Timeline for Is 2**x faster to compute than exp(x)?
Current License: CC BY-SA 3.0
8 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Feb 6, 2017 at 22:18 | comment | added | Pseudonym♦ | Incidentally, I did oversimplify the "typical method" just a little bit. It's possible to implement exp2() and exp() with ulp accuracy using just one small (and quite easy to understand) extension to the method presented here, but an explanation of the small easy-to-understand extension would probably double the length of the answer! | |
| Jun 30, 2014 at 1:59 | history | edited | Pseudonym♦ | CC BY-SA 3.0 |
Fixing a typo
|
| Jun 19, 2014 at 21:47 | history | edited | Pseudonym♦ | CC BY-SA 3.0 |
added 89 characters in body
|
| Jun 19, 2014 at 13:45 | comment | added | Paul Manta | @isomorphismes Here, 'subnormal' is with respect to how floats are implemented. See denormal numbers on Wikipedia. | |
| Jun 19, 2014 at 13:41 | history | edited | Pseudonym♦ | CC BY-SA 3.0 |
Explanation of the new 2008-era floating point jargon
|
| Jun 19, 2014 at 10:22 | comment | added | isomorphismes | when you say "subnormal" – you clearly don't mean "sub-Gaussian"; do you mean "worse than [some benchmark of typicality]"? | |
| Jun 19, 2014 at 8:05 | history | edited | Pseudonym♦ | CC BY-SA 3.0 |
added 22 characters in body
|
| Jun 19, 2014 at 7:57 | history | answered | Pseudonym♦ | CC BY-SA 3.0 |