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 07:05:37 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_twiddle.native -fma -simd -compact -variables 4 -pipeline-latency 8 -n 4 -name t1sv_4 -include dft/simd/ts.h */
29 
30 /*
31  * This function contains 22 FP additions, 12 FP multiplications,
32  * (or, 16 additions, 6 multiplications, 6 fused multiply/add),
33  * 15 stack variables, 0 constants, and 16 memory accesses
34  */
35 #include "dft/simd/ts.h"
36 
t1sv_4(R * ri,R * ii,const R * W,stride rs,INT mb,INT me,INT ms)37 static void t1sv_4(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39      {
40 	  INT m;
41 	  for (m = mb, W = W + (mb * 6); m < me; m = m + (2 * VL), ri = ri + ((2 * VL) * ms), ii = ii + ((2 * VL) * ms), W = W + ((2 * VL) * 6), MAKE_VOLATILE_STRIDE(8, rs)) {
42 	       V T1, Tv, T7, Tu, Te, To, Tk, Tq;
43 	       T1 = LD(&(ri[0]), ms, &(ri[0]));
44 	       Tv = LD(&(ii[0]), ms, &(ii[0]));
45 	       {
46 		    V T3, T6, T4, Tt, T2, T5;
47 		    T3 = LD(&(ri[WS(rs, 2)]), ms, &(ri[0]));
48 		    T6 = LD(&(ii[WS(rs, 2)]), ms, &(ii[0]));
49 		    T2 = LDW(&(W[TWVL * 2]));
50 		    T4 = VMUL(T2, T3);
51 		    Tt = VMUL(T2, T6);
52 		    T5 = LDW(&(W[TWVL * 3]));
53 		    T7 = VFMA(T5, T6, T4);
54 		    Tu = VFNMS(T5, T3, Tt);
55 	       }
56 	       {
57 		    V Ta, Td, Tb, Tn, T9, Tc;
58 		    Ta = LD(&(ri[WS(rs, 1)]), ms, &(ri[WS(rs, 1)]));
59 		    Td = LD(&(ii[WS(rs, 1)]), ms, &(ii[WS(rs, 1)]));
60 		    T9 = LDW(&(W[0]));
61 		    Tb = VMUL(T9, Ta);
62 		    Tn = VMUL(T9, Td);
63 		    Tc = LDW(&(W[TWVL * 1]));
64 		    Te = VFMA(Tc, Td, Tb);
65 		    To = VFNMS(Tc, Ta, Tn);
66 	       }
67 	       {
68 		    V Tg, Tj, Th, Tp, Tf, Ti;
69 		    Tg = LD(&(ri[WS(rs, 3)]), ms, &(ri[WS(rs, 1)]));
70 		    Tj = LD(&(ii[WS(rs, 3)]), ms, &(ii[WS(rs, 1)]));
71 		    Tf = LDW(&(W[TWVL * 4]));
72 		    Th = VMUL(Tf, Tg);
73 		    Tp = VMUL(Tf, Tj);
74 		    Ti = LDW(&(W[TWVL * 5]));
75 		    Tk = VFMA(Ti, Tj, Th);
76 		    Tq = VFNMS(Ti, Tg, Tp);
77 	       }
78 	       {
79 		    V T8, Tl, Ts, Tw;
80 		    T8 = VADD(T1, T7);
81 		    Tl = VADD(Te, Tk);
82 		    ST(&(ri[WS(rs, 2)]), VSUB(T8, Tl), ms, &(ri[0]));
83 		    ST(&(ri[0]), VADD(T8, Tl), ms, &(ri[0]));
84 		    Ts = VADD(To, Tq);
85 		    Tw = VADD(Tu, Tv);
86 		    ST(&(ii[0]), VADD(Ts, Tw), ms, &(ii[0]));
87 		    ST(&(ii[WS(rs, 2)]), VSUB(Tw, Ts), ms, &(ii[0]));
88 	       }
89 	       {
90 		    V Tm, Tr, Tx, Ty;
91 		    Tm = VSUB(T1, T7);
92 		    Tr = VSUB(To, Tq);
93 		    ST(&(ri[WS(rs, 3)]), VSUB(Tm, Tr), ms, &(ri[WS(rs, 1)]));
94 		    ST(&(ri[WS(rs, 1)]), VADD(Tm, Tr), ms, &(ri[WS(rs, 1)]));
95 		    Tx = VSUB(Tv, Tu);
96 		    Ty = VSUB(Te, Tk);
97 		    ST(&(ii[WS(rs, 1)]), VSUB(Tx, Ty), ms, &(ii[WS(rs, 1)]));
98 		    ST(&(ii[WS(rs, 3)]), VADD(Ty, Tx), ms, &(ii[WS(rs, 1)]));
99 	       }
100 	  }
101      }
102      VLEAVE();
103 }
104 
105 static const tw_instr twinstr[] = {
106      VTW(0, 1),
107      VTW(0, 2),
108      VTW(0, 3),
109      { TW_NEXT, (2 * VL), 0 }
110 };
111 
112 static const ct_desc desc = { 4, XSIMD_STRING("t1sv_4"), twinstr, &GENUS, { 16, 6, 6, 0 }, 0, 0, 0 };
113 
XSIMD(codelet_t1sv_4)114 void XSIMD(codelet_t1sv_4) (planner *p) {
