Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/simd/common/t2sv_4.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 07:39:26 EST 2012 */
23
24 #include "codelet-dft.h"
25
26 #ifdef HAVE_FMA
27
28 /* Generated by: ../../../genfft/gen_twiddle.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -n 4 -name t2sv_4 -include ts.h */
29
30 /*
31 * This function contains 24 FP additions, 16 FP multiplications,
32 * (or, 16 additions, 8 multiplications, 8 fused multiply/add),
33 * 37 stack variables, 0 constants, and 16 memory accesses
34 */
35 #include "ts.h"
36
37 static void t2sv_4(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 {
40 INT m;
41 for (m = mb, W = W + (mb * 4); m < me; m = m + (2 * VL), ri = ri + ((2 * VL) * ms), ii = ii + ((2 * VL) * ms), W = W + ((2 * VL) * 4), MAKE_VOLATILE_STRIDE(8, rs)) {
42 V T2, T6, T3, T5, T1, Tx, T8, Tc, Tf, Ta, T4, Th, Tj, Tl;
43 T2 = LDW(&(W[0]));
44 T6 = LDW(&(W[TWVL * 3]));
45 T3 = LDW(&(W[TWVL * 2]));
46 T5 = LDW(&(W[TWVL * 1]));
47 T1 = LD(&(ri[0]), ms, &(ri[0]));
48 Tx = LD(&(ii[0]), ms, &(ii[0]));
49 T8 = LD(&(ri[WS(rs, 2)]), ms, &(ri[0]));
50 Tc = LD(&(ii[WS(rs, 2)]), ms, &(ii[0]));
51 Tf = LD(&(ri[WS(rs, 1)]), ms, &(ri[WS(rs, 1)]));
52 Ta = VMUL(T2, T6);
53 T4 = VMUL(T2, T3);
54 Th = LD(&(ii[WS(rs, 1)]), ms, &(ii[WS(rs, 1)]));
55 Tj = LD(&(ri[WS(rs, 3)]), ms, &(ri[WS(rs, 1)]));
56 Tl = LD(&(ii[WS(rs, 3)]), ms, &(ii[WS(rs, 1)]));
57 {
58 V Tg, Tb, T7, Tp, Tk, Tr, Ti;
59 Tg = VMUL(T2, Tf);
60 Tb = VFNMS(T5, T3, Ta);
61 T7 = VFMA(T5, T6, T4);
62 Tp = VMUL(T2, Th);
63 Tk = VMUL(T3, Tj);
64 Tr = VMUL(T3, Tl);
65 Ti = VFMA(T5, Th, Tg);
66 {
67 V Tv, T9, Tq, Tm, Ts, Tw, Td;
68 Tv = VMUL(T7, Tc);
69 T9 = VMUL(T7, T8);
70 Tq = VFNMS(T5, Tf, Tp);
71 Tm = VFMA(T6, Tl, Tk);
72 Ts = VFNMS(T6, Tj, Tr);
73 Tw = VFNMS(Tb, T8, Tv);
74 Td = VFMA(Tb, Tc, T9);
75 {
76 V Tn, TA, Tu, Tt;
77 Tn = VADD(Ti, Tm);
78 TA = VSUB(Ti, Tm);
79 Tu = VADD(Tq, Ts);
80 Tt = VSUB(Tq, Ts);
81 {
82 V Ty, Tz, Te, To;
83 Ty = VADD(Tw, Tx);
84 Tz = VSUB(Tx, Tw);
85 Te = VADD(T1, Td);
86 To = VSUB(T1, Td);
87 ST(&(ii[WS(rs, 3)]), VADD(TA, Tz), ms, &(ii[WS(rs, 1)]));
88 ST(&(ii[WS(rs, 1)]), VSUB(Tz, TA), ms, &(ii[WS(rs, 1)]));
89 ST(&(ii[WS(rs, 2)]), VSUB(Ty, Tu), ms, &(ii[0]));
90 ST(&(ii[0]), VADD(Tu, Ty), ms, &(ii[0]));
91 ST(&(ri[WS(rs, 1)]), VADD(To, Tt), ms, &(ri[WS(rs, 1)]));
92 ST(&(ri[WS(rs, 3)]), VSUB(To, Tt), ms, &(ri[WS(rs, 1)]));
93 ST(&(ri[0]), VADD(Te, Tn), ms, &(ri[0]));
94 ST(&(ri[WS(rs, 2)]), VSUB(Te, Tn), ms, &(ri[0]));
95 }
96 }
97 }
98 }
99 }
100 }
101 VLEAVE();
102 }
103
104 static const tw_instr twinstr[] = {
105 VTW(0, 1),
106 VTW(0, 3),
107 {TW_NEXT, (2 * VL), 0}
108 };
109
110 static const ct_desc desc = { 4, XSIMD_STRING("t2sv_4"), twinstr, &GENUS, {16, 8, 8, 0}, 0, 0, 0 };
