function WDM_model_10km_queue(options)
%WDM_model_10km_queue
% Job-queue / pipeline version of your script:
% - Main thread runs fiber propagation segment-by-segment
% - At each eval distance: demux once and enqueue per-(l,ri) DSP jobs via parfeval
% - While continuing propagation: periodically harvest completed jobs (non-blocking)
% - End of realization: drain remaining jobs, then save
%
% Save this as a NEW file, e.g. "WDM_model_10km_queue.m"

%%% Run parameters
arguments
    options.num_channels = 16;
    options.channel_spacing = 400e9;
    options.fiber_length_km = 2;
    options.rand_key = 1;
    options.num_realiz = 1;
    options.fwm_mitigation_technique = "co";
end

%% Add the imdd_simulation framework to the path
if ispc
    addpath(genpath('C:\Users\Silas\Documents\MATLAB\imdd_simulation'));
else
    addpath(genpath('/work_beegfs/sutef391/imdd_simulation'));
end

warning('off','MATLAB:datetime:AmbiguousTimeZone');

%% How many workers?
cpus = str2double(getenv('SLURM_CPUS_PER_TASK'));
if ~isfinite(cpus) || cpus < 1, cpus = max(1, feature('numcores')); end

% Use a per-job, node-local JobStorageLocation to avoid stale locks on $HOME
tmpbase = getenv('TMPDIR');
if isempty(tmpbase), tmpbase = tempdir; end
jsl = fullfile(tmpbase, sprintf('matlab_jobstorage_%s_%s', ...
    getenv('USER'), getenv('SLURM_JOB_ID')));
if ~exist(jsl,'dir'); mkdir(jsl); end

% Start pool once
c = parcluster('local');
c.NumWorkers = cpus;
c.JobStorageLocation = jsl;

p = gcp('nocreate');
if isempty(p) || p.NumWorkers ~= cpus
    if ~isempty(p), delete(p); end
    p = parpool(c, cpus);
end
fprintf('parpool up with %d workers; JobStorage=%s\n', p.NumWorkers, c.JobStorageLocation);

%% result filename (timestamp + optional job id)
t      = datetime('now','TimeZone','local','Format','yyyyMMdd_HHmmss');
jobid  = getenv('SLURM_JOB_ID'); if isempty(jobid), jobid = 'nojid'; end
host   = getenv('HOSTNAME'); if isempty(host), host = 'localhost'; end

foldname = sprintf('%dkm_%dch_%dghz_%s', options.fiber_length_km(end), options.num_channels, options.channel_spacing.*1e-9, options.fwm_mitigation_technique);
if ispc
    output_root = fullfile('C:\Users\Silas\Documents\MATLAB\Datensätze\FWM_2025\',foldname,'\');
else
    output_root = fullfile('/work_beegfs/sutef391/results_WDM',foldname,'\');
end
if ~exist(output_root,'dir'), mkdir(output_root); end

fname = sprintf('WDM_%s_%s_%s_%dkm_%dch_%dghz_%s.mat', char(t), host, jobid, options.fiber_length_km(end), options.num_channels, options.channel_spacing.*1e-9, options.fwm_mitigation_technique);

%% Settings
s.num_realiz        = options.num_realiz;
s.wavelengthplan    = calcWavelengthPlan(options.num_channels, options.channel_spacing, 1310);
link_length         = options.fiber_length_km;
s.pmd               = 0.1;
s.gamma             = 0.0023;

s.M                 = 4;
fsym                = 112e9;
fdac                = 2*fsym;
fadc                = 120000000000;
s.random_key        = options.rand_key;

% Laser / s.Modulator
vbias_rel           = 0.5;
u_pi                = 4.6;
vbias               = -vbias_rel*u_pi;
laser_linewidth     = 0e6;

% EQ SETTINGS
dfe_order   = [0 0 0];
len_tr      = 4096*2;
mu_ffe1     = 0.0001;
mu_ffe2     = 0.0008;
mu_ffe3     = 0.001;
mu_dc       = 0.005;
mu_ffe      = [mu_ffe1 mu_ffe3 mu_ffe3];
mu_dfe      = 0.0004;

