[QUOTE=kriesel;531570]..And I say that as a user of DOS from 1982 & Windows from 1993, and linux rarely. (A cpu is a terrible thing to waste, by leaving it idle.) I routinely use cpu and gpu in the SAME 12-hour Colab session. See the third code block in [URL]https://www.mersenneforum.org/showpost.php?p=528073&postcount=8[/URL][/QUOTE]

Me, a DOS user since 1987. I remember having Windows 3.0, but not sure when that came out. It was not much beyond a fancy menu for other DOS based programs. 32 years I've been at this. I am 64 years old so I guess that qualifies me as a senior-citizen of the computer age. Perhaps we both are.

I will check out your link above. I'll guess and say there is a way to change the work-type? I like to rotate once in a while.

:smile:

[QUOTE=kracker;531575]Is the single,(likely memory shared) core worth the time to setup considering how much more powerful the GPU's are?

Well... it's one core sitting around, doing nothing so I guess... :razz:[/QUOTE]It's pretty set-and-forget since it gets assignments from primenet, reports progress, and reports results.
It's no more additional effort than just running sessions for a gpu, once it's set up.

[LEFT][COLOR=#1C1E21][FONT=Helvetica]I started a PRP in 87M on Oct 13 and it reached 45M iterations yesterday, including delays from when I could not get a backend at all. Another was started Oct 18 and is at iteration 52M, projected to finish in 2 weeks (M86939693).[/FONT][/COLOR]
[COLOR=#1C1E21][FONT=Helvetica]
[/FONT][/COLOR][/LEFT]

[QUOTE=storm5510;531585]Me, a DOS user since 1987. I remember having Windows 3.0, but not sure when that came out. It was not much beyond a fancy menu for other DOS based programs. 32 years I've been at this. I am 64 years old so I guess that qualifies me as a senior-citizen of the computer age. Perhaps we both are.[/QUOTE]yes.
[QUOTE]I will check out your link above. I'll guess and say there is a way to change the work-type?[/QUOTE]Yes.

Correction, Windows since 1990. Got a license with an 80386 system. [URL]https://en.wikipedia.org/wiki/Windows_3.0[/URL]
WFWG after that, NT in Oct 1992 prerelease.
VMS in 1981.
Fortran on punch cards from 1968 as a teen.

[QUOTE=kriesel;531590]...Correction, Windows since 1990. Got a license with an 80386 system. [URL]https://en.wikipedia.org/wiki/Windows_3.0[/URL]
WFWG after that, NT in Oct 1992 prerelease.
VMS in 1981.
Fortran on punch cards from 1968 as a teen.[/QUOTE]

When my older brother completed his enlistment in the USAF in 1966, he moved to Indianapolis to be closer to his then girlfriend. He went to a school there, and learned some language on a single computer that filled a room larger than most peoples' garage. It was punch-cards. When he got a job, he had a big briefcase in which he carried his stacks of punch-cards in. I picked it up once. It must have weighed 30 pounds. My brothers' "then girlfriend," he married her in 1972. They are still together. He was 77 this past summer.

I learned COBOL in trade-school. That was on an IBM System 30, I believe. The size of a refrigerator and super noisy.

[I]I think we will get crowned if we go on with this[/I]. :grin::whistle:

[I][U]Reference Post #625[/U][/I]

I have copied the second script out and made the needed changes. This is the one with the extra echo statements. I added my user name, and so on. I was able to understand enough of the code to know what to look for.

It seems that, somewhere near the top, there should be a reference to my Google Drive user name and password. I do not see this anywhere. Then again, I have to be logged in to create a new notebook, so this would not be needed.

At the very bottom, a reference to [I]mfaktc[/I] is made. I am not sure what this is about?

I don't believe I have missed anything. I will give this a try when I get brave enough. :smile:

GPU off with warning
 

When I went to start Kaggle this evening the GPU was off.
When I restarted it I got a warning:

"Custom packages are not supported for GPU instances"

But it still turned on the GPU and it if running mfaktc again.

For how much longer?????

Strange thing on P100.
My script:
[CODE]import os.path
from google.colab import drive

if not os.path.exists('/content/drive/My Drive'):
drive.mount('/content/drive')

%cd '/content/drive/My Drive/cudalucas/'

!cp 'CUDALucas' /usr/local/bin/
!chmod 755 '/usr/local/bin/CUDALucas'

!cd '.' && LD_LIBRARY_PATH="lib:${LD_LIBRARY_PATH}" /usr/local/bin/CUDALucas[/CODE]
My output:
[CODE]/content/drive/My Drive/cudalucas

CUDALucas v2.06 64-bit build, compiled Oct 2 2019 @ 18:44:42

binary compiled for CUDA 10.0
CUDA runtime version 10.0
CUDA driver version 10.10

---------------- DEVICE 0 ----------------
Device Name Tesla P100-PCIE-16GB
ECC Support? Enabled
Compatibility 6.0
clockRate (MHz) 1328
memClockRate (MHz) 715
totalGlobalMem 17071734784
totalConstMem 65536
l2CacheSize 4194304
sharedMemPerBlock 49152
regsPerBlock 65536
warpSize 32
memPitch 2147483647
maxThreadsPerBlock 1024
maxThreadsPerMP 2048
multiProcessorCount 56
maxThreadsDim[3] 1024,1024,64
maxGridSize[3] 2147483647,65535,65535
textureAlignment 512
deviceOverlap 1
pciDeviceID 4
pciBusID 0

You may experience a small delay on 1st startup to due to Just-in-Time Compilation

*** buffer overflow detected ***: /usr/local/bin/CUDALucas terminated[/CODE]
What can be wrong if it working fine on K80?

[QUOTE=Lorenzo;531759]Strange thing on P100.[/QUOTE]

Yep, I posted about this issue before in this thread. I compiled CUDALucas on the P100 but it still did not work. It can run -cufftbench but not a real exponent.

I added gpuowl to my Google Drive and in my script after CUDALucas, so if I get a P100 then CUDALucas fails and script moves on and run gpuowl PRP test instead on another exponent.

[QUOTE=ATH;531760]Yep, I posted about this issue before in this thread. I compiled CUDALucas on the P100 but it still did not work. It can run -cufftbench but not a real exponent.

I added gpuowl to my Google Drive and in my script after CUDALucas, so if I get a P100 then CUDALucas fails and script moves on and run gpuowl PRP test instead on another exponent.[/QUOTE]

ohhh, I see (( Thank you for the advice. I will try this approach :smile:

Anyone can explain how to compile gpuOwL on Colab?

[QUOTE=ATH;531873]Anyone can explain how to compile gpuOwL on Colab?[/QUOTE]

How to compile, and how to run, in one post.
[URL]https://www.mersenneforum.org/showpost.php?p=527930&postcount=7[/URL]
Since it uses git clone, and google drive, it gets the latest commit, and saves the compile results for efficiency.

There's a whole reference thread on colab in [URL]https://mersenneforum.org/showthread.php?t=24607[/URL]