summaryrefslogtreecommitdiffstats
path: root/lib/msun/src
Commit message (Expand)AuthorAgeFilesLines
* Oops, back out previous commit since it was backwards to a wrong branch.bde2007-06-141-1/+1
* MFC: 1.11: fix the threshold for (not) using the simple Taylor approximation.bde2007-06-141-1/+1
* Fix an aliasing bug which was finally detected by gcc-4.2. fdlibm hasbde2007-06-111-1/+1
* Merge the relevant part of rev.1.14 of s_cbrt.c (a micro-optimizationbde2007-05-291-3/+3
* Don't assume that int is signed 32-bits in one place. Keep assumingbde2007-05-022-8/+11
* Implement modfl().das2007-01-072-1/+102
* Correctly handle NaN.das2007-01-061-0/+2
* Correctly handle inf/nan. This routine is currently unused because wedas2007-01-061-0/+4
* Fixed the threshold for using the simple Taylor approximation.bde2006-07-072-2/+2
* Fixed tanh(-0.0) on ia64 and optimizeed tanh(x) for 2**-55 <= |x| <bde2006-07-051-10/+10
* Backed out rev.1.10. It tried to implement ldexpf() as a weak referencebde2006-07-051-2/+0
* Oops, on amd64 (and probably on all non-i386 systems), the previousbde2006-01-051-6/+11
* Use double precision internally to optimize cbrtf(), and change thebde2006-01-051-28/+13
* Extract the high and low words together. With gcc-3.4 on uniformlybde2005-12-201-8/+6
* Use a minimax polynomial approximation instead of a Pade rationalbde2005-12-192-37/+30
* Fixed code to match comments and the algorithm:bde2005-12-182-20/+42
* Added comments about the apparently-magic rational function used inbde2005-12-151-1/+15
* Optimize by not doing excessive conversions for handling the sign bit.bde2005-12-132-15/+9
* Fixed some especially horrible style bugs (indentation that is neitherbde2005-12-132-12/+14
* Added comments about the magic behindbde2005-12-112-14/+25
* Fixed the unexpectedly large maximum error after the previous commit.bde2005-12-111-2/+2
* Fixed all 502518670 errors of more than 1 ulp for cbrtf() on amd64.bde2005-12-111-1/+13
* Fixed some magic numbers.bde2005-12-111-8/+8
* "Create" ldexpf for non-i386 architectures.obrien2005-12-061-0/+2
* Fixed the approximation to pio4. pio4_hi must be pio2_hi/2 since itbde2005-12-041-1/+1
* For log1pf(), fixed the approximations to sqrt(2), sqrt(2)-1 andbde2005-12-042-8/+22
* Use the usual volatile hack to trick gcc into clipping any extra precisionbde2005-12-041-1/+1
* Fixed fdlibm[+cygnus] logbf() and logb() on denormals. Adjustmentbde2005-12-032-8/+20
* Restored removal of the special handling needed for a result of +-0.bde2005-12-031-0/+2
* Simplified the fix in rev.1.3. Instead of using long double forbde2005-12-031-8/+2
* Fixed rint(x) in the following cases:bde2005-12-031-0/+9
* Fixed roundf(). The following cases never worked in FreeBSD:bde2005-12-023-18/+18
* Rearranged the polynomial evaluation to reduce dependencies, as inbde2005-11-302-9/+13
* Fixed cosf(x) when x is a "negative" NaNs. I broke this in rev.1.10.bde2005-11-301-11/+19
* Fixed the hi+lo approximation to log(2). The normal 17+24 bit decompositionbde2005-11-301-4/+4
* Rearranged the polynomial evaluation some more to reduce dependencies.bde2005-11-281-8/+20
* Fixed about 50 million errors of infinity ulps and about 3 million errorsbde2005-11-281-1/+1
* Exploit skew-symmetry to avoid an operation: -sin(x-A) = sin(A-x). Thisbde2005-11-282-4/+4
* Try to use the "proximity" (~) operator consistently in commentsbde2005-11-282-8/+10
* Changed spelling of the request-to-inline macro name to match the changebde2005-11-282-4/+6
* Use only double precision for "kernel" cosf and sinf (except forbde2005-11-286-71/+44
* Minor cleanups and optimizations:bde2005-11-241-11/+5
* Optimized by eliminating the special case for 0.67434 <= |x| < pi/4.bde2005-11-241-16/+7
* Use only double precision for "kernel" tanf (except for returning float).bde2005-11-233-29/+20
* Simplified setiing up args for __kernel_rem_pio2(). We already have xbde2005-11-231-17/+9
* Quick fix for stack buffer overrun in rev.1.13. Oops. The prec == 1bde2005-11-231-4/+4
* Mess up the "kernel" float trig function .c files with ifdefs so thatbde2005-11-216-0/+25
* Use double precision to simplify and optimize a long division.bde2005-11-211-15/+1
* Restored a cleanup in rev.1.9 tthat was lost in rev.1.10.bde2005-11-201-2/+2
* Moved all the optimizations for |x| <= 9pi/2 frombde2005-11-194-67/+105
OpenPOWER on IntegriCloud