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 07:04:56 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 8 -name n2bv_8 -with-ostride 2 -include dft/simd/n2b.h -store-multiple 2 */
29
30 /*
31 * This function contains 26 FP additions, 10 FP multiplications,
32 * (or, 16 additions, 0 multiplications, 10 fused multiply/add),
33 * 24 stack variables, 1 constants, and 20 memory accesses
34 */
35 #include "dft/simd/n2b.h"
36
n2bv_8(const R * ri,const R * ii,R * ro,R * io,stride is,stride os,INT v,INT ivs,INT ovs)37 static void n2bv_8(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
38 {
39 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
40 {
41 INT i;
42 const R *xi;
43 R *xo;
44 xi = ii;
45 xo = io;
46 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(16, is), MAKE_VOLATILE_STRIDE(16, os)) {
47 V T3, Tj, Te, Tk, Ta, Tn, Tf, Tm, Tr, Tu;
48 {
49 V T1, T2, Tc, Td;
50 T1 = LD(&(xi[0]), ivs, &(xi[0]));
51 T2 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
52 T3 = VSUB(T1, T2);
53 Tj = VADD(T1, T2);
54 Tc = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
55 Td = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
56 Te = VSUB(Tc, Td);
57 Tk = VADD(Tc, Td);
58 {
59 V T4, T5, T6, T7, T8, T9;
60 T4 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
61 T5 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
62 T6 = VSUB(T4, T5);
63 T7 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
64 T8 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
65 T9 = VSUB(T7, T8);
66 Ta = VADD(T6, T9);
67 Tn = VADD(T7, T8);
68 Tf = VSUB(T6, T9);
69 Tm = VADD(T4, T5);
70 }
71 }
72 {
73 V Ts, Tb, Tg, Tp, Tq, Tt;
74 Tb = VFNMS(LDK(KP707106781), Ta, T3);
75 Tg = VFNMS(LDK(KP707106781), Tf, Te);
76 Tr = VFNMSI(Tg, Tb);
77 STM2(&(xo[6]), Tr, ovs, &(xo[2]));
78 Ts = VFMAI(Tg, Tb);
79 STM2(&(xo[10]), Ts, ovs, &(xo[2]));
80 Tp = VADD(Tj, Tk);
81 Tq = VADD(Tm, Tn);
82 Tt = VSUB(Tp, Tq);
83 STM2(&(xo[8]), Tt, ovs, &(xo[0]));
84 STN2(&(xo[8]), Tt, Ts, ovs);
85 Tu = VADD(Tp, Tq);
86 STM2(&(xo[0]), Tu, ovs, &(xo[0]));
87 }
88 {
89 V Tw, Th, Ti, Tv;
90 Th = VFMA(LDK(KP707106781), Ta, T3);
91 Ti = VFMA(LDK(KP707106781), Tf, Te);
92 Tv = VFMAI(Ti, Th);
93 STM2(&(xo[2]), Tv, ovs, &(xo[2]));
94 STN2(&(xo[0]), Tu, Tv, ovs);
95 Tw = VFNMSI(Ti, Th);
96 STM2(&(xo[14]), Tw, ovs, &(xo[2]));
97 {
98 V Tl, To, Tx, Ty;
99 Tl = VSUB(Tj, Tk);
100 To = VSUB(Tm, Tn);
101 Tx = VFNMSI(To, Tl);
102 STM2(&(xo[12]), Tx, ovs, &(xo[0]));
103 STN2(&(xo[12]), Tx, Tw, ovs);
104 Ty = VFMAI(To, Tl);
105 STM2(&(xo[4]), Ty, ovs, &(xo[0]));
106 STN2(&(xo[4]), Ty, Tr, ovs);
107 }
108 }
109 }
110 }
111 VLEAVE();
112 }
113
114 static const kdft_desc desc = { 8, XSIMD_STRING("n2bv_8"), { 16, 0, 10, 0 }, &GENUS, 0, 2, 0, 0 };
115
XSIMD(codelet_n2bv_8)116 void XSIMD(codelet_n2bv_8) (planner *p) { X(kdft_register) (p, n2bv_8, &desc);
117 }
118
119 #else
