GSoC 2020 Weekly Reports

June 12, 2020

Work done till Jun 12, 2020

Pull Requests

#27: Modernizes the travis syntax
#28: Adds missing files to the geodesic_library
#31: Adds a job which enables Windows testing
#33: Replaces auto_ptr by unique_ptr as auto_ptr is deprecated
#35: Adds tests for the Cython library
#37: Fix build_ext warnings
#38: Add local_gdist_matrix tests and fix Bug 21
#39: pyproject.toml: Set up minimum build system

Issues Fixed

#19: Numpy requirement in setup is causing installation problems
#21: local distances return faulty output

Weekly Report for Week 3 (Jun 13 - Jun 19)

This week I mostly focused on issue #11 where we planned to use ctypes instead of the current implementation in cython. PR #40 addressed issue #11 and as of now implementation in Linux and macOS is completed.

I also worked on issue #22 and the surface data datasets in the current implementation worked fine for all the datasets except cortex 2x120k. I am currently investigating why the code failed for that dataset.

Pull Requests

#40: Use ctypes instead of cython

Issues Addressed

#11: Commit C source
#22: Crash on small meshes

Weekly Report for Week 4 (Jun 20 - Jun 26)

This week I finished PR #40 (moving from cython to ctypes) which was made last week. Also, I figured out what was causing the failure of cortex 2x120k mesh and fixed that in PR #44.

Pull Requests

#40: Use ctypes instead of cython
#44: Add ability to specify if faces’ vertices are 1-indexed

Issues Fixed

#11: Commit C source
#36: Add Coverage report
#41: Failing on cortex 2x120k

Weekly Report for Week 5 (Jun 27 - Jul 3)

This week I faced an issue on windows where the current implementation of ctypes is throwing segmentation fault occasionally. I am still investigating what’s causing the segmentation fault. I also wrote a test for local_gdist_matrix for inner_skull_642 surface.

Pull Requests

#40: Use ctypes instead of cython
#46: Test equality with stable (local_gdist_matrix test for inner_skull_642 surface)

Issues Addressed

#11: Commit C source

Weekly Report for Week 6 & 7 (Jul 4 - Jul 17)

Last couple of weeks I continued my investigation on the Ctypes segmentation fault issue. At the same time, I have also started working on the Cython implementation.

Pull Requests

#48: Add ability to specify if faces’ vertices are 1-indexed (Cython implementation)
#49: Cython parallel

Issues Addressed

#41: Failing on cortex 2x120k
#10: Use Cython parallel

Weekly Report for Week 8 (Jul 18 - Jul 24)

This week a user reported an issue regarding memory leak. I fixed this issue and also added an API to address a new use case.

Pull Requests

#50: Improve documentation
#53: Fix memory leak issue
#54: gdist.pyx: New API distance_matrix_of_selected_points

Issues Fixed

#43: Add documentation
#51: Possible memory leak
#52: Add API for new use case

Weekly Report for Week 9 (Jul 25 - Jul 31)

This week I worked on adding coverage report functionality and OpenMP for parallelization. I also added a PR removing 2 redundant files.

Pull Requests

#49: Use OpenMP parallel
#55: Add coverage report functionality
#56: Remove 2 files

Issues Fixed

#10: Use Cython parallel
#36: Add coverage report

Weekly Report for Week 10 (Aug 1 - Aug 7)

This week I continued to improve previous week’s OpenMP PR and also made 2 new PRs releated to linting and more rigorous testing.

Pull Requests

#49: Use OpenMP parallel
#58: Add 100% test coverage
#62: .travis.yml: Test on multiple versions of python
#63: lint: Use same convention everywhere

Issues Addressed

#10: Use Cython parallel
#57: Improve code coverage
#59: Run different versions of python in windows
#60: Lint: Use same convention everywhere

Weekly Report for Week 11 & 12 (Aug 8 - Aug 21)

In the last couple of weeks, I wrote the final report and also worked on improving test coverage and the parallelization PR.

Pull Requests

#64: Add googletest set-up
#65: Add parallelism using OpenMP

Issues Addressed

#57: Improve code coverage
#10: Use Cython parallel