Chris@10
|
1 /*
|
Chris@10
|
2 * Copyright (c) 2003, 2007-11 Matteo Frigo
|
Chris@10
|
3 * Copyright (c) 2003, 2007-11 Massachusetts Institute of Technology
|
Chris@10
|
4 *
|
Chris@10
|
5 * This program is free software; you can redistribute it and/or modify
|
Chris@10
|
6 * it under the terms of the GNU General Public License as published by
|
Chris@10
|
7 * the Free Software Foundation; either version 2 of the License, or
|
Chris@10
|
8 * (at your option) any later version.
|
Chris@10
|
9 *
|
Chris@10
|
10 * This program is distributed in the hope that it will be useful,
|
Chris@10
|
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
|
Chris@10
|
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
Chris@10
|
13 * GNU General Public License for more details.
|
Chris@10
|
14 *
|
Chris@10
|
15 * You should have received a copy of the GNU General Public License
|
Chris@10
|
16 * along with this program; if not, write to the Free Software
|
Chris@10
|
17 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
Chris@10
|
18 *
|
Chris@10
|
19 */
|
Chris@10
|
20
|
Chris@10
|
21 /* This file was automatically generated --- DO NOT EDIT */
|
Chris@10
|
22 /* Generated on Sun Nov 25 07:42:29 EST 2012 */
|
Chris@10
|
23
|
Chris@10
|
24 #include "codelet-rdft.h"
|
Chris@10
|
25
|
Chris@10
|
26 #ifdef HAVE_FMA
|
Chris@10
|
27
|
Chris@10
|
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 2 -dit -name hc2cfdftv_2 -include hc2cfv.h */
|
Chris@10
|
29
|
Chris@10
|
30 /*
|
Chris@10
|
31 * This function contains 5 FP additions, 6 FP multiplications,
|
Chris@10
|
32 * (or, 3 additions, 4 multiplications, 2 fused multiply/add),
|
Chris@10
|
33 * 9 stack variables, 1 constants, and 4 memory accesses
|
Chris@10
|
34 */
|
Chris@10
|
35 #include "hc2cfv.h"
|
Chris@10
|
36
|
Chris@10
|
37 static void hc2cfdftv_2(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
|
Chris@10
|
38 {
|
Chris@10
|
39 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
|
Chris@10
|
40 {
|
Chris@10
|
41 INT m;
|
Chris@10
|
42 for (m = mb, W = W + ((mb - 1) * ((TWVL / VL) * 2)); m < me; m = m + VL, Rp = Rp + (VL * ms), Ip = Ip + (VL * ms), Rm = Rm - (VL * ms), Im = Im - (VL * ms), W = W + (TWVL * 2), MAKE_VOLATILE_STRIDE(8, rs)) {
|
Chris@10
|
43 V T1, T2, T4, T3, T5, T7, T6;
|
Chris@10
|
44 T1 = LD(&(Rp[0]), ms, &(Rp[0]));
|
Chris@10
|
45 T2 = LD(&(Rm[0]), -ms, &(Rm[0]));
|
Chris@10
|
46 T4 = LDW(&(W[0]));
|
Chris@10
|
47 T3 = VFMACONJ(T2, T1);
|
Chris@10
|
48 T5 = VZMULIJ(T4, VFNMSCONJ(T2, T1));
|
Chris@10
|
49 T7 = VCONJ(VMUL(LDK(KP500000000), VADD(T3, T5)));
|
Chris@10
|
50 T6 = VMUL(LDK(KP500000000), VSUB(T3, T5));
|
Chris@10
|
51 ST(&(Rm[0]), T7, -ms, &(Rm[0]));
|
Chris@10
|
