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 Dec 10 07:04:44 EST 2020 */
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_notw_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 10 -name n1bv_10 -include dft/simd/n1b.h */
29
30 /*
31 * This function contains 42 FP additions, 22 FP multiplications,
32 * (or, 24 additions, 4 multiplications, 18 fused multiply/add),
33 * 33 stack variables, 4 constants, and 20 memory accesses
34 */
35 #include "dft/simd/n1b.h"
36
n1bv_10(const R * ri,const R * ii,R * ro,R * io,stride is,stride os,INT v,INT ivs,INT ovs)37 static void n1bv_10(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
38 {
39 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
40 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
41 DVK(KP618033988, +0.618033988749894848204586834365638117720309180);
42 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
43 {
44 INT i;
45 const R *xi;
46 R *xo;
47 xi = ii;
48 xo = io;
49 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(20, is), MAKE_VOLATILE_STRIDE(20, os)) {
50 V T3, Tr, Tm, Tn, TD, TC, Tu, Tx, Ty, Ta, Th, Ti, T1, T2;
51 T1 = LD(&(xi[0]), ivs, &(xi[0]));
52 T2 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
53 T3 = VSUB(T1, T2);
54 Tr = VADD(T1, T2);
55 {
56 V T6, Ts, Tg, Tw, T9, Tt, Td, Tv;
57 {
58 V T4, T5, Te, Tf;
59 T4 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
60 T5 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
61 T6 = VSUB(T4, T5);
62 Ts = VADD(T4, T5);
63 Te = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
64 Tf = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
65 Tg = VSUB(Te, Tf);
66 Tw = VADD(Te, Tf);
67 }
68 {
69 V T7, T8, Tb, Tc;
70 T7 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
71 T8 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
72 T9 = VSUB(T7, T8);
73 Tt = VADD(T7, T8);
74 Tb = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
75 Tc = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
76 Td = VSUB(Tb, Tc);
77 Tv = VADD(Tb, Tc);
78 }
79 Tm = VSUB(T6, T9);
80 Tn = VSUB(Td, Tg);
81 TD = VSUB(Ts, Tt);
82 TC = VSUB(Tv, Tw);
83 Tu = VADD(Ts, Tt);
84 Tx = VADD(Tv, Tw);
85 Ty = VADD(Tu, Tx);
86 Ta = VADD(T6, T9);
87 Th = VADD(Td, Tg);
88 Ti = VADD(Ta, Th);
89 }
90 ST(&(xo[WS(os, 5)]), VADD(T3, Ti), ovs, &(xo[WS(os, 1)]));
91 ST(&(xo[0]), VADD(Tr, Ty), ovs, &(xo[0]));
92 {
93 V To, Tq, Tl, Tp, Tj, Tk;
94 To = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), Tn, Tm));
95 Tq = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), Tm, Tn));
96 Tj = VFNMS(LDK(KP250000000), Ti, T3);
97 Tk = VSUB(Ta, Th);
98 Tl = VFMA(LDK(KP559016994), Tk, Tj);
99 Tp = VFNMS(LDK(KP559016994), Tk, Tj);
100 ST(&(xo[WS(os, 1)]), VFMAI(To, Tl), ovs, &(xo[WS(os, 1)]));
101 ST(&(xo[WS(os, 7)]), VFNMSI(Tq, Tp), ovs, &(xo[WS(os, 1)]));
102 ST(&(xo[WS(os, 9)]), VFNMSI(To, Tl), ovs, &(xo[WS(os, 1)]));
103 ST(&(xo[WS(os, 3)]), VFMAI(Tq, Tp), ovs, &(xo[WS(os, 1)]));
104 }
105 {
106 V TE, TG, TB, TF, Tz, TA;
107 TE = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), TD, TC));
108 TG = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), TC, TD));
109 Tz = VFNMS(LDK(KP250000000), Ty, Tr);
110 TA = VSUB(Tu, Tx);
111 TB = VFNMS(LDK(KP559016994), TA, Tz);
112 TF = VFMA(LDK(KP559016994), TA, Tz);
113 ST(&(xo[WS(os, 2)]), VFNMSI(TE, TB), ovs, &(xo[0]));
114 ST(&(xo[WS(os, 6)]), VFMAI(TG, TF), ovs, &(xo[0]));
115 ST(&(xo[WS(os, 8)]), VFMAI(TE, TB), ovs, &(xo[0]));
116 ST(&(xo[WS(os, 4)]), VFNMSI(TG, TF), ovs, &(xo[0]));
117 }
118 }
119 }
120 VLEAVE();
121 }
122
123 static const kdft_desc desc = { 10, XSIMD_STRING("n1bv_10"), { 24, 4, 18, 0 }, &GENUS, 0, 0, 0, 0 };
124
XSIMD(codelet_n1bv_10)125 void XSIMD(codelet_n1bv_10) (planner *p) { X(kdft_register) (p, n1bv_10, &desc);
