nrg_model confusion

Created by: cdleonard

The "integrating a new board in LISA" page claims only a few fields are required but trying to run with such a bare config results in confusing problems (python exceptions).

In particular it seems the TestEnv.platform dict needs an "nrg_model" which seems to partially duplicate EAS energy-costs from DTS? In TestEnv._init_target there is apparently an attempt to read the energy model from target. Unfortunately TestEnv.nrg_model seems to be different from TestEnv.platform['nrg_model'] and the latter can only be loaded from json inside _init_platform.

There are various check for 'nrg_model' in TestEnv.platform in libs/utils/trace.py which makes me think this is optional. But looking at TestEnv._init_platform when init fails it just leaves a None value in the platform dict. I had some luck with the following patch:

diff --git libs/utils/env.py libs/utils/env.py
index 013f44f..5ea783b 100644
--- libs/utils/env.py
+++ libs/utils/env.py
@@ -797,6 +797,9 @@ class TestEnv(ShareState):
         else:
             self.platform['nrg_model'] = self._load_em(self.conf['board'])
 
+        if self.platform['nrg_model'] is None:
+            del self.platform['nrg_model']
+
         # Adding topology information
         self.platform['topology'] = self.topology.get_level("cluster")

This seems like an ugly hack.

Perhaps the EnergyModel object fetched from the target should be converted to the format that is expected inside platform['nrg_model']? But ideally when no nrg_model is available the whole framework should still run, yes?