Slides from lectures

Slides are not included in the repository in order to keep the size down. Below are links to pdf files of the slides for each lecture for which they are available.

In each case 3 versions are provided:

• nup1 means one slide per page, as projected in lecture,
• nup3 has 3 slides per page with room for notes, portrait mode
• nup4 has 4 slides per page, landscape mode

Week 1

• Lecture 1: March 28, 2011 ... nup1 ... nup3 ... nup4
  • Overview of class, large-scale computational science, Virtual Machine
• Lecture 2: March 30, 2011 ... nup1 ... nup3 ... nup4
  • Version control
• Lecture 3: April 1, 2011 ... nup1 ... nup3 ... nup4
  • Mercurial examples

Week 2

• Lecture 4: April 4, 2011 ... nup1 ... nup3 ... nup4
  • Fortran
• Lecture 5: April 6, 2011 ... nup1 ... nup3 ... nup4
  • Fortran arrays, dynamic memory
  • Storing numbers on the computer
  • Floating point exceptions
• Lecture 6: April 8, 2011 ... nup1 ... nup3 ... nup4
  • Floating point arithmetic
  • Functions and subroutines in Fortran

Week 3

• Lecture 7: April 11, 2011 ... nup1 ... nup3 ... nup4
  • Fortran array storage
  • Multi-file Fortran codes, modules
• Lecture 8: April 13, 2011 ... nup1 ... nup3 ... nup4
  • Makefiles
• Lecture 9: April 15, 2011 ... nup1 ... nup3 ... nup4
  • Computer Architecture
  • Cache considerations

Week 4

• Lecture 10: April 18, 2011 ... nup1 ... nup3 ... nup4
  • Fortran optimization
  • Fortran debugging
• Lecture 11: April 20, 2011 ... nup1 ... nup3 ... nup4
  • Fortran debugging
  • Software packages
• Lecture 12: April 22, 2011 ... nup1 ... nup3 ... nup4
  • LAPACK, BLAS
  • Intro to parallel computing

Week 5

• Lecture 13: April 25, 2011 ... nup1 ... nup3 ... nup4
  • Amdahl’s law, speed up
  • Intro to OpenMP
• Lecture 14: April 27, 2011 ... nup1 ... nup3 ... nup4
  • OpenMP: parallel do, private variables, nested loops, reductions
• Lecture 15: April 29, 2011 ... nup1 ... nup3 ... nup4
  • OpenMP: loop dependencies, thread-safe functions, parallel sections, barriers,
  • Coarse vs.fine-grain parallelism