Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/simd/common/t3fv_10.c @ 10:37bf6b4a2645

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add FFTW3
author Chris Cannam
date Wed, 20 Mar 2013 15:35:50 +0000
parents
children
comparison
equal deleted inserted replaced
9:c0fb53affa76 10:37bf6b4a2645
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., 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 Sun Nov 25 07:38:55 EST 2012 */
23
24 #include "codelet-dft.h"
25
26 #ifdef HAVE_FMA
27
28 /* Generated by: ../../../genfft/gen_twiddle_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 10 -name t3fv_10 -include t3f.h */
29
30 /*
31 * This function contains 57 FP additions, 52 FP multiplications,
32 * (or, 39 additions, 34 multiplications, 18 fused multiply/add),
33 * 57 stack variables, 4 constants, and 20 memory accesses
34 */
35 #include "t3f.h"
36
37 static void t3fv_10(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
40 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
41 DVK(KP618033988, +0.618033988749894848204586834365638117720309180);
42 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
43 {
44 INT m;
45 R *x;
46 x = ri;
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 T1, T7, Th, Tx, Tr, Td, Tp, T6, Tv, Tc, Te, Ti, Tl, T2, T3;
49 V T5;
50 T2 = LDW(&(W[0]));
51 T3 = LDW(&(W[TWVL * 2]));
52 T5 = LDW(&(W[TWVL * 4]));
53 T1 = LD(&(x[0]), ms, &(x[0]));
54 T7 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
55 {
56 V To, Tw, Tq, Tu, Ta, T4, Tt, Tk, Tb;
57 To = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
58 Tw = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
59 Tq = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
60 Tu = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
61 Ta = VZMULJ(T2, T3);
62 T4 = VZMUL(T2, T3);
63 Th = VZMULJ(T2, T5);
64 Tt = VZMULJ(T3, T5);
65 Tb = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
66 Tx = VZMULJ(T2, Tw);
67 Tr = VZMULJ(T5, Tq);
68 Tk = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
69 Td = VZMULJ(Ta, T5);
70 Tp = VZMULJ(T4, To);
71 T6 = VZMULJ(T4, T5);
72 Tv = VZMULJ(Tt, Tu);
73 Tc = VZMULJ(Ta, Tb);
74 Te = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
75 Ti = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
76 Tl = VZMULJ(T3, Tk);
77 }
78 {
79 V TN, Ts, T8, Ty, TO, Tf, Tj;
80 TN = VADD(Tp, Tr);
81 Ts = VSUB(Tp, Tr);
82 T8 = VZMULJ(T6, T7);
83 Ty = VSUB(Tv, Tx);
84 TO = VADD(Tv, Tx);
85 Tf = VZMULJ(Td, Te);
86 Tj = VZMULJ(Th, Ti);
87 {
88 V T9, TJ, TP, TU, Tz, TF, Tg, TK, Tm, TL;
89 T9 = VSUB(T1, T8);
90 TJ = VADD(T1, T8);
91 TP = VADD(TN, TO);
92 TU = VSUB(TN, TO);
93 Tz = VADD(Ts, Ty);
94 TF = VSUB(Ts, Ty);
95 Tg = VSUB(Tc, Tf);
96 TK = VADD(Tc, Tf);
97 Tm = VSUB(Tj, Tl);
98 TL = VADD(Tj, Tl);
99 {
100 V TM, TV, Tn, TE;
101 TM = VADD(TK, TL);
102 TV = VSUB(TK, TL);
103 Tn = VADD(Tg, Tm);
104 TE = VSUB(Tg, Tm);
105 {
106 V TW, TY, TS, TQ, TG, TI, TC, TA, TR, TB;
