diff options
Diffstat (limited to 'sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_ceil-vis3.S')
-rw-r--r-- | sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_ceil-vis3.S | 78 |
1 files changed, 78 insertions, 0 deletions
diff --git a/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_ceil-vis3.S b/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_ceil-vis3.S new file mode 100644 index 0000000000..be41219cef --- /dev/null +++ b/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_ceil-vis3.S @@ -0,0 +1,78 @@ +/* ceil function, sparc32 v9 vis3 version. + Copyright (C) 2012 Free Software Foundation, Inc. + This file is part of the GNU C Library. + Contributed by David S. Miller <davem@davemloft.net>, 2012. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + <http://www.gnu.org/licenses/>. */ + +#include <sysdep.h> + + /* Since changing the rounding mode is extremely expensive, we + try to round up using a method that is rounding mode + agnostic. + + We add then subtract (or subtract than add if the initial + value was negative) 2**23 to the value, then subtract it + back out. + + This will clear out the fractional portion of the value. + One of two things will happen for non-whole initial values. + Either the rounding mode will round it up, or it will be + rounded down. If the value started out whole, it will be + equal after the addition and subtraction. This means we + can accurately detect with one test whether we need to add + another 1.0 to round it up properly. + + VIS instructions are used to facilitate the formation of + easier constants, and the propagation of the sign bit. */ + +#define TWO_FIFTYTWO 0x43300000 /* 2**52 */ +#define ONE_DOT_ZERO 0x3ff00000 /* 1.0 */ + +#define ZERO %f10 /* 0.0 */ +#define SIGN_BIT %f12 /* -0.0 */ + +ENTRY (__ceil_vis3) + sethi %hi(TWO_FIFTYTWO), %o2 + sllx %o0, 32, %o0 + sethi %hi(ONE_DOT_ZERO), %o3 + or %o0, %o1, %o0 + movxtod %o0, %f0 + sllx %o2, 32, %o2 + fzero ZERO + sllx %o3, 32, %o3 + + fnegd ZERO, SIGN_BIT + + movxtod %o2, %f16 + fabsd %f0, %f14 + + fcmpd %fcc3, %f14, %f16 + + fmovduge %fcc3, ZERO, %f16 + fand %f0, SIGN_BIT, SIGN_BIT + + for %f16, SIGN_BIT, %f16 + faddd %f0, %f16, %f18 + fsubd %f18, %f16, %f18 + fcmpd %fcc2, %f18, %f0 + movxtod %o3, %f20 + + fmovduge %fcc2, ZERO, %f20 + faddd %f18, %f20, %f0 + fabsd %f0, %f0 + retl + for %f0, SIGN_BIT, %f0 +END (__ceil_vis3) |