Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/rdft/simd/common/hc2cfdftv_8.c @ 95:89f5e221ed7b

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add FFTW3
author Chris Cannam <cannam@all-day-breakfast.com>
date Wed, 20 Mar 2013 15:35:50 +0000
parents
children
comparison
equal deleted inserted replaced
94:d278df1123f9 95:89f5e221ed7b
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 8 -dit -name hc2cfdftv_8 -include hc2cfv.h */
29
30 /*
31 * This function contains 41 FP additions, 40 FP multiplications,
32 * (or, 23 additions, 22 multiplications, 18 fused multiply/add),
33 * 52 stack variables, 2 constants, and 16 memory accesses
34 */
35 #include "hc2cfv.h"
36
37 static void hc2cfdftv_8(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
40 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
41 {
42 INT m;
43 for (m = mb, W = W + ((mb - 1) * ((TWVL / VL) * 14)); m < me; m = m + VL, Rp = Rp + (VL * ms), Ip = Ip + (VL * ms), Rm = Rm - (VL * ms), Im = Im - (VL * ms), W = W + (TWVL * 14), MAKE_VOLATILE_STRIDE(32, rs)) {
44 V T3, Tc, Tl, Ts, Tf, Tg, Te, Tp, T7, Ta, T1, T2, Tb, Tj, Tk;
45 V Ti, Tr, T5, T6, T4, T9, Th, Tq, TC, T8, Td, TF, Tm, TG, TD;
46 V Tt, Tu, Tn, TH, TL, TE, TK, Tz, Tv, Ty, To, TJ, TI, TN, TM;
47 V TB, TA, Tx, Tw;
48 T1 = LD(&(Rp[0]), ms, &(Rp[0]));
49 T2 = LD(&(Rm[0]), -ms, &(Rm[0]));
50 Tb = LDW(&(W[0]));
51 Tj = LD(&(Rp[WS(rs, 3)]), ms, &(Rp[WS(rs, 1)]));
52 Tk = LD(&(Rm[WS(rs, 3)]), -ms, &(Rm[WS(rs, 1)]));
53 Ti = LDW(&(W[TWVL * 12]));
54 Tr = LDW(&(W[TWVL * 10]));
55 T5 = LD(&(Rp[WS(rs, 2)]), ms, &(Rp[0]));
56 T6 = LD(&(Rm[WS(rs, 2)]), -ms, &(Rm[0]));
57 T3 = VFMACONJ(T2, T1);
58 Tc = VZMULIJ(Tb, VFNMSCONJ(T2, T1));
59 T4 = LDW(&(W[TWVL * 6]));
60 T9 = LDW(&(W[TWVL * 8]));
61 Tl = VZMULIJ(Ti, VFNMSCONJ(Tk, Tj));
62 Ts = VZMULJ(Tr, VFMACONJ(Tk, Tj));
63 Tf = LD(&(Rp[WS(rs, 1)]), ms, &(Rp[WS(rs, 1)]));
64 Tg = LD(&(Rm[WS(rs, 1)]), -ms, &(Rm[WS(rs, 1)]));
65 Te = LDW(&(W[TWVL * 4]));
66 Tp = LDW(&(W[TWVL * 2]));
67 T7 = VZMULJ(T4, VFMACONJ(T6, T5));
68 Ta = VZMULIJ(T9, VFNMSCONJ(T6, T5));
69 Th = VZMULIJ(Te, VFNMSCONJ(Tg, Tf));
70 Tq = VZMULJ(Tp, VFMACONJ(Tg, Tf));
71 TC = VADD(T3, T7);
72 T8 = VSUB(T3, T7);
73 Td = VSUB(Ta, Tc);
74 TF = VADD(Tc, Ta);
75 Tm = VSUB(Th, Tl);
76 TG = VADD(Th, Tl);
77 TD = VADD(Tq, Ts);
78 Tt = VSUB(Tq, Ts);
79 Tu = VSUB(Tm, Td);
80 Tn = VADD(Td, Tm);
81 TH = VSUB(TF, TG);
82 TL = VADD(TF, TG);
83 TE = VSUB(TC, TD);
84 TK = VADD(TC, TD);
85 Tz = VFMA(LDK(KP707106781), Tu, Tt);
86 Tv = VFNMS(LDK(KP707106781), Tu, Tt);
87 Ty = VFNMS(LDK(KP707106781), Tn, T8);
