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 07:05:39 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 -twiddle-log3 -precompute-twiddles -n 4 -name t2sv_4 -include dft/simd/ts.h */
29 
30 /*
31  * This function contains 24 FP additions, 16 FP multiplications,
32  * (or, 16 additions, 8 multiplications, 8 fused multiply/add),
33  * 21 stack variables, 0 constants, and 16 memory accesses
34  */
35 #include "dft/simd/ts.h"
36 
t2sv_4(R * ri,R * ii,const R * W,stride rs,INT mb,INT me,INT ms)37 static void t2sv_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 * 4); m < me; m = m + (2 * VL), ri = ri + ((2 * VL) * ms), ii = ii + ((2 * VL) * ms), W = W + ((2 * VL) * 4), MAKE_VOLATILE_STRIDE(8, rs)) {
42 	       V T2, T6, T3, T5, T7, Tb, T4, Ta;
43 	       T2 = LDW(&(W[0]));
44 	       T6 = LDW(&(W[TWVL * 3]));
45 	       T3 = LDW(&(W[TWVL * 2]));
46 	       T4 = VMUL(T2, T3);
47 	       Ta = VMUL(T2, T6);
48 	       T5 = LDW(&(W[TWVL * 1]));
49 	       T7 = VFMA(T5, T6, T4);
50 	       Tb = VFNMS(T5, T3, Ta);
51 	       {
52 		    V T1, Tx, Td, Tw, Ti, Tq, Tm, Ts;
53 		    T1 = LD(&(ri[0]), ms, &(ri[0]));
54 		    Tx = LD(&(ii[0]), ms, &(ii[0]));
55 		    {
56 			 V T8, T9, Tc, Tv;
57 			 T8 = LD(&(ri[WS(rs, 2)]), ms, &(ri[0]));
58 			 T9 = VMUL(T7, T8);
59 			 Tc = LD(&(ii[WS(rs, 2)]), ms, &(ii[0]));
60 			 Tv = VMUL(T7, Tc);
61 			 Td = VFMA(Tb, Tc, T9);
62 			 Tw = VFNMS(Tb, T8, Tv);
63 		    }
64 		    {
65 			 V Tf, Tg, Th, Tp;
66 			 Tf = LD(&(ri[WS(rs, 1)]), ms, &(ri[WS(rs, 1)]));
67 			 Tg = VMUL(T2, Tf);
68 			 Th = LD(&(ii[WS(rs, 1)]), ms, &(ii[WS(rs, 1)]));
69 			 Tp = VMUL(T2, Th);
70 			 Ti = VFMA(T5, Th, Tg);
71 			 Tq = VFNMS(T5, Tf, Tp);
72 		    }
73 		    {
74 			 V Tj, Tk, Tl, Tr;
75 			 Tj = LD(&(ri[WS(rs, 3)]), ms, &(ri[WS(rs, 1)]));
76 			 Tk = VMUL(T3, Tj);
77 			 Tl = LD(&(ii[WS(rs, 3)]), ms, &(ii[WS(rs, 1)]));
78 			 Tr = VMUL(T3, Tl);
79 			 Tm = VFMA(T6, Tl, Tk);
80 			 Ts = VFNMS(T6, Tj, Tr);
81 		    }
82 		    {
83 			 V Te, Tn, Tu, Ty;
84 			 Te = VADD(T1, Td);
85 			 Tn = VADD(Ti, Tm);
86 			 ST(&(ri[WS(rs, 2)]), VSUB(Te, Tn), ms, &(ri[0]));
87 			 ST(&(ri[0]), VADD(Te, Tn), ms, &(ri[0]));
88 			 Tu = VADD(Tq, Ts);
89 			 Ty = VADD(Tw, Tx);
90 			 ST(&(ii[0]), VADD(Tu, Ty), ms, &(ii[0]));
91 			 ST(&(ii[WS(rs, 2)]), VSUB(Ty, Tu), ms, &(ii[0]));
92 		    }
93 		    {
94 			 V To, Tt, Tz, TA;
95 			 To = VSUB(T1, Td);
96 			 Tt = VSUB(Tq, Ts);
97 			 ST(&(ri[WS(rs, 3)]), VSUB(To, Tt), ms, &(ri[WS(rs, 1)]));
98 			 ST(&(ri[WS(rs, 1)]), VADD(To, Tt), ms, &(ri[WS(rs, 1)]));
99 			 Tz = VSUB(Tx, Tw);
100 			 TA = VSUB(Ti, Tm);
101 			 ST(&(ii[WS(rs, 1)]), VSUB(Tz, TA), ms, &(ii[WS(rs, 1)]));
102 			 ST(&(ii[WS(rs, 3)]), VADD(TA, Tz), ms, &(ii[WS(rs, 1)]));
103 		    }
104 	       }
105 	  }
106      }
107      VLEAVE();
108 }
109 
110 static const tw_instr twinstr[] = {
111      VTW(0, 1),
