Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/rdft/simd/common/hc2cfdftv_6.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 6 -dit -name hc2cfdftv_6 -include hc2cfv.h */
29
30 /*
31 * This function contains 29 FP additions, 30 FP multiplications,
32 * (or, 17 additions, 18 multiplications, 12 fused multiply/add),
33 * 38 stack variables, 2 constants, and 12 memory accesses
34 */
35 #include "hc2cfv.h"
36
37 static void hc2cfdftv_6(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(KP866025403, +0.866025403784438646763723170752936183471402627);
41 {
42 INT m;
43 for (m = mb, W = W + ((mb - 1) * ((TWVL / VL) * 10)); m < me; m = m + VL, Rp = Rp + (VL * ms), Ip = Ip + (VL * ms), Rm = Rm - (VL * ms), Im = Im - (VL * ms), W = W + (TWVL * 10), MAKE_VOLATILE_STRIDE(24, rs)) {
44 V T5, T6, T3, Tj, T4, T9, Te, Th, T1, T2, Ti, Tc, Td, Tb, Tg;
45 V T7, Ta, Tt, Tk, Tr, T8, Ts, Tf, Tx, Tu, To, Tl, Tw, Tv, Tn;
46 V Tm, Tz, Ty, Tp, Tq;
47 T1 = LD(&(Rp[0]), ms, &(Rp[0]));
48 T2 = LD(&(Rm[0]), -ms, &(Rm[0]));
49 Ti = LDW(&(W[0]));
50 Tc = LD(&(Rp[WS(rs, 2)]), ms, &(Rp[0]));
51 Td = LD(&(Rm[WS(rs, 2)]), -ms, &(Rm[0]));
52 Tb = LDW(&(W[TWVL * 8]));
53 Tg = LDW(&(W[TWVL * 6]));
54 T5 = LD(&(Rp[WS(rs, 1)]), ms, &(Rp[WS(rs, 1)]));
55 T6 = LD(&(Rm[WS(rs, 1)]), -ms, &(Rm[WS(rs, 1)]));
56 T3 = VFMACONJ(T2, T1);
57 Tj = VZMULIJ(Ti, VFNMSCONJ(T2, T1));
58 T4 = LDW(&(W[TWVL * 4]));
59 T9 = LDW(&(W[TWVL * 2]));
60 Te = VZMULIJ(Tb, VFNMSCONJ(Td, Tc));
61 Th = VZMULJ(Tg, VFMACONJ(Td, Tc));
62 T7 = VZMULIJ(T4, VFNMSCONJ(T6, T5));
63 Ta = VZMULJ(T9, VFMACONJ(T6, T5));
64 Tt = VADD(Tj, Th);
65 Tk = VSUB(Th, Tj);
66 Tr = VADD(T3, T7);
67 T8 = VSUB(T3, T7);
68 Ts = VADD(Ta, Te);
69 Tf = VSUB(Ta, Te);
70 Tx = VMUL(LDK(KP866025403), VSUB(Tt, Ts));
71 Tu = VADD(Ts, Tt);
72 To = VMUL(LDK(KP866025403), VSUB(Tk, Tf));
73 Tl = VADD(Tf, Tk);
74 Tw = VFNMS(LDK(KP500000000), Tu, Tr);
75 Tv = VCONJ(VMUL(LDK(KP500000000), VADD(Tr, Tu)));
76 Tn = VFNMS(LDK(KP500000000), Tl, T8);
77 Tm = VMUL(LDK(KP500000000), VADD(T8, Tl));
78 Tz = VMUL(LDK(KP500000000), VFMAI(Tx, Tw));
79 Ty = VCONJ(VMUL(LDK(KP500000000), VFNMSI(Tx, Tw)));
80 ST(&(Rm[WS(rs, 2)]), Tv, -ms, &(Rm[0]));
81 Tp = VMUL(LDK(KP500000000), VFNMSI(To, Tn));
82 Tq = VCONJ(VMUL(LDK(KP500000000), VFMAI(To, Tn)));
83 ST(&(Rp[0]), Tm, ms, &(Rp[0]));
84 ST(&(Rp[WS(rs, 1)]), Tz, ms, &(Rp[WS(rs, 1)]));
85 ST(&(Rm[0]), Ty, -ms, &(Rm[0]));
86 ST(&(Rm[WS(rs, 1)]), Tq, -ms, &(Rm[WS(rs, 1)]));
87 ST(&(Rp[WS(rs, 2)]), Tp, ms, &(Rp[0]));
88 }
89 }
90 VLEAVE();
91 }
92
93 static const tw_instr twinstr[] = {
94 VTW(1, 1),
95 VTW(1, 2),
96 VTW(1, 3),
97 VTW(1, 4),
98 VTW(1, 5),
99 {TW_NEXT, VL, 0}
100 };
101
102 static const hc2c_desc desc = { 6, XSIMD_STRING("hc2cfdftv_6"), twinstr, &GENUS, {17, 18, 12, 0} };
103
104 void XSIMD(codelet_hc2cfdftv_6) (planner *p) {
105 X(khc2c_register) (p, hc2cfdftv_6, &desc, HC2C_VIA_DFT);
106 }
107 #else /* HAVE_FMA */
108
109 /* Generated by: ../../../genfft/gen_hc2cdft_c.native -simd -compact -variables 4 -pipeline-latency 8 -trivial-stores -variables 32 -no-generate-bytw -n 6 -dit -name hc2cfdftv_6 -include hc2cfv.h */
