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 07:04:56 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 10 -name n2bv_10 -with-ostride 2 -include dft/simd/n2b.h -store-multiple 2 */
29
30 /*
31 * This function contains 42 FP additions, 22 FP multiplications,
32 * (or, 24 additions, 4 multiplications, 18 fused multiply/add),
33 * 36 stack variables, 4 constants, and 25 memory accesses
34 */
35 #include "dft/simd/n2b.h"
36
n2bv_10(const R * ri,const R * ii,R * ro,R * io,stride is,stride os,INT v,INT ivs,INT ovs)37 static void n2bv_10(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(KP250000000, +0.250000000000000000000000000000000000000000000);
41 DVK(KP618033988, +0.618033988749894848204586834365638117720309180);
42 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
43 {
44 INT i;
45 const R *xi;
46 R *xo;
47 xi = ii;
48 xo = io;
49 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(20, is), MAKE_VOLATILE_STRIDE(20, os)) {
50 V T3, Tr, Tm, Tn, TD, TC, Tu, Tx, Ty, Ta, Th, Ti, T1, T2;
51 T1 = LD(&(xi[0]), ivs, &(xi[0]));
52 T2 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
53 T3 = VSUB(T1, T2);
54 Tr = VADD(T1, T2);
55 {
56 V T6, Ts, Tg, Tw, T9, Tt, Td, Tv;
57 {
58 V T4, T5, Te, Tf;
59 T4 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
60 T5 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
61 T6 = VSUB(T4, T5);
62 Ts = VADD(T4, T5);
63 Te = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
64 Tf = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
65 Tg = VSUB(Te, Tf);
66 Tw = VADD(Te, Tf);
67 }
68 {
69 V T7, T8, Tb, Tc;
70 T7 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
71 T8 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
72 T9 = VSUB(T7, T8);
73 Tt = VADD(T7, T8);
74 Tb = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
75 Tc = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
76 Td = VSUB(Tb, Tc);
77 Tv = VADD(Tb, Tc);
78 }
79 Tm = VSUB(T6, T9);
80 Tn = VSUB(Td, Tg);
81 TD = VSUB(Ts, Tt);
82 TC = VSUB(Tv, Tw);
83 Tu = VADD(Ts, Tt);
84 Tx = VADD(Tv, Tw);
85 Ty = VADD(Tu, Tx);
86 Ta = VADD(T6, T9);
87 Th = VADD(Td, Tg);
88 Ti = VADD(Ta, Th);
89 }
90 {
91 V TH, TI, TK, TL, TM;
92 TH = VADD(T3, Ti);
93 STM2(&(xo[10]), TH, ovs, &(xo[2]));
94 TI = VADD(Tr, Ty);
95 STM2(&(xo[0]), TI, ovs, &(xo[0]));
96 {
97 V To, Tq, Tl, Tp, Tj, Tk, TJ;
98 To = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), Tn, Tm));
99 Tq = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), Tm, Tn));
100 Tj = VFNMS(LDK(KP250000000), Ti, T3);
101 Tk = VSUB(Ta, Th);
102 Tl = VFMA(LDK(KP559016994), Tk, Tj);
103 Tp = VFNMS(LDK(KP559016994), Tk, Tj);
104 TJ = VFMAI(To, Tl);
105 STM2(&(xo[2]), TJ, ovs, &(xo[2]));
106 STN2(&(xo[0]), TI, TJ, ovs);
107 TK = VFNMSI(Tq, Tp);
108 STM2(&(xo[14]), TK, ovs, &(xo[2]));
109 TL = VFNMSI(To, Tl);
110 STM2(&(xo[18]), TL, ovs, &(xo[2]));
111 TM = VFMAI(Tq, Tp);
112 STM2(&(xo[6]), TM, ovs, &(xo[2]));
113 }
114 {
115 V TE, TG, TB, TF, Tz, TA;
116 TE = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), TD, TC));
117 TG = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), TC, TD));
118 Tz = VFNMS(LDK(KP250000000), Ty, Tr);
119 TA = VSUB(Tu, Tx);
120 TB = VFNMS(LDK(KP559016994), TA, Tz);
121 TF = VFMA(LDK(KP559016994), TA, Tz);
122 {
123 V TN, TO, TP, TQ;
124 TN = VFNMSI(TE, TB);
125 STM2(&(xo[4]), TN, ovs, &(xo[0]));
126 STN2(&(xo[4]), TN, TM, ovs);
127 TO = VFMAI(TG, TF);
128 STM2(&(xo[12]), TO, ovs, &(xo[0]));
129 STN2(&(xo[12]), TO, TK, ovs);
130 TP = VFMAI(TE, TB);
131 STM2(&(xo[16]), TP, ovs, &(xo[0]));
132 STN2(&(xo[16]), TP, TL, ovs);
133 TQ = VFNMSI(TG, TF);
134 STM2(&(xo[8]), TQ, ovs, &(xo[0]));
135 STN2(&(xo[8]), TQ, TH, ovs);
136 }
137 }
138 }
139 }
140 }
141 VLEAVE();
142 }
143
144 static const kdft_desc desc = { 10, XSIMD_STRING("n2bv_10"), { 24, 4, 18, 0 }, &GENUS, 0, 2, 0, 0 };
145
XSIMD(codelet_n2bv_10)146 void XSIMD(codelet_n2bv_10) (planner *p) { X(kdft_register) (p, n2bv_10, &desc);
147 }
148
149 #else
150
