Multithreading on server

[XristiG]

Currently i am processing one receiver and the processing is sitting on one thread at a time, what is the correct setting to speed up the processing, this is my current setting but it is still sitting on a single thread

chunking:
by_receiver: false # Split large filter and measurement matrices blockwise by receiver ID to improve processing speed
size: 5

Using the Yaml inspector i can't see a setting that chunks the satellite data.

[seballgeyer]

For a single receiver, we would not expect to see significant benefits from chunking or multithreading. The chunking functionality is only available when receivers can be processed independently, i.e. when there are no states shared between receivers. In that case, the filter can be partitioned into independent receiver blocks and processed in parallel.

For a single receiver, the measurement update and filter computations are typically quite small, so most of the processing remains effectively serial and the overhead of parallelisation often outweighs any potential gain. As a result, seeing only one thread actively used is expected behaviour.

There is currently no YAML option to chunk observations by satellite. A satellite-based chunking implementation exists in the codebase but is not functional and is scheduled to be removed in an upcoming release. Therefore, for a single-receiver solution, there is currently no chunking configuration that would provide meaningful additional parallelism.

[XristiG]

Thanks Sebastien,
How long would you expect a single day for a single receiver to take.
I am running a Linux server with 66GB of Memory, running one day gave a real time of 75m and user of 89m - is this within the expected bounds.

[XristiG]

So with extensive troubleshooting of my inputs and yaml file and help from Amir Allahvirdizadeh, i now have this down to
real 4m59.822s
user 23m1.167s
sys 0m6.971s

the server is using multiple cores when running