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 07:05:37 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_twiddle_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 10 -name t3bv_10 -include dft/simd/t3b.h -sign 1 */
29
30 /*
31 * This function contains 57 FP additions, 52 FP multiplications,
32 * (or, 39 additions, 34 multiplications, 18 fused multiply/add),
33 * 41 stack variables, 4 constants, and 20 memory accesses
34 */
35 #include "dft/simd/t3b.h"
36
t3bv_10(R * ri,R * ii,const R * W,stride rs,INT mb,INT me,INT ms)37 static void t3bv_10(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
40 DVK(KP618033988, +0.618033988749894848204586834365638117720309180);
41 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
42 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
43 {
44 INT m;
45 R *x;
46 x = ii;
47 for (m = mb, W = W + (mb * ((TWVL / VL) * 6)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 6), MAKE_VOLATILE_STRIDE(10, rs)) {
48 V T2, T3, T4, Ta, T5, T6, Tt, Td, Th;
49 T2 = LDW(&(W[0]));
50 T3 = LDW(&(W[TWVL * 2]));
51 T4 = VZMUL(T2, T3);
52 Ta = VZMULJ(T2, T3);
53 T5 = LDW(&(W[TWVL * 4]));
54 T6 = VZMULJ(T4, T5);
55 Tt = VZMULJ(T3, T5);
56 Td = VZMULJ(Ta, T5);
57 Th = VZMULJ(T2, T5);
58 {
59 V T9, TJ, Ts, Ty, Tz, TN, TO, TP, Tg, Tm, Tn, TK, TL, TM, T1;
60 V T8, T7;
61 T1 = LD(&(x[0]), ms, &(x[0]));
62 T7 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
63 T8 = VZMUL(T6, T7);
64 T9 = VSUB(T1, T8);
65 TJ = VADD(T1, T8);
66 {
67 V Tp, Tx, Tr, Tv;
68 {
69 V To, Tw, Tq, Tu;
70 To = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
71 Tp = VZMUL(T4, To);
72 Tw = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
73 Tx = VZMUL(T2, Tw);
74 Tq = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
75 Tr = VZMUL(T5, Tq);
76 Tu = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
77 Tv = VZMUL(Tt, Tu);
78 }
79 Ts = VSUB(Tp, Tr);
80 Ty = VSUB(Tv, Tx);
81 Tz = VADD(Ts, Ty);
82 TN = VADD(Tp, Tr);
83 TO = VADD(Tv, Tx);
84 TP = VADD(TN, TO);
85 }
86 {
87 V Tc, Tl, Tf, Tj;
88 {
89 V Tb, Tk, Te, Ti;
90 Tb = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
91 Tc = VZMUL(Ta, Tb);
92 Tk = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
93 Tl = VZMUL(T3, Tk);
94 Te = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
95 Tf = VZMUL(Td, Te);
96 Ti = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
97 Tj = VZMUL(Th, Ti);
98 }
99 Tg = VSUB(Tc, Tf);
100 Tm = VSUB(Tj, Tl);
101 Tn = VADD(Tg, Tm);
102 TK = VADD(Tc, Tf);
103 TL = VADD(Tj, Tl);
104 TM = VADD(TK, TL);
105 }
106 {
107 V TC, TA, TB, TG, TI, TE, TF, TH, TD;
108 TC = VSUB(Tn, Tz);
109 TA = VADD(Tn, Tz);
110 TB = VFNMS(LDK(KP250000000), TA, T9);
111 TE = VSUB(Tg, Tm);
112 TF = VSUB(Ts, Ty);
113 TG = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), TF, TE));
114 TI = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), TE, TF));
