2
* Copyright (c) 2003, 2007-8 Matteo Frigo
3
* Copyright (c) 2003, 2007-8 Massachusetts Institute of Technology
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.
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.
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., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
21
/* This file was automatically generated --- DO NOT EDIT */
22
/* Generated on Sun Jul 12 06:39:53 EDT 2009 */
24
#include "codelet-dft.h"
28
/* Generated by: ../../../genfft/gen_notw_c -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -n 2 -name n2fv_2 -with-ostride 2 -include n2f.h -store-multiple 2 */
31
* This function contains 2 FP additions, 0 FP multiplications,
32
* (or, 2 additions, 0 multiplications, 0 fused multiply/add),
33
* 7 stack variables, 0 constants, and 5 memory accesses
37
static void n2fv_2(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
44
for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(is), MAKE_VOLATILE_STRIDE(os)) {
46
T1 = LD(&(xi[0]), ivs, &(xi[0]));
47
T2 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
49
STM2(&(xo[0]), T3, ovs, &(xo[0]));
51
STM2(&(xo[2]), T4, ovs, &(xo[2]));
52
STN2(&(xo[0]), T3, T4, ovs);
56
static const kdft_desc desc = { 2, "n2fv_2", {2, 0, 0, 0}, &GENUS, 0, 2, 0, 0 };
58
void X(codelet_n2fv_2) (planner *p) {
59
X(kdft_register) (p, n2fv_2, &desc);
64
/* Generated by: ../../../genfft/gen_notw_c -simd -compact -variables 4 -pipeline-latency 8 -n 2 -name n2fv_2 -with-ostride 2 -include n2f.h -store-multiple 2 */
67
* This function contains 2 FP additions, 0 FP multiplications,
68
* (or, 2 additions, 0 multiplications, 0 fused multiply/add),
69
* 7 stack variables, 0 constants, and 5 memory accesses
73
static void n2fv_2(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
80
for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(is), MAKE_VOLATILE_STRIDE(os)) {
82
T1 = LD(&(xi[0]), ivs, &(xi[0]));
83
T2 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
85
STM2(&(xo[2]), T3, ovs, &(xo[2]));
87
STM2(&(xo[0]), T4, ovs, &(xo[0]));
88
STN2(&(xo[0]), T4, T3, ovs);
92
static const kdft_desc desc = { 2, "n2fv_2", {2, 0, 0, 0}, &GENUS, 0, 2, 0, 0 };
94
void X(codelet_n2fv_2) (planner *p) {
95
X(kdft_register) (p, n2fv_2, &desc);