Fix ADIOS Trait for bool

[ax3l]

For default I/O and checkpointing via ADIOS, simulations using our in situ radiation plugin with selected/subsets of particle species (by stride or gamma) did not compile since ADIOS' C API does not define a bool type for radiationFlag.

This adds a trait assuming it can be casted to one byte. It's nasty but I have no idea what to use instead since ADIOS has nothing on bools in its manual.

To Do

• RT test
• we should add a radiation test where RAD_MARK_PARTICLE is larger 1 or RAD_ACTIVATE_GAMMA_FILTER is non-zero to cover that case in the compile suite

[PrometheusPi]

Looks good to me, just a question on what might be an optimization.

[PrometheusPi]

I assume ADIOS_DATATYPES is a type for data-types - is this correct?
If so could you not write something like

constexpr ADIOS_DATATYPES type = adios_unsigned_byte;

PICToAdios() { }

[ax3l]

Thanks for the question. It's an enum, or in-code an int.
There is no need to modify that, since this is in a non-sensitive part of the code. The main sensitive part on the host side is large scale data movement / data reorganization and processing, besides that we needs such tunings on the device. For all other places, like here it is just a minor meta data preparation, usability and maintainability first since it does not matter for performance (which in the ADIOS case is the actual write that is scheduled after all the preparations).

It would also unnecessarily refactor the already existing interface we have right now. It is sufficient.

[ax3l]

@PrometheusPi in case that is interesting for you, you can check with HDFCompass, pip install adios or bpls/bpdump if the data representation of the flag makes some kind of sense to you?

[PrometheusPi]

@ax3l What is the status of the ToDo list?

[ax3l]

I did hope a little that we talk how to test it xD

[ax3l]

Tested in hypnos:/bigdata/hplsim/development/huebl/bunch-adios-radflag-002/simOutput/bp

At least from

bpls simData_0.bp -d "/data/0/particles/e/radiationFlag" | less
bpls simData_2500.bp -d "/data/2500/particles/e/radiationFlag" | less

there seem to be zeroes (at t=0) and ones (for all later stes) in it for radiationFlag as expected.

Same picture with python

import adios as ad

f = ad.File("simData_2500.bp")
print( f["/data/2500/particles/e/radiationFlag"][42] )
#1

f = ad.File("simData_0.bp")
print( f["/data/0/particles/e/radiationFlag"][42] )
#0

so imo we are good to go.