Compilers compile computational graphs to optimized source code for diverse platforms

function compile(graphs::AbstractVector{<:AbstractGraph}; root::AbstractVector{Int}=[id(g) for g in graphs])

Compile a list of graphs into a julia static function. The function takes two arguments: root and leaf. root is a vector of the root node ids of the graphs, and leaf is a vector of the leaf node ids of the graphs. This function calls tojuliastr and generate a defined function using RuntimeGeneratedFunctions. Comparing to eval(Meta.parse(tojuliastr(...))), this function does not leak out the function name into global scope.

Example:

factor = 1.5
V1 = [𝑓⁺(1)𝑓⁻(2), 𝑓⁺(3)𝑓⁻(4)]
subgraphs = [external_vertex(V1[1]), external_vertex(V1[2])]
g = FeynmanGraph(subgraphs; factor=factor)
# println(g)
eval_graph! = Compilers.compile([g,])
root = [0.0,]
leaf = [1.0, 2.0]

@assert eval_graph!(root, leaf) ≈ (leaf[1] + leaf[2]) * factor

function compile_C(graphs::AbstractVector{<:AbstractGraph}, filename::String; 
    datatype::DataType=_dtype.weight, root::AbstractVector{Int}=[id(g) for g in graphs], func_name="eval_graph")

Compiles a set of graphs into C language code and append the generated code to a specified file.

Arguments

• datatype::DataType: This type is used for variables types in the generated C code.
• graphs::AbstractVector{<:AbstractGraph}: An array of graph objects. These graphs are processed to generate Julia code.
• filename::String: The name of the file to which the generated code will be appended. The file is created if it does not exist.
• root::AbstractVector{Int} (keyword): An array of integers representing root nodes for each graph in graphs. By default, it is an array of IDs obtained by calling id(g) for each graph g in graphs.
• func_name::String (keyword): The base name for the function(s) to be generated. Defaults to "eval_graph".

Returns

• A dictionary (leafmap) that maps the index of the leaf weight's table leafVal to the leaf graph.

function compile_Julia(graphs::AbstractVector{<:AbstractGraph}, filename::String; 
    root::AbstractVector{Int}=[id(g) for g in graphs], func_name="eval_graph!")

Compiles a set of graphs into Julia code and append the generated code to a specified file.

Arguments

• graphs::AbstractVector{<:AbstractGraph}: An array of graph objects. These graphs are processed to generate Julia code.
• filename::String: The name of the file to which the generated code will be appended. The file is created if it does not exist.
• root::AbstractVector{Int} (keyword): An array of integers representing root nodes for each graph in graphs. By default, it is an array of IDs obtained by calling id(g) for each graph g in graphs.
• func_name::String (keyword): The base name for the function(s) to be generated. Defaults to "eval_graph!".

Returns

• A dictionary (leafmap) that maps the index of the leaf weight's table leafVal to the leaf graph.

function to_dot_str(graphs::AbstractVector{<:AbstractGraph}, name::String="")

Compile a list of graphs into a string for dot language.

# Arguments:
- `graphs`  vector of computational graphs
- `title`   The name of the compiled function (defaults to nothing)

function to_julia_str(graphs::AbstractVector{<:AbstractGraph}, leafMap::Dict{Int,Int}; root::AbstractVector{Int}=[id(g) for g in graphs],
    name::String="eval_graph!")

Compile a list of Feynman graphs into a string for a julia static function. The complied function takes two arguments: root and leafVal. root is a vector of the root node ids of the graphs, and leafVal is a vector of the leaf nodes' weights of the graphs.

Arguments:

• graphs (AbstractVector{G}): The vector object representing the Feynman graphs,
• root (AbstractVector{Int}, optional): The vector of the root node ids of the graphs (defaults to [id(g) for g in graphs]).
• name (String,optional): The name of the complied function (defaults to "eval_graph!").

Returns:

• A String representing the compiled Julia function.
• leafMap (Dict{Int,G}): A dictionary that maps the index of the leaf weight's table leafVal to the leaf graph.

function to_python_str(graphs::AbstractVector{<:AbstractGraph})

Compile a list of graphs into a string for a python static function and output a python script which support the mindspore and jax framework.

Arguments:

• graphs vector of computational graphs
• framework the type of the python frameworks, including :jax and mindspore.

function to_static(operator::Type, subgraphs::AbstractVector{<:AbstractGraph}, subgraph_factors::AbstractVector)

Returns the static representation of a computational graph node `g` with operator `operator`, subgraphs `subgraphs`, and subgraph factors `subgraph_factors`.