You need to have gmp installed on the instance first. Either you can compile it from source, or you can run


[CODE]!apt install libgmp-dev[/CODE]
and then you should have gmp.h in the place gpuowl expects.

[QUOTE=kriesel;528911]Has anyone found a factor yet by Colab or Kaggle computing?
I haven't yet, and am no longer trying P-1 since the selftest failed.
Dozens of TF no factor found, but no factors found yet. Need a bigger sample size.[/QUOTE]Finally got one TF factor myself, in the last 20 reported from Colab.

[QUOTE=kriesel;529127]Finally got one TF factor myself, in the last 20 reported from Colab.[/QUOTE]

Good to hear! It can be disappointing initially to not see any "kills".

And just to share... There is a bit of disagreement around here how many factors should be found (in an appropriately sized sample set).

The "theorists" say approximately 1/[TF] level.

The "empiricists" say approximately 1/100.

There will, of course, be a slight variance in the probability if a candidate has already been P-1'ed.

Apparently Kaggle does not always shut down the GPU instance when the 30 hours is up.

I started a new instance when I had used ~ 26-27 hours and it ran for the full 9 hours, so now I have used 35h29m this week.

[QUOTE=Fan Ming;528390]So why my attachment disappeared without notification... Did it violate any rules? Here I reupload it in zip.[/QUOTE]
There's one odd thing in the results record from this gpuowl executable, a null version. [CODE]"program":{"name":"gpuowl", "version":[B]""[/B]}[/CODE]

Things that make you go "Hmmm...", Colab Style...
 

So, new behavior from Colab today. At least as far as I've seen.

My two Barbados based accounts received a GPU instance each no problem at around 1800 UTC. Both running in the same browser context; different tabs (of course). Absolutely no access during the morning nor afternoon.

My two SOCKS tunneled environments (the browsers appear to be running in machine rooms somewhere in the "States") have been getting GPU instances since this morning, but they survive exactly an hour and are then ungracefully killed.

Weirdly, no problem re-attaching and getting another instance. Which lasts exactly another hour...

[QUOTE=chalsall;529146]So, new behavior from Colab today. At least as far as I've seen.

My two Barbados based accounts received a GPU instance each no problem at around 1800 UTC. Both running in the same browser context; different tabs (of course). Absolutely no access during the morning nor afternoon.

My two SOCKS tunneled environments (the browsers appear to be running in machine rooms somewhere in the "States") have been getting GPU instances since this morning, but they survive exactly an hour and are then ungracefully killed.

Weirdly, no problem re-attaching and getting another instance. Which lasts exactly another hour...[/QUOTE]
We might be timesharing instances. Lately I get terminations at ~11 hours in. They mainly are launchable around 0620 UTC (0120 local time). Sometimes instances only last several minutes. (Too short to reach an mprime checkpoint. So iterations get repeated.)

[QUOTE=kriesel;529109]How did you get that to build? When I tried make gpuowl I got
[CODE]g++ -MT GmpUtil.o -MMD -MP -MF .d/GmpUtil.Td -Wall -O2 -std=c++17 -c -o GmpUtil.o GmpUtil.cpp
GmpUtil.cpp:3:10: fatal error: gmp.h: No such file or directory
#include <gmp.h>
^~~~~~~
compilation terminated.
Makefile:30: recipe for target 'GmpUtil.o' failed
make: *** [GmpUtil.o] Error 1[/CODE][/QUOTE]
I just initized git ([B]!git init[/B]) and installed libgmp3-dev([B]!apt-get install libgmp3-dev[/B]), and then I changed all [B]<filesystem>[/B] to [B]<experimental/filesystem>[/B], [B]fs = std::filesystem[/B] to [B]fs = std::experimental::filesystem[/B] in .h and .cpp files which contains them.

[QUOTE=Fan Ming;529161]I just initized git ([B]!git init[/B]) and installed libgmp3-dev([B]!apt-get install libgmp3-dev[/B]), and then I changed all [B]<filesystem>[/B] to [B]<experimental/filesystem>[/B], [B]fs = std::filesystem[/B] to [B]fs = std::experimental::filesystem[/B] in .h and .cpp files which contains them.[/QUOTE]

I say instead of doing that, just install a newer version of GCC.

Factors: in average 'bout 1/(bitlevel-log(p)) if no P-1 done. This never failed me, so I guess I am on the lucky set of 'hunters'. Or was it fishermen? The log part is related somehow to the fact that you won't find 25-bits factors for 25-bits exponents... You have to start "counting" from the bitlevel where the factors are possible. Of course, YMMV.
And to be on topic, yes, we found about 10 factors on both colab and kaggle in total since we started this adventure.

[QUOTE=xx005fs;529166]I say instead of doing that, just install a newer version of GCC.[/QUOTE]
That would be better. I know almost nothing about linux so I thought this would be the fail-safe way at that time.