115 ST(&(x[WS(rs, 5)]), VADD(T9, TA), ms, &(x[WS(rs, 1)]));
116 TH = VFNMS(LDK(KP559016994), TC, TB);
117 ST(&(x[WS(rs, 3)]), VFMAI(TI, TH), ms, &(x[WS(rs, 1)]));
118 ST(&(x[WS(rs, 7)]), VFNMSI(TI, TH), ms, &(x[WS(rs, 1)]));
119 TD = VFMA(LDK(KP559016994), TC, TB);
120 ST(&(x[WS(rs, 1)]), VFMAI(TG, TD), ms, &(x[WS(rs, 1)]));
121 ST(&(x[WS(rs, 9)]), VFNMSI(TG, TD), ms, &(x[WS(rs, 1)]));
122 }
123 {
124 V TS, TQ, TR, TW, TY, TU, TV, TX, TT;
125 TS = VSUB(TM, TP);
126 TQ = VADD(TM, TP);
127 TR = VFNMS(LDK(KP250000000), TQ, TJ);
128 TU = VSUB(TN, TO);
129 TV = VSUB(TK, TL);
130 TW = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), TV, TU));
131 TY = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), TU, TV));
132 ST(&(x[0]), VADD(TJ, TQ), ms, &(x[0]));
133 TX = VFMA(LDK(KP559016994), TS, TR);
134 ST(&(x[WS(rs, 4)]), VFNMSI(TY, TX), ms, &(x[0]));
135 ST(&(x[WS(rs, 6)]), VFMAI(TY, TX), ms, &(x[0]));
136 TT = VFNMS(LDK(KP559016994), TS, TR);
137 ST(&(x[WS(rs, 2)]), VFNMSI(TW, TT), ms, &(x[0]));
138 ST(&(x[WS(rs, 8)]), VFMAI(TW, TT), ms, &(x[0]));
139 }
140 }
141 }
142 }
143 VLEAVE();
144 }
145
146 static const tw_instr twinstr[] = {
147 VTW(0, 1),
148 VTW(0, 3),
149 VTW(0, 9),
150 { TW_NEXT, VL, 0 }
151 };
152
153 static const ct_desc desc = { 10, XSIMD_STRING("t3bv_10"), twinstr, &GENUS, { 39, 34, 18, 0 }, 0, 0, 0 };
154
XSIMD(codelet_t3bv_10)155 void XSIMD(codelet_t3bv_10) (planner *p) {
156 X(kdft_dit_register) (p, t3bv_10, &desc);
157 }
158 #else
159
160 /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 10 -name t3bv_10 -include dft/simd/t3b.h -sign 1 */
161
162 /*
163 * This function contains 57 FP additions, 42 FP multiplications,
164 * (or, 51 additions, 36 multiplications, 6 fused multiply/add),
165 * 41 stack variables, 4 constants, and 20 memory accesses
166 */
167 #include "dft/simd/t3b.h"
168
t3bv_10(R * ri,R * ii,const R * W,stride rs,INT mb,INT me,INT ms)169 static void t3bv_10(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
170 {
171 DVK(KP587785252, +0.587785252292473129168705954639072768597652438);
172 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
173 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
174 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
175 {
176 INT m;
177 R *x;
178 x = ii;
179 for (m = mb, W = W + (mb * ((TWVL / VL) * 6)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 6), MAKE_VOLATILE_STRIDE(10, rs)) {
180 V T1, T2, T3, Ti, T6, T7, TA, Tb, To;
181 T1 = LDW(&(W[0]));
182 T2 = LDW(&(W[TWVL * 2]));
183 T3 = VZMULJ(T1, T2);
184 Ti = VZMUL(T1, T2);
185 T6 = LDW(&(W[TWVL * 4]));
186 T7 = VZMULJ(T3, T6);
187 TA = VZMULJ(Ti, T6);
188 Tb = VZMULJ(T1, T6);
189 To = VZMULJ(T2, T6);
190 {
191 V TD, TQ, Tn, Tt, Tx, TM, TN, TS, Ta, Tg, Tw, TJ, TK, TR, Tz;
192 V TC, TB;
193 Tz = LD(&(x[0]), ms, &(x[0]));
194 TB = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
195 TC = VZMUL(TA, TB);
196 TD = VSUB(Tz, TC);
