Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/rdft/simd/common/hc2cbdftv_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 07:42:29 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 8 -dif -sign 1 -name hc2cbdftv_8 -include hc2cbv.h */
29
30 /*
31 * This function contains 41 FP additions, 32 FP multiplications,
32 * (or, 23 additions, 14 multiplications, 18 fused multiply/add),
33 * 51 stack variables, 1 constants, and 16 memory accesses
34 */
35 #include "hc2cbv.h"
36
37 static void hc2cbdftv_8(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
40 {
41 INT m;
42 for (m = mb, W = W + ((mb - 1) * ((TWVL / VL) * 14)); m < me; m = m + VL, Rp = Rp + (VL * ms), Ip = Ip + (VL * ms), Rm = Rm - (VL * ms), Im = Im - (VL * ms), W = W + (TWVL * 14), MAKE_VOLATILE_STRIDE(32, rs)) {
43 V TJ, T4, Tf, TB, TD, TE, Tm, T1, Tj, TF, Tp, Tb, Tg, Tt, Tx;
44 V T2, T3, Td, Te, T5, T6, T8, T9, Tn, T7, To, Ta, Tk, Tl, TG;
45 V TL, Tq, Tc, Tu, Th, Tv, Ty, Tw, TC, Ti, TK, TA, Tz, TI, TH;
46 V Ts, Tr, TN, TM;
47 T2 = LD(&(Rp[0]), ms, &(Rp[0]));
48 T3 = LD(&(Rm[WS(rs, 3)]), -ms, &(Rm[WS(rs, 1)]));
49 Td = LD(&(Rp[WS(rs, 2)]), ms, &(Rp[0]));
50 Te = LD(&(Rm[WS(rs, 1)]), -ms, &(Rm[WS(rs, 1)]));
51 T5 = LD(&(Rp[WS(rs, 1)]), ms, &(Rp[WS(rs, 1)]));
52 T6 = LD(&(Rm[WS(rs, 2)]), -ms, &(Rm[0]));
53 T8 = LD(&(Rp[WS(rs, 3)]), ms, &(Rp[WS(rs, 1)]));
54 T9 = LD(&(Rm[0]), -ms, &(Rm[0]));
55 TJ = LDW(&(W[0]));
56 Tk = VFMACONJ(T3, T2);
57 T4 = VFNMSCONJ(T3, T2);
58 Tl = VFMACONJ(Te, Td);
59 Tf = VFNMSCONJ(Te, Td);
60 Tn = VFMACONJ(T6, T5);
61 T7 = VFNMSCONJ(T6, T5);
62 To = VFMACONJ(T9, T8);
63 Ta = VFMSCONJ(T9, T8);
64 TB = LDW(&(W[TWVL * 8]));
65 TD = LDW(&(W[TWVL * 6]));
66 TE = VADD(Tk, Tl);
67 Tm = VSUB(Tk, Tl);
68 T1 = LDW(&(W[TWVL * 12]));
69 Tj = LDW(&(W[TWVL * 10]));
70 TF = VADD(Tn, To);
71 Tp = VSUB(Tn, To);
72 Tb = VADD(T7, Ta);
73 Tg = VSUB(T7, Ta);
74 Tt = LDW(&(W[TWVL * 4]));
75 Tx = LDW(&(W[TWVL * 2]));
76 TG = VZMUL(TD, VSUB(TE, TF));
77 TL = VADD(TE, TF);
78 Tq = VZMUL(Tj, VFNMSI(Tp, Tm));
79 Tc = VFMA(LDK(KP707106781), Tb, T4);
80 Tu = VFNMS(LDK(KP707106781), Tb, T4);
81 Th = VFMA(LDK(KP707106781), Tg, Tf);
82 Tv = VFNMS(LDK(KP707106781), Tg, Tf);
83 Ty = VZMUL(Tx, VFMAI(Tp, Tm));
84 Tw = VZMULI(Tt, VFNMSI(Tv, Tu));
85 TC = VZMULI(TB, VFMAI(Tv, Tu));
86 Ti = VZMULI(T1, VFNMSI(Th, Tc));