88 To = VFMA(LDK(KP707106781), Tn, T8);
89 TJ = VCONJ(VMUL(LDK(KP500000000), VFNMSI(TH, TE)));
90 TI = VMUL(LDK(KP500000000), VFMAI(TH, TE));
91 TN = VCONJ(VMUL(LDK(KP500000000), VADD(TL, TK)));
92 TM = VMUL(LDK(KP500000000), VSUB(TK, TL));
93 TB = VMUL(LDK(KP500000000), VFMAI(Tz, Ty));
94 TA = VCONJ(VMUL(LDK(KP500000000), VFNMSI(Tz, Ty)));
95 Tx = VCONJ(VMUL(LDK(KP500000000), VFMAI(Tv, To)));
96 Tw = VMUL(LDK(KP500000000), VFNMSI(Tv, To));
97 ST(&(Rm[WS(rs, 1)]), TJ, -ms, &(Rm[WS(rs, 1)]));
98 ST(&(Rp[WS(rs, 2)]), TI, ms, &(Rp[0]));
99 ST(&(Rm[WS(rs, 3)]), TN, -ms, &(Rm[WS(rs, 1)]));
100 ST(&(Rp[0]), TM, ms, &(Rp[0]));
101 ST(&(Rp[WS(rs, 3)]), TB, ms, &(Rp[WS(rs, 1)]));
102 ST(&(Rm[WS(rs, 2)]), TA, -ms, &(Rm[0]));
103 ST(&(Rm[0]), Tx, -ms, &(Rm[0]));
104 ST(&(Rp[WS(rs, 1)]), Tw, ms, &(Rp[WS(rs, 1)]));
105 }
106 }
107 VLEAVE();
108 }
109
110 static const tw_instr twinstr[] = {
111 VTW(1, 1),
112 VTW(1, 2),
113 VTW(1, 3),
114 VTW(1, 4),
115 VTW(1, 5),
116 VTW(1, 6),
117 VTW(1, 7),
118 {TW_NEXT, VL, 0}
119 };
120
121 static const hc2c_desc desc = { 8, XSIMD_STRING("hc2cfdftv_8"), twinstr, &GENUS, {23, 22, 18, 0} };
122
123 void XSIMD(codelet_hc2cfdftv_8) (planner *p) {
124 X(khc2c_register) (p, hc2cfdftv_8, &desc, HC2C_VIA_DFT);
125 }
126 #else /* HAVE_FMA */
127
128 /* Generated by: ../../../genfft/gen_hc2cdft_c.native -simd -compact -variables 4 -pipeline-latency 8 -trivial-stores -variables 32 -no-generate-bytw -n 8 -dit -name hc2cfdftv_8 -include hc2cfv.h */
129
130 /*
131 * This function contains 41 FP additions, 23 FP multiplications,
132 * (or, 41 additions, 23 multiplications, 0 fused multiply/add),
133 * 57 stack variables, 3 constants, and 16 memory accesses
134 */
135 #include "hc2cfv.h"
136
137 static void hc2cfdftv_8(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
138 {
139 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
140 DVK(KP353553390, +0.353553390593273762200422181052424519642417969);
141 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
142 {
143 INT m;
144 for (m = mb, W = W + ((mb - 1) * ((TWVL / VL) * 14)); m < me; m = m + VL, Rp = Rp + (VL * ms), Ip = Ip + (VL * ms), Rm = Rm - (VL * ms), Im = Im - (VL * ms), W = W + (TWVL * 14), MAKE_VOLATILE_STRIDE(32, rs)) {
145 V Ta, TE, Tr, TF, Tl, TK, Tw, TG, T1, T6, T3, T8, T2, T7, T4;
146 V T9, T5, To, Tq, Tn, Tp, Tc, Th, Te, Tj, Td, Ti, Tf, Tk, Tb;
147 V Tg, Tt, Tv, Ts, Tu, Ty, Tz, Tm, Tx, TC, TD, TA, TB, TI, TO;
148 V TL, TP, TH, TJ, TM, TR, TN, TQ;
149 T1 = LD(&(Rp[0]), ms, &(Rp[0]));
150 T6 = LD(&(Rp[WS(rs, 2)]), ms, &(Rp[0]));
