The meaning of the struct thread

[Kim1230]

Hello,
I recently study the source code of coreneuron. And when I try to understand the meaning of NrnThread. I can't understand some variables. Even I read the book neuron and the nrn_lhs/nrn_rhs functions. Can anybody explain or give me some detailed documents to understand it ?

For example,
I want to know the meaning of ncell, _ndata, _nvdata, _nidata?
Thanks for your help

[pramodk]

Hello @Kim1230,

Here is a brief summary:

• NrnThread struct represents a collection of cells
• ncell is a number of cells in a NrnThread
• _ndata is a number of variables of type double needed to represent ncell cells. This include all channels, synapses, compartments and all of their properties.
• _nidata is like _ndata but these are variables of type int
• _nvdata is like _ndata but these are pointer variables that neuron internally need for book keeping.

I wouldn't say there is a good developer documentation available (except some fragmented internal material). But if you are interested, we will be happy to discuss such details.

As the implementation scope of neuron / coreneuron is quite large, I would also curious to understand which specific aspects you are interested in and what are your goals.

By the way, we have monthly neuron developer meetings where we discuss technical/implementation details. If you are interested, @alexsavulescu can send you the invite link. (this month its happening on Friday 16th Sept)

[Kim1230]

Hello@pramodk,

I have a short question and hope for your response. What is the meaning of the fixed_play_continuous and nt->_vecplay[i]? And the meaning of the related NetCvode。 And please send me the invite link. I will try my best to participate. Thanks a lot!

[alexsavulescu]

Hello @Kim1230, could you please provide us your name and email address? Thanks.

[Kim1230]

Hello@alexsavulescu, my name is Kate Yuan, And the email address will be yuanjingya1021@gmail.com

[alexsavulescu]

Thanks. You should have received an invitation.

[Kim1230]

Thanks. I am very intereted. I want to understand it and try to combine it with the thought of AI for science. So, I need to know the detailed implementation of mathmatical euqations. By the way, The meaning of cell in your answer is the neuron not the compartment right?
When I refer the euqations in the book to the code functions, some of them are not clear. Such as in the /home/yjy/nrn_Debug/nrn/external/coreneuron/coreneuron/sim/treeset_core.cpp: 125. I see the mechanism code such as DetAMPANMDA.cpp. I can't see the function of nrn_jacob. So, I don't understand why it exists? And also it always has condition about nrn_gap. what is that meaning? And also it has the variables cell_permute? What is it ? Hope for your explaining

[Kim1230]

And also there are important functions that needs to be understood. Such as handle_forward_skip() and nrn_thread_table_check(), and the fixed_play_continuous() and the variable Netcvode. What are those functions and variables meaning ?
And what is the difference between Presyn and InputPresyn and PresynHelper(I see there has a variable flag). Is one presyn related to one cell?
And what is the relationship between Presyn and netcon. I see the array of netcon_in_presyn_order. And why variable point_process has nrn_thread id? What is the relationship between point process and nrn_thread?
And I also see the netcon it has send function and deliver function. What is the difference between them?
And what does TQItem mean? why it has the structure (left, right and parent)?
And Why there are different types of events ? self event, netcon , condition event and netparevent? what are those meaning?
And what is the meaning of ThreadDatum and Datum, what is the difference between them?
I really wants to know the detailed implementation and really hope your help.
Hope for your answer

[Kim1230]

Hello again, I really need help to understand one thing for the building the matrix for one cell. In my opinion, in the program ./nrn/external/coreneuron/coreneuron/sim/treeset_core.cpp, the function rhs or lhs is to build the matrix for one cell. And nt->end describes the number of node. And nt->ncell describes the number of cell. The relatoinship between them is one cell have many sections, and each of which has many nodes. But when I read the code, I can't understand the code :

can anybody just give me some hints?
Thanks a lot

[nrnhines]

This fragment sets up the matrix for a second derivative with respect to distance part of the branched cable equation (tree topology).
a[I+1]*v[I+1] - (a[I+1] + b[I])*v[I] + b[I]*v[I-1] = total membrane curent. (forgive me If I have an even number of sign errors and have reversed a and b :) d stands for the diagonal of the matrix. b stands for below the diagonal and a stands for above the diagonal. And I hate it when this editor replaces lower case i with upper case I which stands for the ith node of the tree. The additions into the diagonal have to be atomic to avoid thread race conditions where two numbers are added into the same diagonal at the same time. Also for a branch node, there can be many child nodes adjacent to the parent (I.e. branch) node and in this case the parent node gets many "effect of parent voltage on the current into the parent from the child" terms (one per child). There is a fairly long and elaborate comment in nrn/src/nrnoc/treeset.cpp that discusses this.

[Kim1230]

Hello@nrnhines, Thanks for your response. I studied the cable equation hardly and understand the meaning of it. what I understood from the code is that this function is devoted to set up the matrix for one cell(which consists of a tree topology) with its different branches and nodes. But in the code, the i2 means nrn->ncell which means different cells(a group of cells) but i3 means nrn->end which means the number of nodes(a group of nodes from a cell? ) . I just get confused why the function for one cell involves nrn->ncell? Can you explain it clearly? Thanks again

[nrnhines]

It is not just for one cell, It is all cells. The vector is organized so that the root nodes of the n cells are in the first n slots. The loop starts at ncell and the d[0] through d[ncell-1] are filled in by vec_d[parent_index[i]] -= vec_a[I];