WIP on DAG generation

This commit is contained in:
Rubydragon 2024-07-17 15:45:54 +02:00
parent a647801f12
commit 4f0da3dffb
3 changed files with 509 additions and 48 deletions

View File

@ -9,6 +9,8 @@
"name": "stderr",
"output_type": "stream",
"text": [
"WARNING: Method definition (::Type{QEDcore.ParticleStateful{DIR, SPECIES, ELEMENT} where ELEMENT<:QEDbase.AbstractFourMomentum})(QEDbase.AbstractFourMomentum) where {DIR<:QEDbase.ParticleDirection, SPECIES<:QEDbase.AbstractParticleType} in module QEDcore at /home/antonr/.julia/packages/QEDcore/uVldP/src/phase_spaces/create.jl:7 overwritten in module MetagraphOptimization at /home/antonr/.julia/packages/MetagraphOptimization/mvCVq/src/QEDprocesses_patch.jl:15.\n",
"ERROR: Method overwriting is not permitted during Module precompilation. Use `__precompile__(false)` to opt-out of precompilation.\n",
"WARNING: Method definition (::Type{QEDcore.ParticleStateful{DIR, SPECIES, ELEMENT} where ELEMENT<:QEDbase.AbstractFourMomentum})(QEDbase.AbstractFourMomentum) where {DIR<:QEDbase.ParticleDirection, SPECIES<:QEDbase.AbstractParticleType} in module QEDcore at /home/antonr/.julia/packages/QEDcore/uVldP/src/phase_spaces/create.jl:7 overwritten in module MetagraphOptimization at /home/antonr/.julia/packages/MetagraphOptimization/mvCVq/src/QEDprocesses_patch.jl:15.\n",
"ERROR: Method overwriting is not permitted during Module precompilation. Use `__precompile__(false)` to opt-out of precompilation.\n"
]
@ -21,9 +23,40 @@
},
{
"cell_type": "code",
"execution_count": null,
"execution_count": 2,
"metadata": {},
"outputs": [],
"outputs": [
{
"data": {
"text/plain": [
"14-element Vector{VirtualParticle{GenericQEDProcess{Tuple{Photon, Photon, Photon, Electron}, Tuple{Photon, Electron}, Tuple{AllPolarization, AllPolarization, AllPolarization, AllSpin}, Tuple{AllPolarization, AllSpin}}, PT, 4, 2} where PT<:AbstractParticleType}:\n",
" positron: \t0000 | 11\n",
" electron: \t0001 | 10\n",
" positron: \t0010 | 01\n",
" electron: \t0011 | 00\n",
" positron: \t0100 | 01\n",
" electron: \t0101 | 00\n",
" positron: \t1000 | 01\n",
" electron: \t1001 | 00\n",
" positron: \t1000 | 11\n",
" electron: \t1001 | 10\n",
" positron: \t1010 | 01\n",
" electron: \t1011 | 00\n",
" positron: \t1100 | 01\n",
" electron: \t1101 | 00"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"WARNING: both QEDcore and QEDbase export \"mul\"; uses of it in module FeynmanDiagramGenerator must be qualified\n"
]
}
],
"source": [
"proc = GenericQEDProcess(3, 1, 1, 1, 0, 0)\n",
"all_particles = Set()\n",
@ -36,27 +69,103 @@
},
{
"cell_type": "code",
"execution_count": null,
"execution_count": 3,
"metadata": {},
"outputs": [],
"outputs": [
{
"data": {
"text/plain": [
"OrderedCollections.OrderedDict{VirtualParticle, Vector{Tuple{VirtualParticle, VirtualParticle}}} with 14 entries:\n",
" positron: \t0000 | 11 => [(positron: \t0000 | 01, photon: \t0000 | 10)]\n",
" electron: \t0001 | 10 => [(photon: \t0000 | 10, electron: \t0001 | 00)]\n",
" positron: \t0010 | 01 => [(positron: \t0000 | 01, photon: \t0010 | 00)]\n",
" electron: \t0011 | 00 => [(electron: \t0001 | 00, photon: \t0010 | 00)]\n",
" positron: \t0100 | 01 => [(positron: \t0000 | 01, photon: \t0100 | 00)]\n",
" electron: \t0101 | 00 => [(electron: \t0001 | 00, photon: \t0100 | 00)]\n",
