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 07:04:45 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 12 -name n1bv_12 -include dft/simd/n1b.h */
29
30 /*
31 * This function contains 48 FP additions, 20 FP multiplications,
32 * (or, 30 additions, 2 multiplications, 18 fused multiply/add),
33 * 27 stack variables, 2 constants, and 24 memory accesses
34 */
35 #include "dft/simd/n1b.h"
36
n1bv_12(const R * ri,const R * ii,R * ro,R * io,stride is,stride os,INT v,INT ivs,INT ovs)37 static void n1bv_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 = ii;
46 xo = io;
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, TJ, TB, Tq, Tp, Tg, Tl, TG, Ty, Tt, Ts;
49 {
50 V T1, T6, T4, Tz, T9, TA;
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 Tz = VSUB(T2, T3);
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 TA = VSUB(T7, T8);
63 }
64 T5 = VADD(T1, T4);
65 Ta = VADD(T6, T9);
66 TJ = VSUB(Tz, TA);
67 TB = VADD(Tz, TA);
68 Tq = VFNMS(LDK(KP500000000), T9, T6);
69 Tp = VFNMS(LDK(KP500000000), T4, T1);
70 }
71 {
72 V Tc, Th, Tf, Tw, Tk, Tx;
73 Tc = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
74 Th = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
75 {
76 V Td, Te, Ti, Tj;
77 Td = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
78 Te = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
79 Tf = VADD(Td, Te);
80 Tw = VSUB(Td, Te);
81 Ti = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
82 Tj = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
83 Tk = VADD(Ti, Tj);
84 Tx = VSUB(Tj, Ti);
85 }
86 Tg = VADD(Tc, Tf);
87 Tl = VADD(Th, Tk);
88 TG = VADD(Tw, Tx);
89 Ty = VSUB(Tw, Tx);
90 Tt = VFNMS(LDK(KP500000000), Tk, Th);
91 Ts = VFNMS(LDK(KP500000000), Tf, Tc);
92 }
93 {
94 V Tb, Tm, Tn, To;
95 Tb = VSUB(T5, Ta);
96 Tm = VSUB(Tg, Tl);
97 ST(&(xo[WS(os, 3)]), VFNMSI(Tm, Tb), ovs, &(xo[WS(os, 1)]));
98 ST(&(xo[WS(os, 9)]), VFMAI(Tm, Tb), ovs, &(xo[WS(os, 1)]));
99 Tn = VADD(T5, Ta);
100 To = VADD(Tg, Tl);
101 ST(&(xo[WS(os, 6)]), VSUB(Tn, To), ovs, &(xo[0]));
102 ST(&(xo[0]), VADD(Tn, To), ovs, &(xo[0]));
103 }
104 {
105 V TC, TE, Tv, TD, Tr, Tu;
106 TC = VMUL(LDK(KP866025403), VSUB(Ty, TB));
107 TE = VMUL(LDK(KP866025403), VADD(TB, Ty));
108 Tr = VADD(Tp, Tq);
109 Tu = VADD(Ts, Tt);
110 Tv = VSUB(Tr, Tu);
111 TD = VADD(Tr, Tu);
112 ST(&(xo[WS(os, 10)]), VFNMSI(TC, Tv), ovs, &(xo[0]));
113 ST(&(xo[WS(os, 4)]), VFMAI(TE, TD), ovs, &(xo[0]));
114 ST(&(xo[WS(os, 2)]), VFMAI(TC, Tv), ovs, &(xo[0]));
115 ST(&(xo[WS(os, 8)]), VFNMSI(TE, TD), ovs, &(xo[0]));
116 }
117 {
118 V TH, TL, TK, TM, TF, TI;
119 TF = VSUB(Tp, Tq);
120 TH = VFNMS(LDK(KP866025403), TG, TF);
121 TL = VFMA(LDK(KP866025403), TG, TF);
122 TI = VSUB(Ts, Tt);
123 TK = VFMA(LDK(KP866025403), TJ, TI);
124 TM = VFNMS(LDK(KP866025403), TJ, TI);
125 ST(&(xo[WS(os, 1)]), VFMAI(TK, TH), ovs, &(xo[WS(os, 1)]));
126 ST(&(xo[WS(os, 7)]), VFNMSI(TM, TL), ovs, &(xo[WS(os, 1)]));
127 ST(&(xo[WS(os, 11)]), VFNMSI(TK, TH), ovs, &(xo[WS(os, 1)]));
128 ST(&(xo[WS(os, 5)]), VFMAI(TM, TL), ovs, &(xo[WS(os, 1)]));
129 }
130 }
131 }
132 VLEAVE();
133 }
134
135 static const kdft_desc desc = { 12, XSIMD_STRING("n1bv_12"), { 30, 2, 18, 0 }, &GENUS, 0, 0, 0, 0 };
136
XSIMD(codelet_n1bv_12)137 void XSIMD(codelet_n1bv_12) (planner *p) { X(kdft_register) (p, n1bv_12, &desc);
138 }
139
140 #else
141
142 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 12 -name n1bv_12 -include dft/simd/n1b.h */
143
144 /*
145 * This function contains 48 FP additions, 8 FP multiplications,
146 * (or, 44 additions, 4 multiplications, 4 fused multiply/add),