111
112 void XSIMD(codelet_t2sv_4) (planner *p) {
113 X(kdft_dit_register) (p, t2sv_4, &desc);
114 }
115 #else /* HAVE_FMA */
116
117 /* Generated by: ../../../genfft/gen_twiddle.native -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -n 4 -name t2sv_4 -include ts.h */
118
119 /*
120 * This function contains 24 FP additions, 16 FP multiplications,
121 * (or, 16 additions, 8 multiplications, 8 fused multiply/add),
122 * 21 stack variables, 0 constants, and 16 memory accesses
123 */
124 #include "ts.h"
125
126 static void t2sv_4(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
127 {
128 {
129 INT m;
130 for (m = mb, W = W + (mb * 4); m < me; m = m + (2 * VL), ri = ri + ((2 * VL) * ms), ii = ii + ((2 * VL) * ms), W = W + ((2 * VL) * 4), MAKE_VOLATILE_STRIDE(8, rs)) {
131 V T2, T4, T3, T5, T6, T8;
132 T2 = LDW(&(W[0]));
133 T4 = LDW(&(W[TWVL * 1]));
134 T3 = LDW(&(W[TWVL * 2]));
135 T5 = LDW(&(W[TWVL * 3]));
136 T6 = VFMA(T2, T3, VMUL(T4, T5));
137 T8 = VFNMS(T4, T3, VMUL(T2, T5));
138 {
139 V T1, Tp, Ta, To, Te, Tk, Th, Tl, T7, T9;
140 T1 = LD(&(ri[0]), ms, &(ri[0]));
141 Tp = LD(&(ii[0]), ms, &(ii[0]));
142 T7 = LD(&(ri[WS(rs, 2)]), ms, &(ri[0]));
143 T9 = LD(&(ii[WS(rs, 2)]), ms, &(ii[0]));
144 Ta = VFMA(T6, T7, VMUL(T8, T9));
145 To = VFNMS(T8, T7, VMUL(T6, T9));
146 {
147 V Tc, Td, Tf, Tg;
148 Tc = LD(&(ri[WS(rs, 1)]), ms, &(ri[WS(rs, 1)]));
149 Td = LD(&(ii[WS(rs, 1)]), ms, &(ii[WS(rs, 1)]));
150 Te = VFMA(T2, Tc, VMUL(T4, Td));
151 Tk = VFNMS(T4, Tc, VMUL(T2, Td));
152 Tf = LD(&(ri[WS(rs, 3)]), ms, &(ri[WS(rs, 1)]));
153 Tg = LD(&(ii[WS(rs, 3)]), ms, &(ii[WS(rs, 1)]));
154 Th = VFMA(T3, Tf, VMUL(T5, Tg));
155 Tl = VFNMS(T5, Tf, VMUL(T3, Tg));
156 }
157 {
158 V Tb, Ti, Tn, Tq;
159 Tb = VADD(T1, Ta);
160 Ti = VADD(Te, Th);
161 ST(&(ri[WS(rs, 2)]), VSUB(Tb, Ti), ms, &(ri[0]));
162 ST(&(ri[0]), VADD(Tb, Ti), ms, &(ri[0]));
163 Tn = VADD(Tk, Tl);
164 Tq = VADD(To, Tp);
165 ST(&(ii[0]), VADD(Tn, Tq), ms, &(ii[0]));
166 ST(&(ii[WS(rs, 2)]), VSUB(Tq, Tn), ms, &(ii[0]));
167 }
168 {
169 V Tj, Tm, Tr, Ts;
170 Tj = VSUB(T1, Ta);
171 Tm = VSUB(Tk, Tl);
172 ST(&(ri[WS(rs, 3)]), VSUB(Tj, Tm), ms, &(ri[WS(rs, 1)]));
173 ST(&(ri[WS(rs, 1)]), VADD(Tj, Tm), ms, &(ri[WS(rs, 1)]));
174 Tr = VSUB(Tp, To);
175 Ts = VSUB(Te, Th);
176 ST(&(ii[WS(rs, 1)]), VSUB(Tr, Ts), ms, &(ii[WS(rs, 1)]));
177 ST(&(ii[WS(rs, 3)]), VADD(Ts, Tr), ms, &(ii[WS(rs, 1)]));
178 }
179 }
180 }
181 }
182 VLEAVE();
183 }
184
185 static const tw_instr twinstr[] = {
186 VTW(0, 1),
187 VTW(0, 3),
188 {TW_NEXT, (2 * VL), 0}
189 };
190
191 static const ct_desc desc = { 4, XSIMD_STRING("t2sv_4"), twinstr, &GENUS, {16, 8, 8, 0}, 0, 0, 0 };
192
193 void XSIMD(codelet_t2sv_4) (planner *p) {
194 X(kdft_dit_register) (p, t2sv_4, &desc);
195 }
196 #endif /* HAVE_FMA */