% DB Stuff
db_precode      = 0;
db_encode       = 0;
duob_mode       = db_mode.no_db;
apply_pulsef    = 0;

rcalpha     = 0.05;
Pform       = Pulseformer("fsym",fsym,"fdac",4*fsym,"pulse","rc","pulselength",16,"alpha",rcalpha);

N = numel(s.wavelengthplan);
f_plan = physconst('lightspeed')./(s.wavelengthplan.*1e-9);
margin = 25e12; % some THz left and right
f_span = (max(f_plan)+margin)-(min(f_plan)-margin);
f_nyq  = f_span/2;

kover = 4;
upsample_required = f_nyq./(fdac*kover/2);
upsample_pow      = 2^nextpow2(upsample_required);

s.f_opt     = fdac*kover*upsample_pow;
s.f_opt_nyq = s.f_opt/2;

s.rop = -12:0.75:-0.75;

%% ---------- Intermediate evaluation points (distance dimension) ----------
segment_length = 1; % km (must match fiber loop below)

eval_dist_km = [2 10];
eval_dist_km = eval_dist_km(eval_dist_km <= link_length);

if link_length > 0
    if isempty(eval_dist_km) || eval_dist_km(end) ~= link_length
        eval_dist_km = unique([eval_dist_km link_length], 'stable');
    end
end

eval_seg = eval_dist_km ./ segment_length;
if any(abs(eval_seg - round(eval_seg)) > 1e-12)
    error('eval_dist_km must be integer multiples of segment_length=%g km.', segment_length);
end
eval_seg = round(eval_seg);
nEval    = numel(eval_seg);
% -------------------------------------------------------------------------

%% Preallocate outputs (eval distance as 4th dimension)
output_ffe  = cell(N, length(s.rop), s.num_realiz, nEval);
output_dfe  = cell(N, length(s.rop), s.num_realiz, nEval);
output_vnle = cell(N, length(s.rop), s.num_realiz, nEval);
output_mlse = cell(N, length(s.rop), s.num_realiz, nEval);
output_dbt  = cell(N, length(s.rop), s.num_realiz, nEval);

s.p_launch = 3;
s.p = options.fwm_mitigation_technique;

switch s.p
    case "co"
        pol_rot = 100.*ones(1,N);
        d_local = 0;
    case "pair"
        pol_rot = repmat([100,100,0,0],1,N/4);
        d_local = 0;
    case "alt"
        pol_rot = repmat([100,0,100,0],1,N/4);
        d_local = 0;
    case "seg"
        pol_rot = 100.*ones(1,N);
        d_local = 3;
    otherwise
        error('Unknown fwm_mitigation_technique: %s', string(s.p));
end

for realiz = 1:s.num_realiz

    % Per-realization TX storage
    signal_cell = cell(1,N);
    Symbols     = cell(1,N);
    Tx_bits     = cell(1,N);

    % -------- Job queue containers --------
    F = parallel.FevalFuture.empty(0,1);
    meta = struct('l',{},'ri',{},'realiz',{},'eval_ptr',{});
    % -------------------------------------

    %% ---------- TX per channel ----------
    for l = 1:N

        [Digi_sig,Symbols{l},Tx_bits{l}] = PAMsource( ...
            "fsym",fsym,"M",s.M,"order",18,"useprbs",0, ...
            "fs_out",fdac, ...
            "applyclipping",0,"clipfactor",1.5, ...
            "applypulseform",apply_pulsef,"pulseformer",Pform, ...
            "randkey",s.random_key+l+realiz, ...
            "db_precode",db_precode,"db_encode",db_encode, ...
            "mrds_code",0,"mrds_blocklength",512,"duobinary_mode",duob_mode ...
        ).process();

        Lp_awg = Filter('filtdegree',3,"f_cutoff",56e9,"fs",fdac*kover, ...
            "filterType",filtertypes.gaussian,"active",true);
        El_sig = AWG("fdac",fdac,"f_cutoff",fsym,"lpf_active",1,"kover",kover, ...
            "bit_resolution",6,"upsampling_method","samplehold","precomp_sinc_rolloff",0, ...
            "H_lpf",Lp_awg,"dac_max",0.6,"dac_min",-0.6).process(Digi_sig);

