Type-driven data layouts for improved vectorisation
- Submitting institution
-
Heriot-Watt University
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 15136307
- Type
- D - Journal article
- DOI
-
10.1002/cpe.3501
- Title of journal
- Concurrency and Computation: Practice and Experience
- Article number
- -
- First page
- 2092
- Volume
- 28
- Issue
- -
- ISSN
- 1532-0626
- Open access status
- Out of scope for open access requirements
- Month of publication
- May
- Year of publication
- 2015
- URL
-
-
- Supplementary information
-
-
- Request cross-referral to
- -
- Output has been delayed by COVID-19
- No
- COVID-19 affected output statement
- -
- Forensic science
- No
- Criminology
- No
- Interdisciplinary
- No
- Number of additional authors
-
1
- Research group(s)
-
-
- Citation count
- 1
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- This paper takes a novel approach towards compiler introduced vectorisation. Rather than identifying
vectorisation possibilities in existing code it focuses on refactoring the data layouts of high-level programs to enable
vectorisation. The novelty in this work lies in leveraging a type inference system to steer the data layout
transformations.
The paper reflects the main contribution of the PhD thesis of the first author; it is an invited paper within a special
issue based on the conference HPCS'13. The ideas that led to this paper also led to the EPSRC grant EP/L00058X/1:
“Exploiting Parallelism through Type Transformations for Hybrid Manycore Systems”.
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -