1 |
On Sun, 26 Jun 2011 01:54:53 +0200 |
2 |
Marc Joliet <marcec@×××.de> wrote: |
3 |
|
4 |
> |
5 |
> Then according to the results above my machine is apparently OK. (I guess I |
6 |
> really should finally read about IEEE floating point more thoroughly, I've been |
7 |
> planning to do that for a while now.) |
8 |
> |
9 |
|
10 |
These tests are extreme. The accuracy is supposedly less than 1 ulp |
11 |
(unit in the last place). |
12 |
|
13 |
You can check the *output files in the results directory to see |
14 |
more information on the error. For example, the COS(X) error |
15 |
file is ccos_DP.output and contains the maximum positive and |
16 |
maximum negative errors in ulps. On my machine these are 1.567 |
17 |
and -3.74 ulps, so the error is small. |
18 |
|
19 |
In contrast, the LOG(X) function passes the test but it still shows |
20 |
errors of 0.532 and -0.531 ulps. |
21 |
|
22 |
Frank Peters |