It's important to recognize some of these components exist outside CESM. But it really depends how the init routine is implemented and handles thru the cap, how many coupling modes have to be supported, what the requirements are in different coupled systems, whether there is a stand-alone mode, and many other things. Yes, some components are implemented as you suggest. This works because the default iulog (the clm log unit) is initialized when declared and because there is no stand-alone mode anymore, at least as far as I could see.Ĭam: same as clm, but it has a single file/module just to carry around the logunit.Īnyway, I think the point is that it's not as clean across components as you think. In mct, calls the log initialization near the start of the mct init routine. mosart doesn't have a standalone capability and this implementation has been driven by cesm.Ĭlm: initializes the logging in Advertise and calls the clm intialization in Realize. Mosart: updates the logging file unit in the cap in both mct and nuopc. Figure: Future Architecture of Functions. Note that the inventory control and scheduling system are not covered. This means each cap must reproduce the initialization sequence independently and when it changes, all caps have to be modified. The function view is depicted in Future Architecture of Functions. ![]() Mom: achieves the log unit management by calling into multiple initialization routines FROM the cap and does the log unit setup in the middle of those. Pop: source/io_types.F90 where a flag exists to avoid the call to shr_file_setIO and has near identical implementation as cice. And it's not that different from other models, the implementation is actually clean, it's the fact that nuopc wants to define some model internals via nuopc interfaces. So I don't understand why the old implementation was broken while retaining the _modelio.nml files. The _modelio nml file still exists for the pio stuff. The EIB uses CBA as the default methodology to estimate a projects economic rate of return (ERR) that accounts for broader project benefits and costs to. And I think it also makes sense to move the _modelio stuff back into the. guarantee media viability and political independence in a longer perspective. That is useful in this case and possibly for other future needs. scores of 0 were rated 3 and scores of 100 were rated 97 by default. Provide a way to call shr_file_setIO with an actual return code (maybe as an optional argument) instead of an internal abort. ![]() It seems some components are having to awkwardly code around this issue. Either the nconfig _modelio values should be migrated back to the *_modelio.nml file OR the shr_file_setIO should provide a way to be called and return gracefully if the modelio namelist does not exist.Īs things are currently implemented, the component has to know whether the modelio stuff is set by nuopc in nconfig in the cap or by _modelio.nml approach which is often called in the component itself. With nuopc coming in, shr_file_setIO logic is currently problematic.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |