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 Dec 10 07:04:44 EST 2020 */
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 9 -name n1bv_9 -include dft/simd/n1b.h */
29
30 /*
31 * This function contains 46 FP additions, 38 FP multiplications,
32 * (or, 12 additions, 4 multiplications, 34 fused multiply/add),
33 * 50 stack variables, 19 constants, and 18 memory accesses
34 */
35 #include "dft/simd/n1b.h"
36
n1bv_9(const R * ri,const R * ii,R * ro,R * io,stride is,stride os,INT v,INT ivs,INT ovs)37 static void n1bv_9(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
38 {
39 DVK(KP666666666, +0.666666666666666666666666666666666666666666667);
40 DVK(KP852868531, +0.852868531952443209628250963940074071936020296);
41 DVK(KP898197570, +0.898197570222573798468955502359086394667167570);
42 DVK(KP673648177, +0.673648177666930348851716626769314796000375677);
43 DVK(KP879385241, +0.879385241571816768108218554649462939872416269);
44 DVK(KP984807753, +0.984807753012208059366743024589523013670643252);
45 DVK(KP939692620, +0.939692620785908384054109277324731469936208134);
46 DVK(KP826351822, +0.826351822333069651148283373230685203999624323);
47 DVK(KP420276625, +0.420276625461206169731530603237061658838781920);
48 DVK(KP907603734, +0.907603734547952313649323976213898122064543220);
49 DVK(KP347296355, +0.347296355333860697703433253538629592000751354);
50 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
51 DVK(KP968908795, +0.968908795874236621082202410917456709164223497);
52 DVK(KP726681596, +0.726681596905677465811651808188092531873167623);
53 DVK(KP586256827, +0.586256827714544512072145703099641959914944179);
54 DVK(KP152703644, +0.152703644666139302296566746461370407999248646);
55 DVK(KP203604859, +0.203604859554852403062088995281827210665664861);
56 DVK(KP439692620, +0.439692620785908384054109277324731469936208134);
57 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
58 {
59 INT i;
60 const R *xi;
61 R *xo;
62 xi = ii;
63 xo = io;
64 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(18, is), MAKE_VOLATILE_STRIDE(18, os)) {
65 V T5, TF, Tp, Te, Td, TG, TH, Ta, Tm, Tu, Tr, Th, Ti, Tv, Ts;
66 V TK, TI, TJ;
67 {
68 V T1, T2, T3, T4;
69 T1 = LD(&(xi[0]), ivs, &(xi[0]));
70 T2 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
71 T3 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
72 T4 = VADD(T2, T3);
73 T5 = VFNMS(LDK(KP500000000), T4, T1);
74 TF = VADD(T1, T4);
75 Tp = VSUB(T2, T3);
76 }
77 {
78 V T6, Tf, T9, Tg;
79 T6 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
80 Tf = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
81 {
82 V T7, T8, Tb, Tc;
83 T7 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
84 T8 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
85 T9 = VADD(T7, T8);
86 Te = VSUB(T8, T7);
87 Tb = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
88 Tc = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
89 Td = VSUB(Tb, Tc);
90 Tg = VADD(Tb, Tc);
91 }
92 TG = VADD(Tf, Tg);
93 TH = VADD(T6, T9);
94 Ta = VFNMS(LDK(KP500000000), T9, T6);
95 Tm = VFNMS(LDK(KP439692620), Td, Ta);
96 Tu = VFMA(LDK(KP203604859), Ta, Te);
97 Tr = VFNMS(LDK(KP152703644), Te, Ta);
98 Th = VFNMS(LDK(KP500000000), Tg, Tf);