87 TK = VZMULI(TJ, VFMAI(Th, Tc));
88 TA = VCONJ(VSUB(Ty, Tw));
89 Tz = VADD(Tw, Ty);
90 TI = VCONJ(VSUB(TG, TC));
91 TH = VADD(TC, TG);
92 Ts = VCONJ(VSUB(Tq, Ti));
93 Tr = VADD(Ti, Tq);
94 TN = VCONJ(VSUB(TL, TK));
95 TM = VADD(TK, TL);
96 ST(&(Rm[WS(rs, 1)]), TA, -ms, &(Rm[WS(rs, 1)]));
97 ST(&(Rp[WS(rs, 1)]), Tz, ms, &(Rp[WS(rs, 1)]));
98 ST(&(Rm[WS(rs, 2)]), TI, -ms, &(Rm[0]));
99 ST(&(Rp[WS(rs, 2)]), TH, ms, &(Rp[0]));
100 ST(&(Rm[WS(rs, 3)]), Ts, -ms, &(Rm[WS(rs, 1)]));
101 ST(&(Rp[WS(rs, 3)]), Tr, ms, &(Rp[WS(rs, 1)]));
102 ST(&(Rm[0]), TN, -ms, &(Rm[0]));
103 ST(&(Rp[0]), TM, ms, &(Rp[0]));
104 }
105 }
106 VLEAVE();
107 }
108
109 static const tw_instr twinstr[] = {
110 VTW(1, 1),
111 VTW(1, 2),
112 VTW(1, 3),
113 VTW(1, 4),
114 VTW(1, 5),
115 VTW(1, 6),
116 VTW(1, 7),
117 {TW_NEXT, VL, 0}
118 };
119
120 static const hc2c_desc desc = { 8, XSIMD_STRING("hc2cbdftv_8"), twinstr, &GENUS, {23, 14, 18, 0} };
121
122 void XSIMD(codelet_hc2cbdftv_8) (planner *p) {
123 X(khc2c_register) (p, hc2cbdftv_8, &desc, HC2C_VIA_DFT);
124 }
125 #else /* HAVE_FMA */
126
127 /* Generated by: ../../../genfft/gen_hc2cdft_c.native -simd -compact -variables 4 -pipeline-latency 8 -trivial-stores -variables 32 -no-generate-bytw -n 8 -dif -sign 1 -name hc2cbdftv_8 -include hc2cbv.h */
128
129 /*
130 * This function contains 41 FP additions, 16 FP multiplications,
131 * (or, 41 additions, 16 multiplications, 0 fused multiply/add),
132 * 55 stack variables, 1 constants, and 16 memory accesses
133 */
134 #include "hc2cbv.h"
135
136 static void hc2cbdftv_8(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
137 {
138 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
139 {
140 INT m;
141 for (m = mb, W = W + ((mb - 1) * ((TWVL / VL) * 14)); m < me; m = m + VL, Rp = Rp + (VL * ms), Ip = Ip + (VL * ms), Rm = Rm - (VL * ms), Im = Im - (VL * ms), W = W + (TWVL * 14), MAKE_VOLATILE_STRIDE(32, rs)) {
142 V T5, Tj, Tq, TI, Te, Tk, Tt, TJ, T2, Tg, T4, Ti, T3, Th, To;
143 V Tp, T6, Tc, T8, Tb, T7, Ta, T9, Td, Tr, Ts, TP, Tu, Tm, TO;
144 V Tn, Tf, Tl, T1, TN, Tv, TR, Tw, TQ, TC, TK, TA, TG, TB, TH;
145 V Ty, Tz, Tx, TF, TD, TM, TE, TL;
146 T2 = LD(&(Rp[0]), ms, &(Rp[0]));
147 Tg = LD(&(Rp[WS(rs, 2)]), ms, &(Rp[0]));
148 T3 = LD(&(Rm[WS(rs, 3)]), -ms, &(Rm[WS(rs, 1)]));
149 T4 = VCONJ(T3);
