Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/simd/common/t2fv_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 07:38:41 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 -n 10 -name t2fv_10 -include t2f.h */
29
30 /*
31 * This function contains 51 FP additions, 40 FP multiplications,
32 * (or, 33 additions, 22 multiplications, 18 fused multiply/add),
33 * 43 stack variables, 4 constants, and 20 memory accesses
34 */
35 #include "t2f.h"
36
37 static void t2fv_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) * 18)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 18), MAKE_VOLATILE_STRIDE(10, rs)) {
48 V Td, TA, T4, Ta, Tk, TE, Tp, TF, TB, T9, T1, T2, Tb;
49 T1 = LD(&(x[0]), ms, &(x[0]));
50 T2 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
51 {
52 V Tg, Tn, Ti, Tl;
53 Tg = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
54 Tn = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
55 Ti = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
56 Tl = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
57 {
58 V T6, T8, T5, Tc;
59 T5 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
60 Tc = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
61 {
62 V T3, Th, To, Tj, Tm, T7;
63 T7 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
64 T3 = BYTWJ(&(W[TWVL * 8]), T2);
65 Th = BYTWJ(&(W[TWVL * 6]), Tg);
66 To = BYTWJ(&(W[0]), Tn);
67 Tj = BYTWJ(&(W[TWVL * 16]), Ti);
68 Tm = BYTWJ(&(W[TWVL * 10]), Tl);
69 T6 = BYTWJ(&(W[TWVL * 2]), T5);
70 Td = BYTWJ(&(W[TWVL * 4]), Tc);
71 T8 = BYTWJ(&(W[TWVL * 12]), T7);
72 TA = VADD(T1, T3);
73 T4 = VSUB(T1, T3);
74 Ta = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
75 Tk = VSUB(Th, Tj);
76 TE = VADD(Th, Tj);
77 Tp = VSUB(Tm, To);
78 TF = VADD(Tm, To);
79 }
80 TB = VADD(T6, T8);
81 T9 = VSUB(T6, T8);
82 }
83 }
84 Tb = BYTWJ(&(W[TWVL * 14]), Ta);
85 {
86 V TL, TG, Tw, Tq, TC, Te;
87 TL = VSUB(TE, TF);
88 TG = VADD(TE, TF);
89 Tw = VSUB(Tk, Tp);
90 Tq = VADD(Tk, Tp);
91 TC = VADD(Tb, Td);
92 Te = VSUB(Tb, Td);
93 {
94 V TM, TD, Tv, Tf;
95 TM = VSUB(TB, TC);
96 TD = VADD(TB, TC);
97 Tv = VSUB(T9, Te);
98 Tf = VADD(T9, Te);
99 {
100 V TP, TN, TH, TJ, Tz, Tx, Tr, Tt, TI, Ts;
101 TP = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), TL, TM));
102 TN = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), TM, TL));
103 TH = VADD(TD, TG);
104 TJ = VSUB(TD, TG);
105 Tz = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), Tv, Tw));
106 Tx = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), Tw, Tv));
107 Tr = VADD(Tf, Tq);
108 Tt = VSUB(Tf, Tq);
109 ST(&(x[0]), VADD(TA, TH), ms, &(x[0]));
110 TI = VFNMS(LDK(KP250000000), TH, TA);
