4 // Copyright (c) 2000 - 2003, Intel Corporation
5 // All rights reserved.
7 // Contributed 2000 by the Intel Numerics Group, Intel Corporation
9 // Redistribution and use in source and binary forms, with or without
10 // modification, are permitted provided that the following conditions are
13 // * Redistributions of source code must retain the above copyright
14 // notice, this list of conditions and the following disclaimer.
16 // * Redistributions in binary form must reproduce the above copyright
17 // notice, this list of conditions and the following disclaimer in the
18 // documentation and/or other materials provided with the distribution.
20 // * The name of Intel Corporation may not be used to endorse or promote
21 // products derived from this software without specific prior written
24 // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
25 // "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
26 // LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
27 // A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL INTEL OR ITS
28 // CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
29 // EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
30 // PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
31 // PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
32 // OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY OR TORT (INCLUDING
33 // NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
34 // SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36 // Intel Corporation is the author of this code, and requests that all
37 // problem reports or change requests be submitted to it directly at
38 // http://www.intel.com/software/products/opensource/libraries/num.htm.
41 //==============================================================
42 // 02/02/00 Initial version
43 // 06/13/00 Improved speed
44 // 06/27/00 Eliminated incorrect invalid flag setting
45 // 05/20/02 Cleaned up namespace and sf0 syntax
46 // 01/28/03 Improved performance
47 //==============================================================
50 //==============================================================
51 // long double ceill(long double x)
52 //==============================================================
54 // general input registers:
64 // floating-point registers:
73 // predicate registers used:
76 // Overview of operation
77 //==============================================================
78 // long double ceill(long double x)
79 // Return an integer value (represented as a long double) that is the smallest
80 // value not less than x
81 // This is x rounded toward +infinity to an integral value.
82 // Inexact is set if x != ceill(x)
83 //==============================================================
86 // if the exponent is > 1003e => 3F(true) = 63(decimal)
87 // we have a significand of 64 bits 1.63-bits.
88 // If we multiply by 2^63, we no longer have a fractional part
89 // So input is an integer value already.
92 // if the exponent is >= 10033 => 34(true) = 52(decimal)
94 // we have a significand of 53 bits 1.52-bits. (implicit 1)
95 // If we multiply by 2^52, we no longer have a fractional part
96 // So input is an integer value already.
99 // if the exponent is > 10016 => 17(true) = 23(decimal)
100 // we have a significand of 24 bits 1.23-bits. (implicit 1)
101 // If we multiply by 2^23, we no longer have a fractional part
102 // So input is an integer value already.
106 GLOBAL_LIBM_ENTRY(ceill)
109 getf.exp rSignexp = f8 // Get signexp, recompute if unorm
110 fclass.m p7,p0 = f8, 0x0b // Test x unorm
111 addl rBigexp = 0x1003e, r0 // Set exponent at which is integer
114 mov rM1 = -1 // Set all ones
115 fcvt.fx.trunc.s1 fXInt = f8 // Convert to int in significand
116 mov rExpMask = 0x1FFFF // Form exponent mask
121 mov rSignexpM1 = 0x2FFFF // Form signexp of -1
122 fcmp.lt.s1 p8,p9 = f8, f0 // Test x < 0
126 setf.sig fTmp = rM1 // Make const for setting inexact
127 fnorm.s1 fNormX = f8 // Normalize input
128 (p7) br.cond.spnt CEIL_UNORM // Branch if x unorm
133 // Return here from CEIL_UNORM
136 fclass.m p6,p0 = f8, 0x1e7 // Test x natval, nan, inf, 0
141 .pred.rel "mutex",p8,p9
144 (p8) fma.s1 fAdj = f0, f0, f0 // If x < 0, adjustment is 0
149 (p9) fma.s1 fAdj = f1, f1, f0 // If x > 0, adjustment is +1
156 fcvt.xf fPreResult = fXInt // trunc(x)
161 (p6) fma.s0 f8 = f8, f1, f0 // Result if x natval, nan, inf, 0
162 (p6) br.ret.spnt b0 // Exit if x natval, nan, inf, 0
167 and rExp = rSignexp, rExpMask // Get biased exponent
169 cmp.ge p7,p6 = rExp, rBigexp // Is |x| >= 2^63?
170 (p8) cmp.lt.unc p10,p0 = rSignexp, rSignexpM1 // Is -1 < x < 0?
174 // If -1 < x < 0, we turn off p6 and compute result as -0
176 (p10) cmp.ne p6,p0 = r0,r0
177 (p10) fmerge.s f8 = fNormX, f0
182 .pred.rel "mutex",p6,p7
185 (p6) fma.s0 f8 = fPreResult, f1, fAdj // Result if !int, |x| < 2^63
190 (p7) fma.s0 f8 = fNormX, f1, f0 // Result, if |x| >= 2^63
191 (p10) cmp.eq p6,p0 = r0,r0 // If -1 < x < 0, turn on p6 again
197 (p6) fcmp.eq.unc.s1 p8, p9 = fPreResult, fNormX // Is trunc(x) = x ?
204 (p9) fmpy.s0 fTmp = fTmp, fTmp // Dummy to set inexact
209 (p8) fma.s0 f8 = fNormX, f1, f0 // If x int, result normalized x
210 br.ret.sptk b0 // Exit main path, 0 < |x| < 2^63
218 getf.exp rSignexp = fNormX // Get signexp, recompute if unorm
219 fcmp.eq.s0 p7,p0 = f8, f0 // Dummy op to set denormal flag
220 br.cond.sptk CEIL_COMMON // Return to main path
224 GLOBAL_LIBM_END(ceill)