Open CL, and or increasing solver speed

So as a bit of an explanation as to what I’m doing, I have fluorescence data from an experiment, and to get a better idea of concentrations of a fluorescent marker thats used I’ve created a calibration curve. now the response is not linear (which would be ever so nice and simple), and therefore I’m testing polynomial and double exponential fits to see which is the better for the dataset.

currently I’m using the average of 10 triplicates across the range to determine this, and test the fit using squared residuals, I then use solver to adjust the polynomial or exponential values accordingly to minimise the residuals (and use the best model for further calculations)

the problem I have is that this is incredibly slow as of typing this message solver has been running for nearly 10 minutes and hasnt yet finished.

Now I know that the default solver is of little use here (being that this is a non-linear problem) and therefore am using the DEPs and SCO engines, however these seem to be only single threaded, and dont seem to be very fast.

now I have an i7 4710MQ (so 8 threads) of which solver hits only one thread (also as far s I can see, openCL isn’t available in linux on the HD4600 iGPU), and I also have an Nvidia GTX 860M which is OpenCL 1.1 compliant.

so essentially wondering what I can do to speed up analysis (I notice that under the AMD options of OpenCL it lists SSE2 and AVX,does the different versions of the intel spec (SSE4.1/4.2, AVX 2. mean these aren’t able to be used.

(also on a sort of side note, does anyone know how to copy graphs across a spreadsheet and have their ranges change automatically (ie, its a graph for 6 columns, and if I want to add more columns I just copy them over and the graph auto adjusts to the ranges below it).)