← Back to branch summary

~ubuntu-branches/ubuntu/utopic/fftw3/utopic

« back to all changes in this revision

Viewing changes to dft/simd/common/n1fv_20.c

  • Committer: Package Import Robot
  • Author(s): Matthias Klose
  • Date: 2011-12-14 13:21:22 UTC
  • mfrom: (3.1.5 sid)
  • Revision ID: package-import@ubuntu.com-20111214132122-l4avyl2kkr7vq5aj
Tags: 3.3-1ubuntu1
* Merge with Debian; remaining changes:
  - Revert the ARM workaround.

Show diffs side-by-side

added added

removed removed

Lines of Context:
dft/simd/common/n1fv_20.c
 
1
/*
 
2
 * Copyright (c) 2003, 2007-11 Matteo Frigo
 
3
 * Copyright (c) 2003, 2007-11 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., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 
18
 *
 
19
 */
 
20
 
 
21
/* This file was automatically generated --- DO NOT EDIT */
 
22
/* Generated on Wed Jul 27 06:13:34 EDT 2011 */
 
23
 
 
24
#include "codelet-dft.h"
 
25
 
 
26
#ifdef HAVE_FMA
 
27
 
 
28
/* Generated by: ../../../genfft/gen_notw_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -n 20 -name n1fv_20 -include n1f.h */
 
29
 
 
30
/*
 
31
 * This function contains 104 FP additions, 50 FP multiplications,
 
32
 * (or, 58 additions, 4 multiplications, 46 fused multiply/add),
 
33
 * 71 stack variables, 4 constants, and 40 memory accesses
 
34
 */
 
35
#include "n1f.h"
 
36
 
 
37
static void n1fv_20(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
 
38
{
 
39
     DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
 
40
     DVK(KP618033988, +0.618033988749894848204586834365638117720309180);
 
41
     DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
 
42
     DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
 
43
     {
 
44
          INT i;
 
45
          const R *xi;
 
46
          R *xo;
 
47
          xi = ri;
 
48
          xo = ro;
 
49
          for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(is), MAKE_VOLATILE_STRIDE(os)) {
 
