Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/simd/common/t3fv_8.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 07:38:49 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 8 -name t3fv_8 -include t3f.h */
29
30 /*
31 * This function contains 37 FP additions, 32 FP multiplications,
32 * (or, 27 additions, 22 multiplications, 10 fused multiply/add),
33 * 43 stack variables, 1 constants, and 16 memory accesses
34 */
35 #include "t3f.h"
36
37 static void t3fv_8(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
40 {
41 INT m;
42 R *x;
43 x = ri;
44 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(8, rs)) {
45 V T2, T3, Tb, T1, T5, Tn, Tq, T8, Td, T4, Ta, Tp, Tg, Ti, T9;
46 T2 = LDW(&(W[0]));
47 T3 = LDW(&(W[TWVL * 2]));
48 Tb = LDW(&(W[TWVL * 4]));
49 T1 = LD(&(x[0]), ms, &(x[0]));
50 T5 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
51 Tn = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
52 Tq = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
53 T8 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
54 Td = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
55 T4 = VZMUL(T2, T3);
56 Ta = VZMULJ(T2, T3);
57 Tp = VZMULJ(T2, Tb);
58 Tg = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
59 Ti = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
60 T9 = VZMULJ(T2, T8);
61 {
62 V T6, To, Tc, Tr, Th, Tj;
63 T6 = VZMULJ(T4, T5);
64 To = VZMULJ(Ta, Tn);
65 Tc = VZMULJ(Ta, Tb);
66 Tr = VZMULJ(Tp, Tq);
67 Th = VZMULJ(Tb, Tg);
68 Tj = VZMULJ(T3, Ti);
69 {
70 V Tx, T7, Te, Ts, Ty, Tk, TB;
71 Tx = VADD(T1, T6);
72 T7 = VSUB(T1, T6);
73 Te = VZMULJ(Tc, Td);
74 Ts = VSUB(To, Tr);
75 Ty = VADD(To, Tr);
76 Tk = VSUB(Th, Tj);
77 TB = VADD(Th, Tj);
78 {
79 V Tf, TA, Tz, TD;
80 Tf = VSUB(T9, Te);
81 TA = VADD(T9, Te);
82 Tz = VADD(Tx, Ty);
83 TD = VSUB(Tx, Ty);
84 {
85 V TC, TE, Tl, Tt;
86 TC = VADD(TA, TB);
87 TE = VSUB(TB, TA);
88 Tl = VADD(Tf, Tk);
89 Tt = VSUB(Tk, Tf);
90 {
91 V Tu, Tw, Tm, Tv;
92 ST(&(x[WS(rs, 2)]), VFMAI(TE, TD), ms, &(x[0]));
93 ST(&(x[WS(rs, 6)]), VFNMSI(TE, TD), ms, &(x[0]));
94 ST(&(x[0]), VADD(Tz, TC), ms, &(x[0]));
95 ST(&(x[WS(rs, 4)]), VSUB(Tz, TC), ms, &(x[0]));
96 Tu = VFNMS(LDK(KP707106781), Tt, Ts);
97 Tw = VFMA(LDK(KP707106781), Tt, Ts);
98 Tm = VFMA(LDK(KP707106781), Tl, T7);
99 Tv = VFNMS(LDK(KP707106781), Tl, T7);
100 ST(&(x[WS(rs, 5)]), VFNMSI(Tw, Tv), ms, &(x[WS(rs, 1)]));
101 ST(&(x[WS(rs, 3)]), VFMAI(Tw, Tv), ms, &(x[WS(rs, 1)]));
102 ST(&(x[WS(rs, 7)]), VFMAI(Tu, Tm), ms, &(x[WS(rs, 1)]));
103 ST(&(x[WS(rs, 1)]), VFNMSI(Tu, Tm), ms, &(x[WS(rs, 1)]));
104 }
105 }
106 }
107 }
108 }
109 }
110 }
111 VLEAVE();
112 }
113
114 static const tw_instr twinstr[] = {
115 VTW(0, 1),
116 VTW(0, 3),
117 VTW(0, 7),
118 {TW_NEXT, VL, 0}
119 };
120
121 static const ct_desc desc = { 8, XSIMD_STRING("t3fv_8"), twinstr, &GENUS, {27, 22, 10, 0}, 0, 0, 0 };
122
123 void XSIMD(codelet_t3fv_8) (planner *p) {
124 X(kdft_dit_register) (p, t3fv_8, &desc);
125 }
126 #else /* HAVE_FMA */
127
