2 // Copyright (c) 2000, Intel Corporation
3 // All rights reserved.
5 // Contributed 2/15/2000 by Marius Cornea, John Harrison, Cristina Iordache,
6 // Ted Kubaska, Bob Norin, and Shane Story of the Computational Software Lab,
11 // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
12 // "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
13 // LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
14 // A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL INTEL OR ITS
15 // CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
16 // EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
17 // PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
18 // PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
19 // OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY OR TORT (INCLUDING
20 // NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
21 // SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
23 // Intel Corporation is the author of this code, and requests that all
24 // problem reports or change requests be submitted to it directly at
25 // http://developer.intel.com/opensource.
28 #include <machine/asm.h>
29 __FBSDID("$FreeBSD$");
38 // predicate registers used: p6
39 // floating-point registers used: f6, f7, f8, f9, f10, f11
41 // load a, the first argument, in f6
46 // load b, the second argument, in f7
52 // BEGIN DOUBLE PRECISION LATENCY-OPTIMIZED DIVIDE ALGORITHM
63 (p6) fma.s1 f9=f6,f8,f0
68 // e0 = 1 - b * y0 in f10
69 (p6) fnma.s1 f10=f7,f8,f1
74 // q1 = q0 + e0 * q0 in f9
75 (p6) fma.s1 f9=f10,f9,f9
80 // e1 = e0 * e0 in f11
81 (p6) fma.s1 f11=f10,f10,f0
86 // y1 = y0 + e0 * y0 in f8
87 (p6) fma.s1 f8=f10,f8,f8
92 // q2 = q1 + e1 * q1 in f9
93 (p6) fma.s1 f9=f11,f9,f9
98 // e2 = e1 * e1 in f10
99 (p6) fma.s1 f10=f11,f11,f0
104 // y2 = y1 + e1 * y1 in f8
105 (p6) fma.s1 f8=f11,f8,f8
110 // q3 = q2 + e2 * q2 in f9
111 (p6) fma.d.s1 f9=f10,f9,f9
116 // y3 = y2 + e2 * y2 in f8
117 (p6) fma.s1 f8=f10,f8,f8
122 // r0 = a - b * q3 in f6
123 (p6) fnma.d.s1 f6=f7,f9,f6
128 // q4 = q3 + r0 * y3 in f8
129 (p6) fma.d.s0 f8=f6,f8,f9
132 // END DOUBLE PRECISION LATENCY-OPTIMIZED DIVIDE ALGORITHM