FERN: Fast Explicit Reaction Networks SVNCode

Packed integrateNetwork arguments into structs, added debugParams kernel for debugging NetworkParams on the device

Converted all serial routines to parallel, fixed reduceMax and reduceSum. Parallel and serial versions now give bit-wise similar answers.

Fixed bug in Fplus and Fminus parallel computation

Adding alpha network input files for the previous commit

Tested alpha network T9=7 rho=1e8 for intervals log t = -6 to log t = -5.8 and log t = -5 to log t = -4.9, generated comparisons to Java of serial/parallel CUDA on abd@...:~/Dev/fern/. Comparisons of mass fractions are nearly visually perfect.

Added printf statements to produce Gnuplot-readable data

Moved shared variables in integrateNetwork kernel to global memory

Configured CMake to build both parallel and serial binary, syncronized serial and parallel code to produce similar results, and enabled shared variables for storing timestepping-related data.

Added two logical variables, diagnosticPrint and printSteps, to control the amount of diagnostic printout from the kernel execution. If the first is set to 1 very detailed info is printed for each integration step. If the second is set to 1, at each integration step a summary of network variables is printed (setting diagnosticPrint to 1 generates much more additional output than setting printSteps to 1).

Small changes in explicitgpuDouble.cu