" positron: \t1000 | 01 => [(positron: \t0000 | 01, photon: \t1000 | 00)]\n",
" electron: \t1001 | 00 => [(electron: \t0001 | 00, photon: \t1000 | 00)]\n",
" positron: \t1000 | 11 => [(photon: \t0000 | 10, positron: \t1000 | 01), (photon: \t10…\n",
" electron: \t1001 | 10 => [(photon: \t0000 | 10, electron: \t1001 | 00), (photon: \t10…\n",
" positron: \t1010 | 01 => [(photon: \t0010 | 00, positron: \t1000 | 01), (photon: \t10…\n",
" electron: \t1011 | 00 => [(photon: \t0010 | 00, electron: \t1001 | 00), (photon: \t10…\n",
" positron: \t1100 | 01 => [(photon: \t0100 | 00, positron: \t1000 | 01), (photon: \t10…\n",
" electron: \t1101 | 00 => [(photon: \t0100 | 00, electron: \t1001 | 00), (photon: \t10…"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"pairs = FeynmanDiagramGenerator.particle_pairs(all_particles)"
"pairs = sort(FeynmanDiagramGenerator.particle_pairs(all_particles))"
]
},
{
"cell_type": "code",
"execution_count": null,
"execution_count": 4,
"metadata": {},
"outputs": [],
"outputs": [
{
"data": {
"text/plain": [
"12-element Vector{Tuple{VirtualParticle, VirtualParticle, VirtualParticle}}:\n",
" (photon: \t0000 | 10, electron: \t0011 | 00, positron: \t1100 | 01)\n",
" (photon: \t0000 | 10, electron: \t0101 | 00, positron: \t1010 | 01)\n",
" (photon: \t0000 | 10, electron: \t1101 | 00, positron: \t0010 | 01)\n",
" (photon: \t0000 | 10, electron: \t1011 | 00, positron: \t0100 | 01)\n",
" (photon: \t0010 | 00, electron: \t0001 | 10, positron: \t1100 | 01)\n",
" (photon: \t0010 | 00, electron: \t0101 | 00, positron: \t1000 | 11)\n",
" (photon: \t0010 | 00, electron: \t1101 | 00, positron: \t0000 | 11)\n",
" (photon: \t0010 | 00, electron: \t1001 | 10, positron: \t0100 | 01)\n",
" (photon: \t0100 | 00, electron: \t0001 | 10, positron: \t1010 | 01)\n",
" (photon: \t0100 | 00, electron: \t0011 | 00, positron: \t1000 | 11)\n",
" (photon: \t0100 | 00, electron: \t1011 | 00, positron: \t0000 | 11)\n",
" (photon: \t0100 | 00, electron: \t1001 | 10, positron: \t0010 | 01)"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"triples = FeynmanDiagramGenerator.total_particle_triples(all_particles)"
]
},
{
"cell_type": "code",
"execution_count": null,
"execution_count": 5,
"metadata": {},
"outputs": [],
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"s: 24, should be: 24\n",
"number of triples: 12\n"
]
},
{
"data": {
"text/plain": [
"12-element Vector{Tuple{VirtualParticle, VirtualParticle, VirtualParticle}}:\n",
" (photon: \t0000 | 10, electron: \t0011 | 00, positron: \t1100 | 01)\n",
" (photon: \t0000 | 10, electron: \t0101 | 00, positron: \t1010 | 01)\n",
" (photon: \t0000 | 10, electron: \t1011 | 00, positron: \t0100 | 01)\n",
" (photon: \t0000 | 10, electron: \t1101 | 00, positron: \t0010 | 01)\n",
" (photon: \t0010 | 00, electron: \t0001 | 10, positron: \t1100 | 01)\n",
" (photon: \t0010 | 00, electron: \t0101 | 00, positron: \t1000 | 11)\n",
" (photon: \t0010 | 00, electron: \t1001 | 10, positron: \t0100 | 01)\n",
" (photon: \t0010 | 00, electron: \t1101 | 00, positron: \t0000 | 11)\n",
" (photon: \t0100 | 00, electron: \t0001 | 10, positron: \t1010 | 01)\n",
" (photon: \t0100 | 00, electron: \t0011 | 00, positron: \t1000 | 11)\n",
" (photon: \t0100 | 00, electron: \t1001 | 10, positron: \t0010 | 01)\n",
