Final covers all lectures during the semester
See midterm review for materials in Lecture 1 to Lecture 11


Lecture 12: Interconnection network topology
   - network topology metrics: diameter, nodal degree, bisection bandwidth
   - the construction of regular topologies: linear array, ring, 
     2D array (mesh), 2D-torus, d-dimensional array/torus, hypercubes, 
     k-ary n-cube, trees, irregular topology, 
   - direct and indirect networks
   - multistage networks, Clos Network, fat-tree
   
Lecture 14: Switching, Flow control and Routing
   - packet switching, circuit switching, store-and-forward, cut-through
     switching
   - packet delay analysis
   - Wormhole routing
   - tree saturation problem in lossless network
   - deadlock 
   - dimension order routing for k-ary n-cube
   - deadlock free routing

Lecture 15: Network interface design 
   - What is the problem?
   - messaging layer
   
Lecture 16: Case study IBM Bluegene and InfiniBand
   - BlueGene nodes
   - BlueGene networks
   - InfiniBand architecture components
   - Infiniband link layer, transport layer, and Verbs

Lecture 17/18: MPI

   - Write basic MPI programs
   - MPI p2p and collective routines

Lecture 19: GPU overview
   - GPGPU concept
   - Difference between CPU and GPU and why GPU has large computing power
   - Fermi architecture
   - Warp and SIMT
   - CPU-GPU system bandwidth limits

Lecture 20/21: CUDA programming
   - CUDA model
   - CUDA program structure
   - CUDA C extensions: specifying kernel, kernel launch, 
   - special variables: threadIdx, blockIdx, blockDim, gridDim
   - Cuda thread organization: 
      * grid/block structure
      * computing global ID
   - CUDA memory hierarchy: where are the registers, shared memory, 
     local memory, constant memory, global memory
   - CUDA synchronization: __syncthreads()

Lecture 22: CUDA programming best practices
   - technique to deal with long host/device memory transfer
   - Memory coallescing
   - shared memory bank conflicts
   - control flow divergence
   - occupancy

Lecture 23: PRAM

   - Describe the model
   - EREW, CREW, CRCW (common), CRCW (random), CRCW (priority)
   - developing PRAM algorithms

Lecture 24: LogP and BSP

   - Describe the models
   - Analyze communication algorithm with LogP
   - Relation between PRAM, LogP, BSP