)
Which file is that? and who are you asking? as your message was posted to the thread, and not any message in particular.
All the files for the app_info.xml and the app_info.xml itself need to go into the project folder of the project that the app is for, ie BOINC/Data/Projects/Einstein
Claggy
Thanks Claggy, have tried but
)
Thanks Claggy, have tried but failed as listed...
Will check that ALL the files are present and try again.
Thanks for your patience, and valuable input to the forum.
dunx
P.S. Finally succeeded ! You are a star fella ! Looking forward to sorting PC number two ! THANKS !
Potentially stating the
)
Potentially stating the obvious here, but...
Don't forget to change the 1.000000 tags in Claggy's example to the appropriate fraction (i.e. 0.5 for 2 tasks at once, 0.333 for 3 tasks, etc.) or the app_info.xml file won't be of any benefit.
-- Tony D.
Thanks to Claggy. Working
)
Thanks to Claggy. Working fine now.
Hello, I have a question
)
Hello,
I have a question concerning the value
220200960.000000
in the file app_info.xml.
What does this entry mean?
Do I have to specify the total memory of the video card or the memory which should be used for each GPU task or the minimum requirements for a GPU task?
i second this question about
)
i second this question about gpu_ram.
playing with it give me nothing useful
RE: i second this question
)
Hi,
let me guess: 220200960.000000 informs BOINC how much GPU RAM a WU needs. Setting it to low and starting several WUs per GPU might crash some WUs due to missing RAM, so do not mess around with this value.
Michael
Team Linux Users Everywhere
some WUs require 230 mb, some
)
some WUs require 230 mb, some 400 mb, so which value have we to set?
RE: some WUs require 230
)
Set it high and let the units that don't need as much enjoy the excess.
RE: some WUs require 230
)
No WU requires 400 mb RAM, this was a misunderstanding. When running several WUs in parallel, the log would indeed sometimes indicate a memory consumption of 400+ MB. This is because the debugging code that prints out this number assumed only one task running at a time. Basically it records the free mem when the task starts and during the task after memory has been allocated, as there is no other way to query the CUDA driver for the memory that THIS process is consuming. If more than one app starts at the same time, the log info will therefore be misleading, and the new versions have a log messages that explicitly states "assuming one task at a time".
CU
HB