Why is my LSTM network taking 3 hours to train on a single CPU?

I have been training a LSTM network that contains 5000 samples of numerical data (each containing 100 time-steps). I am perplexed as to why this network is training so much slower than other networks I have experimented with in patternnet. Is this just the nature of the LSTM structure? Could there be something in my code that is causing the network to run slower than normal? Thanks. My slow code is below:

RTrain = Sample5000; %100*5000 double where each of 5000 samples
%has 100 time-steps. There are 4 classes represented in this sample (1250
%of each).

% Assign Categories to the 4 cases

m = [1:5000];
% placing a 1 in row 1 for 0-1250, 
m(1, 1:1250) = 1;
% a 1 in row 2 for 1251-2500,
m(1, 1251:2500) = 2;
% and so on
m(1, 2501:3750) = 3;
m(1, 3751:5000) = 4;
m = transpose(m);

Categories = categorical(m);

%%Define LSTM Network Architecture

inputSize = 1;
outputSize = 100; %Number of time steps 
outputMode = 'last';
numClasses = 4; 

layers = [ ...
    sequenceInputLayer(inputSize)
    lstmLayer(outputSize,'OutputMode',outputMode)
    fullyConnectedLayer(numClasses)
    softmaxLayer
    classificationLayer]

% Adjust training options

maxEpochs = 150;

options = trainingOptions('sgdm', ...
    'MaxEpochs',maxEpochs);

%%Train LSTM Network 

net = trainNetwork(RTrain,Categories,layers,options);