112      VTW(0, 3),
113      { TW_NEXT, (2 * VL), 0 }
114 };
115 
116 static const ct_desc desc = { 4, XSIMD_STRING("t2sv_4"), twinstr, &GENUS, { 16, 8, 8, 0 }, 0, 0, 0 };
117 
XSIMD(codelet_t2sv_4)118 void XSIMD(codelet_t2sv_4) (planner *p) {
119      X(kdft_dit_register) (p, t2sv_4, &desc);
120 }
121 #else
122 
123 /* Generated by: ../../../genfft/gen_twiddle.native -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -n 4 -name t2sv_4 -include dft/simd/ts.h */
124 
125 /*
126  * This function contains 24 FP additions, 16 FP multiplications,
127  * (or, 16 additions, 8 multiplications, 8 fused multiply/add),
128  * 21 stack variables, 0 constants, and 16 memory accesses
129  */
130 #include "dft/simd/ts.h"
131 
t2sv_4(R * ri,R * ii,const R * W,stride rs,INT mb,INT me,INT ms)132 static void t2sv_4(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
133 {
134      {
135 	  INT m;
136 	  for (m = mb, W = W + (mb * 4); m < me; m = m + (2 * VL), ri = ri + ((2 * VL) * ms), ii = ii + ((2 * VL) * ms), W = W + ((2 * VL) * 4), MAKE_VOLATILE_STRIDE(8, rs)) {
137 	       V T2, T4, T3, T5, T6, T8;
138 	       T2 = LDW(&(W[0]));
139 	       T4 = LDW(&(W[TWVL * 1]));
140 	       T3 = LDW(&(W[TWVL * 2]));
141 	       T5 = LDW(&(W[TWVL * 3]));
142 	       T6 = VFMA(T2, T3, VMUL(T4, T5));
143 	       T8 = VFNMS(T4, T3, VMUL(T2, T5));
144 	       {
145 		    V T1, Tp, Ta, To, Te, Tk, Th, Tl, T7, T9;
146 		    T1 = LD(&(ri[0]), ms, &(ri[0]));
147 		    Tp = LD(&(ii[0]), ms, &(ii[0]));
148 		    T7 = LD(&(ri[WS(rs, 2)]), ms, &(ri[0]));
149 		    T9 = LD(&(ii[WS(rs, 2)]), ms, &(ii[0]));
150 		    Ta = VFMA(T6, T7, VMUL(T8, T9));
151 		    To = VFNMS(T8, T7, VMUL(T6, T9));
152 		    {
153 			 V Tc, Td, Tf, Tg;
154 			 Tc = LD(&(ri[WS(rs, 1)]), ms, &(ri[WS(rs, 1)]));
155 			 Td = LD(&(ii[WS(rs, 1)]), ms, &(ii[WS(rs, 1)]));
156 			 Te = VFMA(T2, Tc, VMUL(T4, Td));
157 			 Tk = VFNMS(T4, Tc, VMUL(T2, Td));
158 			 Tf = LD(&(ri[WS(rs, 3)]), ms, &(ri[WS(rs, 1)]));
159 			 Tg = LD(&(ii[WS(rs, 3)]), ms, &(ii[WS(rs, 1)]));
160 			 Th = VFMA(T3, Tf, VMUL(T5, Tg));
161 			 Tl = VFNMS(T5, Tf, VMUL(T3, Tg));
162 		    }
163 		    {
164 			 V Tb, Ti, Tn, Tq;
165 			 Tb = VADD(T1, Ta);
166 			 Ti = VADD(Te, Th);
167 			 ST(&(ri[WS(rs, 2)]), VSUB(Tb, Ti), ms, &(ri[0]));
168 			 ST(&(ri[0]), VADD(Tb, 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, Ta);
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(Te, 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      }
188      VLEAVE();
189 }
190 
191 static const tw_instr twinstr[] = {
192      VTW(0, 1),
193      VTW(0, 3),
194      { TW_NEXT, (2 * VL), 0 }
195 };
196 
197 static const ct_desc desc = { 4, XSIMD_STRING("t2sv_4"), twinstr, &GENUS, { 16, 8, 8, 0 }, 0, 0, 0 };
198 
XSIMD(codelet_t2sv_4)199 void XSIMD(codelet_t2sv_4) (planner *p) {
200      X(kdft_dit_register) (p, t2sv_4, &desc);
201 }
202 #endif
203