107 TW = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), TV, TU));
108 TY = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), TU, TV));
109 TS = VSUB(TM, TP);
110 TQ = VADD(TM, TP);
111 TG = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), TF, TE));
112 TI = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), TE, TF));
113 TC = VSUB(Tn, Tz);
114 TA = VADD(Tn, Tz);
115 ST(&(x[0]), VADD(TJ, TQ), ms, &(x[0]));
116 TR = VFNMS(LDK(KP250000000), TQ, TJ);
117 ST(&(x[WS(rs, 5)]), VADD(T9, TA), ms, &(x[WS(rs, 1)]));
118 TB = VFNMS(LDK(KP250000000), TA, T9);
119 {
120 V TX, TT, TH, TD;
121 TX = VFMA(LDK(KP559016994), TS, TR);
122 TT = VFNMS(LDK(KP559016994), TS, TR);
123 TH = VFNMS(LDK(KP559016994), TC, TB);
124 TD = VFMA(LDK(KP559016994), TC, TB);
125 ST(&(x[WS(rs, 8)]), VFNMSI(TW, TT), ms, &(x[0]));
126 ST(&(x[WS(rs, 2)]), VFMAI(TW, TT), ms, &(x[0]));
127 ST(&(x[WS(rs, 6)]), VFNMSI(TY, TX), ms, &(x[0]));
128 ST(&(x[WS(rs, 4)]), VFMAI(TY, TX), ms, &(x[0]));
129 ST(&(x[WS(rs, 9)]), VFMAI(TG, TD), ms, &(x[WS(rs, 1)]));
130 ST(&(x[WS(rs, 1)]), VFNMSI(TG, TD), ms, &(x[WS(rs, 1)]));
131 ST(&(x[WS(rs, 7)]), VFMAI(TI, TH), ms, &(x[WS(rs, 1)]));
132 ST(&(x[WS(rs, 3)]), VFNMSI(TI, TH), ms, &(x[WS(rs, 1)]));
133 }
134 }
135 }
136 }
137 }
138 }
139 }
140 VLEAVE();
141 }
142
143 static const tw_instr twinstr[] = {
144 VTW(0, 1),
145 VTW(0, 3),
146 VTW(0, 9),
147 {TW_NEXT, VL, 0}
148 };
149
150 static const ct_desc desc = { 10, XSIMD_STRING("t3fv_10"), twinstr, &GENUS, {39, 34, 18, 0}, 0, 0, 0 };
151
152 void XSIMD(codelet_t3fv_10) (planner *p) {
153 X(kdft_dit_register) (p, t3fv_10, &desc);
154 }
155 #else /* HAVE_FMA */
156
157 /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 10 -name t3fv_10 -include t3f.h */
158
159 /*
160 * This function contains 57 FP additions, 42 FP multiplications,
161 * (or, 51 additions, 36 multiplications, 6 fused multiply/add),
162 * 41 stack variables, 4 constants, and 20 memory accesses
163 */
164 #include "t3f.h"
165
166 static void t3fv_10(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
167 {
168 DVK(KP587785252, +0.587785252292473129168705954639072768597652438);
169 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
170 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
171 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
172 {
173 INT m;
174 R *x;
175 x = ri;
176 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)) {
177 V T1, T2, T3, Ti, T6, T7, Tx, Tb, To;
178 T1 = LDW(&(W[0]));
179 T2 = LDW(&(W[TWVL * 2]));
180 T3 = VZMULJ(T1, T2);
181 Ti = VZMUL(T1, T2);
182 T6 = LDW(&(W[TWVL * 4]));
183 T7 = VZMULJ(T3, T6);
184 Tx = VZMULJ(Ti, T6);
185 Tb = VZMULJ(T1, T6);
186 To = VZMULJ(T2, T6);
187 {
188 V TA, TQ, Tn, Tt, Tu, TJ, TK, TS, Ta, Tg, Th, TM, TN, TR, Tw;
189 V Tz, Ty;
190 Tw = LD(&(x[0]), ms, &(x[0]));
191 Ty = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
192 Tz = VZMULJ(Tx, Ty);
