Documentation URL (in readme) is offline

[LangdalP]

The readme states

Developer activities and documentation are centralized in http://nanos.ac.upc.edu
This page seems to be offline.

The reason I am looking for more detailed documentation is the following paragraph, also from the readme:

Some benchmarks are coded following a recursive model which makes an intensive use of stack
frames. It is usually needed to increase default stack sizes when working with high recursion
levels. Main stack size can be increased using ulimit -s' linux command. Thread stacks are usually controlled through runtime specific environment variable (e.g. Intel C Compiler uses KMP_STACKSIZE=value' for this purpose).

I am experiencing this problem for the uts kernel. It segfaults unless I increase the main stack size. What I am wondering is whether there are any estimates as to how large the main stack and thread stacks need to be for a given input size.

[alcides]

My repository predated the public availability of the official repository (available at https://github.com/bsc-pm/bots and https://pm.bsc.es/gitlab/benchmarks/bots).

I have started a fork from that repository, and applied all my patches on to it. This repository is now based on the same initial commit as the official repo, and not my "initial import".

The official repository has the right URLs. I would suggest you to start a new fork from either mine (which has your commit) or the official and go on from that one.

[alcides]

Regarding your question, I had the same issue with UTS. I am not sure whether I actually fixed it on one of my machines, but the solution for the others was to set:

ulimit -s unlimited

[LangdalP]

An update on the UTS segfaults - I think they are supposed to be set by the run script, but it doesn't seem to work, at least not by default.
In config/classes/uts, one may see:

inputs=test.input
ompenv=OMP_STACKSIZE="2M"

[small]
inputs=small.input
ompenv=OMP_STACKSIZE="16M"

[medium]
inputs=medium.input
ompenv=OMP_STACKSIZE="55M"

[large]
inputs=large.input
ompenv=OMP_STACKSIZE="150M"

By the way, do you know how to tell the run script of a particular program to run with one of these classes? Let's say I want to run large input. For instance, I have tried:

./run-fft.sh -C large

Because the -h flag says that the -C flag can be used to provide a "list of execution classes to use". I took a quick look at run.common to try to figure out how it works, but it contained too much bash magic for me. I might take another look at it later.

[alcides]

It should be either -C or -i