" (photon: \t0100 | 00, electron: \t1011 | 00, positron: \t0000 | 11)"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"function n(vp::VirtualParticle)\n",
" if !haskey(pairs, vp)\n",
@ -80,13 +189,77 @@
"sort(triples)"
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"Graph:\n",
" Nodes: Total: 2320, FeynmanDiagramGenerator.ComputeTask_CollectTriples: 64, MetagraphOptimization.DataTask: 1173, \n",
" FeynmanDiagramGenerator.ComputeTask_CollectPairs: 80, FeynmanDiagramGenerator.ComputeTask_SpinPolCumulation: 1, FeynmanDiagramGenerator.ComputeTask_Propagator: 14, \n",
" FeynmanDiagramGenerator.ComputeTask_Triple: 768, FeynmanDiagramGenerator.ComputeTask_BaseState: 12, FeynmanDiagramGenerator.ComputeTask_PropagatePairs: 80, \n",
" FeynmanDiagramGenerator.ComputeTask_Pair: 128\n",
" Edges: 4853\n",
" Total Compute Effort: 0.0\n",
" Total Data Transfer: 0.0\n",
" Total Compute Intensity: 0.0\n"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"graph = generate_DAG(proc)"
]
},
{
"cell_type": "code",
"execution_count": 16,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"compute__2e0e67fe_4441_11ef_36f2_5fda31178519 (generic function with 1 method)"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"using MetagraphOptimization\n",
"using UUIDs\n",
"\n",
"function mock_machine()\n",
" return Machine(\n",
" [\n",
" MetagraphOptimization.NumaNode(\n",
" 0,\n",
" 1,\n",
" MetagraphOptimization.default_strategy(MetagraphOptimization.NumaNode),\n",
" -1.0,\n",
" UUIDs.uuid1(),\n",
" ),\n",
" ],\n",
" [-1.0;;],\n",
" )\n",
"end\n",
"\n",
"func = get_compute_function(graph, proc, mock_machine())"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"generate_DAG(proc)"
"psp = PhaseSpacePoint(proc, PerturbativeQED(), PhasespaceDefinition(SphericalCoordinateSystem(), ElectronRestFrame()), [rand(SFourMomentum) for _ in 1:number_incoming_particles(proc)], [rand(SFourMomentum) for _ in 1:number_outgoing_particles(proc)])"
]
}
],

View File

@ -1,13 +1,20 @@
struct ComputeTask_BaseState <: AbstractComputeTask end # calculate the base state of an external particle
struct ComputeTask_Propagator <: AbstractComputeTask end # calculate the propagator term of a virtual particle
struct ComputeTask_Pair <: AbstractComputeTask end # from a pair of virtual particle currents, calculate the product
struct ComputeTask_CollectPairs <: AbstractComputeTask end # for a list of virtual particle current pair products, sum
struct ComputeTask_PropagatePairs <: AbstractComputeTask end # for the result of a CollectPairs compute task and a propagator, propagate the sum
struct ComputeTask_Triple <: AbstractComputeTask end # from a triple of virtual particle currents, calculate the diagram result
struct ComputeTask_CollectTriples <: AbstractComputeTask end # sum over triples results and
struct ComputeTask_BaseState <: AbstractComputeTask end # calculate the base state of an external particle
struct ComputeTask_Propagator <: AbstractComputeTask end # calculate the propagator term of a virtual particle
struct ComputeTask_Pair <: AbstractComputeTask end # from a pair of virtual particle currents, calculate the product
struct ComputeTask_CollectPairs <: AbstractComputeTask # for a list of virtual particle current pair products, sum
children::Int
end