        El_sig = El_sig.normalize("mode","oneone");
        scaling = 0.6*(u_pi/2-abs(vbias-u_pi/2));
        El_sig = El_sig .* scaling;

        Eml_out = EML("mode",eml_mode.im_cosinus,"power",3,"fsimu",El_sig.fs, ...
            "lambda",s.wavelengthplan(l),"bias",vbias,"u_pi",u_pi, ...
            "linewidth",laser_linewidth,"randomkey",s.random_key+l+realiz).process(El_sig);

        signal_cell{l} = Polarization_Controller("mode","rot_power","desired_power",pol_rot(l)).process(Eml_out);
    end

    %% ---------- WDM mux + launch ----------
    Opt_sig_wdm = Optical_Multiplex("fs_in",fdac*kover,"fs_out",upsample_pow*fdac*kover, ...
        "lambda_center",1310,"random_key",0,"filtype",1,"B",120e9).process(signal_cell);

    Opt_sig_wdm = Amplifier("amp_mode","ideal_no_noise","gain_mode","output_power", ...
        "amplification_db",s.p_launch+10*log10(N)).process(Opt_sig_wdm);

    %% ---------- Fiber propagation ----------
    Opt_sig_wdm_fib = Opt_sig_wdm;

    nSegments = link_length/segment_length;
    if abs(nSegments - round(nSegments)) > 1e-12
        error('fiber_length_km=%g must be an integer multiple of segment_length=%g km.', link_length, segment_length);
    end
    nSegments = round(nSegments);

    zdw = 1310;
    randomize_D = true;
    Dvec = getDispersionVector(nSegments, d_local, zdw, randomize_D, s.random_key+realiz);

    eval_ptr = 1;

    for seg = 1:nSegments

        fprintf('Realiz %d/%d: Segment %d/%d \n', realiz, s.num_realiz, seg, nSegments);

        if 1
        Opt_sig_wdm_fib = DP_Fiber("L",segment_length,"D",Dvec(seg),"Dpmd",s.pmd,"Ds",0.07, ...
            "beat_len",10,"corr_len",100,"dz",1,"manakov",0, ...
            "gamma",s.gamma,"lambda",zdw,"n_waveplates",10,"SS_dphimax",0.01, ...
            "SS_dzmax",50,"SS_dzmin",10,"X_alpha",0.3,"X_beta",0,"rng",1).process(Opt_sig_wdm_fib);
        end

        % -------- Evaluate at intermediate distance (enqueue jobs) --------
        if eval_ptr <= nEval && seg == eval_seg(eval_ptr)

            Opt_sig_wdm_demux = Optical_Demultiplex("attenuation",0,"B",200e9,"filtype",1, ...
                "fs_out",fdac*kover,"fs_in",fdac*kover*upsample_pow,"lambda_center",1310).process(Opt_sig_wdm_fib);

            for ri = 1:length(s.rop)
                for l = 1:N
                    % Enqueue one job per (l,ri)
                    F(end+1,1) = parfeval(p, @rx_job, 5, ...
                        Opt_sig_wdm_demux{l}, ...   % channel signal at this eval point
                        s.rop(ri), ...             % rop [dB]
                        Symbols{l}, Tx_bits{l}, ...% reference + bits
                        s, l, realiz, ...          % settings + indices
                        fdac, kover, fadc, fsym, ...
                        len_tr, mu_dc, mu_ffe, mu_dfe, dfe_order, duob_mode);

                    meta(end+1) = struct('l',l,'ri',ri,'realiz',realiz,'eval_ptr',eval_ptr);
                end
            end

            eval_ptr = eval_ptr + 1;
        end
        % ----------------------------------------------------------------

