1/* Function sin vectorized with SSE4.
2   Copyright (C) 2014-2022 Free Software Foundation, Inc.
3   This file is part of the GNU C Library.
4
5   The GNU C Library is free software; you can redistribute it and/or
6   modify it under the terms of the GNU Lesser General Public
7   License as published by the Free Software Foundation; either
8   version 2.1 of the License, or (at your option) any later version.
9
10   The GNU C Library is distributed in the hope that it will be useful,
11   but WITHOUT ANY WARRANTY; without even the implied warranty of
12   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
13   Lesser General Public License for more details.
14
15   You should have received a copy of the GNU Lesser General Public
16   License along with the GNU C Library; if not, see
17   <https://www.gnu.org/licenses/>.  */
18
19#include <sysdep.h>
20#include "svml_d_trig_data.h"
21
22	.text
23ENTRY (_ZGVbN2v_sin_sse4)
24/* ALGORITHM DESCRIPTION:
25
26      ( low accuracy ( < 4ulp ) or enhanced performance
27       ( half of correct mantissa ) implementation )
28
29      Argument representation:
30      arg = N*Pi + R
31
32      Result calculation:
33      sin(arg) = sin(N*Pi + R) = (-1)^N * sin(R)
34      sin(R) is approximated by corresponding polynomial
35 */
36        pushq     %rbp
37        cfi_adjust_cfa_offset (8)
38        cfi_rel_offset (%rbp, 0)
39        movq      %rsp, %rbp
40        cfi_def_cfa_register (%rbp)
41        andq      $-64, %rsp
42        subq      $320, %rsp
43        movaps    %xmm0, %xmm5
44        movq      __svml_d_trig_data@GOTPCREL(%rip), %rax
45        movups __dAbsMask(%rax), %xmm3
46/*
47   ARGUMENT RANGE REDUCTION:
48   X' = |X|
49 */
50        movaps    %xmm3, %xmm4
51
52/* SignX - sign bit of X */
53        andnps    %xmm5, %xmm3
54        movups __dInvPI(%rax), %xmm2
55        andps     %xmm5, %xmm4
56
57/* Y = X'*InvPi + RS : right shifter add */
58        mulpd     %xmm4, %xmm2
59        movups __dRShifter(%rax), %xmm6
60
61/* R = X' - N*Pi1 */
62        movaps    %xmm4, %xmm0
63        addpd     %xmm6, %xmm2
64        cmpnlepd __dRangeVal(%rax), %xmm4
65
66/* N = Y - RS : right shifter sub */
67        movaps    %xmm2, %xmm1
68
69/* SignRes = Y<<63 : shift LSB to MSB place for result sign */
70        psllq     $63, %xmm2
71        subpd     %xmm6, %xmm1
72        movmskpd  %xmm4, %ecx
73        movups __dPI1(%rax), %xmm7
74        mulpd     %xmm1, %xmm7
75        movups __dPI2(%rax), %xmm6
76
77/* R = R - N*Pi2 */
78        mulpd     %xmm1, %xmm6
79        subpd     %xmm7, %xmm0
80        movups __dPI3(%rax), %xmm7
81
82/* R = R - N*Pi3 */
83        mulpd     %xmm1, %xmm7
84        subpd     %xmm6, %xmm0
85        movups __dPI4(%rax), %xmm6
86
87/* R = R - N*Pi4 */
88        mulpd     %xmm6, %xmm1
89        subpd     %xmm7, %xmm0
90        subpd     %xmm1, %xmm0
91
92/*
93   POLYNOMIAL APPROXIMATION:
94   R2 = R*R
95 */
96        movaps    %xmm0, %xmm1
97        mulpd     %xmm0, %xmm1
98
99/* R = R^SignRes : update sign of reduced argument */
100        xorps     %xmm2, %xmm0
101        movups    __dC7_sin(%rax), %xmm2
102        mulpd     %xmm1, %xmm2
103        addpd     __dC6_sin(%rax), %xmm2
104        mulpd     %xmm1, %xmm2
105        addpd     __dC5_sin(%rax), %xmm2
106        mulpd     %xmm1, %xmm2
107        addpd     __dC4_sin(%rax), %xmm2
108
109/* Poly = C3+R2*(C4+R2*(C5+R2*(C6+R2*C7))) */