111 ST(&(x[WS(rs, 5)]), VADD(T4, Tr), ms, &(x[WS(rs, 1)]));
112 Ts = VFNMS(LDK(KP250000000), Tr, T4);
113 {
114 V TK, TO, Tu, Ty;
115 TK = VFNMS(LDK(KP559016994), TJ, TI);
116 TO = VFMA(LDK(KP559016994), TJ, TI);
117 Tu = VFMA(LDK(KP559016994), Tt, Ts);
118 Ty = VFNMS(LDK(KP559016994), Tt, Ts);
119 ST(&(x[WS(rs, 8)]), VFNMSI(TN, TK), ms, &(x[0]));
120 ST(&(x[WS(rs, 2)]), VFMAI(TN, TK), ms, &(x[0]));
121 ST(&(x[WS(rs, 6)]), VFNMSI(TP, TO), ms, &(x[0]));
122 ST(&(x[WS(rs, 4)]), VFMAI(TP, TO), ms, &(x[0]));
123 ST(&(x[WS(rs, 9)]), VFMAI(Tx, Tu), ms, &(x[WS(rs, 1)]));
124 ST(&(x[WS(rs, 1)]), VFNMSI(Tx, Tu), ms, &(x[WS(rs, 1)]));
125 ST(&(x[WS(rs, 7)]), VFMAI(Tz, Ty), ms, &(x[WS(rs, 1)]));
126 ST(&(x[WS(rs, 3)]), VFNMSI(Tz, Ty), ms, &(x[WS(rs, 1)]));
127 }
128 }
129 }
130 }
131 }
132 }
133 VLEAVE();
134 }
135
136 static const tw_instr twinstr[] = {
137 VTW(0, 1),
138 VTW(0, 2),
139 VTW(0, 3),
140 VTW(0, 4),
141 VTW(0, 5),
142 VTW(0, 6),
143 VTW(0, 7),
144 VTW(0, 8),
145 VTW(0, 9),
146 {TW_NEXT, VL, 0}
147 };
148
149 static const ct_desc desc = { 10, XSIMD_STRING("t2fv_10"), twinstr, &GENUS, {33, 22, 18, 0}, 0, 0, 0 };
150
151 void XSIMD(codelet_t2fv_10) (planner *p) {
152 X(kdft_dit_register) (p, t2fv_10, &desc);
153 }
154 #else /* HAVE_FMA */
155
156 /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 10 -name t2fv_10 -include t2f.h */
157
158 /*
159 * This function contains 51 FP additions, 30 FP multiplications,
160 * (or, 45 additions, 24 multiplications, 6 fused multiply/add),
161 * 32 stack variables, 4 constants, and 20 memory accesses
162 */
163 #include "t2f.h"
164
165 static void t2fv_10(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
166 {
167 DVK(KP587785252, +0.587785252292473129168705954639072768597652438);
168 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
169 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
170 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
171 {
172 INT m;
173 R *x;
174 x = ri;
175 for (m = mb, W = W + (mb * ((TWVL / VL) * 18)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 18), MAKE_VOLATILE_STRIDE(10, rs)) {
176 V Tr, TH, Tg, Tl, Tm, TA, TB, TJ, T5, Ta, Tb, TD, TE, TI, To;
177 V Tq, Tp;
178 To = LD(&(x[0]), ms, &(x[0]));
179 Tp = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
180 Tq = BYTWJ(&(W[TWVL * 8]), Tp);
181 Tr = VSUB(To, Tq);
182 TH = VADD(To, Tq);
183 {
184 V Td, Tk, Tf, Ti;
185 {
186 V Tc, Tj, Te, Th;
187 Tc = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
188 Td = BYTWJ(&(W[TWVL * 6]), Tc);
189 Tj = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
190 Tk = BYTWJ(&(W[0]), Tj);
191 Te = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
192 Tf = BYTWJ(&(W[TWVL * 16]), Te);
