Skip to content

Latest commit

 

History

History
125 lines (100 loc) · 3.78 KB

schedule.md

File metadata and controls

125 lines (100 loc) · 3.78 KB
layout title description
page
CME 213 Course Schedule
Detailed class schedule

Week 1

Monday, April 1

  • Introduction to the class; syllabus.
  • Homework, class material

Wednesday April 3

  • Why we need parallelism; example of parallel program: summing up numbers
  • Shared memory and multicore processors
  • Introduction to Pthreads; creating and joining threads
  • Homework 1 out

Week 2

Monday April 8

  • Example: multiplication of two matrices
  • Mutexes; example: piazza restaurant delivery
  • Condition variables; example with codes

Wednesday April 10

  • OpenMP; introduction; parallel regions
  • Parallel for loops; matrix multiplication; sections; tasks; taskwait; using tasks for tree traversal and list traversal
  • Homework 1 due
  • Homework 2 out

Week 3

Monday April 15

  • OpenMP wrap-up: reduction, atomic, critical, single; master; barrier, etc.

Wednesday April 17

  • Colfax lecture 1

Friday April 19

  • Colfax lecture 2
  • Homework 3 out

Week 4

Monday April 22

  • CUDA introduction; moving data between CPU and GPU memory; execution of a skeleton CUDA code (malloc, memcpy, kernel); threading model, basic commands, simple example programs
  • Homework 2 due

Wednesday April 24

  • Threads, warps, blocks; execution model for a kernel; resident blocks; executing an instruction for a warp; synchronizations between threads; SIMT model
  • Execution model for a kernel; resident blocks; resources required by a grid block; resources available on an SM; memory and caches; occupancy
  • Warps; coalescing and performance impact; memory access pattern; caching

Week 5

Monday April 29

  • Example of matrix transpose
  • Shared memory; bank conflicts

Wednesday May 1

  • Lecture on homework 3; finite-difference stencil; algorithm based on global memory; loop over y to increase data-reuse and arithmetic intensity; shared memory algorithm
  • Homework 3 due
  • Homework 4 out

Week 6

Monday May 6

  • Students were asked to group in team and find an efficient procedure to quickly calculate the cumulative sum of many numbers.
  • Reduction algorithm; warp; shared memory; thread divergence; bank conflict; thread-block; use of atomics.

Wednesday May 8

  • Discussion of Thrust; segmented algorithms; examples of problems that can be broken into Thrust algorithms; lambda functions and placeholders
  • Homework 4 due
  • Homework 5 out

Friday May 10

  • Lecture on the final project
  • Final project out; neural network for digit recognition

Week 7

Monday May 13

  • OpenACC, NVIDIA

Wednesday May 15

  • CUDA optimization, flops and mems, NVIDIA

Week 8

Monday May 20

  • CUDA profiling, nvvp, optimization guidelines, NVIDIA

Wednesday May 22

  • Message passing; introduction; MPI; collective communications; collective communications for final project
  • Homework 5 due

Friday May 24

  • Point-to-point communication; Deadlocks; ring communication; blocking vs. non-blocking; sample sort

Week 9

Monday May 27

  • Memorial Day; no class

Wednesday May 29

  • Matrix-vector product; groups, communicators
  • Cartesian topologies; application to matrix-vector products with 2D partitioning
  • Communications using graph topologies; graph; neighbor_all_to_all comms; FEM applications

Friday May 31

  • One-page interim report due

Week 10

Monday June 3

  • Performance metrics; speed-up, efficiency; Amdahl’s law; example: dot-product; efficiency and isoefficiency

Wednesday June 5

  • Alphago, board games, deep learning, and Monte-Carlo tree searches; algorithm and implementation
  • Matrix-vector product with 1D and 2D partitioning; matrix-matrix products; Cannon and DNS algorithms; LU factorization algorithms
  • Conclusion, and wrap-up

Sunday June 9

  • Projects are due, deadline: 11PM