Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/dft/simd/common/t3bv_8.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 08:06:06 EDT 2018 */
23
24 #include "dft/codelet-dft.h"
25
26 #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)
27
28 /* Generated by: ../../../genfft/gen_twiddle_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 8 -name t3bv_8 -include dft/simd/t3b.h -sign 1 */
29
30 /*
31 * This function contains 37 FP additions, 32 FP multiplications,
32 * (or, 27 additions, 22 multiplications, 10 fused multiply/add),
33 * 31 stack variables, 1 constants, and 16 memory accesses
34 */
35 #include "dft/simd/t3b.h"
36
37 static void t3bv_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 = ii;
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, Ta, T4, Tb, Tc, Tp;
46 T2 = LDW(&(W[0]));
47 T3 = LDW(&(W[TWVL * 2]));
48 Ta = VZMULJ(T2, T3);
49 T4 = VZMUL(T2, T3);
50 Tb = LDW(&(W[TWVL * 4]));
51 Tc = VZMULJ(Ta, Tb);
52 Tp = VZMULJ(T2, Tb);
53 {
54 V T7, Tx, Ts, Ty, Tf, TA, Tk, TB, T1, T6, T5;
55 T1 = LD(&(x[0]), ms, &(x[0]));
56 T5 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
57 T6 = VZMUL(T4, T5);
58 T7 = VSUB(T1, T6);
59 Tx = VADD(T1, T6);
60 {
61 V To, Tr, Tn, Tq;
62 Tn = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
63 To = VZMUL(Ta, Tn);
64 Tq = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
65 Tr = VZMUL(Tp, Tq);
66 Ts = VSUB(To, Tr);
67 Ty = VADD(To, Tr);
68 }
69 {
70 V T9, Te, T8, Td;
71 T8 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
72 T9 = VZMUL(T2, T8);
73 Td = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
74 Te = VZMUL(Tc, Td);
75 Tf = VSUB(T9, Te);
76 TA = VADD(T9, Te);
77 }
78 {
79 V Th, Tj, Tg, Ti;
80 Tg = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
81 Th = VZMUL(Tb, Tg);
82 Ti = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
83 Tj = VZMUL(T3, Ti);
84 Tk = VSUB(Th, Tj);
85 TB = VADD(Th, Tj);
86 }
87 {
88 V Tz, TC, TD, TE;
89 Tz = VSUB(Tx, Ty);
90 TC = VSUB(TA, TB);
91 ST(&(x[WS(rs, 6)]), VFNMSI(TC, Tz), ms, &(x[0]));
92 ST(&(x[WS(rs, 2)]), VFMAI(TC, Tz), ms, &(x[0]));
93 TD = VADD(Tx, Ty);
94 TE = VADD(TA, TB);
95 ST(&(x[WS(rs, 4)]), VSUB(TD, TE), ms, &(x[0]));
96 ST(&(x[0]), VADD(TD, TE), ms, &(x[0]));
97 {
98 V Tm, Tv, Tu, Tw, Tl, Tt;
99 Tl = VADD(Tf, Tk);
100 Tm = VFNMS(LDK(KP707106781), Tl, T7);
101 Tv = VFMA(LDK(KP707106781), Tl, T7);
102 Tt = VSUB(Tf, Tk);
103 Tu = VFNMS(LDK(KP707106781), Tt, Ts);
104 Tw = VFMA(LDK(KP707106781), Tt, Ts);
105 ST(&(x[WS(rs, 3)]), VFNMSI(Tu, Tm), ms, &(x[WS(rs, 1)]));
106 ST(&(x[WS(rs, 7)]), VFNMSI(Tw, Tv), ms, &(x[WS(rs, 1)]));
107 ST(&(x[WS(rs, 5)]), VFMAI(Tu, Tm), ms, &(x[WS(rs, 1)]));
108 ST(&(x[WS(rs, 1)]), VFMAI(Tw, Tv), ms, &(x[WS(rs, 1)]));
109 }
110 }
111 }
112 }
113 }
114 VLEAVE();
115 }
116
117 static const tw_instr twinstr[] = {
118 VTW(0, 1),
119 VTW(0, 3),
120 VTW(0, 7),
121 {TW_NEXT, VL, 0}
122 };
123
124 static const ct_desc desc = { 8, XSIMD_STRING("t3bv_8"), twinstr, &GENUS, {27, 22, 10, 0}, 0, 0, 0 };
125
126 void XSIMD(codelet_t3bv_8) (planner *p) {
