Looks like you’re on some kind of Unix or Linux? I’d check the system logs (on “modern” Linux, with the journalctl command). The message – and your python test – means the darktable process received the SIGKILL signal. I think one way that might happen is when the Out Of Memory Killer springs into action
All those numbers have to be divided by 1024 to give you ratios.
As you can see, mine is set to large, where:
the 700 means darktable can take about 70% (700/1024) of the RAM;
64 as the 2nd number tells us that the minimum tiling buffer size is about 6.5% (64/1024) of RAM. According to the manual:
Note that this setting is largely historic and is no longer of much practical use – you are advised to leave it at its default value.
about 13% (128/1024) can be used for the thumbnail cache
for OpenCL (my card has 6 GB), about 90% (900/1024) of (6 GB - 400 MB) VRAM may be used (according to nvidia-smi, I currently have 1182 MB in use, so that’s quite at the limit, I could lower the 900 if I ran into problems).