struct ComputeTask_PropagatePairs <: AbstractComputeTask end # for the result of a CollectPairs compute task and a propagator, propagate the sum
struct ComputeTask_Triple <: AbstractComputeTask end # from a triple of virtual particle currents, calculate the diagram result
struct ComputeTask_CollectTriples <: AbstractComputeTask # sum over triples results and
children::Int
end
struct ComputeTask_SpinPolCumulation <: AbstractComputeTask # abs2 sum over all spin/pol configs
children::Int
end
# import compute so we don't have to repeat it all the time
import MetagraphOptimization: compute, compute_effort
import MetagraphOptimization: compute, compute_effort, children
compute_effort(::ComputeTask_BaseState) = 0
compute_effort(::ComputeTask_Propagator) = 0
@ -16,6 +23,16 @@ compute_effort(::ComputeTask_CollectPairs) = 0
compute_effort(::ComputeTask_PropagatePairs) = 0
compute_effort(::ComputeTask_Triple) = 0
compute_effort(::ComputeTask_CollectTriples) = 0
compute_effort(::ComputeTask_SpinPolCumulation) = 0
children(::ComputeTask_BaseState) = 1
children(::ComputeTask_Propagator) = 1
children(::ComputeTask_Pair) = 2
children(t::ComputeTask_CollectPairs) = t.children
children(::ComputeTask_PropagatePairs) = 2
children(::ComputeTask_Triple) = 3
children(t::ComputeTask_CollectTriples) = t.children
children(t::ComputeTask_SpinPolCumulation) = t.children
struct BaseStateInput{PS_T<:AbstractParticleStateful,SPIN_POL_T<:AbstractSpinOrPolarization}
particle::PS_T
@ -142,3 +159,10 @@ end
# use a summation algorithm with more accuracy and/or parallelization
@inline compute(::ComputeTask_CollectPairs, args::Vararg{N,T}) where {N,T} = sum(args)
@inline compute(::ComputeTask_CollectTriples, args::Vararg{N,T}) where {N,T} = sum(args)
@inline function compute(::ComputeTask_SpinPolCumulation, args::Vararg{N,T}) where {N,T}
sum = 0.0
for arg in args
sum += abs2(arg)
end
return sum
end

View File

@ -22,7 +22,7 @@ function _parse_particle(name::String)
throw(InvalidInputError("failed to parse particle direction from \"$name\""))
end
name = name[4:end]
name = name[5:end]
local species
if startswith(name, "el")
@ -35,7 +35,7 @@ function _parse_particle(name::String)
throw(InvalidInputError("failed to parse particle species from name \"$name\""))
end
name = name[3:end]
name = name[4:end]
local spin_pol
if startswith(name, "su")
@ -54,13 +54,15 @@ function _parse_particle(name::String)
)
end
name = name[3:end]
name = name[4:end]
index = parse(Int, name)
return (dir, species, spin_pol, index)
end
function input_expr(instance::GenericQEDProcess, name::String, psp_symbol::Symbol)
function MetagraphOptimization.input_expr(
proc::AbstractProcessDefinition, name::String, psp_symbol::Symbol
)
if startswith(name, "bs_")
(dir, species, spin_pol, index) = _parse_particle(name[4:end])
dir_str = _construction_string(dir)
@ -81,7 +83,7 @@ function input_expr(instance::GenericQEDProcess, name::String, psp_symbol::Symbo
return Meta.parse("PropagatorInput(
VirtualParticle(
process($psp_symbol),
$species_str,
$(_species_str(particle_species(vp))),
$(vp.in_particle_contributions),
$(vp.out_particle_contributions)
),
@ -92,6 +94,19 @@ function input_expr(instance::GenericQEDProcess, name::String, psp_symbol::Symbo
end
end
function MetagraphOptimization.input_type(p::AbstractProcessDefinition)
