Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/rdft/simd/common/hc2cbdftv_6.c @ 167:bd3cc4d1df30

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add FFTW 3.3.8 source, and a Linux build
author Chris Cannam <cannam@all-day-breakfast.com>
date Tue, 19 Nov 2019 14:52:55 +0000
parents
children
comparison
equal deleted inserted replaced
166:cbd6d7e562c7 167:bd3cc4d1df30
1 /*
2 * Copyright (c) 2003, 2007-14 Matteo Frigo
3 * Copyright (c) 2003, 2007-14 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 Thu May 24 08:08:11 EDT 2018 */
23
24 #include "rdft/codelet-rdft.h"
25
26 #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)
27
28 /* Generated by: ../../../genfft/gen_hc2cdft_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -trivial-stores -variables 32 -no-generate-bytw -n 6 -dif -sign 1 -name hc2cbdftv_6 -include rdft/simd/hc2cbv.h */
29
30 /*
31 * This function contains 29 FP additions, 24 FP multiplications,
32 * (or, 17 additions, 12 multiplications, 12 fused multiply/add),
33 * 38 stack variables, 2 constants, and 12 memory accesses
34 */
35 #include "rdft/simd/hc2cbv.h"
36
37 static void hc2cbdftv_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 T4, Te, Tj, Tp, Tb, To, Th, Ti, Ta, Tg, T7, Tf, T2, T3, T8;
45 V T9, T5, T6, Tx, Tw, Tv, Ty, Tz, Tq, Ts, Tn, Tr, Tt, Tu, Tc;
46 V Tk, T1, Td, Tl, Tm;
47 T2 = LD(&(Rp[0]), ms, &(Rp[0]));
48 T3 = LD(&(Rm[WS(rs, 2)]), -ms, &(Rm[0]));
49 T4 = VFNMSCONJ(T3, T2);
50 Te = VFMACONJ(T3, T2);
51 T8 = LD(&(Rp[WS(rs, 1)]), ms, &(Rp[WS(rs, 1)]));
52 T9 = LD(&(Rm[WS(rs, 1)]), -ms, &(Rm[WS(rs, 1)]));
53 Ta = VFMSCONJ(T9, T8);
54 Tg = VFMACONJ(T9, T8);
55 T5 = LD(&(Rp[WS(rs, 2)]), ms, &(Rp[0]));
56 T6 = LD(&(Rm[0]), -ms, &(Rm[0]));
57 T7 = VFNMSCONJ(T6, T5);
58 Tf = VFMACONJ(T6, T5);
59 Tj = VMUL(LDK(KP866025403), VSUB(Tf, Tg));
60 Tp = VMUL(LDK(KP866025403), VSUB(T7, Ta));
61 Tb = VADD(T7, Ta);
62 To = VFNMS(LDK(KP500000000), Tb, T4);
63 Th = VADD(Tf, Tg);
64 Ti = VFNMS(LDK(KP500000000), Th, Te);
65 Tx = VADD(Te, Th);
66 Tv = LDW(&(W[0]));
67 Tw = VZMULI(Tv, VFMAI(Tp, To));
68 Ty = VADD(Tw, Tx);
69 ST(&(Rp[0]), Ty, ms, &(Rp[0]));
70 Tz = VCONJ(VSUB(Tx, Tw));
71 ST(&(Rm[0]), Tz, -ms, &(Rm[0]));
72 Tn = LDW(&(W[TWVL * 8]));
73 Tq = VZMULI(Tn, VFNMSI(Tp, To));
74 Tr = LDW(&(W[TWVL * 6]));
75 Ts = VZMUL(Tr, VFMAI(Tj, Ti));
76 Tt = VADD(Tq, Ts);
77 ST(&(Rp[WS(rs, 2)]), Tt, ms, &(Rp[0]));
78 Tu = VCONJ(VSUB(Ts, Tq));
79 ST(&(Rm[WS(rs, 2)]), Tu, -ms, &(Rm[0]));
80 T1 = LDW(&(W[TWVL * 4]));
81 Tc = VZMULI(T1, VADD(T4, Tb));
82 Td = LDW(&(W[TWVL * 2]));
83 Tk = VZMUL(Td, VFNMSI(Tj, Ti));
84 Tl = VADD(Tc, Tk);
85 ST(&(Rp[WS(rs, 1)]), Tl, ms, &(Rp[WS(rs, 1)]));
86 Tm = VCONJ(VSUB(Tk, Tc));
87 ST(&(Rm[WS(rs, 1)]), Tm, -ms, &(Rm[WS(rs, 1)]));
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("hc2cbdftv_6"), twinstr, &GENUS, {17, 12, 12, 0} };
103