50
               V TU, TI, TP, TX, TM, TW, TT, TF;
 
51
               {
 
52
                    V T3, Tm, T1r, T13, Ta, TN, TH, TA, TG, Tt, Th, TO, T1u, T1C, T1n;
 
53
                    V T1a, T1m, T1h, T1x, T1D, TE, Ti;
 
54
                    {
 
55
                         V T1, T2, Tk, Tl;
 
56
                         T1 = LD(&(xi[0]), ivs, &(xi[0]));
 
57
                         T2 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
 
58
                         Tk = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
 
59
                         Tl = LD(&(xi[WS(is, 15)]), ivs, &(xi[WS(is, 1)]));
 
60
                         {
 
61
                              V T14, T6, T1c, Tw, Tn, T1f, Tz, T17, T9, To, Tq, T1b, Td, Tr, Te;
 
62
                              V Tf, T15, Tp;
 
63
                              {
 
64
                                   V Tx, Ty, T7, T8, Tb, Tc;
 
65
                                   {
 
66
                                        V T4, T5, Tu, Tv, T11, T12;
 
67
                                        T4 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
 
68
                                        T5 = LD(&(xi[WS(is, 14)]), ivs, &(xi[0]));
 
69
                                        Tu = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
 
70
                                        Tv = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
 
71
                                        Tx = LD(&(xi[WS(is, 17)]), ivs, &(xi[WS(is, 1)]));
 
72
                                        T3 = VSUB(T1, T2);
 
73
                                        T11 = VADD(T1, T2);
 
74
                                        Tm = VSUB(Tk, Tl);
 
75
                                        T12 = VADD(Tk, Tl);
 
76
                                        T14 = VADD(T4, T5);
 
77
                                        T6 = VSUB(T4, T5);
 
78
                                        T1c = VADD(Tu, Tv);
 
79
                                        Tw = VSUB(Tu, Tv);
 
80
                                        Ty = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
 
81
                                        T7 = LD(&(xi[WS(is, 16)]), ivs, &(xi[0]));
 
82
                                        T8 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
 
83
                                        T1r = VADD(T11, T12);
 
84
                                        T13 = VSUB(T11, T12);
 
85
                                   }
 
86
                                   Tb = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
 
87
                                   Tc = LD(&(xi[WS(is, 18)]), ivs, &(xi[0]));
 
88
                                   Tn = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
 
89
                                   T1f = VADD(Tx, Ty);
 
90
                                   Tz = VSUB(Tx, Ty);
 
91
                                   T17 = VADD(T7, T8);
 
92
                                   T9 = VSUB(T7, T8);
 
93
                                   To = LD(&(xi[WS(is, 19)]), ivs, &(xi[WS(is, 1)]));
 
94
                                   Tq = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
 
95
                                   T1b = VADD(Tb, Tc);
 
96
                                   Td = VSUB(Tb, Tc);
 
97
                                   Tr = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
 
98
                                   Te = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
 
99
                                   Tf = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
 
100
                              }
 
101
                              Ta = VADD(T6, T9);
 
102
                              TN = VSUB(T6, T9);
 
103
                              T15 = VADD(Tn, To);
 
104
                              Tp = VSUB(Tn, To);
 
105
                              TH = VSUB(Tz, Tw);
 
106
                              TA = VADD(Tw, Tz);
 
107
                              {
 
108
                                   V T1d, T1v, T18, Ts, T1e, Tg, T16, T1s;
 
109
                                   T1d = VSUB(T1b, T1c);
 
110
                                   T1v = VADD(T1b, T1c);
 
111
                                   T18 = VADD(Tq, Tr);
 
112
                                   Ts = VSUB(Tq, Tr);
 
113
                                   T1e = VADD(Te, Tf);
 
114
                                   Tg = VSUB(Te, Tf);
 
115
                                   T16 = VSUB(T14, T15);
 
116
                                   T1s = VADD(T14, T15);
 
117
                                   {
 
118
                                        V T1t, T19, T1w, T1g;
 
119
                                        T1t = VADD(T17, T18);
 
120
                                        T19 = VSUB(T17, T18);
 
121
                                        TG = VSUB(Ts, Tp);
 
122
                                        Tt = VADD(Tp, Ts);
 
123
                                        T1w = VADD(T1e, T1f);
 
124
                                        T1g = VSUB(T1e, T1f);
 
125
                                        Th = VADD(Td, Tg);
 
126
                                        TO = VSUB(Td, Tg);
 
127
                                        T1u = VADD(T1s, T1t);
 
128
                                        T1C = VSUB(T1s, T1t);
 
129
                                        T1n = VSUB(T16, T19);
 
