1 /* Function sin vectorized with AVX2.
2 Copyright (C) 2014-2022 Free Software Foundation, Inc.
3 This file is part of the GNU C Library.
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.
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.
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/>. */
20 #include "svml_d_trig_data.h"
23 ENTRY (_ZGVdN4v_sin_avx2)
24 /* ALGORITHM DESCRIPTION:
26 ( low accuracy ( < 4ulp ) or enhanced performance
27 ( half of correct mantissa ) implementation )
29 Argument representation:
33 sin(arg) = sin(N*Pi + R) = (-1)^N * sin(R)
34 sin(R) is approximated by corresponding polynomial
37 cfi_adjust_cfa_offset (8)
38 cfi_rel_offset (%rbp, 0)
40 cfi_def_cfa_register (%rbp)
43 movq __svml_d_trig_data@GOTPCREL(%rip), %rax
45 vmovupd __dAbsMask(%rax), %ymm2
46 vmovupd __dInvPI(%rax), %ymm6
47 vmovupd __dRShifter(%rax), %ymm5
48 vmovupd __dPI1_FMA(%rax), %ymm7
50 ARGUMENT RANGE REDUCTION:
53 vandpd %ymm2, %ymm4, %ymm3
55 /* Y = X'*InvPi + RS : right shifter add */
56 vfmadd213pd %ymm5, %ymm3, %ymm6
58 /* N = Y - RS : right shifter sub */
59 vsubpd %ymm5, %ymm6, %ymm1
61 /* SignRes = Y<<63 : shift LSB to MSB place for result sign */
62 vpsllq $63, %ymm6, %ymm5
66 vfnmadd231pd %ymm1, %ymm7, %ymm0
67 vcmpnle_uqpd __dRangeVal(%rax), %ymm3, %ymm3
70 vfnmadd231pd __dPI2_FMA(%rax), %ymm1, %ymm0
73 vfnmadd132pd __dPI3_FMA(%rax), %ymm0, %ymm1
76 POLYNOMIAL APPROXIMATION:
79 vmulpd %ymm1, %ymm1, %ymm0
81 /* R = R^SignRes : update sign of reduced argument */
82 vxorpd %ymm5, %ymm1, %ymm6
83 vmovupd __dC7_sin(%rax), %ymm1
84 vfmadd213pd __dC6_sin(%rax), %ymm0, %ymm1
85 vfmadd213pd __dC5_sin(%rax), %ymm0, %ymm1
86 vfmadd213pd __dC4_sin(%rax), %ymm0, %ymm1
88 /* Poly = C3+R2*(C4+R2*(C5+R2*(C6+R2*C7))) */
89 vfmadd213pd __dC3_sin(%rax), %ymm0, %ymm1
91 /* Poly = R2*(C1+R2*(C2+R2*Poly)) */
92 vfmadd213pd __dC2_sin(%rax), %ymm0, %ymm1
93 vfmadd213pd __dC1_sin(%rax), %ymm0, %ymm1
95 /* SignX - sign bit of X */
96 vandnpd %ymm4, %ymm2, %ymm7
97 vmulpd %ymm0, %ymm1, %ymm2
99 /* Poly = Poly*R + R */
100 vfmadd213pd %ymm6, %ymm6, %ymm2
101 vmovmskpd %ymm3, %ecx
105 Final sign setting: Res = Poly^SignX
107 vxorpd %ymm7, %ymm2, %ymm0
114 cfi_def_cfa_register (%rsp)
116 cfi_adjust_cfa_offset (-8)
122 vmovupd %ymm4, 320(%rsp)
123 vmovupd %ymm0, 384(%rsp)
128 vmovups %ymm8, 224(%rsp)
129 vmovups %ymm9, 192(%rsp)
130 vmovups %ymm10, 160(%rsp)
131 vmovups %ymm11, 128(%rsp)
132 vmovups %ymm12, 96(%rsp)
133 vmovups %ymm13, 64(%rsp)
134 vmovups %ymm14, 32(%rsp)
135 vmovups %ymm15, (%rsp)
139 cfi_offset_rel_rsp (12, 296)
142 cfi_offset_rel_rsp (13, 288)
145 cfi_offset_rel_rsp (14, 280)
148 cfi_offset_rel_rsp (15, 272)
166 vmovups 224(%rsp), %ymm8
167 vmovups 192(%rsp), %ymm9
168 vmovups 160(%rsp), %ymm10
169 vmovups 128(%rsp), %ymm11
170 vmovups 96(%rsp), %ymm12
171 vmovups 64(%rsp), %ymm13
172 vmovups 32(%rsp), %ymm14
173 vmovups (%rsp), %ymm15
174 vmovupd 384(%rsp), %ymm0
191 vmovsd 328(%rsp,%r15), %xmm0
196 vmovsd %xmm0, 392(%rsp,%r15)
202 vmovsd 320(%rsp,%r15), %xmm0
207 vmovsd %xmm0, 384(%rsp,%r15)
210 END (_ZGVdN4v_sin_avx2)