104 void XSIMD(codelet_hc2cbdftv_6) (planner *p) {
105 X(khc2c_register) (p, hc2cbdftv_6, &desc, HC2C_VIA_DFT);
106 }
107 #else
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 -dif -sign 1 -name hc2cbdftv_6 -include rdft/simd/hc2cbv.h */
110
111 /*
112 * This function contains 29 FP additions, 14 FP multiplications,
113 * (or, 27 additions, 12 multiplications, 2 fused multiply/add),
114 * 41 stack variables, 2 constants, and 12 memory accesses
115 */
116 #include "rdft/simd/hc2cbv.h"
117
118 static void hc2cbdftv_6(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
119 {
120 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
121 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
122 {
123 INT m;
124 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)) {
125 V T5, Th, Te, Ts, Tk, Tm, T2, T4, T3, T6, Tc, T8, Tb, T7, Ta;
126 V T9, Td, Ti, Tj, TA, Tf, Tn, Tv, Tt, Tz, T1, Tl, Tg, Tu, Tr;
127 V Tq, Ty, To, Tp, TC, TB, Tx, Tw;
128 T2 = LD(&(Rp[0]), ms, &(Rp[0]));
129 T3 = LD(&(Rm[WS(rs, 2)]), -ms, &(Rm[0]));
130 T4 = VCONJ(T3);
131 T5 = VSUB(T2, T4);
132 Th = VADD(T2, T4);
133 T6 = LD(&(Rp[WS(rs, 2)]), ms, &(Rp[0]));
134 Tc = LD(&(Rp[WS(rs, 1)]), ms, &(Rp[WS(rs, 1)]));
135 T7 = LD(&(Rm[0]), -ms, &(Rm[0]));
136 T8 = VCONJ(T7);
137 Ta = LD(&(Rm[WS(rs, 1)]), -ms, &(Rm[WS(rs, 1)]));
138 Tb = VCONJ(Ta);
139 T9 = VSUB(T6, T8);
140 Td = VSUB(Tb, Tc);
141 Te = VADD(T9, Td);
142 Ts = VBYI(VMUL(LDK(KP866025403), VSUB(T9, Td)));
143 Ti = VADD(T6, T8);
144 Tj = VADD(Tb, Tc);
145 Tk = VADD(Ti, Tj);
146 Tm = VBYI(VMUL(LDK(KP866025403), VSUB(Ti, Tj)));
147 TA = VADD(Th, Tk);
148 T1 = LDW(&(W[TWVL * 4]));
149 Tf = VZMULI(T1, VADD(T5, Te));
150 Tl = VFNMS(LDK(KP500000000), Tk, Th);
151 Tg = LDW(&(W[TWVL * 2]));
152 Tn = VZMUL(Tg, VSUB(Tl, Tm));
153 Tu = LDW(&(W[TWVL * 6]));
154 Tv = VZMUL(Tu, VADD(Tm, Tl));
155 Tr = VFNMS(LDK(KP500000000), Te, T5);
156 Tq = LDW(&(W[TWVL * 8]));
157 Tt = VZMULI(Tq, VSUB(Tr, Ts));
158 Ty = LDW(&(W[0]));
159 Tz = VZMULI(Ty, VADD(Ts, Tr));
160 To = VADD(Tf, Tn);
161 ST(&(Rp[WS(rs, 1)]), To, ms, &(Rp[WS(rs, 1)]));
162 Tp = VCONJ(VSUB(Tn, Tf));
163 ST(&(Rm[WS(rs, 1)]), Tp, -ms, &(Rm[WS(rs, 1)]));
164 TC = VCONJ(VSUB(TA, Tz));
165 ST(&(Rm[0]), TC, -ms, &(Rm[0]));
166 TB = VADD(Tz, TA);
167 ST(&(Rp[0]), TB, ms, &(Rp[0]));
168 Tx = VCONJ(VSUB(Tv, Tt));
169 ST(&(Rm[WS(rs, 2)]), Tx, -ms, &(Rm[0]));
170 Tw = VADD(Tt, Tv);
171 ST(&(Rp[WS(rs, 2)]), Tw, ms, &(Rp[0]));
172 }
173 }
174 VLEAVE();
175 }
176
177 static const tw_instr twinstr[] = {
178 VTW(1, 1),
179 VTW(1, 2),
180 VTW(1, 3),
181 VTW(1, 4),
182 VTW(1, 5),
183 {TW_NEXT, VL, 0}
184 };
185
186 static const hc2c_desc desc = { 6, XSIMD_STRING("hc2cbdftv_6"), twinstr, &GENUS, {27, 12, 2, 0} };
187
188 void XSIMD(codelet_hc2cbdftv_6) (planner *p) {
189 X(khc2c_register) (p, hc2cbdftv_6, &desc, HC2C_VIA_DFT);
190 }
191 #endif