147 * 27 stack variables, 2 constants, and 24 memory accesses
148 */
149 #include "dft/simd/n1b.h"
150
n1bv_12(const R * ri,const R * ii,R * ro,R * io,stride is,stride os,INT v,INT ivs,INT ovs)151 static void n1bv_12(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
152 {
153 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
154 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
155 {
156 INT i;
157 const R *xi;
158 R *xo;
159 xi = ii;
160 xo = io;
161 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)) {
162 V T5, Ta, TG, TF, Ty, Tm, Ti, Tp, TJ, TI, Tx, Ts;
163 {
164 V T1, T6, T4, Tk, T9, Tl;
165 T1 = LD(&(xi[0]), ivs, &(xi[0]));
166 T6 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
167 {
168 V T2, T3, T7, T8;
169 T2 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
170 T3 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
171 T4 = VADD(T2, T3);
172 Tk = VSUB(T2, T3);
173 T7 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
174 T8 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
175 T9 = VADD(T7, T8);
176 Tl = VSUB(T7, T8);
177 }
178 T5 = VFNMS(LDK(KP500000000), T4, T1);
179 Ta = VFNMS(LDK(KP500000000), T9, T6);
180 TG = VADD(T6, T9);
181 TF = VADD(T1, T4);
182 Ty = VADD(Tk, Tl);
183 Tm = VMUL(LDK(KP866025403), VSUB(Tk, Tl));
184 }
185 {
186 V Tn, Tq, Te, To, Th, Tr;
187 Tn = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
188 Tq = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
189 {
190 V Tc, Td, Tf, Tg;
191 Tc = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
192 Td = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
193 Te = VSUB(Tc, Td);
194 To = VADD(Tc, Td);
195 Tf = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
196 Tg = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
197 Th = VSUB(Tf, Tg);
198 Tr = VADD(Tf, Tg);
199 }
200 Ti = VMUL(LDK(KP866025403), VSUB(Te, Th));
201 Tp = VFNMS(LDK(KP500000000), To, Tn);
202 TJ = VADD(Tq, Tr);
203 TI = VADD(Tn, To);
204 Tx = VADD(Te, Th);
205 Ts = VFNMS(LDK(KP500000000), Tr, Tq);
206 }
207 {
208 V TH, TK, TL, TM;
209 TH = VSUB(TF, TG);
210 TK = VBYI(VSUB(TI, TJ));
211 ST(&(xo[WS(os, 3)]), VSUB(TH, TK), ovs, &(xo[WS(os, 1)]));
212 ST(&(xo[WS(os, 9)]), VADD(TH, TK), ovs, &(xo[WS(os, 1)]));
213 TL = VADD(TF, TG);
214 TM = VADD(TI, TJ);
215 ST(&(xo[WS(os, 6)]), VSUB(TL, TM), ovs, &(xo[0]));
216 ST(&(xo[0]), VADD(TL, TM), ovs, &(xo[0]));
217 }
218 {
219 V Tj, Tv, Tu, Tw, Tb, Tt;
220 Tb = VSUB(T5, Ta);
221 Tj = VSUB(Tb, Ti);
222 Tv = VADD(Tb, Ti);
223 Tt = VSUB(Tp, Ts);
224 Tu = VBYI(VADD(Tm, Tt));
225 Tw = VBYI(VSUB(Tt, Tm));
226 ST(&(xo[WS(os, 11)]), VSUB(Tj, Tu), ovs, &(xo[WS(os, 1)]));
227 ST(&(xo[WS(os, 5)]), VADD(Tv, Tw), ovs, &(xo[WS(os, 1)]));
228 ST(&(xo[WS(os, 1)]), VADD(Tj, Tu), ovs, &(xo[WS(os, 1)]));
229 ST(&(xo[WS(os, 7)]), VSUB(Tv, Tw), ovs, &(xo[WS(os, 1)]));
230 }
231 {
232 V Tz, TD, TC, TE, TA, TB;
233 Tz = VBYI(VMUL(LDK(KP866025403), VSUB(Tx, Ty)));
234 TD = VBYI(VMUL(LDK(KP866025403), VADD(Ty, Tx)));
235 TA = VADD(T5, Ta);
236 TB = VADD(Tp, Ts);
237 TC = VSUB(TA, TB);
238 TE = VADD(TA, TB);
239 ST(&(xo[WS(os, 2)]), VADD(Tz, TC), ovs, &(xo[0]));
240 ST(&(xo[WS(os, 8)]), VSUB(TE, TD), ovs, &(xo[0]));
241 ST(&(xo[WS(os, 10)]), VSUB(TC, Tz), ovs, &(xo[0]));
242 ST(&(xo[WS(os, 4)]), VADD(TD, TE), ovs, &(xo[0]));
243 }
244 }
245 }
246 VLEAVE();
247 }
248
249 static const kdft_desc desc = { 12, XSIMD_STRING("n1bv_12"), { 44, 4, 4, 0 }, &GENUS, 0, 0, 0, 0 };
250
XSIMD(codelet_n1bv_12)251 void XSIMD(codelet_n1bv_12) (planner *p) { X(kdft_register) (p, n1bv_12, &desc);
252 }
253
254 #endif
255