Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/rdft/simd/common/hc2cfdftv_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 07:42:29 EST 2012 */
23
24 #include "codelet-rdft.h"
25
26 #ifdef HAVE_FMA
27
28 /* Generated by: ../../../genfft/gen_hc2cdft_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -trivial-stores -variables 32 -no-generate-bytw -n 10 -dit -name hc2cfdftv_10 -include hc2cfv.h */
29
30 /*
31 * This function contains 61 FP additions, 60 FP multiplications,
32 * (or, 33 additions, 32 multiplications, 28 fused multiply/add),
33 * 77 stack variables, 5 constants, and 20 memory accesses
34 */
35 #include "hc2cfv.h"
36
37 static void hc2cfdftv_10(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
40 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
41 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
42 DVK(KP618033988, +0.618033988749894848204586834365638117720309180);
43 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
44 {
45 INT m;
46 for (m = mb, W = W + ((mb - 1) * ((TWVL / VL) * 18)); m < me; m = m + VL, Rp = Rp + (VL * ms), Ip = Ip + (VL * ms), Rm = Rm - (VL * ms), Im = Im - (VL * ms), W = W + (TWVL * 18), MAKE_VOLATILE_STRIDE(40, rs)) {
47 V T5, T6, Tw, Tr, Tc, Tj, Tl, Tm, Tk, Ts, Tg, Ty, T3, T4, T1;
48 V T2, Tv, Tq, Ta, Tb, T9, Ti, Te, Tf, Td, Tx, Tn, Tt, Th, TQ;
49 V TT, Tz, T7, TR, To, Tu, TU;
50 T1 = LD(&(Rp[0]), ms, &(Rp[0]));
51 T2 = LD(&(Rm[0]), -ms, &(Rm[0]));
52 Tv = LDW(&(W[0]));
53 T5 = LD(&(Rp[WS(rs, 2)]), ms, &(Rp[0]));
54 T6 = LD(&(Rm[WS(rs, 2)]), -ms, &(Rm[0]));
55 Tq = LDW(&(W[TWVL * 6]));
56 Ta = LD(&(Rp[WS(rs, 1)]), ms, &(Rp[WS(rs, 1)]));
57 Tb = LD(&(Rm[WS(rs, 1)]), -ms, &(Rm[WS(rs, 1)]));
58 T9 = LDW(&(W[TWVL * 2]));
59 Ti = LDW(&(W[TWVL * 4]));
60 Tw = VZMULIJ(Tv, VFNMSCONJ(T2, T1));
61 Te = LD(&(Rp[WS(rs, 3)]), ms, &(Rp[WS(rs, 1)]));
62 Tf = LD(&(Rm[WS(rs, 3)]), -ms, &(Rm[WS(rs, 1)]));
63 Tr = VZMULJ(Tq, VFMACONJ(T6, T5));
64 Td = LDW(&(W[TWVL * 12]));
65 Tx = LDW(&(W[TWVL * 10]));
66 Tc = VZMULJ(T9, VFMACONJ(Tb, Ta));
67 Tj = VZMULIJ(Ti, VFNMSCONJ(Tb, Ta));
68 Tl = LD(&(Rp[WS(rs, 4)]), ms, &(Rp[0]));
69 Tm = LD(&(Rm[WS(rs, 4)]), -ms, &(Rm[0]));
70 Tk = LDW(&(W[TWVL * 14]));
71 Ts = LDW(&(W[TWVL * 16]));
72 Tg = VZMULIJ(Td, VFNMSCONJ(Tf, Te));
73 Ty = VZMULJ(Tx, VFMACONJ(Tf, Te));
74 T3 = VFMACONJ(T2, T1);
75 T4 = LDW(&(W[TWVL * 8]));
76 Tn = VZMULJ(Tk, VFMACONJ(Tm, Tl));
77 Tt = VZMULIJ(Ts, VFNMSCONJ(Tm, Tl));
78 Th = VSUB(Tc, Tg);
79 TQ = VADD(Tc, Tg);
80 TT = VADD(Tw, Ty);
81 Tz = VSUB(Tw, Ty);
82 T7 = VZMULIJ(T4, VFNMSCONJ(T6, T5));
83 TR = VADD(Tj, Tn);