in_t = QEDcore._assemble_tuple_type(incoming_particles(p), Incoming(), SFourMomentum)
out_t = QEDcore._assemble_tuple_type(outgoing_particles(p), Outgoing(), SFourMomentum)
return PhaseSpacePoint{
typeof(p),
PerturbativeQED,
PhasespaceDefinition{SphericalCoordinateSystem,ElectronRestFrame},
Tuple{in_t...},
Tuple{out_t...},
SFourMomentum,
}
end
_species_str(::Photon) = "ph"
_species_str(::Electron) = "el"
_species_str(::Positron) = "po"
@ -101,6 +116,22 @@ _spin_pol_str(::SpinDown) = "sd"
_spin_pol_str(::PolX) = "px"
_spin_pol_str(::PolY) = "py"
function Base.parse(::Type{AbstractSpinOrPolarization}, s::AbstractString)
if s == "su"
return SpinUp()
end
if s == "sd"
return SpinDown()
end
if s == "px"
return PolX()
end
if s == "py"
return PolY()
end
throw(InvalidInputError("invalid string \"$s\" to parse to AbstractSpinOrPolarization"))
end
_dir_str(::Incoming) = "inc"
_dir_str(::Outgoing) = "out"
@ -113,6 +144,49 @@ _spin_pols(::PolY) = (PolY(),)
_is_external(p::VirtualParticle) = number_contributions(p) == 1
function _total_index(
proc::AbstractProcessDefinition,
dir::ParticleDirection,
species::AbstractParticleType,
n::Int,
)
# find particle index of all particles given n-th particle of dir and species (inverse of _species_index)
total_index = 0
species_count = 0
for p in particles(proc, dir)
total_index += 1
if species == p
species_count += 1
end
if species_count == n
return if dir == Incoming()
total_index
else
number_incoming_particles(proc) + total_index
end
end
end
throw("did not find $n-th $dir $species")
end
function _species_index(
proc::AbstractProcessDefinition,
dir::ParticleDirection,
species::AbstractParticleType,
n::Int,
)
# find particle index of n-th particle of *this species and dir*
species_index = 0
for i in 1:n
if particles(proc, dir)[i] == species
species_index += 1
end
end
return species_index
end
function _base_state_name(p::VirtualParticle)
proc = process(p)
@ -130,13 +204,7 @@ function _base_state_name(p::VirtualParticle)
species = particles(proc, dir)[index]
# find particle index of *this species*
species_index = 0
for i in 1:index
if particles(proc, dir)[i] == species
species_index += 1
end
end
species_index = _species_index(proc, dir, species, index)
spin_pol = spin_or_pol(proc, dir, species, species_index)
@ -147,11 +215,70 @@ function _base_state_name(p::VirtualParticle)
)
end
# from two or three node names like "1_su-2-px"... return a single tuple of the indices and spin/pols in sorted
function _parse_node_names(name1::String, name2::String)
split_strings_1 = split.(split(name1, "-"), "_")
split_strings_2 = split.(split(name2, "-"), "_")
return tuple(
# TODO: could use merge sort since the sub lists are sorted already
sort([
tuple.(
parse.(Int, getindex.(split_strings_1, 1)),
parse.(AbstractSpinOrPolarization, getindex.(split_strings_1, 2)),
)...,
tuple.(
parse.(Int, getindex.(split_strings_2, 1)),
parse.(AbstractSpinOrPolarization, getindex.(split_strings_2, 2)),
)...,
])...,
)
end
function _parse_node_names(name1::String, name2::String, name3::String)
split_strings_1 = split.(split(name1, "-"), "_")
split_strings_2 = split.(split(name2, "-"), "_")
split_strings_3 = split.(split(name3, "-"), "_")
return tuple(
# TODO: could use merge sort since the sub lists are sorted already
sort([
tuple.(