130
                                        T1a = VADD(T16, T19);
 
131
                                        T1m = VSUB(T1d, T1g);
 
132
                                        T1h = VADD(T1d, T1g);
 
133
                                        T1x = VADD(T1v, T1w);
 
134
                                        T1D = VSUB(T1v, T1w);
 
135
                                   }
 
136
                              }
 
137
                         }
 
138
                    }
 
139
                    TE = VSUB(Ta, Th);
 
140
                    Ti = VADD(Ta, Th);
 
141
                    {
 
142
                         V TL, T1k, T1A, Tj, TD, T1E, T1G, TK, TC, T1j, T1z, T1i, T1y, TB;
 
143
                         TL = VSUB(TA, Tt);
 
144
                         TB = VADD(Tt, TA);
 
145
                         T1i = VADD(T1a, T1h);
 
146
                         T1k = VSUB(T1a, T1h);
 
147
                         T1y = VADD(T1u, T1x);
 
148
                         T1A = VSUB(T1u, T1x);
 
149
                         Tj = VADD(T3, Ti);
 
150
                         TD = VFNMS(LDK(KP250000000), Ti, T3);
 
151
                         T1E = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1D, T1C));
 
152
                         T1G = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1C, T1D));
 
153
                         TK = VFNMS(LDK(KP250000000), TB, Tm);
 
154
                         TC = VADD(Tm, TB);
 
155
                         T1j = VFNMS(LDK(KP250000000), T1i, T13);
 
156
                         ST(&(xo[0]), VADD(T1r, T1y), ovs, &(xo[0]));
 
157
                         T1z = VFNMS(LDK(KP250000000), T1y, T1r);
 
158
                         ST(&(xo[WS(os, 10)]), VADD(T13, T1i), ovs, &(xo[0]));
 
159
                         {
 
160
                              V T1p, T1l, T1o, T1q, T1F, T1B;
 
161
                              TU = VFNMS(LDK(KP618033988), TG, TH);
 
162
                              TI = VFMA(LDK(KP618033988), TH, TG);
 
163
                              TP = VFMA(LDK(KP618033988), TO, TN);
 
164
                              TX = VFNMS(LDK(KP618033988), TN, TO);
 
165
                              ST(&(xo[WS(os, 15)]), VFMAI(TC, Tj), ovs, &(xo[WS(os, 1)]));
 
166
                              ST(&(xo[WS(os, 5)]), VFNMSI(TC, Tj), ovs, &(xo[WS(os, 1)]));
 
167
                              T1p = VFMA(LDK(KP559016994), T1k, T1j);
 
168
                              T1l = VFNMS(LDK(KP559016994), T1k, T1j);
 
169
                              T1o = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1n, T1m));
 
170
                              T1q = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1m, T1n));
 
171
                              T1F = VFNMS(LDK(KP559016994), T1A, T1z);
 
172
                              T1B = VFMA(LDK(KP559016994), T1A, T1z);
 
173
                              ST(&(xo[WS(os, 14)]), VFMAI(T1q, T1p), ovs, &(xo[0]));
 
174
                              ST(&(xo[WS(os, 6)]), VFNMSI(T1q, T1p), ovs, &(xo[0]));
 
175
                              ST(&(xo[WS(os, 18)]), VFNMSI(T1o, T1l), ovs, &(xo[0]));
 
176
                              ST(&(xo[WS(os, 2)]), VFMAI(T1o, T1l), ovs, &(xo[0]));
 
177
                              ST(&(xo[WS(os, 16)]), VFNMSI(T1E, T1B), ovs, &(xo[0]));
 
178
                              ST(&(xo[WS(os, 4)]), VFMAI(T1E, T1B), ovs, &(xo[0]));
 
179
                              ST(&(xo[WS(os, 12)]), VFMAI(T1G, T1F), ovs, &(xo[0]));
 
180
                              ST(&(xo[WS(os, 8)]), VFNMSI(T1G, T1F), ovs, &(xo[0]));
 
181
                              TM = VFNMS(LDK(KP559016994), TL, TK);
 
182
                              TW = VFMA(LDK(KP559016994), TL, TK);
 
183
                              TT = VFNMS(LDK(KP559016994), TE, TD);
 
184
                              TF = VFMA(LDK(KP559016994), TE, TD);
 
185
                         }
 
186
                    }
 
187
               }
 
188
               {
 
189
                    V T10, TY, TQ, TS, TJ, TR, TZ, TV;
 
190
                    T10 = VFMA(LDK(KP951056516), TX, TW);
 
191
                    TY = VFNMS(LDK(KP951056516), TX, TW);
 
192
                    TQ = VFMA(LDK(KP951056516), TP, TM);
 
193
                    TS = VFNMS(LDK(KP951056516), TP, TM);
 
194
                    TJ = VFMA(LDK(KP951056516), TI, TF);
 
195
                    TR = VFNMS(LDK(KP951056516), TI, TF);
 
196
                    TZ = VFMA(LDK(KP951056516), TU, TT);
 
197
                    TV = VFNMS(LDK(KP951056516), TU, TT);
 
198
                    ST(&(xo[WS(os, 11)]), VFMAI(TS, TR), ovs, &(xo[WS(os, 1)]));
 
199
                    ST(&(xo[WS(os, 9)]), VFNMSI(TS, TR), ovs, &(xo[WS(os, 1)]));
 
200
                    ST(&(xo[WS(os, 19)]), VFMAI(TQ, TJ), ovs, &(xo[WS(os, 1)]));
 
201
                    ST(&(xo[WS(os, 1)]), VFNMSI(TQ, TJ), ovs, &(xo[WS(os, 1)]));
 
202
                    ST(&(xo[WS(os, 3)]), VFMAI(TY, TV), ovs, &(xo[WS(os, 1)]));
 
203
                    ST(&(xo[WS(os, 17)]), VFNMSI(TY, TV), ovs, &(xo[WS(os, 1)]));
 
204
                    ST(&(xo[WS(os, 7)]), VFMAI(T10, TZ), ovs, &(xo[WS(os, 1)]));
 
205
                    ST(&(xo[WS(os, 13)]), VFNMSI(T10, TZ), ovs, &(xo[WS(os, 1)]));
 
206
               }
 
207
          }
 
208
     }
 
209
     VLEAVE();
 
210
}
 
211
 
 
212
static const kdft_desc desc = { 20, XSIMD_STRING("n1fv_20"), {58, 4, 46, 0}, &GENUS, 0, 0, 0, 0 };
 
