| 1 |
>>>>> "FP" == Frank Peters <frank.peters@×××××××.net> writes: |
| 2 |
|
| 3 |
FP> All I need now is for someone to come forward and say that he is using |
| 4 |
FP> gcc-4.5.2 and has no problem. Then I'll really start to bang my head. |
| 5 |
|
| 6 |
Using gcc-4.6.0 (compiled with +graphite), using: |
| 7 |
|
| 8 |
-O3 -ggdb3 -march=amdfam10 -ftree-vectorize |
| 9 |
|
| 10 |
a number of the cabs(3) and cpow(3) tests returned (0,0) (for each of |
| 11 |
float and double, and each of c and fortran) instead of the correct results. |
| 12 |
|
| 13 |
The sin and cos tests had max errors in the +- 5 ulp range for doubles |
| 14 |
and the +- 1.5 range for floats. |
| 15 |
|
| 16 |
Your obsercation that graphite may be to blame is likely correct. |
| 17 |
|
| 18 |
You may be interested in crlibm (correctly rounded lib m): |
| 19 |
|
| 20 |
http://lipforge.ens-lyon.fr/projects/crlibm |
| 21 |
http://lipforge.ens-lyon.fr/www/crlibm/ |
| 22 |
|
| 23 |
-JimC |
| 24 |
-- |
| 25 |
James Cloos <cloos@×××××××.com> OpenPGP: 1024D/ED7DAEA6 |
Archives