Week 1-2 Recap
###Up-to-date plan for the remainder of the week
- (Thursday) Main goal: Identify speed slow downs
- Compare current version to version used to make plots for PR
- Continue benchmarking with ProfileView
- (Friday) Main goal: Further Progress in Optimizing
ode113
- (Saturday/Sunday) Main goal: Finish optimization of
0de113
Main Goals for Weeks 10-17
- Finish polishing Adam Bashforth PR (weeks 9-10)
- Finish bench marking for in place or out of place arrays 2.
- Reopen pull request
- Notebook version of IVPTestSuite (Weeks 11-13) 1. 2. 3.
- Scrubbing code for end of GSoC (Week 13)
- Finalize documentation on new Adam Methods
- Finalize documentation on new IVPTestSuite notebook version
- Open Thorough Pull Request for RadauIIA Solver (Weeks 14-17)
- Get Newton Iteration Step working (Week 14/15)
- Step size control (Week 15)
- Error control (Week 16)
- Documentation on RADAU solver and end of the Summer tasks
Journal Entry (Thursday/Friday)
- After profiling and noticing that using the Polynomials
package for computing $\gamma_i$ is a significant part of the runtime of ode113
, I realized that I only needed to compute $\gamma_i$ for i = 1:16! I precalculated these, and have these values hard coded into the algorithm, which lead to an significant decrease in runtime.
- I also used ProfileView to determine which version of array assignments is quicker: in-place or out-of-place. Contrary to what we expected, in-place took longer and more memory.
Journal Entry