diff options
Diffstat (limited to 'sysdeps/powerpc/powerpc64/fpu/s_nearbyintl.S')
| -rw-r--r-- | sysdeps/powerpc/powerpc64/fpu/s_nearbyintl.S | 8 |
1 files changed, 4 insertions, 4 deletions
diff --git a/sysdeps/powerpc/powerpc64/fpu/s_nearbyintl.S b/sysdeps/powerpc/powerpc64/fpu/s_nearbyintl.S index 9cf03ccd7a..b235d9b99c 100644 --- a/sysdeps/powerpc/powerpc64/fpu/s_nearbyintl.S +++ b/sysdeps/powerpc/powerpc64/fpu/s_nearbyintl.S @@ -26,16 +26,16 @@ .section ".text" /* long double [fp1,fp2] nearbyintl (long double x [fp1,fp2]) - IEEE 1003.1 nearbyintl function. nearbyintl is simular to the rintl + IEEE 1003.1 nearbyintl function. nearbyintl is similar to the rintl but does raise the "inexact" exception. This implementation is - based on rintl but explicitly maskes the inexact exception on entry + based on rintl but explicitly masks the inexact exception on entry and clears any pending inexact before restoring the exception mask on exit. PowerPC64 long double uses the IBM extended format which is represented two 64-floating point double values. The values are non-overlapping giving an effective precision of 106 bits. The first - double contains the high order bits of mantisa and is always rounded + double contains the high order bits of mantissa and is always rounded to represent a normal rounding of long double to double. Since the long double value is sum of the high and low values, the low double normally has the opposite sign to compensate for the this rounding. @@ -44,7 +44,7 @@ 1) |x| < 2**52, all the integer bits are in the high double. floor the high double and set the low double to -0.0. 2) |x| >= 2**52, Rounding involves both doubles. - See the comment before lable .L2 for details. + See the comment before label .L2 for details. */ ENTRY (__nearbyintl) mffs fp11 /* Save current FPSCR. */ |