On 07/01/2014 04:08 PM, Sébastien Bourdeauducq wrote: > On 07/01/2014 09:42 PM, Robert Jordens wrote: >> It looks like there are the "truly global parameters" > > Are you talking about long lists of constants like those in > IonProperties_inc.dc from your current code?
Many of those, yes. But that file is already a machine generated header that is included. The parameters are managed in a little database with history etc. > We could simply use the regular Python mechanism for that, and import > another Python module defining them. The current inline transform > already supports globals (and treats them as constants), as exemplified > by the "my_range" global in compiler_test.py (it also works with integers). Absolutely. But we should look beyond machine generating it. >> That sounds cumbersome. The top level would have to retreive and pass >> down parameters (and channels) through many layers to the lower kernels. >> Introducing a new paramter in a low level kernel would require making >> changes at all intermediate levels. > > Sets of parameters can be grouped in dictionaries, named tuples, or > SimpleNamespaces that are common to all the experiment hierarchy. Would > that be ok? Yes. Robert. _______________________________________________ ARTIQ mailing list https://ssl.serverraum.org/lists/listinfo/artiq Migen/MiSoC: please use de...@lists.m-labs.hk instead.