197 TQ = VADD(Tz, TC);
198 {
199 V Tk, Ts, Tm, Tq;
200 {
201 V Tj, Tr, Tl, Tp;
202 Tj = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
203 Tk = VZMUL(Ti, Tj);
204 Tr = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
205 Ts = VZMUL(T1, Tr);
206 Tl = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
207 Tm = VZMUL(T6, Tl);
208 Tp = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
209 Tq = VZMUL(To, Tp);
210 }
211 Tn = VSUB(Tk, Tm);
212 Tt = VSUB(Tq, Ts);
213 Tx = VADD(Tn, Tt);
214 TM = VADD(Tk, Tm);
215 TN = VADD(Tq, Ts);
216 TS = VADD(TM, TN);
217 }
218 {
219 V T5, Tf, T9, Td;
220 {
221 V T4, Te, T8, Tc;
222 T4 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
223 T5 = VZMUL(T3, T4);
224 Te = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
225 Tf = VZMUL(T2, Te);
226 T8 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
227 T9 = VZMUL(T7, T8);
228 Tc = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
229 Td = VZMUL(Tb, Tc);
230 }
231 Ta = VSUB(T5, T9);
232 Tg = VSUB(Td, Tf);
233 Tw = VADD(Ta, Tg);
234 TJ = VADD(T5, T9);
235 TK = VADD(Td, Tf);
236 TR = VADD(TJ, TK);
237 }
238 {
239 V Ty, TE, TF, Tv, TI, Th, Tu, TH, TG;
240 Ty = VMUL(LDK(KP559016994), VSUB(Tw, Tx));
241 TE = VADD(Tw, Tx);
242 TF = VFNMS(LDK(KP250000000), TE, TD);
243 Th = VSUB(Ta, Tg);
244 Tu = VSUB(Tn, Tt);
245 Tv = VBYI(VFMA(LDK(KP951056516), Th, VMUL(LDK(KP587785252), Tu)));
246 TI = VBYI(VFNMS(LDK(KP951056516), Tu, VMUL(LDK(KP587785252), Th)));
247 ST(&(x[WS(rs, 5)]), VADD(TD, TE), ms, &(x[WS(rs, 1)]));
248 TH = VSUB(TF, Ty);
249 ST(&(x[WS(rs, 3)]), VSUB(TH, TI), ms, &(x[WS(rs, 1)]));
250 ST(&(x[WS(rs, 7)]), VADD(TI, TH), ms, &(x[WS(rs, 1)]));
251 TG = VADD(Ty, TF);
252 ST(&(x[WS(rs, 1)]), VADD(Tv, TG), ms, &(x[WS(rs, 1)]));
253 ST(&(x[WS(rs, 9)]), VSUB(TG, Tv), ms, &(x[WS(rs, 1)]));
254 }
255 {
256 V TV, TT, TU, TP, TY, TL, TO, TX, TW;
257 TV = VMUL(LDK(KP559016994), VSUB(TR, TS));
258 TT = VADD(TR, TS);
259 TU = VFNMS(LDK(KP250000000), TT, TQ);
260 TL = VSUB(TJ, TK);
261 TO = VSUB(TM, TN);
262 TP = VBYI(VFNMS(LDK(KP951056516), TO, VMUL(LDK(KP587785252), TL)));
263 TY = VBYI(VFMA(LDK(KP951056516), TL, VMUL(LDK(KP587785252), TO)));
264 ST(&(x[0]), VADD(TQ, TT), ms, &(x[0]));
265 TX = VADD(TV, TU);
266 ST(&(x[WS(rs, 4)]), VSUB(TX, TY), ms, &(x[0]));
267 ST(&(x[WS(rs, 6)]), VADD(TY, TX), ms, &(x[0]));
268 TW = VSUB(TU, TV);
269 ST(&(x[WS(rs, 2)]), VADD(TP, TW), ms, &(x[0]));
270 ST(&(x[WS(rs, 8)]), VSUB(TW, TP), ms, &(x[0]));
271 }
272 }
273 }
274 }
275 VLEAVE();
276 }
277
278 static const tw_instr twinstr[] = {
279 VTW(0, 1),
280 VTW(0, 3),
281 VTW(0, 9),
282 { TW_NEXT, VL, 0 }
283 };
284
285 static const ct_desc desc = { 10, XSIMD_STRING("t3bv_10"), twinstr, &GENUS, { 51, 36, 6, 0 }, 0, 0, 0 };
286
XSIMD(codelet_t3bv_10)287 void XSIMD(codelet_t3bv_10) (planner *p) {
288 X(kdft_dit_register) (p, t3bv_10, &desc);
289 }
290 #endif
291