151 T2 = LD(&(Rm[0]), -ms, &(Rm[0]));
152 T3 = VCONJ(T2);
153 T7 = LD(&(Rm[WS(rs, 2)]), -ms, &(Rm[0]));
154 T8 = VCONJ(T7);
155 T4 = VADD(T1, T3);
156 T5 = LDW(&(W[TWVL * 6]));
157 T9 = VZMULJ(T5, VADD(T6, T8));
158 Ta = VADD(T4, T9);
159 TE = VMUL(LDK(KP500000000), VSUB(T4, T9));
160 Tn = LDW(&(W[0]));
161 To = VZMULIJ(Tn, VSUB(T3, T1));
162 Tp = LDW(&(W[TWVL * 8]));
163 Tq = VZMULIJ(Tp, VSUB(T8, T6));
164 Tr = VADD(To, Tq);
165 TF = VSUB(To, Tq);
166 Tc = LD(&(Rp[WS(rs, 1)]), ms, &(Rp[WS(rs, 1)]));
167 Th = LD(&(Rp[WS(rs, 3)]), ms, &(Rp[WS(rs, 1)]));
168 Td = LD(&(Rm[WS(rs, 1)]), -ms, &(Rm[WS(rs, 1)]));
169 Te = VCONJ(Td);
170 Ti = LD(&(Rm[WS(rs, 3)]), -ms, &(Rm[WS(rs, 1)]));
171 Tj = VCONJ(Ti);
172 Tb = LDW(&(W[TWVL * 2]));
173 Tf = VZMULJ(Tb, VADD(Tc, Te));
174 Tg = LDW(&(W[TWVL * 10]));
175 Tk = VZMULJ(Tg, VADD(Th, Tj));
176 Tl = VADD(Tf, Tk);
177 TK = VSUB(Tf, Tk);
178 Ts = LDW(&(W[TWVL * 4]));
179 Tt = VZMULIJ(Ts, VSUB(Te, Tc));
180 Tu = LDW(&(W[TWVL * 12]));
181 Tv = VZMULIJ(Tu, VSUB(Tj, Th));
182 Tw = VADD(Tt, Tv);
183 TG = VSUB(Tv, Tt);
184 Tm = VADD(Ta, Tl);
185 Tx = VADD(Tr, Tw);
186 Ty = VCONJ(VMUL(LDK(KP500000000), VSUB(Tm, Tx)));
187 Tz = VMUL(LDK(KP500000000), VADD(Tm, Tx));
188 ST(&(Rm[WS(rs, 3)]), Ty, -ms, &(Rm[WS(rs, 1)]));
189 ST(&(Rp[0]), Tz, ms, &(Rp[0]));
190 TA = VSUB(Ta, Tl);
191 TB = VBYI(VSUB(Tw, Tr));
192 TC = VCONJ(VMUL(LDK(KP500000000), VSUB(TA, TB)));
193 TD = VMUL(LDK(KP500000000), VADD(TA, TB));
194 ST(&(Rm[WS(rs, 1)]), TC, -ms, &(Rm[WS(rs, 1)]));
195 ST(&(Rp[WS(rs, 2)]), TD, ms, &(Rp[0]));
196 TH = VMUL(LDK(KP353553390), VADD(TF, TG));
197 TI = VADD(TE, TH);
198 TO = VSUB(TE, TH);
199 TJ = VMUL(LDK(KP707106781), VSUB(TG, TF));
200 TL = VMUL(LDK(KP500000000), VBYI(VSUB(TJ, TK)));
201 TP = VMUL(LDK(KP500000000), VBYI(VADD(TK, TJ)));
202 TM = VCONJ(VSUB(TI, TL));
203 ST(&(Rm[0]), TM, -ms, &(Rm[0]));
204 TR = VADD(TO, TP);
205 ST(&(Rp[WS(rs, 3)]), TR, ms, &(Rp[WS(rs, 1)]));
206 TN = VADD(TI, TL);
207 ST(&(Rp[WS(rs, 1)]), TN, ms, &(Rp[WS(rs, 1)]));
208 TQ = VCONJ(VSUB(TO, TP));
209 ST(&(Rm[WS(rs, 2)]), TQ, -ms, &(Rm[0]));
210 }
211 }
212 VLEAVE();
213 }
214
215 static const tw_instr twinstr[] = {
216 VTW(1, 1),
217 VTW(1, 2),
218 VTW(1, 3),
219 VTW(1, 4),
220 VTW(1, 5),
221 VTW(1, 6),
222 VTW(1, 7),
223 {TW_NEXT, VL, 0}
224 };
225
226 static const hc2c_desc desc = { 8, XSIMD_STRING("hc2cfdftv_8"), twinstr, &GENUS, {41, 23, 0, 0} };
227
228 void XSIMD(codelet_hc2cfdftv_8) (planner *p) {
229 X(khc2c_register) (p, hc2cfdftv_8, &desc, HC2C_VIA_DFT);
230 }
231 #endif /* HAVE_FMA */