parse.(Int, getindex.(split_strings_1, 1)),
parse.(AbstractSpinOrPolarization, getindex.(split_strings_1, 2)),
)...,
tuple.(
parse.(Int, getindex.(split_strings_2, 1)),
parse.(AbstractSpinOrPolarization, getindex.(split_strings_2, 2)),
)...,
tuple.(
parse.(Int, getindex.(split_strings_3, 1)),
parse.(AbstractSpinOrPolarization, getindex.(split_strings_3, 2)),
)...,
])...,
)
end
function _make_node_name(
spin_pols::NTuple{N,Tuple{Int,AbstractSpinOrPolarization}}
) where {N}
node_name = ""
first = true
for spin_pol_tuple in spin_pols
if !first
node_name *= "-"
else
first = false
end
node_name *= "$(spin_pol_tuple[1])_$(_spin_pol_str(spin_pol_tuple[2]))"
end
return node_name
end
function generate_DAG(proc::GenericQEDProcess)
external_particles = _pseudo_virtual_particles(proc) # external particles that will be input to base_state tasks
particles = virtual_particles(proc) # virtual particles that will be input to propagator tasks
pairs = particle_pairs(particles) # pairs to generate the pair tasks
triples = total_particle_triples(particles) # triples to generate the triple tasks
pairs = sort(particle_pairs(particles)) # pairs to generate the pair tasks
triples = sort(total_particle_triples(particles)) # triples to generate the triple tasks
graph = DAG()
@ -181,7 +308,13 @@ function generate_DAG(proc::GenericQEDProcess)
)
data_out = insert_node!(
graph, make_node(DataTask(0)); track=false, invalidate_cache=false
graph,
make_node(
DataTask(0),
"$(_total_index(proc, dir, species, index))_$(_spin_pol_str(spin_pol))",
);
track=false,
invalidate_cache=false,
)
insert_edge!(graph, data_in, compute_base_state)
@ -202,7 +335,10 @@ function generate_DAG(proc::GenericQEDProcess)
data_node_name = "pr_$vp_index"
data_in = insert_node!(
graph, make_node(DataTask(0)); track=false, invalidate_cache=false
graph,
make_node(DataTask(0), data_node_name);
track=false,
invalidate_cache=false,
)
compute_vp_propagator = insert_node!(
graph, make_node(ComputeTask_Propagator()); track=false, invalidate_cache=false
@ -214,15 +350,19 @@ function generate_DAG(proc::GenericQEDProcess)
insert_edge!(graph, data_in, compute_vp_propagator)
insert_edge!(graph, compute_vp_propagator, data_out)
propagator_task_outputs[data_node_name] = data_out
propagator_task_outputs[vp] = data_out
end
# -- Pair Tasks --
pair_task_outputs = Dict()
pair_task_outputs = Dict{VirtualParticle,Vector{Node}}()
for (product_particle, input_particle_vector) in pairs
# for all spins/pols of particles in product_particles do ...
pair_task_outputs[product_particle] = Vector{Node}()
pair_task_outputs[product_particle] = Vector()
# make a dictionary of vectors to collect the outputs depending on spin/pol configs of the input particles
N = number_contributions(product_particle)
pair_output_nodes_by_spin_pol = Dict{
NTuple{N,Tuple{Int,AbstractSpinOrPolarization}},Vector{DataTaskNode}
}()
for input_particles in input_particle_vector
particles_data_out_nodes = (Vector(), Vector())
@ -237,28 +377,152 @@ function generate_DAG(proc::GenericQEDProcess)
)
else
# grab from propagated particles
push!(particles_date_out_nodes[c], pair_task_outputs[p])
append!(particles_data_out_nodes[c], pair_task_outputs[p])
end
end
for in_nodes in Iterators.product(input_particles...)
for in_nodes in Iterators.product(particles_data_out_nodes...)
