[Bug 298260] [libm] fmaf() mishandled some subnormal input

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=298260

--- Comment #1 from Steve Kargl <kargl@FreeBSD.org> ---
Created attachment 274517
  --> https://bugs.freebsd.org/bugzilla/attachment.cgi?id=274517&action=edit
Possible fix for fmaf handling of subnormal

The attached diff converts the implemention for fma() into an
implementation for fmaf().  Note, there are a few magic 
numbers used in the manipulation of exponents.  Someone else
needs to confirm correctness.

The patch has been tested against the code in the original
comment for this bug.  I have not run msun/tests/fma_test.c
as I don't have the test framework installed.

Buyer beware.

-- 
You are receiving this mail because:
You are the assignee for the bug.