Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/dft/simd/common/n2bv_8.c @ 167:bd3cc4d1df30

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add FFTW 3.3.8 source, and a Linux build
author Chris Cannam <cannam@all-day-breakfast.com>
date Tue, 19 Nov 2019 14:52:55 +0000
parents
children
comparison
equal deleted inserted replaced
166:cbd6d7e562c7 167:bd3cc4d1df30
1 /*
2 * Copyright (c) 2003, 2007-14 Matteo Frigo
3 * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology
4 *
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by
7 * the Free Software Foundation; either version 2 of the License, or
8 * (at your option) any later version.
9 *
10 * This program 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
13 * GNU General Public License for more details.
14 *
15 * You should have received a copy of the GNU General Public License
16 * along with this program; if not, write to the Free Software
17 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
18 *
19 */
20
21 /* This file was automatically generated --- DO NOT EDIT */
22 /* Generated on Thu May 24 08:05:11 EDT 2018 */
23
24 #include "dft/codelet-dft.h"
25
26 #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)
27
28 /* Generated by: ../../../genfft/gen_notw_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 8 -name n2bv_8 -with-ostride 2 -include dft/simd/n2b.h -store-multiple 2 */
29
30 /*
31 * This function contains 26 FP additions, 10 FP multiplications,
32 * (or, 16 additions, 0 multiplications, 10 fused multiply/add),
33 * 24 stack variables, 1 constants, and 20 memory accesses
34 */
35 #include "dft/simd/n2b.h"
36
37 static void n2bv_8(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
38 {
39 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
40 {
41 INT i;
42 const R *xi;
43 R *xo;
44 xi = ii;
45 xo = io;
46 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(16, is), MAKE_VOLATILE_STRIDE(16, os)) {
47 V T3, Tj, Te, Tk, Ta, Tn, Tf, Tm, Tr, Tu;
48 {
49 V T1, T2, Tc, Td;
50 T1 = LD(&(xi[0]), ivs, &(xi[0]));
51 T2 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
52 T3 = VSUB(T1, T2);
53 Tj = VADD(T1, T2);
54 Tc = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
55 Td = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
56 Te = VSUB(Tc, Td);
57 Tk = VADD(Tc, Td);
58 {
59 V T4, T5, T6, T7, T8, T9;
60 T4 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
61 T5 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
62 T6 = VSUB(T4, T5);
63 T7 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
64 T8 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
65 T9 = VSUB(T7, T8);
66 Ta = VADD(T6, T9);
67 Tn = VADD(T7, T8);
68 Tf = VSUB(T6, T9);
69 Tm = VADD(T4, T5);
70 }
71 }
72 {
73 V Ts, Tb, Tg, Tp, Tq, Tt;
74 Tb = VFNMS(LDK(KP707106781), Ta, T3);
75 Tg = VFNMS(LDK(KP707106781), Tf, Te);
76 Tr = VFNMSI(Tg, Tb);
77 STM2(&(xo[6]), Tr, ovs, &(xo[2]));
78 Ts = VFMAI(Tg, Tb);
79 STM2(&(xo[10]), Ts, ovs, &(xo[2]));
80 Tp = VADD(Tj, Tk);
81 Tq = VADD(Tm, Tn);
82 Tt = VSUB(Tp, Tq);
83 STM2(&(xo[8]), Tt, ovs, &(xo[0]));
84 STN2(&(xo[8]), Tt, Ts, ovs);
85 Tu = VADD(Tp, Tq);
86 STM2(&(xo[0]), Tu, ovs, &(xo[0]));
87 }
88 {
89 V Tw, Th, Ti, Tv;
90 Th = VFMA(LDK(KP707106781), Ta, T3);
91 Ti = VFMA(LDK(KP707106781), Tf, Te);
92 Tv = VFMAI(Ti, Th);
93 STM2(&(xo[2]), Tv, ovs, &(xo[2]));
94 STN2(&(xo[0]), Tu, Tv, ovs);
