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 07:04:53 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 -n 12 -name n2fv_12 -with-ostride 2 -include dft/simd/n2f.h -store-multiple 2 */
29
30 /*
31 * This function contains 48 FP additions, 20 FP multiplications,
32 * (or, 30 additions, 2 multiplications, 18 fused multiply/add),
33 * 33 stack variables, 2 constants, and 30 memory accesses
34 */
35 #include "dft/simd/n2f.h"
36
n2fv_12(const R * ri,const R * ii,R * ro,R * io,stride is,stride os,INT v,INT ivs,INT ovs)37 static void n2fv_12(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
38 {
39 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
40 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
41 {
42 INT i;
43 const R *xi;
44 R *xo;
45 xi = ri;
46 xo = ro;
47 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(24, is), MAKE_VOLATILE_STRIDE(24, os)) {
48 V T5, Ta, TG, TF, TB, Tt, Ti, Tm, TJ, TI, TA, Tp;
49 {
50 V T1, T6, T4, Tr, T9, Ts;
51 T1 = LD(&(xi[0]), ivs, &(xi[0]));
52 T6 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
53 {
54 V T2, T3, T7, T8;
55 T2 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
56 T3 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
57 T4 = VADD(T2, T3);
58 Tr = VSUB(T3, T2);
59 T7 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
60 T8 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
61 T9 = VADD(T7, T8);
62 Ts = VSUB(T8, T7);
63 }
64 T5 = VFNMS(LDK(KP500000000), T4, T1);
65 Ta = VFNMS(LDK(KP500000000), T9, T6);
66 TG = VADD(T6, T9);
67 TF = VADD(T1, T4);
68 TB = VADD(Tr, Ts);
69 Tt = VSUB(Tr, Ts);
70 }
71 {
72 V Tk, Tn, Te, Tl, Th, To;
73 Tk = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
74 Tn = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
75 {
76 V Tc, Td, Tf, Tg;
77 Tc = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
78 Td = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
79 Te = VSUB(Tc, Td);
80 Tl = VADD(Td, Tc);
81 Tf = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
82 Tg = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
83 Th = VSUB(Tf, Tg);
84 To = VADD(Tf, Tg);
85 }
86 Ti = VADD(Te, Th);
87 Tm = VFNMS(LDK(KP500000000), Tl, Tk);
88 TJ = VADD(Tn, To);
89 TI = VADD(Tk, Tl);
90 TA = VSUB(Te, Th);
91 Tp = VFNMS(LDK(KP500000000), To, Tn);
92 }
93 {
94 V TN, TO, TP, TQ, TT, TU;
95 {
96 V TH, TK, TL, TM;
97 TH = VSUB(TF, TG);
98 TK = VSUB(TI, TJ);
99 TN = VFNMSI(TK, TH);
100 STM2(&(xo[18]), TN, ovs, &(xo[2]));
101 TO = VFMAI(TK, TH);
102 STM2(&(xo[6]), TO, ovs, &(xo[2]));
103 TL = VADD(TF, TG);
104 TM = VADD(TI, TJ);
105 TP = VSUB(TL, TM);
106 STM2(&(xo[12]), TP, ovs, &(xo[0]));
107 TQ = VADD(TL, TM);
108 STM2(&(xo[0]), TQ, ovs, &(xo[0]));
109 }
110 {
111 V Tj, Tv, Tu, Tw, Tb, Tq, TR, TS;
112 Tb = VSUB(T5, Ta);
113 Tj = VFMA(LDK(KP866025403), Ti, Tb);
