annotate src/fftw-3.3.5/rdft/scalar/r2cb/hc2cbdft_2.c @ 148:b4bfdf10c4b3

Update Win64 capnp builds to v0.6
author Chris Cannam <cannam@all-day-breakfast.com>
date Mon, 22 May 2017 18:56:49 +0100
parents 7867fa7e1b6b
children
rev   line source
cannam@127 1 /*
cannam@127 2 * Copyright (c) 2003, 2007-14 Matteo Frigo
cannam@127 3 * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology
cannam@127 4 *
cannam@127 5 * This program is free software; you can redistribute it and/or modify
cannam@127 6 * it under the terms of the GNU General Public License as published by
cannam@127 7 * the Free Software Foundation; either version 2 of the License, or
cannam@127 8 * (at your option) any later version.
cannam@127 9 *
cannam@127 10 * This program is distributed in the hope that it will be useful,
cannam@127 11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
cannam@127 12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
cannam@127 13 * GNU General Public License for more details.
cannam@127 14 *
cannam@127 15 * You should have received a copy of the GNU General Public License
cannam@127 16 * along with this program; if not, write to the Free Software
cannam@127 17 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
cannam@127 18 *
cannam@127 19 */
cannam@127 20
cannam@127 21 /* This file was automatically generated --- DO NOT EDIT */
cannam@127 22 /* Generated on Sat Jul 30 16:51:54 EDT 2016 */
cannam@127 23
cannam@127 24 #include "codelet-rdft.h"
cannam@127 25
cannam@127 26 #ifdef HAVE_FMA
cannam@127 27
cannam@127 28 /* Generated by: ../../../genfft/gen_hc2cdft.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -sign 1 -n 2 -dif -name hc2cbdft_2 -include hc2cb.h */
cannam@127 29
cannam@127 30 /*
cannam@127 31 * This function contains 10 FP additions, 4 FP multiplications,
cannam@127 32 * (or, 8 additions, 2 multiplications, 2 fused multiply/add),
cannam@127 33 * 11 stack variables, 0 constants, and 8 memory accesses
cannam@127 34 */
cannam@127 35 #include "hc2cb.h"
cannam@127 36
cannam@127 37 static void hc2cbdft_2(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
cannam@127 38 {
cannam@127 39 {
cannam@127 40 INT m;
cannam@127 41 for (m = mb, W = W + ((mb - 1) * 2); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 2, MAKE_VOLATILE_STRIDE(8, rs)) {
cannam@127 42 E T9, Ta, T3, Tc, T7, T4;
cannam@127 43 {
cannam@127 44 E T1, T2, T5, T6;
cannam@127 45 T1 = Ip[0];
cannam@127 46 T2 = Im[0];
cannam@127 47 T5 = Rp[0];
cannam@127 48 T6 = Rm[0];
cannam@127 49 T9 = W[1];
cannam@127 50 Ta = T1 + T2;
cannam@127 51 T3 = T1 - T2;
cannam@127 52 Tc = T5 + T6;
cannam@127 53 T7 = T5 - T6;
cannam@127 54 T4 = W[0];
cannam@127 55 }
cannam@127 56 {
cannam@127 57 E Td, T8, Te, Tb;
cannam@127 58 Td = T9 * T7;
cannam@127 59 T8 = T4 * T7;
cannam@127 60 Te = FMA(T4, Ta, Td);
cannam@127 61 Tb = FNMS(T9, Ta, T8);
cannam@127 62 Rm[0] = Tc + Te;
cannam@127 63 Rp[0] = Tc - Te;
cannam@127 64 Im[0] = Tb - T3;
cannam@127 65 Ip[0] = T3 + Tb;
cannam@127 66 }
cannam@127 67 }
cannam@127 68 }
cannam@127 69 }
cannam@127 70
cannam@127 71 static const tw_instr twinstr[] = {
cannam@127 72 {TW_FULL, 1, 2},
cannam@127 73 {TW_NEXT, 1, 0}
cannam@127 74 };
cannam@127 75
cannam@127 76 static const hc2c_desc desc = { 2, "hc2cbdft_2", twinstr, &GENUS, {8, 2, 2, 0} };
cannam@127 77
cannam@127 78 void X(codelet_hc2cbdft_2) (planner *p) {
cannam@127 79 X(khc2c_register) (p, hc2cbdft_2, &desc, HC2C_VIA_DFT);
cannam@127 80 }
cannam@127 81 #else /* HAVE_FMA */
cannam@127 82
cannam@127 83 /* Generated by: ../../../genfft/gen_hc2cdft.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 2 -dif -name hc2cbdft_2 -include hc2cb.h */
cannam@127 84
cannam@127 85 /*
cannam@127 86 * This function contains 10 FP additions, 4 FP multiplications,
cannam@127 87 * (or, 8 additions, 2 multiplications, 2 fused multiply/add),
cannam@127 88 * 9 stack variables, 0 constants, and 8 memory accesses
cannam@127 89 */
cannam@127 90 #include "hc2cb.h"
cannam@127 91
cannam@127 92 static void hc2cbdft_2(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
cannam@127 93 {
cannam@127 94 {
cannam@127 95 INT m;
cannam@127 96 for (m = mb, W = W + ((mb - 1) * 2); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 2, MAKE_VOLATILE_STRIDE(8, rs)) {
cannam@127 97 E T3, T9, T7, Tb;
cannam@127 98 {
cannam@127 99 E T1, T2, T5, T6;
cannam@127 100 T1 = Ip[0];
cannam@127 101 T2 = Im[0];
cannam@127 102 T3 = T1 - T2;
cannam@127 103 T9 = T1 + T2;
cannam@127 104 T5 = Rp[0];
cannam@127 105 T6 = Rm[0];
cannam@127 106 T7 = T5 - T6;
cannam@127 107 Tb = T5 + T6;
cannam@127 108 }
cannam@127 109 {
cannam@127 110 E Ta, Tc, T4, T8;
cannam@127 111 T4 = W[0];
cannam@127 112 T8 = W[1];
cannam@127 113 Ta = FNMS(T8, T9, T4 * T7);
cannam@127 114 Tc = FMA(T8, T7, T4 * T9);
cannam@127 115 Ip[0] = T3 + Ta;
cannam@127 116 Rp[0] = Tb - Tc;
cannam@127 117 Im[0] = Ta - T3;
cannam@127 118 Rm[0] = Tb + Tc;
cannam@127 119 }
cannam@127 120 }
cannam@127 121 }
cannam@127 122 }
cannam@127 123
cannam@127 124 static const tw_instr twinstr[] = {
cannam@127 125 {TW_FULL, 1, 2},
cannam@127 126 {TW_NEXT, 1, 0}
cannam@127 127 };
cannam@127 128
cannam@127 129 static const hc2c_desc desc = { 2, "hc2cbdft_2", twinstr, &GENUS, {8, 2, 2, 0} };
cannam@127 130
cannam@127 131 void X(codelet_hc2cbdft_2) (planner *p) {
cannam@127 132 X(khc2c_register) (p, hc2cbdft_2, &desc, HC2C_VIA_DFT);
cannam@127 133 }
cannam@127 134 #endif /* HAVE_FMA */