84 To = VSUB(Tj, Tn);
85 Tu = VSUB(Tr, Tt);
86 TU = VADD(Tr, Tt);
87 {
88 V TP, T8, TS, T11, Tp, TH, TA, TG, TV, T12, TE, TB, TM, TI, TZ;
89 V TW, T17, T13, TD, TC, TY, TX, TL, TF, T10, T16, TN, TO, TK, TJ;
90 V T18, T19, T15, T14;
91 TP = VADD(T3, T7);
92 T8 = VSUB(T3, T7);
93 TS = VADD(TQ, TR);
94 T11 = VSUB(TQ, TR);
95 Tp = VSUB(Th, To);
96 TH = VADD(Th, To);
97 TA = VSUB(Tu, Tz);
98 TG = VADD(Tz, Tu);
99 TV = VADD(TT, TU);
100 T12 = VSUB(TU, TT);
101 TE = VSUB(Tp, TA);
102 TB = VADD(Tp, TA);
103 TM = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), TG, TH));
104 TI = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), TH, TG));
105 TZ = VSUB(TS, TV);
106 TW = VADD(TS, TV);
107 T17 = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T11, T12));
108 T13 = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T12, T11));
109 TD = VFNMS(LDK(KP250000000), TB, T8);
110 TC = VMUL(LDK(KP500000000), VADD(T8, TB));
111 TY = VFNMS(LDK(KP250000000), TW, TP);
112 TX = VCONJ(VMUL(LDK(KP500000000), VADD(TP, TW)));
113 TL = VFMA(LDK(KP559016994), TE, TD);
114 TF = VFNMS(LDK(KP559016994), TE, TD);
115 ST(&(Rp[0]), TC, ms, &(Rp[0]));
116 T10 = VFMA(LDK(KP559016994), TZ, TY);
117 T16 = VFNMS(LDK(KP559016994), TZ, TY);
118 ST(&(Rm[WS(rs, 4)]), TX, -ms, &(Rm[0]));
119 TN = VCONJ(VMUL(LDK(KP500000000), VFNMSI(TM, TL)));
120 TO = VMUL(LDK(KP500000000), VFMAI(TM, TL));
121 TK = VMUL(LDK(KP500000000), VFMAI(TI, TF));
122 TJ = VCONJ(VMUL(LDK(KP500000000), VFNMSI(TI, TF)));
123 T18 = VMUL(LDK(KP500000000), VFNMSI(T17, T16));
124 T19 = VCONJ(VMUL(LDK(KP500000000), VFMAI(T17, T16)));
125 T15 = VCONJ(VMUL(LDK(KP500000000), VFMAI(T13, T10)));
126 T14 = VMUL(LDK(KP500000000), VFNMSI(T13, T10));
127 ST(&(Rm[WS(rs, 3)]), TN, -ms, &(Rm[WS(rs, 1)]));
128 ST(&(Rp[WS(rs, 4)]), TO, ms, &(Rp[0]));
129 ST(&(Rp[WS(rs, 2)]), TK, ms, &(Rp[0]));
130 ST(&(Rm[WS(rs, 1)]), TJ, -ms, &(Rm[WS(rs, 1)]));
131 ST(&(Rp[WS(rs, 3)]), T18, ms, &(Rp[WS(rs, 1)]));
132 ST(&(Rm[WS(rs, 2)]), T19, -ms, &(Rm[0]));
133 ST(&(Rm[0]), T15, -ms, &(Rm[0]));
134 ST(&(Rp[WS(rs, 1)]), T14, ms, &(Rp[WS(rs, 1)]));
135 }
136 }
137 }
138 VLEAVE();
139 }
140
141 static const tw_instr twinstr[] = {
142 VTW(1, 1),
143 VTW(1, 2),
144 VTW(1, 3),
145 VTW(1, 4),
146 VTW(1, 5),
147 VTW(1, 6),
148 VTW(1, 7),
149 VTW(1, 8),
150 VTW(1, 9),
151 {TW_NEXT, VL, 0}
152 };
153
154 static const hc2c_desc desc = { 10, XSIMD_STRING("hc2cfdftv_10"), twinstr, &GENUS, {33, 32, 28, 0} };
155
156 void XSIMD(codelet_hc2cfdftv_10) (planner *p) {
157 X(khc2c_register) (p, hc2cfdftv_10, &desc, HC2C_VIA_DFT);
158 }