213
 
 
214
void XSIMD(codelet_n1fv_20) (planner *p) {
 
215
     X(kdft_register) (p, n1fv_20, &desc);
 
216
}
 
217
 
 
218
#else                           /* HAVE_FMA */
 
219
 
 
220
/* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 20 -name n1fv_20 -include n1f.h */
 
221
 
 
222
/*
 
223
 * This function contains 104 FP additions, 24 FP multiplications,
 
224
 * (or, 92 additions, 12 multiplications, 12 fused multiply/add),
 
225
 * 53 stack variables, 4 constants, and 40 memory accesses
 
226
 */
 
227
#include "n1f.h"
 
228
 
 
229
static void n1fv_20(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
 
230
{
 
231
     DVK(KP587785252, +0.587785252292473129168705954639072768597652438);
 
232
     DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
 
233
     DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
 
234
     DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
 
235
     {
 
236
          INT i;
 
237
          const R *xi;
 
238
          R *xo;
 
239
          xi = ri;
 
240
          xo = ro;
 
241
          for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(is), MAKE_VOLATILE_STRIDE(os)) {
 
242
               V T3, T1B, Tm, T1i, TG, TN, TO, TH, T13, T16, T1k, T1u, T1v, T1z, T1r;
 
243
               V T1s, T1y, T1a, T1d, T1j, Ti, TD, TB, TL, Tj, TC;
 
244
               {
 
245
                    V T1, T2, T1g, Tk, Tl, T1h;
 
246
                    T1 = LD(&(xi[0]), ivs, &(xi[0]));
 
247
                    T2 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
 
248
                    T1g = VADD(T1, T2);
 
249
                    Tk = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
 
250
                    Tl = LD(&(xi[WS(is, 15)]), ivs, &(xi[WS(is, 1)]));
 
251
                    T1h = VADD(Tk, Tl);
 
252
                    T3 = VSUB(T1, T2);
 
253
                    T1B = VADD(T1g, T1h);
 
254
                    Tm = VSUB(Tk, Tl);
 
255
                    T1i = VSUB(T1g, T1h);
 
256
               }
 
257
               {
 
258
                    V T6, T18, Tw, T12, Tz, T15, T9, T1b, Td, T11, Tp, T19, Ts, T1c, Tg;
 
259
                    V T14;
 
260
                    {
 
261
                         V T4, T5, Tu, Tv;
 
262
                         T4 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
 
263
                         T5 = LD(&(xi[WS(is, 14)]), ivs, &(xi[0]));
 
264
                         T6 = VSUB(T4, T5);
 
265
                         T18 = VADD(T4, T5);
 
266
                         Tu = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
 
267
                         Tv = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
 
268
                         Tw = VSUB(Tu, Tv);
 
269
                         T12 = VADD(Tu, Tv);
 
270
                    }
 
271
                    {
 
272
                         V Tx, Ty, T7, T8;
 
273
                         Tx = LD(&(xi[WS(is, 17)]), ivs, &(xi[WS(is, 1)]));
 
274
                         Ty = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
 
275
                         Tz = VSUB(Tx, Ty);
 
276
                         T15 = VADD(Tx, Ty);
 
277
                         T7 = LD(&(xi[WS(is, 16)]), ivs, &(xi[0]));
 
278
                         T8 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
 
279
                         T9 = VSUB(T7, T8);
 
280
                         T1b = VADD(T7, T8);
 
281
                    }
 
282
                    {
 
283
                         V Tb, Tc, Tn, To;
 
284
                         Tb = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
 
285
                         Tc = LD(&(xi[WS(is, 18)]), ivs, &(xi[0]));
 
