| Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | Release notes for Linux Kernel VFP support code |
| 2 | ----------------------------------------------- |
| 3 | |
| 4 | Date: 20 May 2004 |
| 5 | Author: Russell King |
| 6 | |
| 7 | This is the first release of the Linux Kernel VFP support code. It |
| 8 | provides support for the exceptions bounced from VFP hardware found |
| 9 | on ARM926EJ-S. |
| 10 | |
| 11 | This release has been validated against the SoftFloat-2b library by |
| 12 | John R. Hauser using the TestFloat-2a test suite. Details of this |
| 13 | library and test suite can be found at: |
| 14 | |
| Randy Dunlap | 98766fb | 2005-11-21 21:32:31 -0800 | [diff] [blame] | 15 | http://www.jhauser.us/arithmetic/SoftFloat.html |
| Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 16 | |
| 17 | The operations which have been tested with this package are: |
| 18 | |
| 19 | - fdiv |
| 20 | - fsub |
| 21 | - fadd |
| 22 | - fmul |
| 23 | - fcmp |
| 24 | - fcmpe |
| 25 | - fcvtd |
| 26 | - fcvts |
| 27 | - fsito |
| 28 | - ftosi |
| 29 | - fsqrt |
| 30 | |
| 31 | All the above pass softfloat tests with the following exceptions: |
| 32 | |
| 33 | - fadd/fsub shows some differences in the handling of +0 / -0 results |
| 34 | when input operands differ in signs. |
| 35 | - the handling of underflow exceptions is slightly different. If a |
| 36 | result underflows before rounding, but becomes a normalised number |
| 37 | after rounding, we do not signal an underflow exception. |
| 38 | |
| 39 | Other operations which have been tested by basic assembly-only tests |
| 40 | are: |
| 41 | |
| 42 | - fcpy |
| 43 | - fabs |
| 44 | - fneg |
| 45 | - ftoui |
| 46 | - ftosiz |
| 47 | - ftouiz |
| 48 | |
| 49 | The combination operations have not been tested: |
| 50 | |
| 51 | - fmac |
| 52 | - fnmac |
| 53 | - fmsc |
| 54 | - fnmsc |
| 55 | - fnmul |