159 #else /* HAVE_FMA */
160
161 /* Generated by: ../../../genfft/gen_hc2cdft_c.native -simd -compact -variables 4 -pipeline-latency 8 -trivial-stores -variables 32 -no-generate-bytw -n 10 -dit -name hc2cfdftv_10 -include hc2cfv.h */
162
163 /*
164 * This function contains 61 FP additions, 38 FP multiplications,
165 * (or, 55 additions, 32 multiplications, 6 fused multiply/add),
166 * 82 stack variables, 5 constants, and 20 memory accesses
167 */
168 #include "hc2cfv.h"
169
170 static void hc2cfdftv_10(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
171 {
172 DVK(KP125000000, +0.125000000000000000000000000000000000000000000);
173 DVK(KP279508497, +0.279508497187473712051146708591409529430077295);
174 DVK(KP587785252, +0.587785252292473129168705954639072768597652438);
175 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
176 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
177 {
178 INT m;
179 for (m = mb, W = W + ((mb - 1) * ((TWVL / VL) * 18)); m < me; m = m + VL, Rp = Rp + (VL * ms), Ip = Ip + (VL * ms), Rm = Rm - (VL * ms), Im = Im - (VL * ms), W = W + (TWVL * 18), MAKE_VOLATILE_STRIDE(40, rs)) {
180 V Tl, Tt, Tu, TY, TZ, T10, Tz, TE, TF, TV, TW, TX, Ta, TU, TN;
181 V TR, TH, TQ, TK, TL, TM, TI, TG, TJ, TT, TO, TP, TS, T18, T1c;
182 V T12, T1b, T15, T16, T17, T14, T11, T13, T1e, T19, T1a, T1d;
183 {
184 V T1, T3, Ty, T8, T7, TB, Tf, Ts, Tk, Tw, Tq, TD, T2, Tx, T6;
185 V TA, Tc, Te, Td, Tb, Tr, Tj, Ti, Th, Tg, Tv, Tn, Tp, To, Tm;
186 V TC, T4, T9, T5;
187 T1 = LD(&(Rp[0]), ms, &(Rp[0]));
188 T2 = LD(&(Rm[0]), -ms, &(Rm[0]));
189 T3 = VCONJ(T2);
190 Tx = LDW(&(W[0]));
191 Ty = VZMULIJ(Tx, VSUB(T3, T1));
192 T8 = LD(&(Rp[WS(rs, 2)]), ms, &(Rp[0]));
193 T6 = LD(&(Rm[WS(rs, 2)]), -ms, &(Rm[0]));
194 T7 = VCONJ(T6);
195 TA = LDW(&(W[TWVL * 6]));
196 TB = VZMULJ(TA, VADD(T7, T8));
197 Tc = LD(&(Rp[WS(rs, 1)]), ms, &(Rp[WS(rs, 1)]));
198 Td = LD(&(Rm[WS(rs, 1)]), -ms, &(Rm[WS(rs, 1)]));
199 Te = VCONJ(Td);
200 Tb = LDW(&(W[TWVL * 2]));
201 Tf = VZMULJ(Tb, VADD(Tc, Te));
202 Tr = LDW(&(W[TWVL * 4]));
203 Ts = VZMULIJ(Tr, VSUB(Te, Tc));
204 Tj = LD(&(Rp[WS(rs, 3)]), ms, &(Rp[WS(rs, 1)]));
205 Th = LD(&(Rm[WS(rs, 3)]), -ms, &(Rm[WS(rs, 1)]));
206 Ti = VCONJ(Th);
207 Tg = LDW(&(W[TWVL * 12]));
208 Tk = VZMULIJ(Tg, VSUB(Ti, Tj));
209 Tv = LDW(&(W[TWVL * 10]));
210 Tw = VZMULJ(Tv, VADD(Ti, Tj));
211 Tn = LD(&(Rp[WS(rs, 4)]), ms, &(Rp[0]));
212 To = LD(&(Rm[WS(rs, 4)]), -ms, &(Rm[0]));
213 Tp = VCONJ(To);
214 Tm = LDW(&(W[TWVL * 14]));
215 Tq = VZMULJ(Tm, VADD(Tn, Tp));
216 TC = LDW(&(W[TWVL * 16]));
217 TD = VZMULIJ(TC, VSUB(Tp, Tn));