        % Non-blocking harvest: store finished jobs (if any) while we continue propagation
        [F, meta, output_ffe, output_dfe, output_vnle, output_mlse, output_dbt] = ...
            collect_done(F, meta, output_ffe, output_dfe, output_vnle, output_mlse, output_dbt, false);

    end

    % Drain remaining jobs for this realization (blocking until queue is empty)
    while ~isempty(F)
        [F, meta, output_ffe, output_dfe, output_vnle, output_mlse, output_dbt] = ...
            collect_done(F, meta, output_ffe, output_dfe, output_vnle, output_mlse, output_dbt, true);
    end

    %% Save results (per realization)
    
    res = struct();
    res.settings      = s;
    res.eval_dist_km  = eval_dist_km;
    res.ffe           = output_ffe;
    res.dfe           = output_dfe;
    res.vnle          = output_vnle;
    res.mlse          = output_mlse;
    res.dbt           = output_dbt;

    save(fullfile(output_root, fname), 'res', '-v7.3');
    fprintf('Saved results to: %s\n', fullfile(output_root, fname));
    disp(datetime('now','TimeZone','local','Format','yyyyMs.Mdd_HHmmss'));

end

end % end main


%% ========================= Local functions =========================

function [ffe_results, dfe_results, vnle_results, mlse_results, dbt_results] = rx_job( ...
    Opt_sig_chan, rop_db, Symbols_l, Tx_bits_l, s, l, realiz, ...
    fdac, kover, fadc, fsym, len_tr, mu_dc, mu_ffe, mu_dfe, dfe_order, duob_mode)

% NOTE: keep plotting OFF in workers
debug_plots = 0;

%%%%%% ROP %%%%%%
Opt_sig_wdm_rx = Amplifier("amp_mode","ideal_no_noise","gain_mode","output_power", ...
    "amplification_db", rop_db).process(Opt_sig_chan);

%%%%%% PD Square Law %%%%%%
assert(fdac*kover==Opt_sig_wdm_rx.fs,'Sampling Frequencies do not match! Check previous steps');
PD_sig = Photodiode("fsimu",fdac*kover,"dark_current",2e-08,"responsivity",1,"temperature",20, ...
    "nep",1.8e-11,"randomkey",s.random_key + l + realiz).process(Opt_sig_wdm_rx);

%%%%%% Low-pass RX (PD, El. Connectors and Scope) %%%%%%
rx_bwl = 100e9;
PD_sig = Filter('filtdegree',4,"f_cutoff",rx_bwl,"fs",fdac*kover, ...
    "filterType",filtertypes.butterworth,"active",true).process(PD_sig);

%%%%%% Low-pass Scope %%%%%%
Lp_scpe = Filter('filtdegree',4,"f_cutoff",80e9,"fs",fadc, ...
    "filterType",filtertypes.butterworth,"active",true);

%%%%%% Scope %%%%%%
Scpe_sig = Scope("fsimu",fdac*kover,"fadc",fadc, ...
    "delay",0,"fixed_delay",0,"filtertype",filtertypes.butterworth, ...
    "samplingdelay",0,"rand_samplingdelay",0,"freq_offset",0,"samp_jitter",0, ...
    "adcresolution",8,"quantbuffer",0.1,'block_dc',1,'lpf_active',0,'H_lpf',Lp_scpe).process(PD_sig);

Scpe_sig_2sps = Scpe_sig.resample("fs_out",2*fsym);

[~, Scpe_cell, ~, ~] = Scpe_sig_2sps.tsynch("reference", Symbols_l, "fs_ref", fsym, "debug_plots", debug_plots);
Rx_sig = Scpe_cell{1}.normalize("mode","rms");

% -------------------- FFE --------------------
ffe_order   = [50, 0, 0];
eq_ffe      = EQ("Ne",ffe_order,"Nb",[0,0,0], ...
    "training_length",len_tr,"training_loops",5,"dd_loops",5, ...
    "K",2,"DCmu",mu_dc,"DDmu",[mu_ffe mu_dfe],"DFEmu",0.005, ...
    "FFEmu",0,"plotfinal",0,"ideal_dfe",0);

ffe_results = ffe(eq_ffe,s.M,Rx_sig,Symbols_l,Tx_bits_l, ...
    "precode_mode",duob_mode,'showAnalysis',0,"postFFE",[], ...
    "eth_style_symbol_mapping",0);