193 Th = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
194 Ti = BYTWJ(&(W[TWVL * 10]), Th);
195 }
196 Tg = VSUB(Td, Tf);
197 Tl = VSUB(Ti, Tk);
198 Tm = VADD(Tg, Tl);
199 TA = VADD(Td, Tf);
200 TB = VADD(Ti, Tk);
201 TJ = VADD(TA, TB);
202 }
203 {
204 V T2, T9, T4, T7;
205 {
206 V T1, T8, T3, T6;
207 T1 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
208 T2 = BYTWJ(&(W[TWVL * 2]), T1);
209 T8 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
210 T9 = BYTWJ(&(W[TWVL * 4]), T8);
211 T3 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
212 T4 = BYTWJ(&(W[TWVL * 12]), T3);
213 T6 = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
214 T7 = BYTWJ(&(W[TWVL * 14]), T6);
215 }
216 T5 = VSUB(T2, T4);
217 Ta = VSUB(T7, T9);
218 Tb = VADD(T5, Ta);
219 TD = VADD(T2, T4);
220 TE = VADD(T7, T9);
221 TI = VADD(TD, TE);
222 }
223 {
224 V Tn, Ts, Tt, Tx, Tz, Tv, Tw, Ty, Tu;
225 Tn = VMUL(LDK(KP559016994), VSUB(Tb, Tm));
226 Ts = VADD(Tb, Tm);
227 Tt = VFNMS(LDK(KP250000000), Ts, Tr);
228 Tv = VSUB(T5, Ta);
229 Tw = VSUB(Tg, Tl);
230 Tx = VBYI(VFMA(LDK(KP951056516), Tv, VMUL(LDK(KP587785252), Tw)));
231 Tz = VBYI(VFNMS(LDK(KP587785252), Tv, VMUL(LDK(KP951056516), Tw)));
232 ST(&(x[WS(rs, 5)]), VADD(Tr, Ts), ms, &(x[WS(rs, 1)]));
233 Ty = VSUB(Tt, Tn);
234 ST(&(x[WS(rs, 3)]), VSUB(Ty, Tz), ms, &(x[WS(rs, 1)]));
235 ST(&(x[WS(rs, 7)]), VADD(Tz, Ty), ms, &(x[WS(rs, 1)]));
236 Tu = VADD(Tn, Tt);
237 ST(&(x[WS(rs, 1)]), VSUB(Tu, Tx), ms, &(x[WS(rs, 1)]));
238 ST(&(x[WS(rs, 9)]), VADD(Tx, Tu), ms, &(x[WS(rs, 1)]));
239 }
240 {
241 V TM, TK, TL, TG, TO, TC, TF, TP, TN;
242 TM = VMUL(LDK(KP559016994), VSUB(TI, TJ));
243 TK = VADD(TI, TJ);
244 TL = VFNMS(LDK(KP250000000), TK, TH);
245 TC = VSUB(TA, TB);
246 TF = VSUB(TD, TE);
247 TG = VBYI(VFNMS(LDK(KP587785252), TF, VMUL(LDK(KP951056516), TC)));
248 TO = VBYI(VFMA(LDK(KP951056516), TF, VMUL(LDK(KP587785252), TC)));
249 ST(&(x[0]), VADD(TH, TK), ms, &(x[0]));
250 TP = VADD(TM, TL);
251 ST(&(x[WS(rs, 4)]), VADD(TO, TP), ms, &(x[0]));
252 ST(&(x[WS(rs, 6)]), VSUB(TP, TO), ms, &(x[0]));
253 TN = VSUB(TL, TM);
254 ST(&(x[WS(rs, 2)]), VADD(TG, TN), ms, &(x[0]));
255 ST(&(x[WS(rs, 8)]), VSUB(TN, TG), ms, &(x[0]));
256 }
257 }
258 }
259 VLEAVE();
260 }
261
262 static const tw_instr twinstr[] = {
263 VTW(0, 1),
264 VTW(0, 2),
265 VTW(0, 3),
266 VTW(0, 4),
267 VTW(0, 5),
268 VTW(0, 6),
269 VTW(0, 7),
270 VTW(0, 8),
271 VTW(0, 9),
272 {TW_NEXT, VL, 0}
273 };
274
275 static const ct_desc desc = { 10, XSIMD_STRING("t2fv_10"), twinstr, &GENUS, {45, 24, 6, 0}, 0, 0, 0 };
276
277 void XSIMD(codelet_t2fv_10) (planner *p) {
278 X(kdft_dit_register) (p, t2fv_10, &desc);
279 }
280 #endif /* HAVE_FMA */