286
                         Td = VSUB(Tb, Tc);
 
287
                         T11 = VADD(Tb, Tc);
 
288
                         Tn = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
 
289
                         To = LD(&(xi[WS(is, 19)]), ivs, &(xi[WS(is, 1)]));
 
290
                         Tp = VSUB(Tn, To);
 
291
                         T19 = VADD(Tn, To);
 
292
                    }
 
293
                    {
 
294
                         V Tq, Tr, Te, Tf;
 
295
                         Tq = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
 
296
                         Tr = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
 
297
                         Ts = VSUB(Tq, Tr);
 
298
                         T1c = VADD(Tq, Tr);
 
299
                         Te = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
 
300
                         Tf = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
 
301
                         Tg = VSUB(Te, Tf);
 
302
                         T14 = VADD(Te, Tf);
 
303
                    }
 
304
                    TG = VSUB(Ts, Tp);
 
305
                    TN = VSUB(T6, T9);
 
306
                    TO = VSUB(Td, Tg);
 
307
                    TH = VSUB(Tz, Tw);
 
308
                    T13 = VSUB(T11, T12);
 
309
                    T16 = VSUB(T14, T15);
 
310
                    T1k = VADD(T13, T16);
 
311
                    T1u = VADD(T11, T12);
 
312
                    T1v = VADD(T14, T15);
 
313
                    T1z = VADD(T1u, T1v);
 
314
                    T1r = VADD(T18, T19);
 
315
                    T1s = VADD(T1b, T1c);
 
316
                    T1y = VADD(T1r, T1s);
 
317
                    T1a = VSUB(T18, T19);
 
318
                    T1d = VSUB(T1b, T1c);
 
319
                    T1j = VADD(T1a, T1d);
 
320
                    {
 
321
                         V Ta, Th, Tt, TA;
 
322
                         Ta = VADD(T6, T9);
 
323
                         Th = VADD(Td, Tg);
 
324
                         Ti = VADD(Ta, Th);
 
325
                         TD = VMUL(LDK(KP559016994), VSUB(Ta, Th));
 
326
                         Tt = VADD(Tp, Ts);
 
327
                         TA = VADD(Tw, Tz);
 
328
                         TB = VADD(Tt, TA);
 
329
                         TL = VMUL(LDK(KP559016994), VSUB(TA, Tt));
 
330
                    }
 
331
               }
 
332
               Tj = VADD(T3, Ti);
 
333
               TC = VBYI(VADD(Tm, TB));
 
334
               ST(&(xo[WS(os, 5)]), VSUB(Tj, TC), ovs, &(xo[WS(os, 1)]));
 
335
               ST(&(xo[WS(os, 15)]), VADD(Tj, TC), ovs, &(xo[WS(os, 1)]));
 
336
               {
 
337
                    V T1A, T1C, T1D, T1x, T1G, T1t, T1w, T1F, T1E;
 
338
                    T1A = VMUL(LDK(KP559016994), VSUB(T1y, T1z));
 
339
                    T1C = VADD(T1y, T1z);
 
340
                    T1D = VFNMS(LDK(KP250000000), T1C, T1B);
 
341
                    T1t = VSUB(T1r, T1s);
 
342
                    T1w = VSUB(T1u, T1v);
 
343
                    T1x = VBYI(VFMA(LDK(KP951056516), T1t, VMUL(LDK(KP587785252), T1w)));
 
344
                    T1G = VBYI(VFNMS(LDK(KP587785252), T1t, VMUL(LDK(KP951056516), T1w)));
 
345
                    ST(&(xo[0]), VADD(T1B, T1C), ovs, &(xo[0]));
 
346
                    T1F = VSUB(T1D, T1A);
 
347
                    ST(&(xo[WS(os, 8)]), VSUB(T1F, T1G), ovs, &(xo[0]));
 
348
                    ST(&(xo[WS(os, 12)]), VADD(T1G, T1F), ovs, &(xo[0]));
 
349
                    T1E = VADD(T1A, T1D);
 
350
                    ST(&(xo[WS(os, 4)]), VADD(T1x, T1E), ovs, &(xo[0]));
 
351