52 ST(&(Rp[0]), T6, ms, &(Rp[0]));
|
Chris@10
|
53 }
|
Chris@10
|
54 }
|
Chris@10
|
55 VLEAVE();
|
Chris@10
|
56 }
|
Chris@10
|
57
|
Chris@10
|
58 static const tw_instr twinstr[] = {
|
Chris@10
|
59 VTW(1, 1),
|
Chris@10
|
60 {TW_NEXT, VL, 0}
|
Chris@10
|
61 };
|
Chris@10
|
62
|
Chris@10
|
63 static const hc2c_desc desc = { 2, XSIMD_STRING("hc2cfdftv_2"), twinstr, &GENUS, {3, 4, 2, 0} };
|
Chris@10
|
64
|
Chris@10
|
65 void XSIMD(codelet_hc2cfdftv_2) (planner *p) {
|
Chris@10
|
66 X(khc2c_register) (p, hc2cfdftv_2, &desc, HC2C_VIA_DFT);
|
Chris@10
|
67 }
|
Chris@10
|
68 #else /* HAVE_FMA */
|
Chris@10
|
69
|
Chris@10
|
70 /* Generated by: ../../../genfft/gen_hc2cdft_c.native -simd -compact -variables 4 -pipeline-latency 8 -trivial-stores -variables 32 -no-generate-bytw -n 2 -dit -name hc2cfdftv_2 -include hc2cfv.h */
|
Chris@10
|
71
|
Chris@10
|
72 /*
|
Chris@10
|
73 * This function contains 5 FP additions, 4 FP multiplications,
|
Chris@10
|
74 * (or, 5 additions, 4 multiplications, 0 fused multiply/add),
|
Chris@10
|
75 * 10 stack variables, 1 constants, and 4 memory accesses
|
Chris@10
|
76 */
|
Chris@10
|
77 #include "hc2cfv.h"
|
Chris@10
|
78
|
Chris@10
|
79 static void hc2cfdftv_2(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
|
Chris@10
|
80 {
|
Chris@10
|
81 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
|
Chris@10
|
82 {
|
Chris@10
|
83 INT m;
|
Chris@10
|
84 for (m = mb, W = W + ((mb - 1) * ((TWVL / VL) * 2)); m < me; m = m + VL, Rp = Rp + (VL * ms), Ip = Ip + (VL * ms), Rm = Rm - (VL * ms), Im = Im - (VL * ms), W = W + (TWVL * 2), MAKE_VOLATILE_STRIDE(8, rs)) {
|
Chris@10
|
85 V T4, T6, T1, T3, T2, T5, T7, T8;
|
Chris@10
|
86 T1 = LD(&(Rp[0]), ms, &(Rp[0]));
|
Chris@10
|
87 T2 = LD(&(Rm[0]), -ms, &(Rm[0]));
|
Chris@10
|
88 T3 = VCONJ(T2);
|
Chris@10
|
89 T4 = VADD(T1, T3);
|
Chris@10
|
90 T5 = LDW(&(W[0]));
|
Chris@10
|
91 T6 = VZMULIJ(T5, VSUB(T3, T1));
|
Chris@10
|
92 T7 = VCONJ(VMUL(LDK(KP500000000), VSUB(T4, T6)));
|
Chris@10
|
93 ST(&(Rm[0]), T7, -ms, &(Rm[0]));
|
Chris@10
|
94 T8 = VMUL(LDK(KP500000000), VADD(T4, T6));
|
Chris@10
|
95 ST(&(Rp[0]), T8, ms, &(Rp[0]));
|
Chris@10
|
96 }
|
Chris@10
|
97 }
|
Chris@10
|
98 VLEAVE();
|
Chris@10
|
99 }
|
Chris@10
|
100
|
Chris@10
|
101 static const tw_instr twinstr[] = {
|
Chris@10
|
102 VTW(1, 1),
|
Chris@10
|
103 {TW_NEXT, VL, 0}
|
Chris@10
|
104 };
|
Chris@10
|
105
|
Chris@10
|
106 static const hc2c_desc desc = { 2, XSIMD_STRING("hc2cfdftv_2"), twinstr, &GENUS, {5, 4, 0, 0} };
|
Chris@10
|
107
|
Chris@10
|
108 void XSIMD(codelet_hc2cfdftv_2) (planner *p) {
|
Chris@10
|
109 X(khc2c_register) (p, hc2cfdftv_2, &desc, HC2C_VIA_DFT);
|
Chris@10
|
110 }
|
Chris@10
|
111 #endif /* HAVE_FMA */
|