Knowing how to use hsolve can only be interesting when you also know how to setup experiments and how to save the output. Interfacing hsolve to other elements is not always trivial and deserves a special paragraph of attention.
A distinction can again be made between the lower and the higher chanmodes. When using the lower chanmodes - the 'compatibility' chanmodes - everything behaves almost as without hsolve. In chanmode 0, under no circumstance you will experience problems when interfacing hsolve to other objects. Incoming and outgoing messages that have been created before the SETUP action call are handled automatically. In chanmode 1, hsolve puts severe restrictions on the messages that can be send to the tabulated channels. Since hsolve will not always give an appropriate error message for messages that are handled or that are ignored, it is recommended not to use chanmode 1 unless you are sure that you do not have any tabulated channels with messages that are not handled by hsolve (these messages include DOMAINCONC and ADD_GBAR).
The higher chanmodes are the more interesting modes of operation because of their excellent performance. We will concentrate on them from now on and explain how to use them correctly.
As already said, when using the higher chanmodes, hsolve compiles your model into optimized byte-codes. During the simulation these byte-codes are interpreted at each time step. They instruct hsolve to compute the conductances and evaluate the cable equations. To be able to compile a model into byte-codes, hsolve must know the object type. The object types hsolve is currently supporting are compartment, tabchannel, tab2Dchannel, tabcurrent, Ca_concen, concpool, difshell, taupump, mmpump, hillpump, fixbuffer, difbuffer, dif2buffer, fura2, nernst, ghk, channelC2, channelC3, synchan, synchan2, Mg_block, spikegen, neutral.
When using these objects in a model, be sure to create them beneath the subtree of the compartments they reside in. Then you link them together with messages as usual. Afterwards you create hsolve, set the path field (point it to the compartments only), and perform a SETUP and RESET call. If anything is wrong with the model structure, e.g. if a message incoming to a channel is not handled by hsolve, hsolve will give an appropriate error message. If all is right, hsolve is setup correctly. This means that (1) all computations normally done by the modeling elements, are now done by hsolve in a more efficient way and (2) the message passing between these elements is done by hsolve's internal communication facilities. This is all made transparent for a user except ......that the computed values are not stored anymore in the original elements, but they are stored in hsolve's internal data arrays. How to communicate with these data arrays will be explained shortly.
| NOTE: Hsolve implicitly assumes that you are not doing fancy things like sending an AXIAL message from a channel to a compartment. Genesis's flexibility allows you to construct absurd and unrealistic models. It could be that you do not get any error message if you try to have hsolve compute such models. |