% -------------------- DFE --------------------
dfe_order2   = [50, 0, 0];
eq_dfe      = EQ("Ne",dfe_order2,"Nb",[2,0,0], ...
    "training_length",len_tr,"training_loops",5,"dd_loops",5, ...
    "K",2,"DCmu",mu_dc,"DDmu",[mu_ffe mu_dfe],"DFEmu",0.005, ...
    "FFEmu",0,"plotfinal",0,"ideal_dfe",0);

dfe_results = ffe(eq_dfe,s.M,Rx_sig,Symbols_l,Tx_bits_l, ...
    "precode_mode",duob_mode,'showAnalysis',0,"postFFE",[], ...
    "eth_style_symbol_mapping",0);

% -------------------- VNLE + MLSE --------------------
pf_ncoeffs  = 1;
ffe_order3  = [50, 5, 5];
eq_v        = EQ("Ne",ffe_order3,"Nb",dfe_order, ...
    "training_length",len_tr,"training_loops",5,"dd_loops",5, ...
    "K",2,"DCmu",mu_dc,"DDmu",[mu_ffe mu_dfe],"DFEmu",0.005, ...
    "FFEmu",0,"plotfinal",0,"ideal_dfe",1);
pf_         = Postfilter("ncoeff",pf_ncoeffs,"useBurg",1);

mlse_ = MLSE("duobinary_output",0,'M',s.M,'trellis_states',PAMmapper(s.M,0).levels);

[vnle_results, mlse_results] = vnle_postfilter_mlse(eq_v, pf_, mlse_, s.M, Rx_sig, Symbols_l, Tx_bits_l, ...
    "precode_mode", duob_mode, 'showAnalysis', 0, "postFFE", [], "eth_style_symbol_mapping", 0);

% -------------------- DB target --------------------
mlse_db_ = MLSE("DIR",[1,1],"duobinary_output",0,"M",s.M,'trellis_states',PAMmapper(s.M,0).levels);

dbt_results = duobinary_target(eq_v, mlse_db_, s.M, Rx_sig, Symbols_l, Tx_bits_l, ...
    "precode_mode", duob_mode, 'showAnalysis', 0, "postFFE", []);
end


function [F, meta, output_ffe, output_dfe, output_vnle, output_mlse, output_dbt] = ...
    collect_done(F, meta, output_ffe, output_dfe, output_vnle, output_mlse, output_dbt, block)

if nargin < 9, block = false; end
if isempty(F), return; end

% Non-blocking mode: try to fetch as many finished as possible with timeout=0
% Blocking mode: wait for at least one result (timeout=Inf), then keep draining with timeout=0
if block
    timeout_first = Inf;
else
    timeout_first = 0;
end

% Fetch first (maybe blocking), then drain the rest non-blocking
first = true;
while ~isempty(F)
    if first
        timeout = timeout_first;
        first = false;
    else
        timeout = 0;
    end

    try
        [k, ffe_r, dfe_r, vnle_r, mlse_r, dbt_r] = fetchNext(F, timeout);
    catch
        % fetchNext throws if timeout=0 and nothing finished in some releases; treat as "nothing ready"
        break;
    end

    if isempty(k)
        break; % nothing ready
    end

    m = meta(k);

    output_ffe{m.l, m.ri, m.realiz, m.eval_ptr}  = ffe_r;
    output_dfe{m.l, m.ri, m.realiz, m.eval_ptr}  = dfe_r;
    output_vnle{m.l, m.ri, m.realiz, m.eval_ptr} = vnle_r;
    output_mlse{m.l, m.ri, m.realiz, m.eval_ptr} = mlse_r;
    output_dbt{m.l, m.ri, m.realiz, m.eval_ptr}  = dbt_r;

    % Remove completed future & meta entry (important: indices change afterwards)
    F(k) = [];
    meta(k) = [];
end
end