218 Tl = VSUB(Tf, Tk);
219 Tt = VSUB(Tq, Ts);
220 Tu = VADD(Tl, Tt);
221 TY = VADD(Ty, Tw);
222 TZ = VADD(TB, TD);
223 T10 = VADD(TY, TZ);
224 Tz = VSUB(Tw, Ty);
225 TE = VSUB(TB, TD);
226 TF = VADD(Tz, TE);
227 TV = VADD(Tf, Tk);
228 TW = VADD(Ts, Tq);
229 TX = VADD(TV, TW);
230 T4 = VADD(T1, T3);
231 T5 = LDW(&(W[TWVL * 8]));
232 T9 = VZMULIJ(T5, VSUB(T7, T8));
233 Ta = VSUB(T4, T9);
234 TU = VADD(T4, T9);
235 }
236 TL = VSUB(Tl, Tt);
237 TM = VSUB(TE, Tz);
238 TN = VMUL(LDK(KP500000000), VBYI(VFMA(LDK(KP951056516), TL, VMUL(LDK(KP587785252), TM))));
239 TR = VMUL(LDK(KP500000000), VBYI(VFNMS(LDK(KP587785252), TL, VMUL(LDK(KP951056516), TM))));
240 TI = VMUL(LDK(KP279508497), VSUB(Tu, TF));
241 TG = VADD(Tu, TF);
242 TJ = VFNMS(LDK(KP125000000), TG, VMUL(LDK(KP500000000), Ta));
243 TH = VCONJ(VMUL(LDK(KP500000000), VADD(Ta, TG)));
244 TQ = VSUB(TJ, TI);
245 TK = VADD(TI, TJ);
246 ST(&(Rm[WS(rs, 4)]), TH, -ms, &(Rm[0]));
247 TT = VCONJ(VADD(TQ, TR));
248 ST(&(Rm[WS(rs, 2)]), TT, -ms, &(Rm[0]));
249 TO = VSUB(TK, TN);
250 ST(&(Rp[WS(rs, 1)]), TO, ms, &(Rp[WS(rs, 1)]));
251 TP = VCONJ(VADD(TK, TN));
252 ST(&(Rm[0]), TP, -ms, &(Rm[0]));
253 TS = VSUB(TQ, TR);
254 ST(&(Rp[WS(rs, 3)]), TS, ms, &(Rp[WS(rs, 1)]));
255 T16 = VSUB(TZ, TY);
256 T17 = VSUB(TV, TW);
257 T18 = VMUL(LDK(KP500000000), VBYI(VFNMS(LDK(KP587785252), T17, VMUL(LDK(KP951056516), T16))));
258 T1c = VMUL(LDK(KP500000000), VBYI(VFMA(LDK(KP951056516), T17, VMUL(LDK(KP587785252), T16))));
259 T14 = VMUL(LDK(KP279508497), VSUB(TX, T10));
260 T11 = VADD(TX, T10);
261 T13 = VFNMS(LDK(KP125000000), T11, VMUL(LDK(KP500000000), TU));
262 T12 = VMUL(LDK(KP500000000), VADD(TU, T11));
263 T1b = VADD(T14, T13);
264 T15 = VSUB(T13, T14);
265 ST(&(Rp[0]), T12, ms, &(Rp[0]));
266 T1e = VADD(T1b, T1c);
267 ST(&(Rp[WS(rs, 4)]), T1e, ms, &(Rp[0]));
268 T19 = VCONJ(VSUB(T15, T18));
269 ST(&(Rm[WS(rs, 1)]), T19, -ms, &(Rm[WS(rs, 1)]));
270 T1a = VADD(T15, T18);
271 ST(&(Rp[WS(rs, 2)]), T1a, ms, &(Rp[0]));
272 T1d = VCONJ(VSUB(T1b, T1c));
273 ST(&(Rm[WS(rs, 3)]), T1d, -ms, &(Rm[WS(rs, 1)]));
274 }
275 }
276 VLEAVE();
277 }
278
279 static const tw_instr twinstr[] = {
280 VTW(1, 1),
281 VTW(1, 2),
282 VTW(1, 3),
283 VTW(1, 4),
284 VTW(1, 5),
285 VTW(1, 6),
286 VTW(1, 7),
287 VTW(1, 8),
288 VTW(1, 9),
289 {TW_NEXT, VL, 0}
290 };
291
292 static const hc2c_desc desc = { 10, XSIMD_STRING("hc2cfdftv_10"), twinstr, &GENUS, {55, 32, 6, 0} };
293
294 void XSIMD(codelet_hc2cfdftv_10) (planner *p) {
295 X(khc2c_register) (p, hc2cfdftv_10, &desc, HC2C_VIA_DFT);
296 }
297 #endif /* HAVE_FMA */