Basic HTML version of Foils prepared August 7 98

Foil 29 Irregular Mesh:
Partitioning & Communication

From DoD HPF Training -- 5. Parallel Programming in HPF DoD Training and Others -- 1995-98. by Chuck Koelbel -- Rice University
Each iteration of the relaxation uses all the data computed in the previous step, and the edge array
  • No parallelism at this level
  • (Does this sound familiar?)
Instead, use the data-parallel edge and node updates
  • flux(i)=(x(iedge(i,1))-x(iedge(i,2)))/2
    • Independent because edge_val ‚ node_val
  • x(iedge(i,1)) = x(iedge(i,1)) - flux(i);
  • x(iedge(i,2)) = x(iedge(i,2)) + flux(i)
    • Not independent because sometimes iedge(iY,1) = iedge(iX,2)
    • Fortunately, HPF provides the SUM_SCATTER function
Communication needed in both stages
  • Between edges and nodes to compute flux
  • Edge-node and node-node to compute x
  • All communication is static, local with respect to grid, but unstructured with respect to array indices

© Northeast Parallel Architectures Center, Syracuse University, npac@npac.syr.edu

If you have any comments about this server, send e-mail to webmaster@npac.syr.edu.

Page produced by wwwfoil on Sun Aug 16 1998