193 TA = VSUB(Tw, Tz);
194 TQ = VADD(Tw, Tz);
195 {
196 V Tk, Ts, Tm, Tq;
197 {
198 V Tj, Tr, Tl, Tp;
199 Tj = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
200 Tk = VZMULJ(Ti, Tj);
201 Tr = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
202 Ts = VZMULJ(T1, Tr);
203 Tl = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
204 Tm = VZMULJ(T6, Tl);
205 Tp = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
206 Tq = VZMULJ(To, Tp);
207 }
208 Tn = VSUB(Tk, Tm);
209 Tt = VSUB(Tq, Ts);
210 Tu = VADD(Tn, Tt);
211 TJ = VADD(Tk, Tm);
212 TK = VADD(Tq, Ts);
213 TS = VADD(TJ, TK);
214 }
215 {
216 V T5, Tf, T9, Td;
217 {
218 V T4, Te, T8, Tc;
219 T4 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
220 T5 = VZMULJ(T3, T4);
221 Te = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
222 Tf = VZMULJ(T2, Te);
223 T8 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
224 T9 = VZMULJ(T7, T8);
225 Tc = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
226 Td = VZMULJ(Tb, Tc);
227 }
228 Ta = VSUB(T5, T9);
229 Tg = VSUB(Td, Tf);
230 Th = VADD(Ta, Tg);
231 TM = VADD(T5, T9);
232 TN = VADD(Td, Tf);
233 TR = VADD(TM, TN);
234 }
235 {
236 V Tv, TB, TC, TG, TI, TE, TF, TH, TD;
237 Tv = VMUL(LDK(KP559016994), VSUB(Th, Tu));
238 TB = VADD(Th, Tu);
239 TC = VFNMS(LDK(KP250000000), TB, TA);
240 TE = VSUB(Ta, Tg);
241 TF = VSUB(Tn, Tt);
242 TG = VBYI(VFMA(LDK(KP951056516), TE, VMUL(LDK(KP587785252), TF)));
243 TI = VBYI(VFNMS(LDK(KP587785252), TE, VMUL(LDK(KP951056516), TF)));
244 ST(&(x[WS(rs, 5)]), VADD(TA, TB), ms, &(x[WS(rs, 1)]));
245 TH = VSUB(TC, Tv);
246 ST(&(x[WS(rs, 3)]), VSUB(TH, TI), ms, &(x[WS(rs, 1)]));
247 ST(&(x[WS(rs, 7)]), VADD(TI, TH), ms, &(x[WS(rs, 1)]));
248 TD = VADD(Tv, TC);
249 ST(&(x[WS(rs, 1)]), VSUB(TD, TG), ms, &(x[WS(rs, 1)]));
250 ST(&(x[WS(rs, 9)]), VADD(TG, TD), ms, &(x[WS(rs, 1)]));
251 }
252 {
253 V TV, TT, TU, TP, TX, TL, TO, TY, TW;
254 TV = VMUL(LDK(KP559016994), VSUB(TR, TS));
255 TT = VADD(TR, TS);
256 TU = VFNMS(LDK(KP250000000), TT, TQ);
257 TL = VSUB(TJ, TK);
258 TO = VSUB(TM, TN);
259 TP = VBYI(VFNMS(LDK(KP587785252), TO, VMUL(LDK(KP951056516), TL)));
260 TX = VBYI(VFMA(LDK(KP951056516), TO, VMUL(LDK(KP587785252), TL)));
261 ST(&(x[0]), VADD(TQ, TT), ms, &(x[0]));
262 TY = VADD(TV, TU);
263 ST(&(x[WS(rs, 4)]), VADD(TX, TY), ms, &(x[0]));
264 ST(&(x[WS(rs, 6)]), VSUB(TY, TX), ms, &(x[0]));
265 TW = VSUB(TU, TV);
266 ST(&(x[WS(rs, 2)]), VADD(TP, TW), ms, &(x[0]));
267 ST(&(x[WS(rs, 8)]), VSUB(TW, TP), ms, &(x[0]));
268 }
269 }
270 }
271 }
272 VLEAVE();
273 }
274
275 static const tw_instr twinstr[] = {
276 VTW(0, 1),
277 VTW(0, 3),
278 VTW(0, 9),
279 {TW_NEXT, VL, 0}
280 };
281
282 static const ct_desc desc = { 10, XSIMD_STRING("t3fv_10"), twinstr, &GENUS, {51, 36, 6, 0}, 0, 0, 0 };
283
284 void XSIMD(codelet_t3fv_10) (planner *p) {
285 X(kdft_dit_register) (p, t3fv_10, &desc);
286 }
287 #endif /* HAVE_FMA */