120
121 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 8 -name n2bv_8 -with-ostride 2 -include dft/simd/n2b.h -store-multiple 2 */
122
123 /*
124 * This function contains 26 FP additions, 2 FP multiplications,
125 * (or, 26 additions, 2 multiplications, 0 fused multiply/add),
126 * 24 stack variables, 1 constants, and 20 memory accesses
127 */
128 #include "dft/simd/n2b.h"
129
n2bv_8(const R * ri,const R * ii,R * ro,R * io,stride is,stride os,INT v,INT ivs,INT ovs)130 static void n2bv_8(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
131 {
132 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
133 {
134 INT i;
135 const R *xi;
136 R *xo;
137 xi = ii;
138 xo = io;
139 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(16, is), MAKE_VOLATILE_STRIDE(16, os)) {
140 V Ta, Tk, Te, Tj, T7, Tn, Tf, Tm, Tr, Tu;
141 {
142 V T8, T9, Tc, Td;
143 T8 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
144 T9 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
145 Ta = VSUB(T8, T9);
146 Tk = VADD(T8, T9);
147 Tc = LD(&(xi[0]), ivs, &(xi[0]));
148 Td = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
149 Te = VSUB(Tc, Td);
150 Tj = VADD(Tc, Td);
151 {
152 V T1, T2, T3, T4, T5, T6;
153 T1 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
154 T2 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
155 T3 = VSUB(T1, T2);
156 T4 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
157 T5 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
158 T6 = VSUB(T4, T5);
159 T7 = VMUL(LDK(KP707106781), VSUB(T3, T6));
160 Tn = VADD(T4, T5);
161 Tf = VMUL(LDK(KP707106781), VADD(T3, T6));
162 Tm = VADD(T1, T2);
163 }
164 }
165 {
166 V Ts, Tb, Tg, Tp, Tq, Tt;
167 Tb = VBYI(VSUB(T7, Ta));
168 Tg = VSUB(Te, Tf);
169 Tr = VADD(Tb, Tg);
170 STM2(&(xo[6]), Tr, ovs, &(xo[2]));
171 Ts = VSUB(Tg, Tb);
172 STM2(&(xo[10]), Ts, ovs, &(xo[2]));
173 Tp = VADD(Tj, Tk);
174 Tq = VADD(Tm, Tn);
175 Tt = VSUB(Tp, Tq);
176 STM2(&(xo[8]), Tt, ovs, &(xo[0]));
177 STN2(&(xo[8]), Tt, Ts, ovs);
178 Tu = VADD(Tp, Tq);
179 STM2(&(xo[0]), Tu, ovs, &(xo[0]));
180 }
181 {
182 V Tw, Th, Ti, Tv;
183 Th = VBYI(VADD(Ta, T7));
184 Ti = VADD(Te, Tf);
185 Tv = VADD(Th, Ti);
186 STM2(&(xo[2]), Tv, ovs, &(xo[2]));
187 STN2(&(xo[0]), Tu, Tv, ovs);
188 Tw = VSUB(Ti, Th);
189 STM2(&(xo[14]), Tw, ovs, &(xo[2]));
190 {
191 V Tl, To, Tx, Ty;
192 Tl = VSUB(Tj, Tk);
193 To = VBYI(VSUB(Tm, Tn));
194 Tx = VSUB(Tl, To);
195 STM2(&(xo[12]), Tx, ovs, &(xo[0]));
196 STN2(&(xo[12]), Tx, Tw, ovs);
197 Ty = VADD(Tl, To);
198 STM2(&(xo[4]), Ty, ovs, &(xo[0]));
199 STN2(&(xo[4]), Ty, Tr, ovs);
200 }
201 }
202 }
203 }
204 VLEAVE();
205 }
206
207 static const kdft_desc desc = { 8, XSIMD_STRING("n2bv_8"), { 26, 2, 0, 0 }, &GENUS, 0, 2, 0, 0 };
208
XSIMD(codelet_n2bv_8)209 void XSIMD(codelet_n2bv_8) (planner *p) { X(kdft_register) (p, n2bv_8, &desc);
210 }
211
212 #endif
213