114 Tv = VFNMS(LDK(KP866025403), Ti, Tb);
115 Tq = VSUB(Tm, Tp);
116 Tu = VFNMS(LDK(KP866025403), Tt, Tq);
117 Tw = VFMA(LDK(KP866025403), Tt, Tq);
118 TR = VFNMSI(Tu, Tj);
119 STM2(&(xo[2]), TR, ovs, &(xo[2]));
120 STN2(&(xo[0]), TQ, TR, ovs);
121 TS = VFMAI(Tw, Tv);
122 STM2(&(xo[14]), TS, ovs, &(xo[2]));
123 STN2(&(xo[12]), TP, TS, ovs);
124 TT = VFMAI(Tu, Tj);
125 STM2(&(xo[22]), TT, ovs, &(xo[2]));
126 TU = VFNMSI(Tw, Tv);
127 STM2(&(xo[10]), TU, ovs, &(xo[2]));
128 }
129 {
130 V TC, TE, Tz, TD, Tx, Ty;
131 TC = VMUL(LDK(KP866025403), VSUB(TA, TB));
132 TE = VMUL(LDK(KP866025403), VADD(TB, TA));
133 Tx = VADD(T5, Ta);
134 Ty = VADD(Tm, Tp);
135 Tz = VSUB(Tx, Ty);
136 TD = VADD(Tx, Ty);
137 {
138 V TV, TW, TX, TY;
139 TV = VFMAI(TC, Tz);
140 STM2(&(xo[4]), TV, ovs, &(xo[0]));
141 STN2(&(xo[4]), TV, TO, ovs);
142 TW = VFNMSI(TE, TD);
143 STM2(&(xo[16]), TW, ovs, &(xo[0]));
144 STN2(&(xo[16]), TW, TN, ovs);
145 TX = VFNMSI(TC, Tz);
146 STM2(&(xo[20]), TX, ovs, &(xo[0]));
147 STN2(&(xo[20]), TX, TT, ovs);
148 TY = VFMAI(TE, TD);
149 STM2(&(xo[8]), TY, ovs, &(xo[0]));
150 STN2(&(xo[8]), TY, TU, ovs);
151 }
152 }
153 }
154 }
155 }
156 VLEAVE();
157 }
158
159 static const kdft_desc desc = { 12, XSIMD_STRING("n2fv_12"), { 30, 2, 18, 0 }, &GENUS, 0, 2, 0, 0 };
160
XSIMD(codelet_n2fv_12)161 void XSIMD(codelet_n2fv_12) (planner *p) { X(kdft_register) (p, n2fv_12, &desc);
162 }
163
164 #else
165
166 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 12 -name n2fv_12 -with-ostride 2 -include dft/simd/n2f.h -store-multiple 2 */
167
168 /*
169 * This function contains 48 FP additions, 8 FP multiplications,
170 * (or, 44 additions, 4 multiplications, 4 fused multiply/add),
171 * 33 stack variables, 2 constants, and 30 memory accesses
172 */
173 #include "dft/simd/n2f.h"
174
n2fv_12(const R * ri,const R * ii,R * ro,R * io,stride is,stride os,INT v,INT ivs,INT ovs)175 static void n2fv_12(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
176 {
177 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
178 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
179 {
180 INT i;
181 const R *xi;
182 R *xo;
183 xi = ri;
184 xo = ro;
185 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(24, is), MAKE_VOLATILE_STRIDE(24, os)) {
186 V T5, Ta, TJ, Ty, Tq, Tp, Tg, Tl, TI, TA, Tz, Tu;
187 {
188 V T1, T6, T4, Tw, T9, Tx;
189 T1 = LD(&(xi[0]), ivs, &(xi[0]));
190 T6 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
191 {
192 V T2, T3, T7, T8;
193 T2 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
194 T3 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
195 T4 = VADD(T2, T3);
196 Tw = VSUB(T3, T2);
197 T7 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
198 T8 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
199 T9 = VADD(T7, T8);
200 Tx = VSUB(T8, T7);
201 }