127 X(kdft_dit_register) (p, t3bv_8, &desc);
128 }
129 #else
130
131 /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 8 -name t3bv_8 -include dft/simd/t3b.h -sign 1 */
132
133 /*
134 * This function contains 37 FP additions, 24 FP multiplications,
135 * (or, 37 additions, 24 multiplications, 0 fused multiply/add),
136 * 31 stack variables, 1 constants, and 16 memory accesses
137 */
138 #include "dft/simd/t3b.h"
139
140 static void t3bv_8(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
141 {
142 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
143 {
144 INT m;
145 R *x;
146 x = ii;
147 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)) {
148 V T1, T4, T5, Tp, T6, T7, Tj;
149 T1 = LDW(&(W[0]));
150 T4 = LDW(&(W[TWVL * 2]));
151 T5 = VZMULJ(T1, T4);
152 Tp = VZMUL(T1, T4);
153 T6 = LDW(&(W[TWVL * 4]));
154 T7 = VZMULJ(T5, T6);
155 Tj = VZMULJ(T1, T6);
156 {
157 V Ts, Tx, Tm, Ty, Ta, TA, Tf, TB, To, Tr, Tq;
158 To = LD(&(x[0]), ms, &(x[0]));
159 Tq = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
160 Tr = VZMUL(Tp, Tq);
161 Ts = VSUB(To, Tr);
162 Tx = VADD(To, Tr);
163 {
164 V Ti, Tl, Th, Tk;
165 Th = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
166 Ti = VZMUL(T5, Th);
167 Tk = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
168 Tl = VZMUL(Tj, Tk);
169 Tm = VSUB(Ti, Tl);
170 Ty = VADD(Ti, Tl);
171 }
172 {
173 V T3, T9, T2, T8;
174 T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
175 T3 = VZMUL(T1, T2);
176 T8 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
177 T9 = VZMUL(T7, T8);
178 Ta = VSUB(T3, T9);
179 TA = VADD(T3, T9);
180 }
181 {
182 V Tc, Te, Tb, Td;
183 Tb = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
184 Tc = VZMUL(T6, Tb);
185 Td = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
186 Te = VZMUL(T4, Td);
187 Tf = VSUB(Tc, Te);
188 TB = VADD(Tc, Te);
189 }
190 {
191 V Tz, TC, TD, TE;
192 Tz = VSUB(Tx, Ty);
193 TC = VBYI(VSUB(TA, TB));
194 ST(&(x[WS(rs, 6)]), VSUB(Tz, TC), ms, &(x[0]));
195 ST(&(x[WS(rs, 2)]), VADD(Tz, TC), ms, &(x[0]));
196 TD = VADD(Tx, Ty);
197 TE = VADD(TA, TB);
198 ST(&(x[WS(rs, 4)]), VSUB(TD, TE), ms, &(x[0]));
199 ST(&(x[0]), VADD(TD, TE), ms, &(x[0]));
200 {
201 V Tn, Tv, Tu, Tw, Tg, Tt;
202 Tg = VMUL(LDK(KP707106781), VSUB(Ta, Tf));
203 Tn = VBYI(VSUB(Tg, Tm));
204 Tv = VBYI(VADD(Tm, Tg));
205 Tt = VMUL(LDK(KP707106781), VADD(Ta, Tf));
206 Tu = VSUB(Ts, Tt);
207 Tw = VADD(Ts, Tt);
208 ST(&(x[WS(rs, 3)]), VADD(Tn, Tu), ms, &(x[WS(rs, 1)]));
209 ST(&(x[WS(rs, 7)]), VSUB(Tw, Tv), ms, &(x[WS(rs, 1)]));
210 ST(&(x[WS(rs, 5)]), VSUB(Tu, Tn), ms, &(x[WS(rs, 1)]));
211 ST(&(x[WS(rs, 1)]), VADD(Tv, Tw), ms, &(x[WS(rs, 1)]));
212 }
213 }
214 }
215 }
216 }
217 VLEAVE();
218 }
219
220 static const tw_instr twinstr[] = {
221 VTW(0, 1),
222 VTW(0, 3),
223 VTW(0, 7),
224 {TW_NEXT, VL, 0}
225 };
226
227 static const ct_desc desc = { 8, XSIMD_STRING("t3bv_8"), twinstr, &GENUS, {37, 24, 0, 0}, 0, 0, 0 };
228
229 void XSIMD(codelet_t3bv_8) (planner *p) {
230 X(kdft_dit_register) (p, t3bv_8, &desc);
231 }
232 #endif