Automated validation of TPG algorithms

[adam-abed-abud]

In the framework of adding tools to validate the TPG it would be nice to have an automated check. Such tool would be helpful for different tests, such as:

1. validate naive and avx implementation of TPG algorithms
2. provide a way to check that the current AVX implementations of the TPG algorithms provide sensible results against known patterns. Some checks include: time-over-threshold results, identifying channels with values above threshold, etc.
3. Potentially, the validation step should also become an integration test.

Some instructions on how to run such tests would be beneficial

[adam-abed-abud]

As discussed with Ivana, the next steps include making an application that given different known frame patterns it executes the TPG (naive or avx) for a specific amount of time and identifies key parameters such as the total number of hits, channels for the identified hits, etc.
Prepare the frame files to be added in asset list

[hristovaivana]

As discussed in the last TPG technical meeting, I've added in the adc_peak_feature branch the instructions and the relevant app/scripts needed to generate WIBEth test patterns and to validate the TP hits found by the AVX and NAIVE implementation. The basic instructions are provided in the README file of fdreadoutlibs and I'm working on improving the documentation.

[nilic123]

@ShyamB97 @aeoranday @xinyue-uoft what are your thoughts here? Maybe we should discuss in a readout meeting the best way to approach this?

[xinyue-uoft]

I think this is quite similar to what @alessandrothea was proposing for the side by side study of native and avx implementation of processors. With the new plan to refactor the processor API, validation of TPG Processor can be done this way. The main separate work would be in the preparation of the "known frame pattern", then we should be able to automatically validate by comparing native and avx implementation (more automatic), or avx against expected outputs.

Although in relation to the original plans, time-over-threshold results are available at pipeline level, after generation of tp_masks. Therefore, if we would like to stick to that, I would also expect some more development in the pipeline level.

Also, Maybe this issue should be moved to tpglibs now that the libraries are separated?