GitLab

experiments/pyexps + experiments/simbricks/orchestration/nodeconfig : fix experiment environment imports

experiments/simbricks/orchestration/nodeconfig : added ExpEnv variable to AppConfig.config_files method

experiments/simbricks/orchestration/nodeconfig: config_files takes additional env parameter for better path resolution

This is better since now we don't care about the concrete values of the preprocessor variables anymore.

Allow users to specify custom paths to p4.
Co-authored-by: Antoine Kaufmann <antoinek@mpi-sws.org>

When `AXILManager::pending_` only contained a single operation and that one was removed from `AXILManager::pending_` in `step()` on completion, there were still some signals that needed to be reset to their idle values. The temp variables were correctly modified to do so in `step()`, however, `step_apply()` was skipped since `AXILManager::pending_` was now empty.

To-do: Add synchronization support

We didn't clear the package cache again after the additional sudo install, which increased the size of the container.

VS Code started enforcing the locale but they decided to use en_US.UTF-8. It looks like they are going to change this to C.UTF-8, which is available on every POSIX compliant system but this might take some time. So instead, we add the required en_US.UTF-8 locale to the devcontainer for now.

Otherwise, `make documentation` fails in the devcontainer.

For SimBricks messages sent by gem5, we want the simulation to look like it started at tick 0. The reason is that otherwise, after restoring a checkpoint, connected simulators first have to catch up, i.e. simulate until the checkpoint tick before the simulation can continue.

The solution here was quite hacky in that we modified gem5's checkpoint file to have a timestamp of 0. This actually breaks when simulating a Linux host and conducting time measurements. The numbers you get will end up all over the place.

Add width and height attribute in the top-level svg tag. This also fixes the logo not showing up in our documentation.

This prevents gem5 from erroring out after closing and re-opening VS Code.

Our VS Code devcontainer is based on our buildenv container, so we directly modify the latter.

We needed it to ensure that gem5's `PerfKvmCounter::attach()` would work but with simbricks/simbricks#109 this has become unnecessary.

This includes adding a requirements.txt for the Python packages required for linting. In the future, additional packages that our orchestration framework requires also go in here.