150 Th = LD(&(Rm[WS(rs, 1)]), -ms, &(Rm[WS(rs, 1)]));
151 Ti = VCONJ(Th);
152 T5 = VSUB(T2, T4);
153 Tj = VSUB(Tg, Ti);
154 To = VADD(T2, T4);
155 Tp = VADD(Tg, Ti);
156 Tq = VSUB(To, Tp);
157 TI = VADD(To, Tp);
158 T6 = LD(&(Rp[WS(rs, 1)]), ms, &(Rp[WS(rs, 1)]));
159 Tc = LD(&(Rp[WS(rs, 3)]), ms, &(Rp[WS(rs, 1)]));
160 T7 = LD(&(Rm[WS(rs, 2)]), -ms, &(Rm[0]));
161 T8 = VCONJ(T7);
162 Ta = LD(&(Rm[0]), -ms, &(Rm[0]));
163 Tb = VCONJ(Ta);
164 T9 = VSUB(T6, T8);
165 Td = VSUB(Tb, Tc);
166 Te = VMUL(LDK(KP707106781), VADD(T9, Td));
167 Tk = VMUL(LDK(KP707106781), VSUB(T9, Td));
168 Tr = VADD(T6, T8);
169 Ts = VADD(Tb, Tc);
170 Tt = VBYI(VSUB(Tr, Ts));
171 TJ = VADD(Tr, Ts);
172 TP = VADD(TI, TJ);
173 Tn = LDW(&(W[TWVL * 10]));
174 Tu = VZMUL(Tn, VSUB(Tq, Tt));
175 Tf = VADD(T5, Te);
176 Tl = VBYI(VADD(Tj, Tk));
177 T1 = LDW(&(W[TWVL * 12]));
178 Tm = VZMULI(T1, VSUB(Tf, Tl));
179 TN = LDW(&(W[0]));
180 TO = VZMULI(TN, VADD(Tl, Tf));
181 Tv = VADD(Tm, Tu);
182 ST(&(Rp[WS(rs, 3)]), Tv, ms, &(Rp[WS(rs, 1)]));
183 TR = VCONJ(VSUB(TP, TO));
184 ST(&(Rm[0]), TR, -ms, &(Rm[0]));
185 Tw = VCONJ(VSUB(Tu, Tm));
186 ST(&(Rm[WS(rs, 3)]), Tw, -ms, &(Rm[WS(rs, 1)]));
187 TQ = VADD(TO, TP);
188 ST(&(Rp[0]), TQ, ms, &(Rp[0]));
189 TB = LDW(&(W[TWVL * 2]));
190 TC = VZMUL(TB, VADD(Tq, Tt));
191 TH = LDW(&(W[TWVL * 6]));
192 TK = VZMUL(TH, VSUB(TI, TJ));
193 Ty = VBYI(VSUB(Tk, Tj));
194 Tz = VSUB(T5, Te);
195 Tx = LDW(&(W[TWVL * 4]));
196 TA = VZMULI(Tx, VADD(Ty, Tz));
197 TF = LDW(&(W[TWVL * 8]));
198 TG = VZMULI(TF, VSUB(Tz, Ty));
199 TD = VADD(TA, TC);
200 ST(&(Rp[WS(rs, 1)]), TD, ms, &(Rp[WS(rs, 1)]));
201 TM = VCONJ(VSUB(TK, TG));
202 ST(&(Rm[WS(rs, 2)]), TM, -ms, &(Rm[0]));
203 TE = VCONJ(VSUB(TC, TA));
204 ST(&(Rm[WS(rs, 1)]), TE, -ms, &(Rm[WS(rs, 1)]));
205 TL = VADD(TG, TK);
206 ST(&(Rp[WS(rs, 2)]), TL, ms, &(Rp[0]));
207 }
208 }
209 VLEAVE();
210 }
211
212 static const tw_instr twinstr[] = {
213 VTW(1, 1),
214 VTW(1, 2),
215 VTW(1, 3),
216 VTW(1, 4),
217 VTW(1, 5),
218 VTW(1, 6),
219 VTW(1, 7),
220 {TW_NEXT, VL, 0}
221 };
222
223 static const hc2c_desc desc = { 8, XSIMD_STRING("hc2cbdftv_8"), twinstr, &GENUS, {41, 16, 0, 0} };
224
225 void XSIMD(codelet_hc2cbdftv_8) (planner *p) {
226 X(khc2c_register) (p, hc2cbdftv_8, &desc, HC2C_VIA_DFT);
227 }
228 #endif /* HAVE_FMA */