# make the compute pair nodes for every combination of the found input_particle_nodes to get all spin/pol combinations
compute_pair = insert_node!(
graph,
make_node(ComputeTask_Pair());
track=false,
invalidate_cache=false,
)
pair_data_out = insert_node!(
graph, make_node(DataTask(0)); track=false, invalidate_cache=false
)
#insert_node!(graph, )
insert_edge!(graph, in_nodes[1], compute_pair)
insert_edge!(graph, in_nodes[2], compute_pair)
insert_edge!(graph, compute_pair, pair_data_out)
# get the spin/pol config of the input particles from the data_out names
index = _parse_node_names(in_nodes[1].name, in_nodes[2].name)
if !haskey(pair_output_nodes_by_spin_pol, index)
pair_output_nodes_by_spin_pol[index] = Vector()
end
push!(pair_output_nodes_by_spin_pol[index], pair_data_out)
end
# make the collect pair and propagate nodes
end
data_out_propagated = insert_node!(
graph, make_node(DataTask(0)); track=false, invalidate_caches=false
)
propagator_node = propagator_task_outputs[product_particle]
pair_task_outputs[p] = data_out_propagated
for (index, nodes_to_sum) in pair_output_nodes_by_spin_pol
compute_pairs_sum = insert_node!(
graph,
make_node(ComputeTask_CollectPairs(length(nodes_to_sum)));
track=false,
invalidate_cache=false,
)
data_pairs_sum = insert_node!(
graph, make_node(DataTask(0)); track=false, invalidate_cache=false
)
compute_propagated = insert_node!(
graph,
make_node(ComputeTask_PropagatePairs());
track=false,
invalidate_cache=false,
)
# give this out node the correct name
data_out_propagated = insert_node!(
graph,
make_node(DataTask(0), _make_node_name(index));
track=false,
invalidate_cache=false,
)
# ... end do
for node in nodes_to_sum
insert_edge!(graph, node, compute_pairs_sum)
end
insert_edge!(graph, compute_pairs_sum, data_pairs_sum)
insert_edge!(graph, propagator_node, compute_propagated)
insert_edge!(graph, data_pairs_sum, compute_propagated)
insert_edge!(graph, compute_propagated, data_out_propagated)
push!(pair_task_outputs[product_particle], data_out_propagated)
end
end
# -- Triples --
triples_results = Dict()
for (ph, el, po) in triples # for each triple (each "diagram")
photons = if is_external(ph)
getindex.(Ref(base_state_task_outputs), _base_state_name(ph))
else
pair_task_outputs[ph]
end
electrons = if is_external(el)
getindex.(Ref(base_state_task_outputs), _base_state_name(el))
else
pair_task_outputs[el]
end
positrons = if is_external(po)
getindex.(Ref(base_state_task_outputs), _base_state_name(po))
else
pair_task_outputs[po]
end
for (a, b, c) in Iterators.product(photons, electrons, positrons) # for each spin/pol config of each part
compute_triples = insert_node!(
graph, make_node(ComputeTask_Triple()); track=false, invalidate_cache=false
)
data_triples = insert_node!(
graph, make_node(DataTask(0)); track=false, invalidate_cache=false
)
insert_edge!(graph, a, compute_triples)
insert_edge!(graph, b, compute_triples)
insert_edge!(graph, c, compute_triples)
insert_edge!(graph, compute_triples, data_triples)
index = _parse_node_names(a.name, b.name, c.name)
if !haskey(triples_results, index)
triples_results[index] = Vector{DataTaskNode}()
end
push!(triples_results[index], data_triples)
end
end
# -- Collect Triples --
collected_triples = Vector{DataTaskNode}()
for (index, results) in triples_results
compute_collect_triples = insert_node!(
graph,
make_node(ComputeTask_CollectTriples(length(results)));
track=false,
invalidate_cache=false,
)
data_collect_triples = insert_node!(
graph, make_node(DataTask(0)); track=false, invalidate_cache=false
)
for triple in results
insert_edge!(graph, triple, compute_collect_triples)
end
insert_edge!(graph, compute_collect_triples, data_collect_triples)
push!(collected_triples, data_collect_triples)
end
# Finally, abs2 sum over spin/pol configurations
compute_total_result = insert_node!(
graph,
make_node(ComputeTask_SpinPolCumulation(length(collected_triples)));
track=false,
invalidate_cache=false,
)
for finished_triple in collected_triples
insert_edge!(graph, finished_triple, compute_total_result)
end
final_data_out = insert_node!(
graph, make_node(DataTask(0)); track=false, invalidate_cache=false
)
insert_edge!(graph, compute_total_result, final_data_out)
return graph
end