115      X(kdft_dit_register) (p, t1sv_4, &desc);
116 }
117 #else
118 
119 /* Generated by: ../../../genfft/gen_twiddle.native -simd -compact -variables 4 -pipeline-latency 8 -n 4 -name t1sv_4 -include dft/simd/ts.h */
120 
121 /*
122  * This function contains 22 FP additions, 12 FP multiplications,
123  * (or, 16 additions, 6 multiplications, 6 fused multiply/add),
124  * 13 stack variables, 0 constants, and 16 memory accesses
125  */
126 #include "dft/simd/ts.h"
127 
t1sv_4(R * ri,R * ii,const R * W,stride rs,INT mb,INT me,INT ms)128 static void t1sv_4(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
129 {
130      {
131 	  INT m;
132 	  for (m = mb, W = W + (mb * 6); m < me; m = m + (2 * VL), ri = ri + ((2 * VL) * ms), ii = ii + ((2 * VL) * ms), W = W + ((2 * VL) * 6), MAKE_VOLATILE_STRIDE(8, rs)) {
133 	       V T1, Tp, T6, To, Tc, Tk, Th, Tl;
134 	       T1 = LD(&(ri[0]), ms, &(ri[0]));
135 	       Tp = LD(&(ii[0]), ms, &(ii[0]));
136 	       {
137 		    V T3, T5, T2, T4;
138 		    T3 = LD(&(ri[WS(rs, 2)]), ms, &(ri[0]));
139 		    T5 = LD(&(ii[WS(rs, 2)]), ms, &(ii[0]));
140 		    T2 = LDW(&(W[TWVL * 2]));
141 		    T4 = LDW(&(W[TWVL * 3]));
142 		    T6 = VFMA(T2, T3, VMUL(T4, T5));
143 		    To = VFNMS(T4, T3, VMUL(T2, T5));
144 	       }
145 	       {
146 		    V T9, Tb, T8, Ta;
147 		    T9 = LD(&(ri[WS(rs, 1)]), ms, &(ri[WS(rs, 1)]));
148 		    Tb = LD(&(ii[WS(rs, 1)]), ms, &(ii[WS(rs, 1)]));
149 		    T8 = LDW(&(W[0]));
150 		    Ta = LDW(&(W[TWVL * 1]));
151 		    Tc = VFMA(T8, T9, VMUL(Ta, Tb));
152 		    Tk = VFNMS(Ta, T9, VMUL(T8, Tb));
153 	       }
154 	       {
155 		    V Te, Tg, Td, Tf;
156 		    Te = LD(&(ri[WS(rs, 3)]), ms, &(ri[WS(rs, 1)]));
157 		    Tg = LD(&(ii[WS(rs, 3)]), ms, &(ii[WS(rs, 1)]));
158 		    Td = LDW(&(W[TWVL * 4]));
159 		    Tf = LDW(&(W[TWVL * 5]));
160 		    Th = VFMA(Td, Te, VMUL(Tf, Tg));
161 		    Tl = VFNMS(Tf, Te, VMUL(Td, Tg));
162 	       }
163 	       {
164 		    V T7, Ti, Tn, Tq;
165 		    T7 = VADD(T1, T6);
166 		    Ti = VADD(Tc, Th);
167 		    ST(&(ri[WS(rs, 2)]), VSUB(T7, Ti), ms, &(ri[0]));
168 		    ST(&(ri[0]), VADD(T7, Ti), ms, &(ri[0]));
169 		    Tn = VADD(Tk, Tl);
170 		    Tq = VADD(To, Tp);
171 		    ST(&(ii[0]), VADD(Tn, Tq), ms, &(ii[0]));
172 		    ST(&(ii[WS(rs, 2)]), VSUB(Tq, Tn), ms, &(ii[0]));
173 	       }
174 	       {
175 		    V Tj, Tm, Tr, Ts;
176 		    Tj = VSUB(T1, T6);
177 		    Tm = VSUB(Tk, Tl);
178 		    ST(&(ri[WS(rs, 3)]), VSUB(Tj, Tm), ms, &(ri[WS(rs, 1)]));
179 		    ST(&(ri[WS(rs, 1)]), VADD(Tj, Tm), ms, &(ri[WS(rs, 1)]));
180 		    Tr = VSUB(Tp, To);
181 		    Ts = VSUB(Tc, Th);
182 		    ST(&(ii[WS(rs, 1)]), VSUB(Tr, Ts), ms, &(ii[WS(rs, 1)]));
183 		    ST(&(ii[WS(rs, 3)]), VADD(Ts, Tr), ms, &(ii[WS(rs, 1)]));
184 	       }
185 	  }
186      }
187      VLEAVE();
188 }
189 
190 static const tw_instr twinstr[] = {
191      VTW(0, 1),
192      VTW(0, 2),
193      VTW(0, 3),
194      { TW_NEXT, (2 * VL), 0 }
195 };
196 
197 static const ct_desc desc = { 4, XSIMD_STRING("t1sv_4"), twinstr, &GENUS, { 16, 6, 6, 0 }, 0, 0, 0 };
198 
XSIMD(codelet_t1sv_4)199 void XSIMD(codelet_t1sv_4) (planner *p) {
200      X(kdft_dit_register) (p, t1sv_4, &desc);
201 }
202 #endif
203