151 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 10 -name n2bv_10 -with-ostride 2 -include dft/simd/n2b.h -store-multiple 2 */
152
153 /*
154 * This function contains 42 FP additions, 12 FP multiplications,
155 * (or, 36 additions, 6 multiplications, 6 fused multiply/add),
156 * 36 stack variables, 4 constants, and 25 memory accesses
157 */
158 #include "dft/simd/n2b.h"
159
n2bv_10(const R * ri,const R * ii,R * ro,R * io,stride is,stride os,INT v,INT ivs,INT ovs)160 static void n2bv_10(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
161 {
162 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
163 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
164 DVK(KP587785252, +0.587785252292473129168705954639072768597652438);
165 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
166 {
167 INT i;
168 const R *xi;
169 R *xo;
170 xi = ii;
171 xo = io;
172 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(20, is), MAKE_VOLATILE_STRIDE(20, os)) {
173 V Tl, Ty, T7, Te, Tw, Tt, Tz, TA, TB, Tg, Th, Tm, Tj, Tk;
174 Tj = LD(&(xi[0]), ivs, &(xi[0]));
175 Tk = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
176 Tl = VSUB(Tj, Tk);
177 Ty = VADD(Tj, Tk);
178 {
179 V T3, Tr, Td, Tv, T6, Ts, Ta, Tu;
180 {
181 V T1, T2, Tb, Tc;
182 T1 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
183 T2 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
184 T3 = VSUB(T1, T2);
185 Tr = VADD(T1, T2);
186 Tb = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
187 Tc = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
188 Td = VSUB(Tb, Tc);
189 Tv = VADD(Tb, Tc);
190 }
191 {
192 V T4, T5, T8, T9;
193 T4 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
194 T5 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
195 T6 = VSUB(T4, T5);
196 Ts = VADD(T4, T5);
197 T8 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
198 T9 = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
199 Ta = VSUB(T8, T9);
200 Tu = VADD(T8, T9);
201 }
202 T7 = VSUB(T3, T6);
203 Te = VSUB(Ta, Td);
204 Tw = VSUB(Tu, Tv);
205 Tt = VSUB(Tr, Ts);
206 Tz = VADD(Tr, Ts);
207 TA = VADD(Tu, Tv);
208 TB = VADD(Tz, TA);
209 Tg = VADD(T3, T6);
210 Th = VADD(Ta, Td);
211 Tm = VADD(Tg, Th);
212 }
213 {
214 V TH, TI, TK, TL, TM;
215 TH = VADD(Tl, Tm);
216 STM2(&(xo[10]), TH, ovs, &(xo[2]));
217 TI = VADD(Ty, TB);
218 STM2(&(xo[0]), TI, ovs, &(xo[0]));
219 {
220 V Tf, Tq, To, Tp, Ti, Tn, TJ;
221 Tf = VBYI(VFMA(LDK(KP951056516), T7, VMUL(LDK(KP587785252), Te)));
222 Tq = VBYI(VFNMS(LDK(KP951056516), Te, VMUL(LDK(KP587785252), T7)));
223 Ti = VMUL(LDK(KP559016994), VSUB(Tg, Th));
224 Tn = VFNMS(LDK(KP250000000), Tm, Tl);
225 To = VADD(Ti, Tn);
226 Tp = VSUB(Tn, Ti);
227 TJ = VADD(Tf, To);
228 STM2(&(xo[2]), TJ, ovs, &(xo[2]));
229 STN2(&(xo[0]), TI, TJ, ovs);
230 TK = VADD(Tq, Tp);
231 STM2(&(xo[14]), TK, ovs, &(xo[2]));
232 TL = VSUB(To, Tf);
233 STM2(&(xo[18]), TL, ovs, &(xo[2]));
234 TM = VSUB(Tp, Tq);
235 STM2(&(xo[6]), TM, ovs, &(xo[2]));
236 }
237 {
238 V Tx, TG, TE, TF, TC, TD;
239 Tx = VBYI(VFNMS(LDK(KP951056516), Tw, VMUL(LDK(KP587785252), Tt)));
240 TG = VBYI(VFMA(LDK(KP951056516), Tt, VMUL(LDK(KP587785252), Tw)));
241 TC = VFNMS(LDK(KP250000000), TB, Ty);
242 TD = VMUL(LDK(KP559016994), VSUB(Tz, TA));
243 TE = VSUB(TC, TD);
244 TF = VADD(TD, TC);
245 {
246 V TN, TO, TP, TQ;
247 TN = VADD(Tx, TE);
248 STM2(&(xo[4]), TN, ovs, &(xo[0]));
249 STN2(&(xo[4]), TN, TM, ovs);
250 TO = VADD(TG, TF);
251 STM2(&(xo[12]), TO, ovs, &(xo[0]));
252 STN2(&(xo[12]), TO, TK, ovs);
253 TP = VSUB(TE, Tx);
254 STM2(&(xo[16]), TP, ovs, &(xo[0]));
255 STN2(&(xo[16]), TP, TL, ovs);
256 TQ = VSUB(TF, TG);
257 STM2(&(xo[8]), TQ, ovs, &(xo[0]));
258 STN2(&(xo[8]), TQ, TH, ovs);
259 }
260 }
261 }
262 }
263 }
264 VLEAVE();
265 }
266
267 static const kdft_desc desc = { 10, XSIMD_STRING("n2bv_10"), { 36, 6, 6, 0 }, &GENUS, 0, 2, 0, 0 };
268
XSIMD(codelet_n2bv_10)269 void XSIMD(codelet_n2bv_10) (planner *p) { X(kdft_register) (p, n2bv_10, &desc);
270 }
271
272 #endif
273