Optimizing Parallel Performance with Work and Span in the OpenCilk Compiler
OpenCilk is the modern iteration of Cilk, a multithreaded programming environment designed for high-performance multicore computing. OpenCilk consists of a LLVM fork called Tapir and a runtime scheduler, which, together, allow for OpenCilk’s high performance in practice. However, there are many oppo...
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis |
| Published: |
Massachusetts Institute of Technology
2023
|
| Online Access: | https://hdl.handle.net/1721.1/147495 |
| _version_ | 1811079630158823424 |
|---|---|
| author | Reddy, Nikhil |
| author2 | Schardl, Tao B. |
| author_facet | Schardl, Tao B. Reddy, Nikhil |
| author_sort | Reddy, Nikhil |
| collection | MIT |
| description | OpenCilk is the modern iteration of Cilk, a multithreaded programming environment designed for high-performance multicore computing. OpenCilk consists of a LLVM fork called Tapir and a runtime scheduler, which, together, allow for OpenCilk’s high performance in practice. However, there are many opportunities to improve on the implementation of OpenCilk. In particular, current grainsize calculations for OpenCilk rely on a notion of work that is limited in a few ways.
In this thesis, I propose an alternate implementation that creates a first-class notion of work and span within the OpenCilk compiler that is then used to inform optimizations within OpenCilk. I then analyze the current formulas for grainsizes. I identify key scenarios where the compiler is unable to determine a suitable grainsize and use this to suggest improvements. Finally, I construct a benchmark of one of these scenarios using a Twitter follower dataset and empirically analyze optimal grainsizes for it. |
| first_indexed | 2024-09-23T11:18:06Z |
| format | Thesis |
| id | mit-1721.1/147495 |
| institution | Massachusetts Institute of Technology |
| last_indexed | 2024-09-23T11:18:06Z |
| publishDate | 2023 |
| publisher | Massachusetts Institute of Technology |
| record_format | dspace |
| spelling | mit-1721.1/1474952023-01-20T03:40:14Z Optimizing Parallel Performance with Work and Span in the OpenCilk Compiler Reddy, Nikhil Schardl, Tao B. Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science OpenCilk is the modern iteration of Cilk, a multithreaded programming environment designed for high-performance multicore computing. OpenCilk consists of a LLVM fork called Tapir and a runtime scheduler, which, together, allow for OpenCilk’s high performance in practice. However, there are many opportunities to improve on the implementation of OpenCilk. In particular, current grainsize calculations for OpenCilk rely on a notion of work that is limited in a few ways. In this thesis, I propose an alternate implementation that creates a first-class notion of work and span within the OpenCilk compiler that is then used to inform optimizations within OpenCilk. I then analyze the current formulas for grainsizes. I identify key scenarios where the compiler is unable to determine a suitable grainsize and use this to suggest improvements. Finally, I construct a benchmark of one of these scenarios using a Twitter follower dataset and empirically analyze optimal grainsizes for it. M.Eng. 2023-01-19T19:54:12Z 2023-01-19T19:54:12Z 2022-09 2022-09-16T20:24:13.419Z Thesis https://hdl.handle.net/1721.1/147495 In Copyright - Educational Use Permitted Copyright MIT http://rightsstatements.org/page/InC-EDU/1.0/ application/pdf Massachusetts Institute of Technology |
| spellingShingle | Reddy, Nikhil Optimizing Parallel Performance with Work and Span in the OpenCilk Compiler |
| title | Optimizing Parallel Performance with Work and Span in the OpenCilk Compiler |
| title_full | Optimizing Parallel Performance with Work and Span in the OpenCilk Compiler |
| title_fullStr | Optimizing Parallel Performance with Work and Span in the OpenCilk Compiler |
| title_full_unstemmed | Optimizing Parallel Performance with Work and Span in the OpenCilk Compiler |
| title_short | Optimizing Parallel Performance with Work and Span in the OpenCilk Compiler |
| title_sort | optimizing parallel performance with work and span in the opencilk compiler |
| url | https://hdl.handle.net/1721.1/147495 |
| work_keys_str_mv | AT reddynikhil optimizingparallelperformancewithworkandspanintheopencilkcompiler |