                    ST(&(xo[WS(os, 16)]), VSUB(T1E, T1x), ovs, &(xo[0]));
 
352
               }
 
353
               {
 
354
                    V T1n, T1l, T1m, T1f, T1q, T17, T1e, T1p, T1o;
 
355
                    T1n = VMUL(LDK(KP559016994), VSUB(T1j, T1k));
 
356
                    T1l = VADD(T1j, T1k);
 
357
                    T1m = VFNMS(LDK(KP250000000), T1l, T1i);
 
358
                    T17 = VSUB(T13, T16);
 
359
                    T1e = VSUB(T1a, T1d);
 
360
                    T1f = VBYI(VFNMS(LDK(KP587785252), T1e, VMUL(LDK(KP951056516), T17)));
 
361
                    T1q = VBYI(VFMA(LDK(KP951056516), T1e, VMUL(LDK(KP587785252), T17)));
 
362
                    ST(&(xo[WS(os, 10)]), VADD(T1i, T1l), ovs, &(xo[0]));
 
363
                    T1p = VADD(T1n, T1m);
 
364
                    ST(&(xo[WS(os, 6)]), VSUB(T1p, T1q), ovs, &(xo[0]));
 
365
                    ST(&(xo[WS(os, 14)]), VADD(T1q, T1p), ovs, &(xo[0]));
 
366
                    T1o = VSUB(T1m, T1n);
 
367
                    ST(&(xo[WS(os, 2)]), VADD(T1f, T1o), ovs, &(xo[0]));
 
368
                    ST(&(xo[WS(os, 18)]), VSUB(T1o, T1f), ovs, &(xo[0]));
 
369
               }
 
370
               {
 
371
                    V TI, TP, TX, TU, TM, TW, TF, TT, TK, TE;
 
372
                    TI = VFMA(LDK(KP951056516), TG, VMUL(LDK(KP587785252), TH));
 
373
                    TP = VFMA(LDK(KP951056516), TN, VMUL(LDK(KP587785252), TO));
 
374
                    TX = VFNMS(LDK(KP587785252), TN, VMUL(LDK(KP951056516), TO));
 
375
                    TU = VFNMS(LDK(KP587785252), TG, VMUL(LDK(KP951056516), TH));
 
376
                    TK = VFMS(LDK(KP250000000), TB, Tm);
 
377
                    TM = VADD(TK, TL);
 
378
                    TW = VSUB(TL, TK);
 
379
                    TE = VFNMS(LDK(KP250000000), Ti, T3);
 
380
                    TF = VADD(TD, TE);
 
381
                    TT = VSUB(TE, TD);
 
382
                    {
 
383
                         V TJ, TQ, TZ, T10;
 
384
                         TJ = VADD(TF, TI);
 
385
                         TQ = VBYI(VSUB(TM, TP));
 
386
                         ST(&(xo[WS(os, 19)]), VSUB(TJ, TQ), ovs, &(xo[WS(os, 1)]));
 
387
                         ST(&(xo[WS(os, 1)]), VADD(TJ, TQ), ovs, &(xo[WS(os, 1)]));
 
388
                         TZ = VADD(TT, TU);
 
389
                         T10 = VBYI(VADD(TX, TW));
 
390
                         ST(&(xo[WS(os, 13)]), VSUB(TZ, T10), ovs, &(xo[WS(os, 1)]));
 
391
                         ST(&(xo[WS(os, 7)]), VADD(TZ, T10), ovs, &(xo[WS(os, 1)]));
 
392
                    }
 
393
                    {
 
394
                         V TR, TS, TV, TY;
 
395
                         TR = VSUB(TF, TI);
 
396
                         TS = VBYI(VADD(TP, TM));
 
397
                         ST(&(xo[WS(os, 11)]), VSUB(TR, TS), ovs, &(xo[WS(os, 1)]));
 
398
                         ST(&(xo[WS(os, 9)]), VADD(TR, TS), ovs, &(xo[WS(os, 1)]));
 
399
                         TV = VSUB(TT, TU);
 
400
                         TY = VBYI(VSUB(TW, TX));
 
401
                         ST(&(xo[WS(os, 17)]), VSUB(TV, TY), ovs, &(xo[WS(os, 1)]));
 
402
                         ST(&(xo[WS(os, 3)]), VADD(TV, TY), ovs, &(xo[WS(os, 1)]));
 
403
                    }
 
404
               }
 
405
          }
 
406
     }
 
407
     VLEAVE();
 
408
}
 
409
 
 
410
static const kdft_desc desc = { 20, XSIMD_STRING("n1fv_20"), {92, 12, 12, 0}, &GENUS, 0, 0, 0, 0 };
 
411
 
 
412
void XSIMD(codelet_n1fv_20) (planner *p) {
 
413
     X(kdft_register) (p, n1fv_20, &desc);
 
414
}
 
415
 
 
416
#endif                          /* HAVE_FMA */