126 }
127
128 #else
129
130 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 10 -name n1bv_10 -include dft/simd/n1b.h */
131
132 /*
133 * This function contains 42 FP additions, 12 FP multiplications,
134 * (or, 36 additions, 6 multiplications, 6 fused multiply/add),
135 * 33 stack variables, 4 constants, and 20 memory accesses
136 */
137 #include "dft/simd/n1b.h"
138
n1bv_10(const R * ri,const R * ii,R * ro,R * io,stride is,stride os,INT v,INT ivs,INT ovs)139 static void n1bv_10(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
140 {
141 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
142 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
143 DVK(KP587785252, +0.587785252292473129168705954639072768597652438);
144 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
145 {
146 INT i;
147 const R *xi;
148 R *xo;
149 xi = ii;
150 xo = io;
151 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(20, is), MAKE_VOLATILE_STRIDE(20, os)) {
152 V Tl, Ty, T7, Te, Tw, Tt, Tz, TA, TB, Tg, Th, Tm, Tj, Tk;
153 Tj = LD(&(xi[0]), ivs, &(xi[0]));
154 Tk = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
155 Tl = VSUB(Tj, Tk);
156 Ty = VADD(Tj, Tk);
157 {
158 V T3, Tr, Td, Tv, T6, Ts, Ta, Tu;
159 {
160 V T1, T2, Tb, Tc;
161 T1 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
162 T2 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
163 T3 = VSUB(T1, T2);
164 Tr = VADD(T1, T2);
165 Tb = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
166 Tc = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
167 Td = VSUB(Tb, Tc);
168 Tv = VADD(Tb, Tc);
169 }
170 {
171 V T4, T5, T8, T9;
172 T4 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
173 T5 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
174 T6 = VSUB(T4, T5);
175 Ts = VADD(T4, T5);
176 T8 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
177 T9 = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
178 Ta = VSUB(T8, T9);
179 Tu = VADD(T8, T9);
180 }
181 T7 = VSUB(T3, T6);
182 Te = VSUB(Ta, Td);
183 Tw = VSUB(Tu, Tv);
184 Tt = VSUB(Tr, Ts);
185 Tz = VADD(Tr, Ts);
186 TA = VADD(Tu, Tv);
187 TB = VADD(Tz, TA);
188 Tg = VADD(T3, T6);
189 Th = VADD(Ta, Td);
190 Tm = VADD(Tg, Th);
191 }
192 ST(&(xo[WS(os, 5)]), VADD(Tl, Tm), ovs, &(xo[WS(os, 1)]));
193 ST(&(xo[0]), VADD(Ty, TB), ovs, &(xo[0]));
194 {
195 V Tf, Tq, To, Tp, Ti, Tn;
196 Tf = VBYI(VFMA(LDK(KP951056516), T7, VMUL(LDK(KP587785252), Te)));
197 Tq = VBYI(VFNMS(LDK(KP951056516), Te, VMUL(LDK(KP587785252), T7)));
198 Ti = VMUL(LDK(KP559016994), VSUB(Tg, Th));
199 Tn = VFNMS(LDK(KP250000000), Tm, Tl);
200 To = VADD(Ti, Tn);
201 Tp = VSUB(Tn, Ti);
202 ST(&(xo[WS(os, 1)]), VADD(Tf, To), ovs, &(xo[WS(os, 1)]));
203 ST(&(xo[WS(os, 7)]), VADD(Tq, Tp), ovs, &(xo[WS(os, 1)]));
204 ST(&(xo[WS(os, 9)]), VSUB(To, Tf), ovs, &(xo[WS(os, 1)]));
205 ST(&(xo[WS(os, 3)]), VSUB(Tp, Tq), ovs, &(xo[WS(os, 1)]));
206 }
207 {
208 V Tx, TG, TE, TF, TC, TD;
209 Tx = VBYI(VFNMS(LDK(KP951056516), Tw, VMUL(LDK(KP587785252), Tt)));
210 TG = VBYI(VFMA(LDK(KP951056516), Tt, VMUL(LDK(KP587785252), Tw)));
211 TC = VFNMS(LDK(KP250000000), TB, Ty);
212 TD = VMUL(LDK(KP559016994), VSUB(Tz, TA));
213 TE = VSUB(TC, TD);
214 TF = VADD(TD, TC);
215 ST(&(xo[WS(os, 2)]), VADD(Tx, TE), ovs, &(xo[0]));
216 ST(&(xo[WS(os, 6)]), VADD(TG, TF), ovs, &(xo[0]));
217 ST(&(xo[WS(os, 8)]), VSUB(TE, Tx), ovs, &(xo[0]));
218 ST(&(xo[WS(os, 4)]), VSUB(TF, TG), ovs, &(xo[0]));
219 }
220 }
221 }
222 VLEAVE();
223 }
224
225 static const kdft_desc desc = { 10, XSIMD_STRING("n1bv_10"), { 36, 6, 6, 0 }, &GENUS, 0, 0, 0, 0 };
226
XSIMD(codelet_n1bv_10)227 void XSIMD(codelet_n1bv_10) (planner *p) { X(kdft_register) (p, n1bv_10, &desc);
228 }
229
230 #endif
231