110        mulpd     %xmm1, %xmm2
111        addpd     __dC3_sin(%rax), %xmm2
112
113/* Poly = R2*(C1+R2*(C2+R2*Poly)) */
114        mulpd     %xmm1, %xmm2
115        addpd     __dC2_sin(%rax), %xmm2
116        mulpd     %xmm1, %xmm2
117        addpd     __dC1_sin(%rax), %xmm2
118        mulpd     %xmm2, %xmm1
119
120/* Poly = Poly*R + R */
121        mulpd     %xmm0, %xmm1
122        addpd     %xmm1, %xmm0
123
124/*
125   RECONSTRUCTION:
126   Final sign setting: Res = Poly^SignX
127 */
128        xorps     %xmm3, %xmm0
129        testl     %ecx, %ecx
130        jne       .LBL_1_3
131
132.LBL_1_2:
133        cfi_remember_state
134        movq      %rbp, %rsp
135        cfi_def_cfa_register (%rsp)
136        popq      %rbp
137        cfi_adjust_cfa_offset (-8)
138        cfi_restore (%rbp)
139        ret
140
141.LBL_1_3:
142        cfi_restore_state
143        movups    %xmm5, 192(%rsp)
144        movups    %xmm0, 256(%rsp)
145        je        .LBL_1_2
146
147        xorb      %dl, %dl
148        xorl      %eax, %eax
149        movups    %xmm8, 112(%rsp)
150        movups    %xmm9, 96(%rsp)
151        movups    %xmm10, 80(%rsp)
152        movups    %xmm11, 64(%rsp)
153        movups    %xmm12, 48(%rsp)
154        movups    %xmm13, 32(%rsp)
155        movups    %xmm14, 16(%rsp)
156        movups    %xmm15, (%rsp)
157        movq      %rsi, 136(%rsp)
158        movq      %rdi, 128(%rsp)
159        movq      %r12, 168(%rsp)
160        cfi_offset_rel_rsp (12, 168)
161        movb      %dl, %r12b
162        movq      %r13, 160(%rsp)
163        cfi_offset_rel_rsp (13, 160)
164        movl      %ecx, %r13d
165        movq      %r14, 152(%rsp)
166        cfi_offset_rel_rsp (14, 152)
167        movl      %eax, %r14d
168        movq      %r15, 144(%rsp)
169        cfi_offset_rel_rsp (15, 144)
170        cfi_remember_state
171
172.LBL_1_6:
173        btl       %r14d, %r13d
174        jc        .LBL_1_12
175
176.LBL_1_7:
177        lea       1(%r14), %esi
178        btl       %esi, %r13d
179        jc        .LBL_1_10
180
181.LBL_1_8:
182        incb      %r12b
183        addl      $2, %r14d
184        cmpb      $16, %r12b
185        jb        .LBL_1_6
186
187        movups    112(%rsp), %xmm8
188        movups    96(%rsp), %xmm9
189        movups    80(%rsp), %xmm10
190        movups    64(%rsp), %xmm11
191        movups    48(%rsp), %xmm12
192        movups    32(%rsp), %xmm13
193        movups    16(%rsp), %xmm14
194        movups    (%rsp), %xmm15
195        movq      136(%rsp), %rsi
196        movq      128(%rsp), %rdi
197        movq      168(%rsp), %r12
198        cfi_restore (%r12)
199        movq      160(%rsp), %r13
200        cfi_restore (%r13)
201        movq      152(%rsp), %r14
202        cfi_restore (%r14)
203        movq      144(%rsp), %r15
204        cfi_restore (%r15)
205        movups    256(%rsp), %xmm0
206        jmp       .LBL_1_2
207
208.LBL_1_10:
209        cfi_restore_state
210        movzbl    %r12b, %r15d
211        shlq      $4, %r15
212        movsd     200(%rsp,%r15), %xmm0
213
214        call      JUMPTARGET(sin)
215
216        movsd     %xmm0, 264(%rsp,%r15)
217        jmp       .LBL_1_8
218
219.LBL_1_12:
220        movzbl    %r12b, %r15d
221        shlq      $4, %r15
222        movsd     192(%rsp,%r15), %xmm0
223
224        call      JUMPTARGET(sin)
225
226        movsd     %xmm0, 256(%rsp,%r15)
227        jmp       .LBL_1_7
228
229END (_ZGVbN2v_sin_sse4)
230