95 Tw = VFNMSI(Ti, Th);
96 STM2(&(xo[14]), Tw, ovs, &(xo[2]));
97 {
98 V Tl, To, Tx, Ty;
99 Tl = VSUB(Tj, Tk);
100 To = VSUB(Tm, Tn);
101 Tx = VFNMSI(To, Tl);
102 STM2(&(xo[12]), Tx, ovs, &(xo[0]));
103 STN2(&(xo[12]), Tx, Tw, ovs);
104 Ty = VFMAI(To, Tl);
105 STM2(&(xo[4]), Ty, ovs, &(xo[0]));
106 STN2(&(xo[4]), Ty, Tr, ovs);
107 }
108 }
109 }
110 }
111 VLEAVE();
112 }
113
114 static const kdft_desc desc = { 8, XSIMD_STRING("n2bv_8"), {16, 0, 10, 0}, &GENUS, 0, 2, 0, 0 };
115
116 void XSIMD(codelet_n2bv_8) (planner *p) {
117 X(kdft_register) (p, n2bv_8, &desc);
118 }
119
120 #else
121
122 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 8 -name n2bv_8 -with-ostride 2 -include dft/simd/n2b.h -store-multiple 2 */
123
124 /*
125 * This function contains 26 FP additions, 2 FP multiplications,
126 * (or, 26 additions, 2 multiplications, 0 fused multiply/add),
127 * 24 stack variables, 1 constants, and 20 memory accesses
128 */
129 #include "dft/simd/n2b.h"
130
131 static void n2bv_8(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
132 {
133 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
134 {
135 INT i;
136 const R *xi;
137 R *xo;
138 xi = ii;
139 xo = io;
140 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(16, is), MAKE_VOLATILE_STRIDE(16, os)) {
141 V Ta, Tk, Te, Tj, T7, Tn, Tf, Tm, Tr, Tu;
142 {
143 V T8, T9, Tc, Td;
144 T8 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
145 T9 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
146 Ta = VSUB(T8, T9);
147 Tk = VADD(T8, T9);
148 Tc = LD(&(xi[0]), ivs, &(xi[0]));
149 Td = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
150 Te = VSUB(Tc, Td);
151 Tj = VADD(Tc, Td);
152 {
153 V T1, T2, T3, T4, T5, T6;
154 T1 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
155 T2 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
156 T3 = VSUB(T1, T2);
157 T4 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
158 T5 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
159 T6 = VSUB(T4, T5);
160 T7 = VMUL(LDK(KP707106781), VSUB(T3, T6));
161 Tn = VADD(T4, T5);
162 Tf = VMUL(LDK(KP707106781), VADD(T3, T6));
163 Tm = VADD(T1, T2);
164 }
165 }
166 {
167 V Ts, Tb, Tg, Tp, Tq, Tt;
168 Tb = VBYI(VSUB(T7, Ta));
169 Tg = VSUB(Te, Tf);
170 Tr = VADD(Tb, Tg);
171 STM2(&(xo[6]), Tr, ovs, &(xo[2]));
172 Ts = VSUB(Tg, Tb);
173 STM2(&(xo[10]), Ts, ovs, &(xo[2]));
174 Tp = VADD(Tj, Tk);
175 Tq = VADD(Tm, Tn);
176 Tt = VSUB(Tp, Tq);
177 STM2(&(xo[8]), Tt, ovs, &(xo[0]));
178 STN2(&(xo[8]), Tt, Ts, ovs);
179 Tu = VADD(Tp, Tq);
180 STM2(&(xo[0]), Tu, ovs, &(xo[0]));
181 }
182 {
183 V Tw, Th, Ti, Tv;
184 Th = VBYI(VADD(Ta, T7));
185 Ti = VADD(Te, Tf);
186 Tv = VADD(Th, Ti);
187 STM2(&(xo[2]), Tv, ovs, &(xo[2]));
188 STN2(&(xo[0]), Tu, Tv, ovs);
189 Tw = VSUB(Ti, Th);
190 STM2(&(xo[14]), Tw, ovs, &(xo[2]));
191 {
192 V Tl, To, Tx, Ty;
193 Tl = VSUB(Tj, Tk);
194 To = VBYI(VSUB(Tm, Tn));
195 Tx = VSUB(Tl, To);
196 STM2(&(xo[12]), Tx, ovs, &(xo[0]));
197 STN2(&(xo[12]), Tx, Tw, ovs);
198 Ty = VADD(Tl, To);
199 STM2(&(xo[4]), Ty, ovs, &(xo[0]));
200 STN2(&(xo[4]), Ty, Tr, ovs);
201 }
202 }
203 }
204 }
205 VLEAVE();
206 }
207
208 static const kdft_desc desc = { 8, XSIMD_STRING("n2bv_8"), {26, 2, 0, 0}, &GENUS, 0, 2, 0, 0 };
209
210 void XSIMD(codelet_n2bv_8) (planner *p) {
211 X(kdft_register) (p, n2bv_8, &desc);
212 }
213
214 #endif