99 Ti = VFNMS(LDK(KP586256827), Th, Te);
100 Tv = VFNMS(LDK(KP726681596), Td, Th);
101 Ts = VFMA(LDK(KP968908795), Th, Td);
102 }
103 TK = VMUL(LDK(KP866025403), VSUB(TG, TH));
104 TI = VADD(TG, TH);
105 TJ = VFNMS(LDK(KP500000000), TI, TF);
106 ST(&(xo[WS(os, 3)]), VFMAI(TK, TJ), ovs, &(xo[WS(os, 1)]));
107 ST(&(xo[0]), VADD(TI, TF), ovs, &(xo[0]));
108 ST(&(xo[WS(os, 6)]), VFNMSI(TK, TJ), ovs, &(xo[0]));
109 {
110 V Tk, To, Tj, Tn, Tl, Tq;
111 Tj = VFNMS(LDK(KP347296355), Ti, Td);
112 Tk = VFNMS(LDK(KP907603734), Tj, Ta);
113 Tn = VFNMS(LDK(KP420276625), Tm, Te);
114 To = VFNMS(LDK(KP826351822), Tn, Th);
115 Tl = VFNMS(LDK(KP939692620), Tk, T5);
116 Tq = VMUL(LDK(KP984807753), VFNMS(LDK(KP879385241), Tp, To));
117 ST(&(xo[WS(os, 7)]), VFNMSI(Tq, Tl), ovs, &(xo[WS(os, 1)]));
118 ST(&(xo[WS(os, 2)]), VFMAI(Tq, Tl), ovs, &(xo[0]));
119 }
120 {
121 V Tx, TD, TB, TE, Ty, TC;
122 {
123 V Tt, Tw, Tz, TA;
124 Tt = VFNMS(LDK(KP673648177), Ts, Tr);
125 Tw = VFMA(LDK(KP898197570), Tv, Tu);
126 Tx = VFNMS(LDK(KP500000000), Tw, Tt);
127 TD = VFMA(LDK(KP852868531), Tw, T5);
128 Tz = VFNMS(LDK(KP898197570), Tv, Tu);
129 TA = VFMA(LDK(KP673648177), Ts, Tr);
130 TB = VFMA(LDK(KP666666666), TA, Tz);
131 TE = VMUL(LDK(KP984807753), VFMA(LDK(KP879385241), Tp, TA));
132 }
133 ST(&(xo[WS(os, 1)]), VFMAI(TE, TD), ovs, &(xo[WS(os, 1)]));
134 ST(&(xo[WS(os, 8)]), VFNMSI(TE, TD), ovs, &(xo[0]));
135 Ty = VFMA(LDK(KP852868531), Tx, T5);
136 TC = VMUL(LDK(KP866025403), VFNMS(LDK(KP852868531), TB, Tp));
137 ST(&(xo[WS(os, 4)]), VFMAI(TC, Ty), ovs, &(xo[0]));
138 ST(&(xo[WS(os, 5)]), VFNMSI(TC, Ty), ovs, &(xo[WS(os, 1)]));
139 }
140 }
141 }
142 VLEAVE();
143 }
144
145 static const kdft_desc desc = { 9, XSIMD_STRING("n1bv_9"), { 12, 4, 34, 0 }, &GENUS, 0, 0, 0, 0 };
146
XSIMD(codelet_n1bv_9)147 void XSIMD(codelet_n1bv_9) (planner *p) { X(kdft_register) (p, n1bv_9, &desc);
148 }
149
150 #else
151
152 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 9 -name n1bv_9 -include dft/simd/n1b.h */
153
154 /*
155 * This function contains 46 FP additions, 26 FP multiplications,
156 * (or, 30 additions, 10 multiplications, 16 fused multiply/add),
157 * 41 stack variables, 14 constants, and 18 memory accesses
158 */
159 #include "dft/simd/n1b.h"
160
n1bv_9(const R * ri,const R * ii,R * ro,R * io,stride is,stride os,INT v,INT ivs,INT ovs)161 static void n1bv_9(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
162 {
163 DVK(KP342020143, +0.342020143325668733044099614682259580763083368);
164 DVK(KP813797681, +0.813797681349373692844693217248393223289101568);
165 DVK(KP939692620, +0.939692620785908384054109277324731469936208134);
166 DVK(KP296198132, +0.296198132726023843175338011893050938967728390);
167 DVK(KP642787609, +0.642787609686539326322643409907263432907559884);
168 DVK(KP663413948, +0.663413948168938396205421319635891297216863310);
169 DVK(KP556670399, +0.556670399226419366452912952047023132968291906);
170 DVK(KP766044443, +0.766044443118978035202392650555416673935832457);
171 DVK(KP984807753, +0.984807753012208059366743024589523013670643252);
172 DVK(KP150383733, +0.150383733180435296639271897612501926072238258);
173 DVK(KP852868531, +0.852868531952443209628250963940074071936020296);
