Blue Waters is Wide

Blue Waters is wide, I can now compile
On stacks of racks my loops will fly.
Give me a code that will decompose
On many cores with MPI.

There is a node, runs Fortran or C,
She's loaded deep, with memory,
And with each core, some local cache.
I'll address it all universally.

Now TeraGrid has become XSEDE,
From NSF, came an RFP;
And centers all, formed a TEOS new
For petascale, like the Waters Blue.

Blue Waters is wide, I can now compile
On stacks of racks my loops will fly.
Give me a code that will decompose
On many cores with MPI.