[QUOTE=kriesel;533927]Interesting.
I count "Tesla P100-PCIE-16GB" as 20,
and "GeForce GTX 1080 Ti" as 19.[CODE]CUDALucas v2.06beta 64-bit build, compiled May 5 2017 @ 13:02:54

binary compiled for CUDA 8.0
CUDA runtime version 8.0
CUDA driver version 8.0

------- DEVICE 0 -------
name GeForce GTX 1080 Ti
[/CODE]But, "GeForce GTX 1060 3GB" as 20 also, and no problem.[CODE]CUDALucas v2.06beta 64-bit build, compiled May 5 2017 @ 13:00:15

binary compiled for CUDA 6.50
CUDA runtime version 6.50
CUDA driver version 8.0

------- DEVICE 0 -------
name GeForce GTX 1060 3GB
[/CODE]Those are from flashjh's compiles for Windows.[/QUOTE]

Anyway, array out of bounds are undefined behaviours, it's not sure what they'll cause or whether problem occurs.:smile:

[QUOTE=kriesel;533927]Interesting.
I count "Tesla P100-PCIE-16GB" as 20,
and "GeForce GTX 1080 Ti" as 19.[/QUOTE]

[CODE]
char fftfile[32];
sprintf (fftfile, "%s fft.txt", g_dev.name);

char threadfile[32];
sprintf (threadfile, "%s threads.txt", g_dev.name);
[/CODE]
[c]Tesla P100-PCIE-16GB fft.txt[/c] is 28 characters. [c]Tesla P100-PCIE-16GB threads.txt[/c] is however 32 characters and there is no space to store the 0 byte string terminator. That could be it.

[QUOTE=axn;533978][c]Tesla P100-PCIE-16GB fft.txt[/c] is 28 characters. [c]Tesla P100-PCIE-16GB threads.txt[/c] is however 32 characters and there is no space to store the 0 byte string terminator. That could be it.[/QUOTE]

Yeah that is it for sure, because I could run the -cufftbench with the old version without issues and -threadbench all the way through, but then it crashed at the end when it had to save the file.

Successfully completes a LL double check:
[M]50501939[/M]
About ~13 hours compute on colab P100.

[QUOTE=ATH;533985]Yeah that is it for sure...[/QUOTE]

Umm... Isn't this what strncpy() is for? :smile:

Introducing tf1G.py v0.12:
A few bugs came up in running this on Colab:
* The name Error is not defined, which would occasionally cause the Colab interpreter to stop. To fix this, I replaced the raise Error(...) lines with a print statement followed by a sys.exit() (which in turn kills the interpreter).
* In the functions analyze_for_factors and count_exponents_tested, we previously blindly assumed the file exists (namely, results.txt). If we have submitted the results already and we kill and restart the script before it fetches more stuff, we get a FileNotFoundError. This has been rectified.
In addition, since it's a new year, the copyleft has been updated.

[QUOTE=Dylan14;534019]Introducing tf1G.py v0.12[/QUOTE]Not sure if it was intentional, but previous versions had an exponent range of 1000000000-4294967295, v0.12 has a range of 3320000000-3329999999 (lines 41-42).

[QUOTE=James Heinrich;534042]Not sure if it was intentional, but previous versions had an exponent range of 1000000000-4294967295, v0.12 has a range of 3320000000-3329999999 (lines 41-42).[/QUOTE]


That isn't intentional, that is just the settings I have set for my computer. The range is still 1000000000-4294967295.

Introducing tf1G.py v0.13: While fixing the bugs in the previous version, I created another by getting rid of an indent in the count_exponents_tested function, which broke the function. It has been fixed.
Also I set the min and max exponents back to the whole range. It is user adjustable though.

GPU availability seems to have become much more diffuclt in the last day ...

Yeah, during the holidays I got a P100 right away almost every time I tried, also just after my previous session ended.
Many people did not bother using Colab during the holidays I guess.

But now it is back to "normal" or even harder than normal to get a GPU.