128 /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 8 -name t3fv_8 -include t3f.h */
129
130 /*
131 * This function contains 37 FP additions, 24 FP multiplications,
132 * (or, 37 additions, 24 multiplications, 0 fused multiply/add),
133 * 31 stack variables, 1 constants, and 16 memory accesses
134 */
135 #include "t3f.h"
136
137 static void t3fv_8(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
138 {
139 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
140 {
141 INT m;
142 R *x;
143 x = ri;
144 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(8, rs)) {
145 V T2, T3, Ta, T4, Tb, Tc, Tq;
146 T2 = LDW(&(W[0]));
147 T3 = LDW(&(W[TWVL * 2]));
148 Ta = VZMULJ(T2, T3);
149 T4 = VZMUL(T2, T3);
150 Tb = LDW(&(W[TWVL * 4]));
151 Tc = VZMULJ(Ta, Tb);
152 Tq = VZMULJ(T2, Tb);
153 {
154 V T7, Tx, Tt, Ty, Tf, TA, Tk, TB, T1, T6, T5;
155 T1 = LD(&(x[0]), ms, &(x[0]));
156 T5 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
157 T6 = VZMULJ(T4, T5);
158 T7 = VSUB(T1, T6);
159 Tx = VADD(T1, T6);
160 {
161 V Tp, Ts, To, Tr;
162 To = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
163 Tp = VZMULJ(Ta, To);
164 Tr = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
165 Ts = VZMULJ(Tq, Tr);
166 Tt = VSUB(Tp, Ts);
167 Ty = VADD(Tp, Ts);
168 }
169 {
170 V T9, Te, T8, Td;
171 T8 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
172 T9 = VZMULJ(T2, T8);
173 Td = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
174 Te = VZMULJ(Tc, Td);
175 Tf = VSUB(T9, Te);
176 TA = VADD(T9, Te);
177 }
178 {
179 V Th, Tj, Tg, Ti;
180 Tg = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
181 Th = VZMULJ(Tb, Tg);
182 Ti = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
183 Tj = VZMULJ(T3, Ti);
184 Tk = VSUB(Th, Tj);
185 TB = VADD(Th, Tj);
186 }
187 {
188 V Tz, TC, TD, TE;
189 Tz = VADD(Tx, Ty);
190 TC = VADD(TA, TB);
191 ST(&(x[WS(rs, 4)]), VSUB(Tz, TC), ms, &(x[0]));
192 ST(&(x[0]), VADD(Tz, TC), ms, &(x[0]));
193 TD = VSUB(Tx, Ty);
194 TE = VBYI(VSUB(TB, TA));
195 ST(&(x[WS(rs, 6)]), VSUB(TD, TE), ms, &(x[0]));
196 ST(&(x[WS(rs, 2)]), VADD(TD, TE), ms, &(x[0]));
197 {
198 V Tm, Tv, Tu, Tw, Tl, Tn;
199 Tl = VMUL(LDK(KP707106781), VADD(Tf, Tk));
200 Tm = VADD(T7, Tl);
201 Tv = VSUB(T7, Tl);
202 Tn = VMUL(LDK(KP707106781), VSUB(Tk, Tf));
203 Tu = VBYI(VSUB(Tn, Tt));
204 Tw = VBYI(VADD(Tt, Tn));
205 ST(&(x[WS(rs, 7)]), VSUB(Tm, Tu), ms, &(x[WS(rs, 1)]));
206 ST(&(x[WS(rs, 3)]), VADD(Tv, Tw), ms, &(x[WS(rs, 1)]));
207 ST(&(x[WS(rs, 1)]), VADD(Tm, Tu), ms, &(x[WS(rs, 1)]));
208 ST(&(x[WS(rs, 5)]), VSUB(Tv, Tw), ms, &(x[WS(rs, 1)]));
209 }
210 }
211 }
212 }
213 }
214 VLEAVE();
215 }
216
217 static const tw_instr twinstr[] = {
218 VTW(0, 1),
219 VTW(0, 3),
220 VTW(0, 7),
221 {TW_NEXT, VL, 0}
222 };
223
224 static const ct_desc desc = { 8, XSIMD_STRING("t3fv_8"), twinstr, &GENUS, {37, 24, 0, 0}, 0, 0, 0 };
225
226 void XSIMD(codelet_t3fv_8) (planner *p) {
227 X(kdft_dit_register) (p, t3fv_8, &desc);
228 }
229 #endif /* HAVE_FMA */