On 12/3/06, fsenkel@verizon.net <fsenkel@verizon.net> wrote:
Hello,

I'm taking a CFD class, one of the codes I wrote runs very slow. When I look at hotshot is says the function below is the problem. Since this is an explicit step, the for loops are only traversed once, so I think it's caused by memory usage, but I'm not sure if it's the local variables or the loop? I can vectorize the inner loop,  would declaring the data structures in the calling routine and passing them in be a better idea than using local storage?

I'm new at python and numpy, I need to look at how to get profiling information for the lines within a function.


Thank you,

Frank


PS
I tried to post this via google groups, but it didn't seem to go through, sorry if it ends up as multiple postings


def findw(wnext,wprior,phiprior,uprior,vprior):
   #format here is x[i,j] where i's are rows, j's columns, use flipud() to get the
   #print out consistent with the spacial up-down directions

   #assign local names that are more
   #inline with the class notation
   w = wprior
   phi = phiprior
   u = uprior
   v = vprior


   #three of the BC are known so just set them
   #symetry plane
   wnext[0,0:gcols] = 0.0

   #upper wall
   wnext[gN,0:gcols] = 2.0/gdy**2 * (phi[gN,0:gcols] - phi[gN-1,0:gcols])

   #inlet, off the walls
   wnext[1:grows-1,0] = 0.0


   upos = where(u>0)
   vpos = where(v>0)

   Sx = ones_like(u)
   Sx[upos] = 0.0

   Sy = ones_like(v)
   Sy[vpos] = 0.0

   uw = u*w
   vw = v*w

   #interior nodes
   for j in range(1,gasizej-1):
       for i in range(1,gasizei-1):

           wnext[i,j] =( w[i,j] + gnu*gdt/gdx**2 * (w[i,j-1] - 2.0*w[i,j] + w[i,j+1]) +
                         gnu*gdt/gdy**2 * (w[i-1,j] - 2.0*w[i,j] + w[i+1,j]) -
                         (1.0 - Sx[i,j]) * gdt/gdx * (uw[i,j] - uw[i,j-1]) -
                         Sx[i,j] * gdt/gdx * (uw[i,j+1] - uw[i,j]) -
                         (1.0 - Sy[i,j]) * gdt/gdy * (vw[i,j] - vw[i-1,j]) -
                         Sy[i,j] * gdt/gdy * (vw[i+1,j] - vw[i,j]) )

##        print "***wnext****"
##        print "i: ", i, "j: ", j, "wnext[i,j]: ", wnext[i,j]

   #final BC at outlet, off walls
   wnext[1:grows-1,gM] = wnext[1:grows-1,gM-1]
_

Explicit indexing tends to be very slow. I note what looks to be a lot of  differencing in the code, so I suspect what you have here is a PDE.  Your best bet in the short term is to vectorize as many of these operations as possible, but because the expression is so complicated it is a bit of a chore to see just how.  It your CFD class allows it, there are probably tools in scipy that are adapted to this sort of problem, and in particular to CFD. Sandia also puts out PyTrilinos, http://software.sandia.gov/trilinos/packages/pytrilinos/, which provides interfaces to distributed and parallel PDE solvers. It's big iron software for serious problems, so might be a bit of overkill for your applications.

Chuck