202 T5 = VADD(T1, T4);
203 Ta = VADD(T6, T9);
204 TJ = VADD(Tw, Tx);
205 Ty = VMUL(LDK(KP866025403), VSUB(Tw, Tx));
206 Tq = VFNMS(LDK(KP500000000), T9, T6);
207 Tp = VFNMS(LDK(KP500000000), T4, T1);
208 }
209 {
210 V Tc, Th, Tf, Ts, Tk, Tt;
211 Tc = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
212 Th = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
213 {
214 V Td, Te, Ti, Tj;
215 Td = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
216 Te = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
217 Tf = VADD(Td, Te);
218 Ts = VSUB(Te, Td);
219 Ti = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
220 Tj = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
221 Tk = VADD(Ti, Tj);
222 Tt = VSUB(Tj, Ti);
223 }
224 Tg = VADD(Tc, Tf);
225 Tl = VADD(Th, Tk);
226 TI = VADD(Ts, Tt);
227 TA = VFNMS(LDK(KP500000000), Tk, Th);
228 Tz = VFNMS(LDK(KP500000000), Tf, Tc);
229 Tu = VMUL(LDK(KP866025403), VSUB(Ts, Tt));
230 }
231 {
232 V TN, TO, TP, TQ, TR, TS;
233 {
234 V Tb, Tm, Tn, To;
235 Tb = VSUB(T5, Ta);
236 Tm = VBYI(VSUB(Tg, Tl));
237 TN = VSUB(Tb, Tm);
238 STM2(&(xo[18]), TN, ovs, &(xo[2]));
239 TO = VADD(Tb, Tm);
240 STM2(&(xo[6]), TO, ovs, &(xo[2]));
241 Tn = VADD(T5, Ta);
242 To = VADD(Tg, Tl);
243 TP = VSUB(Tn, To);
244 STM2(&(xo[12]), TP, ovs, &(xo[0]));
245 TQ = VADD(Tn, To);
246 STM2(&(xo[0]), TQ, ovs, &(xo[0]));
247 }
248 {
249 V Tv, TE, TC, TD, Tr, TB, TT, TU;
250 Tr = VSUB(Tp, Tq);
251 Tv = VSUB(Tr, Tu);
252 TE = VADD(Tr, Tu);
253 TB = VSUB(Tz, TA);
254 TC = VBYI(VADD(Ty, TB));
255 TD = VBYI(VSUB(Ty, TB));
256 TR = VSUB(Tv, TC);
257 STM2(&(xo[10]), TR, ovs, &(xo[2]));
258 TS = VSUB(TE, TD);
259 STM2(&(xo[22]), TS, ovs, &(xo[2]));
260 TT = VADD(TC, Tv);
261 STM2(&(xo[14]), TT, ovs, &(xo[2]));
262 STN2(&(xo[12]), TP, TT, ovs);
263 TU = VADD(TD, TE);
264 STM2(&(xo[2]), TU, ovs, &(xo[2]));
265 STN2(&(xo[0]), TQ, TU, ovs);
266 }
267 {
268 V TK, TM, TH, TL, TF, TG;
269 TK = VBYI(VMUL(LDK(KP866025403), VSUB(TI, TJ)));
270 TM = VBYI(VMUL(LDK(KP866025403), VADD(TJ, TI)));
271 TF = VADD(Tp, Tq);
272 TG = VADD(Tz, TA);
273 TH = VSUB(TF, TG);
274 TL = VADD(TF, TG);
275 {
276 V TV, TW, TX, TY;
277 TV = VSUB(TH, TK);
278 STM2(&(xo[20]), TV, ovs, &(xo[0]));
279 STN2(&(xo[20]), TV, TS, ovs);
280 TW = VADD(TL, TM);
281 STM2(&(xo[8]), TW, ovs, &(xo[0]));
282 STN2(&(xo[8]), TW, TR, ovs);
283 TX = VADD(TH, TK);
284 STM2(&(xo[4]), TX, ovs, &(xo[0]));
285 STN2(&(xo[4]), TX, TO, ovs);
286 TY = VSUB(TL, TM);
287 STM2(&(xo[16]), TY, ovs, &(xo[0]));
288 STN2(&(xo[16]), TY, TN, ovs);
289 }
290 }
291 }
292 }
293 }
294 VLEAVE();
295 }
296
297 static const kdft_desc desc = { 12, XSIMD_STRING("n2fv_12"), { 44, 4, 4, 0 }, &GENUS, 0, 2, 0, 0 };
298
XSIMD(codelet_n2fv_12)299 void XSIMD(codelet_n2fv_12) (planner *p) { X(kdft_register) (p, n2fv_12, &desc);
300 }
301
302 #endif
303