diff options
author | Russell King <rmk@flint.arm.linux.org.uk> | 2004-06-18 15:33:48 +0100 |
---|---|---|
committer | Russell King <rmk@flint.arm.linux.org.uk> | 2004-06-18 15:33:48 +0100 |
commit | 451436b7bbb20add6d09d158f2ddf1ffcd7cb360 (patch) | |
tree | c4a1a651c7ed2f766164d1ef5a471e5ea204003b /Documentation | |
parent | cda97ee3b4935547f7a547f3d7ad8b0dd07bc8ee (diff) | |
download | history-451436b7bbb20add6d09d158f2ddf1ffcd7cb360.tar.gz |
[ARM] Add support code for ARM hardware vector floating point
This cset adds the code to handle the hardware vector floating point
unit found on some ARM926 and later CPUs. The hardware provides
an implementation for the common cases, and bounces exceptions for
other cases, which have to be handled in software, and signalling
SIGFPE as appropriate.
Diffstat (limited to 'Documentation')
-rw-r--r-- | Documentation/arm/VFP/release-notes.txt | 55 |
1 files changed, 55 insertions, 0 deletions
diff --git a/Documentation/arm/VFP/release-notes.txt b/Documentation/arm/VFP/release-notes.txt new file mode 100644 index 00000000000000..f28e0222f5e548 --- /dev/null +++ b/Documentation/arm/VFP/release-notes.txt @@ -0,0 +1,55 @@ +Release notes for Linux Kernel VFP support code +----------------------------------------------- + +Date: 20 May 2004 +Author: Russell King + +This is the first release of the Linux Kernel VFP support code. It +provides support for the exceptions bounced from VFP hardware found +on ARM926EJ-S. + +This release has been validated against the SoftFloat-2b library by +John R. Hauser using the TestFloat-2a test suite. Details of this +library and test suite can be found at: + + http://www.cs.berkeley.edu/~jhauser/arithmetic/SoftFloat.html + +The operations which have been tested with this package are: + + - fdiv + - fsub + - fadd + - fmul + - fcmp + - fcmpe + - fcvtd + - fcvts + - fsito + - ftosi + - fsqrt + +All the above pass softfloat tests with the following exceptions: + +- fadd/fsub shows some differences in the handling of +0 / -0 results + when input operands differ in signs. +- the handling of underflow exceptions is slightly different. If a + result underflows before rounding, but becomes a normalised number + after rounding, we do not signal an underflow exception. + +Other operations which have been tested by basic assembly-only tests +are: + + - fcpy + - fabs + - fneg + - ftoui + - ftosiz + - ftouiz + +The combination operations have not been tested: + + - fmac + - fnmac + - fmsc + - fnmsc + - fnmul |