The “Chimera”: An Off-The-Shelf CPU/GPGPU/FPGA Hybrid Computing Platform

Table 2

The “Thirteen Dwarves” of Berkeley. Each dwarf represents an “algorithmic method encapsulating a pattern of computation and/or communication,” and this intended to be a comprehensive list of the major requirements of parallel computational problems for now into the short term.

	Dwarf	Examples/Applications
1	Dense Matrix	Linear algebra (dense matrices)
2	Sparse Matrix	Linear algebra (sparse matrices)
3	Spectral	FFT-based methods
4	N-Body	Particle-particle interactions
5	Structured Grid	Fluid dynamics, meteorology
6	Unstructured Grid	Adaptive mesh FEM
7	MapReduce	Monte Carlo integration
8	Combinational Logic	Logic gates (e.g., Toffoli gates)
9	Graph traversal	Searching, selection
10	Dynamic Programming	Tower of Hanoi problem
11	Backtrack/	Global optimization
	Branch-and-Bound
12	Graphical Models	Probabilistic networks
13	Finite State Machine	TTL counter