[HN Gopher] Exploring the native use of 64-bit posit arithmetic ...
___________________________________________________________________
Exploring the native use of 64-bit posit arithmetic in scientific
computing
Author : PaulHoule
Score : 44 points
Date : 2023-05-12 17:57 UTC (5 hours ago)
(HTM) web link (arxiv.org)
(TXT) w3m dump (arxiv.org)
| andrepd wrote:
| Posits are pretty cool. I hacked a relatively high-performance
| implementation in ocaml.
|
| They're really elegant.
| mncharity wrote:
| Posit standard[1]; intro[2]; interactive graphic[4].
|
| [1] https://posithub.org/docs/posit_standard-2.pdf [2]
| https://posithub.org/docs/Posits4.pdf [4]
| https://cse512-19s.github.io/FP-Well-Rounded/
| amelius wrote:
| ACM article:
|
| https://www.sigarch.org/posit-a-potential-replacement-for-ie...
|
| Original paper:
|
| https://dl.acm.org/doi/10.14529/jsfi170206
| mncharity wrote:
| Here's that paper's pdf[1]. ACM... sigh.
|
| [1]
| http://www.johngustafson.net/pdfs/BeatingFloatingPoint.pdf
| PaulHoule wrote:
| That paywall is symptomatic of the ACM being a professional
| society for computer science professors and not software
| practitioners. If you go to the editorial section of CACM
| it is frequently stuck in a loop as every decade or so
| there are violent ups and downs of CS program enrollment
| driven by the realization that "cs students are in demand"
| and then "IT jobs suck", the latter of which could be
| addressed by having a professional society for software
| practitioners...
| g0xA52A2A wrote:
| Slides from this year's CoNGA (Conference on Next Generation
| Arithmetic) are available here [1]. Not sure if there are video
| recordings to come.
|
| [1] https://posithub.org/conga/2023/#schedules
| Zeetah wrote:
| What is quire? I couldn't find anything that made sense in the
| way it is referred to in the paper.
| andrepd wrote:
| Simply put: an accumulator (in fixes-point format) with enough
| bits to calculate dot products with up to 2^31 terms with NO
| loss of precision.
| phkahler wrote:
| I believe it's something like an accumulator that has higher
| precision than the data types being used. It helps reduce
| rounding errors by maintaining extra bits until you store a
| final result.
___________________________________________________________________
(page generated 2023-05-12 23:00 UTC)