174 DVK(KP173648177, +0.173648177666930348851716626769314796000375677);
175 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
176 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
177 {
178 INT i;
179 const R *xi;
180 R *xo;
181 xi = ii;
182 xo = io;
183 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(18, is), MAKE_VOLATILE_STRIDE(18, os)) {
184 V T5, Ty, Tm, Ti, Tw, Th, Tj, To, Tb, Tv, Ta, Tc, Tn;
185 {
186 V T1, T2, T3, T4;
187 T1 = LD(&(xi[0]), ivs, &(xi[0]));
188 T2 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
189 T3 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
190 T4 = VADD(T2, T3);
191 T5 = VFNMS(LDK(KP500000000), T4, T1);
192 Ty = VADD(T1, T4);
193 Tm = VMUL(LDK(KP866025403), VSUB(T2, T3));
194 }
195 {
196 V Td, Tg, Te, Tf;
197 Td = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
198 Te = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
199 Tf = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
200 Tg = VADD(Te, Tf);
201 Ti = VSUB(Te, Tf);
202 Tw = VADD(Td, Tg);
203 Th = VFNMS(LDK(KP500000000), Tg, Td);
204 Tj = VFNMS(LDK(KP852868531), Ti, VMUL(LDK(KP173648177), Th));
205 To = VFMA(LDK(KP150383733), Ti, VMUL(LDK(KP984807753), Th));
206 }
207 {
208 V T6, T9, T7, T8;
209 T6 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
210 T7 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
211 T8 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
212 T9 = VADD(T7, T8);
213 Tb = VSUB(T7, T8);
214 Tv = VADD(T6, T9);
215 Ta = VFNMS(LDK(KP500000000), T9, T6);
216 Tc = VFNMS(LDK(KP556670399), Tb, VMUL(LDK(KP766044443), Ta));
217 Tn = VFMA(LDK(KP663413948), Tb, VMUL(LDK(KP642787609), Ta));
218 }
219 {
220 V Tx, Tz, TA, Tt, Tu;
221 Tx = VBYI(VMUL(LDK(KP866025403), VSUB(Tv, Tw)));
222 Tz = VADD(Tv, Tw);
223 TA = VFNMS(LDK(KP500000000), Tz, Ty);
224 ST(&(xo[WS(os, 3)]), VADD(Tx, TA), ovs, &(xo[WS(os, 1)]));
225 ST(&(xo[0]), VADD(Ty, Tz), ovs, &(xo[0]));
226 ST(&(xo[WS(os, 6)]), VSUB(TA, Tx), ovs, &(xo[0]));
227 Tt = VFMA(LDK(KP852868531), Tb, VFMA(LDK(KP173648177), Ta, VFMA(LDK(KP296198132), Ti, VFNMS(LDK(KP939692620), Th, T5))));
228 Tu = VBYI(VSUB(VFMA(LDK(KP984807753), Ta, VFMA(LDK(KP813797681), Ti, VFNMS(LDK(KP150383733), Tb, VMUL(LDK(KP342020143), Th)))), Tm));
229 ST(&(xo[WS(os, 7)]), VSUB(Tt, Tu), ovs, &(xo[WS(os, 1)]));
230 ST(&(xo[WS(os, 2)]), VADD(Tt, Tu), ovs, &(xo[0]));
231 {
232 V Tl, Ts, Tq, Tr, Tk, Tp;
233 Tk = VADD(Tc, Tj);
234 Tl = VADD(T5, Tk);
235 Ts = VFMA(LDK(KP866025403), VSUB(To, Tn), VFNMS(LDK(KP500000000), Tk, T5));
236 Tp = VADD(Tn, To);
237 Tq = VBYI(VADD(Tm, Tp));
238 Tr = VBYI(VADD(Tm, VFNMS(LDK(KP500000000), Tp, VMUL(LDK(KP866025403), VSUB(Tc, Tj)))));
239 ST(&(xo[WS(os, 8)]), VSUB(Tl, Tq), ovs, &(xo[0]));
240 ST(&(xo[WS(os, 5)]), VSUB(Ts, Tr), ovs, &(xo[WS(os, 1)]));
241 ST(&(xo[WS(os, 1)]), VADD(Tl, Tq), ovs, &(xo[WS(os, 1)]));
242 ST(&(xo[WS(os, 4)]), VADD(Tr, Ts), ovs, &(xo[0]));
243 }
244 }
245 }
246 }
247 VLEAVE();
248 }
249
250 static const kdft_desc desc = { 9, XSIMD_STRING("n1bv_9"), { 30, 10, 16, 0 }, &GENUS, 0, 0, 0, 0 };
251
XSIMD(codelet_n1bv_9)252 void XSIMD(codelet_n1bv_9) (planner *p) { X(kdft_register) (p, n1bv_9, &desc);
253 }
254
255 #endif
256