110
111 /*
112 * This function contains 29 FP additions, 20 FP multiplications,
113 * (or, 27 additions, 18 multiplications, 2 fused multiply/add),
114 * 42 stack variables, 3 constants, and 12 memory accesses
115 */
116 #include "hc2cfv.h"
117
118 static void hc2cfdftv_6(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
119 {
120 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
121 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
122 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
123 {
124 INT m;
125 for (m = mb, W = W + ((mb - 1) * ((TWVL / VL) * 10)); m < me; m = m + VL, Rp = Rp + (VL * ms), Ip = Ip + (VL * ms), Rm = Rm - (VL * ms), Im = Im - (VL * ms), W = W + (TWVL * 10), MAKE_VOLATILE_STRIDE(24, rs)) {
126 V Ta, Tu, Tn, Tw, Ti, Tv, T1, T8, Tg, Tf, T7, T3, Te, T6, T2;
127 V T4, T9, T5, Tk, Tm, Tj, Tl, Tc, Th, Tb, Td, Tr, Tp, Tq, To;
128 V Tt, Ts, TA, Ty, Tz, Tx, TC, TB;
129 T1 = LD(&(Rp[0]), ms, &(Rp[0]));
130 T8 = LD(&(Rp[WS(rs, 1)]), ms, &(Rp[WS(rs, 1)]));
131 Tg = LD(&(Rp[WS(rs, 2)]), ms, &(Rp[0]));
132 Te = LD(&(Rm[WS(rs, 2)]), -ms, &(Rm[0]));
133 Tf = VCONJ(Te);
134 T6 = LD(&(Rm[WS(rs, 1)]), -ms, &(Rm[WS(rs, 1)]));
135 T7 = VCONJ(T6);
136 T2 = LD(&(Rm[0]), -ms, &(Rm[0]));
137 T3 = VCONJ(T2);
138 T4 = VADD(T1, T3);
139 T5 = LDW(&(W[TWVL * 4]));
140 T9 = VZMULIJ(T5, VSUB(T7, T8));
141 Ta = VADD(T4, T9);
142 Tu = VSUB(T4, T9);
143 Tj = LDW(&(W[0]));
144 Tk = VZMULIJ(Tj, VSUB(T3, T1));
145 Tl = LDW(&(W[TWVL * 6]));
146 Tm = VZMULJ(Tl, VADD(Tf, Tg));
147 Tn = VADD(Tk, Tm);
148 Tw = VSUB(Tm, Tk);
149 Tb = LDW(&(W[TWVL * 2]));
150 Tc = VZMULJ(Tb, VADD(T7, T8));
151 Td = LDW(&(W[TWVL * 8]));
152 Th = VZMULIJ(Td, VSUB(Tf, Tg));
153 Ti = VADD(Tc, Th);
154 Tv = VSUB(Tc, Th);
155 Tr = VMUL(LDK(KP500000000), VBYI(VMUL(LDK(KP866025403), VSUB(Tn, Ti))));
156 To = VADD(Ti, Tn);
157 Tp = VMUL(LDK(KP500000000), VADD(Ta, To));
158 Tq = VFNMS(LDK(KP250000000), To, VMUL(LDK(KP500000000), Ta));
159 ST(&(Rp[0]), Tp, ms, &(Rp[0]));
160 Tt = VCONJ(VADD(Tq, Tr));
161 ST(&(Rm[WS(rs, 1)]), Tt, -ms, &(Rm[WS(rs, 1)]));
162 Ts = VSUB(Tq, Tr);
163 ST(&(Rp[WS(rs, 2)]), Ts, ms, &(Rp[0]));
164 TA = VMUL(LDK(KP500000000), VBYI(VMUL(LDK(KP866025403), VSUB(Tw, Tv))));
165 Tx = VADD(Tv, Tw);
166 Ty = VCONJ(VMUL(LDK(KP500000000), VADD(Tu, Tx)));
167 Tz = VFNMS(LDK(KP250000000), Tx, VMUL(LDK(KP500000000), Tu));
168 ST(&(Rm[WS(rs, 2)]), Ty, -ms, &(Rm[0]));
169 TC = VADD(Tz, TA);
170 ST(&(Rp[WS(rs, 1)]), TC, ms, &(Rp[WS(rs, 1)]));
171 TB = VCONJ(VSUB(Tz, TA));
172 ST(&(Rm[0]), TB, -ms, &(Rm[0]));
173 }
174 }
175 VLEAVE();
176 }
177
178 static const tw_instr twinstr[] = {
179 VTW(1, 1),
180 VTW(1, 2),
181 VTW(1, 3),
182 VTW(1, 4),
183 VTW(1, 5),
184 {TW_NEXT, VL, 0}
185 };
186
187 static const hc2c_desc desc = { 6, XSIMD_STRING("hc2cfdftv_6"), twinstr, &GENUS, {27, 18, 2, 0} };
188
189 void XSIMD(codelet_hc2cfdftv_6) (planner *p) {
190 X(khc2c_register) (p, hc2cfdftv_6, &desc, HC2C_VIA_DFT);
191 }
192 #endif /* HAVE_FMA */