Posits as an alternative to floats for weather and climate models
Posit numbers, a recently proposed alternative to floating-point numbers, claim to have smaller arithmetic rounding errors in many applications. By studying weather and climate models of low and medium complexity (the Lorenz system and a shallow water model) we present benefits of posits compared to...
| Egile Nagusiak: | , , |
|---|---|
| Formatua: | Conference item |
| Argitaratua: |
Association for Computing Machinery
2019
|
| _version_ | 1826286549047705600 |
|---|---|
| author | Klöwer, M Düben, P Palmer, T |
| author_facet | Klöwer, M Düben, P Palmer, T |
| author_sort | Klöwer, M |
| collection | OXFORD |
| description | Posit numbers, a recently proposed alternative to floating-point numbers, claim to have smaller arithmetic rounding errors in many applications. By studying weather and climate models of low and medium complexity (the Lorenz system and a shallow water model) we present benefits of posits compared to floats at 16 bit. As a standardised posit processor does not exist yet, we emulate posit arithmetic on a conventional CPU. Using a shallow water model, forecasts based on 16-bit posits with 1 or 2 exponent bits are clearly more accurate than half precision floats. We therefore propose 16 bit with 2 exponent bits as a standard posit format, as its wide dynamic range of 32 orders of magnitude provides a great potential for many weather and climate models. Although the focus is on geophysical fluid simulations, the results are also meaningful and promising for reduced precision posit arithmetic in the wider field of computational fluid dynamics. |
| first_indexed | 2024-03-07T01:45:24Z |
| format | Conference item |
| id | oxford-uuid:984094f0-a1b0-438e-a226-4c5cb252c2f3 |
| institution | University of Oxford |
| last_indexed | 2024-03-07T01:45:24Z |
| publishDate | 2019 |
| publisher | Association for Computing Machinery |
| record_format | dspace |
| spelling | oxford-uuid:984094f0-a1b0-438e-a226-4c5cb252c2f32022-03-27T00:05:41ZPosits as an alternative to floats for weather and climate modelsConference itemhttp://purl.org/coar/resource_type/c_5794uuid:984094f0-a1b0-438e-a226-4c5cb252c2f3Symplectic Elements at OxfordAssociation for Computing Machinery2019Klöwer, MDüben, PPalmer, TPosit numbers, a recently proposed alternative to floating-point numbers, claim to have smaller arithmetic rounding errors in many applications. By studying weather and climate models of low and medium complexity (the Lorenz system and a shallow water model) we present benefits of posits compared to floats at 16 bit. As a standardised posit processor does not exist yet, we emulate posit arithmetic on a conventional CPU. Using a shallow water model, forecasts based on 16-bit posits with 1 or 2 exponent bits are clearly more accurate than half precision floats. We therefore propose 16 bit with 2 exponent bits as a standard posit format, as its wide dynamic range of 32 orders of magnitude provides a great potential for many weather and climate models. Although the focus is on geophysical fluid simulations, the results are also meaningful and promising for reduced precision posit arithmetic in the wider field of computational fluid dynamics. |
| spellingShingle | Klöwer, M Düben, P Palmer, T Posits as an alternative to floats for weather and climate models |
| title | Posits as an alternative to floats for weather and climate models |
| title_full | Posits as an alternative to floats for weather and climate models |
| title_fullStr | Posits as an alternative to floats for weather and climate models |
| title_full_unstemmed | Posits as an alternative to floats for weather and climate models |
| title_short | Posits as an alternative to floats for weather and climate models |
| title_sort | posits as an alternative to floats for weather and climate models |
| work_keys_str_mv | AT klowerm positsasanalternativetofloatsforweatherandclimatemodels AT dubenp positsasanalternativetofloatsforweatherandclimatemodels AT palmert positsasanalternativetofloatsforweatherandclimatemodels |