Opened 3 years ago

Last modified 3 years ago

#20786 needs_info enhancement

Multithreaded option for the benchmarking tool on linear codes

Reported by: dlucas Owned by:
Priority: major Milestone: sage-7.3
Component: coding theory Keywords:
Cc: jsrn, cpernet Merged in:
Authors: Reviewers:
Report Upstream: N/A Work issues:
Branch: u/dlucas/multithreaded_benchmarks (Commits) Commit: e6acad489bdd878b81178746a06b8fdac0ad7e08
Dependencies: #20684 Stopgaps:

Description

This ticket adds support for parallel computation for benchmarks object on linear codes (introduced in #20526)

Change History (8)

comment:1 Changed 3 years ago by dlucas

  • Branch set to u/dlucas/multithreaded_benchmarks

comment:2 Changed 3 years ago by dlucas

  • Commit set to b419d910825e1904a793206798a2dadca9de9b29

I pushed a first version of my patch.

Known issues & remarks:

  • Verbosity option is completely broken
  • Scheduling in task_master method can be improved
  • Code duplication (_perform_experiments_for_single_id and _perform_parallel_experiments_for_single_id.
  • Some names can probably be improved.

I'm leaving this in new state for now.


Last 10 new commits:

bc29f26Refined experimental_data method, which now allows request for specific data
a82a477Fixed syntax errors in documentation
0ab93ecChanges and fixes to docstrings and doctests
d2d6e26Improved data selection in experimental_data method
8e363ddSwitched definition of decoding error and decoding failure
fe0b956Merge branch 't/20601/20601_experimental' into benchmarking_tool
272d251Update to 7.3b1
20523d0Added a stop and start mechanism for benchmarks
c2a7980Merge branch 'stop_and_restart_feature_for_benchmarking' into multithreaded_benchmarks
b419d91First version: working support for parallel computation

comment:3 Changed 3 years ago by git

  • Commit changed from b419d910825e1904a793206798a2dadca9de9b29 to e8340b6de21c00667ef01a5e65925417fff48bd3

Branch pushed to git repo; I updated commit sha1. New commits:

471f8e6Changed the behaviour of verbosity option if parallel is active
8f71eebchunk_size is now number_of_chunks: this parameter now represents the number of tasks a run for a given benchmark will be splitted into
e8340b6Rewrote task_master method to support dynamic chunk size

comment:4 Changed 3 years ago by git

  • Commit changed from e8340b6de21c00667ef01a5e65925417fff48bd3 to a7bf30d3b23eb4c013957804824216029647a5ae

Branch pushed to git repo; I updated commit sha1. New commits:

0984627Fixed a bug which prevented all experiments to be run when using parallel option
a7bf30dImplemented dynamic setting of chunk_size

comment:5 Changed 3 years ago by git

  • Commit changed from a7bf30d3b23eb4c013957804824216029647a5ae to 9d2b7176d1d7889dd57e889589172ddc1c8f7796

Branch pushed to git repo; I updated commit sha1. New commits:

9d2b717New strategy: each task is now sized such that it takes a fixed, predetermined amount of time

comment:6 Changed 3 years ago by git

  • Commit changed from 9d2b7176d1d7889dd57e889589172ddc1c8f7796 to 69f74dcdf9b96196b4b72a1bdab396a94921a2ba

Branch pushed to git repo; I updated commit sha1. New commits:

1dac338Tweaked the previous strategy so it now considers the (very slow) first run of each task to size the aforementioned tasks
1036ee1Fixed stop and restart behaviour with parallel option
69f74dcAdded documentation and some comments

comment:7 Changed 3 years ago by git

  • Commit changed from 69f74dcdf9b96196b4b72a1bdab396a94921a2ba to e6acad489bdd878b81178746a06b8fdac0ad7e08

Branch pushed to git repo; I updated commit sha1. New commits:

e6acad4Fixed speed issue

comment:8 Changed 3 years ago by jsrn

  • Status changed from new to needs_info

This will probably be superseded by the BleacherMark project.

Note: See TracTickets for help on using tickets.