Commit | Line | Data |
---|---|---|
1da177e4 LT |
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 | ||
98766fbe | 15 | http://www.jhauser.us/arithmetic/SoftFloat.html |
1da177e4 LT |
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 |