Notes |
(0000081)
psm
03-11-05 15:56
|
consider dropping the comment/support for rem_pio2f, it was used by the Sun
implementations of sinf/cosf/tanf, but due to latest changes in svn, these
are implemented from their double counterparts
the susv3 patch removes the ceilf and floorf implementations, using
floatwrapper instead |
| |
(0000082)
psm
03-12-05 10:43
|
the ldouble patch adds long double versions (similar to float_wrapper)
corrects the fact, that llrint and llround were not built and corrects the
comment in the float_wrapper, exp2,nearbyint,nexttoward add the last missing
functions as of SuSv3
nexttoward() is an alias to nextafter in glibc, but they are different
implemented as wrapper |
| |
(0000083)
psm
03-12-05 12:36
|
the math_priv patch adds __ieee754_log2 to the list, and removes all those that
are not used. with this patch libm builds cleanly. |
| |
(0000178)
psm
04-20-05 01:51
|
fpmacros and ldouble-option (apply in this order) add UCLIBC_HAS_LONG_DOUBLE_MATH
option, and changes all the references of __NO_LONG_DOUBLE_MATH to this.
As of the current configs, only i386/m68k/sparc64/x86_64 are the archs that
have long double, so the option is limited to these
I have also added/enabled __fpclassifyl,__finitel,__singbitl,__isinfl,__isnanl |
| |
(0000187)
jocke
04-23-05 05:49
|
Had a very quick look at these patches and I think the internal
functions to the lib should be static or have attribute_hidden attached. That
will make the lib smaller, faster and won't pollute the namespace. This is
probably needed in the existing lib as well. |
| |
(0012704)
bernhardf
10-03-08 10:19
|
psm, I think that most if not all of these are done/obsolete on trunk. If something is not ok by now then please respin the patch and else please close it as fixed in 0.9.30. |
| |
math-susv3-no-exp2.tar.bz2 [^] (6,154 bytes) 03-11-05 15:51
