If I have some data, how would I determine the best normal distribution fit?

How do I setup/plot this probability

As a thought example, time waiting in a line at cash register.  Let's say 40% of the time you are waiting for ~1 minute while 35% of the time you are served immediately and the rest of the time trailing off waiting up to 15 minutes.

Is this is like a skewed normal distribution?  and how could I set it up?  And then how can I randomly generate some values from this distribution?

Has anyone made use of the GeM module?  found here http://math.usask.ca/~shevyakov/gem/

I have a fairly old laptop, a Toshiba S10 Tecra, windows 7 64 bit and 4 Gb RAM.  It has an NVidia Quadro NVS 150m video card inside and according to NVidia CUDA capable of a compute level to 1.1 which is ok for float[4] but not float[8].  Inititiating the computecabability in maple does indeed bring a score level of 1.1.

Doing any LinearAlgebra matrix calculations with CUDA enabled is maybe 5x's slower than with it disabled.  I run into BLAS errors and screen blankouts at matrix values of 2000.  The video driver is the latest from Toshiba (2010) but not NVidia (2014) I suppose there would be an increased performace with newer drivers but if I'm going to run into the same issue of slower calculations with CUDA enabled there's no point in even testing the newer drivers on an otherwise fine running machine. 

Since it is a low end video card is CUDA even worthwhile, will I even notice a speed up with updated drivers?

Given the lagrangian from the pendulum physics example on the help page,

or text format.

L := -sin(omega*t-phi(t))*(diff(phi(t), t))*a*l*m*omega+(1/2)*(diff(phi(t), t))^2*l^2*m+cos(phi(t))*g*l*m