annotate fft/fftw/fftw-3.3.4/dft/simd/common/q1bv_2.c @ 40:223f770b5341 kissfft-double tip

Try a double-precision kissfft
author Chris Cannam
date Wed, 07 Sep 2016 10:40:32 +0100
parents 26056e866c29
children
rev   line source
Chris@19 1 /*
Chris@19 2 * Copyright (c) 2003, 2007-14 Matteo Frigo
Chris@19 3 * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology
Chris@19 4 *
Chris@19 5 * This program is free software; you can redistribute it and/or modify
Chris@19 6 * it under the terms of the GNU General Public License as published by
Chris@19 7 * the Free Software Foundation; either version 2 of the License, or
Chris@19 8 * (at your option) any later version.
Chris@19 9 *
Chris@19 10 * This program is distributed in the hope that it will be useful,
Chris@19 11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
Chris@19 12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
Chris@19 13 * GNU General Public License for more details.
Chris@19 14 *
Chris@19 15 * You should have received a copy of the GNU General Public License
Chris@19 16 * along with this program; if not, write to the Free Software
Chris@19 17 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
Chris@19 18 *
Chris@19 19 */
Chris@19 20
Chris@19 21 /* This file was automatically generated --- DO NOT EDIT */
Chris@19 22 /* Generated on Tue Mar 4 13:47:58 EST 2014 */
Chris@19 23
Chris@19 24 #include "codelet-dft.h"
Chris@19 25
Chris@19 26 #ifdef HAVE_FMA
Chris@19 27
Chris@19 28 /* Generated by: ../../../genfft/gen_twidsq_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -n 2 -dif -name q1bv_2 -include q1b.h -sign 1 */
Chris@19 29
Chris@19 30 /*
Chris@19 31 * This function contains 6 FP additions, 4 FP multiplications,
Chris@19 32 * (or, 6 additions, 4 multiplications, 0 fused multiply/add),
Chris@19 33 * 8 stack variables, 0 constants, and 8 memory accesses
Chris@19 34 */
Chris@19 35 #include "q1b.h"
Chris@19 36
Chris@19 37 static void q1bv_2(R *ri, R *ii, const R *W, stride rs, stride vs, INT mb, INT me, INT ms)
Chris@19 38 {
Chris@19 39 {
Chris@19 40 INT m;
Chris@19 41 R *x;
Chris@19 42 x = ii;
Chris@19 43 for (m = mb, W = W + (mb * ((TWVL / VL) * 2)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 2), MAKE_VOLATILE_STRIDE(4, rs), MAKE_VOLATILE_STRIDE(4, vs)) {
Chris@19 44 V T1, T2, T4, T5, T3, T6;
Chris@19 45 T1 = LD(&(x[0]), ms, &(x[0]));
Chris@19 46 T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@19 47 T4 = LD(&(x[WS(vs, 1)]), ms, &(x[WS(vs, 1)]));
Chris@19 48 T5 = LD(&(x[WS(vs, 1) + WS(rs, 1)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
Chris@19 49 ST(&(x[0]), VADD(T1, T2), ms, &(x[0]));
Chris@19 50 T3 = BYTW(&(W[0]), VSUB(T1, T2));
Chris@19 51 ST(&(x[WS(rs, 1)]), VADD(T4, T5), ms, &(x[WS(rs, 1)]));
Chris@19 52 T6 = BYTW(&(W[0]), VSUB(T4, T5));
Chris@19 53 ST(&(x[WS(vs, 1)]), T3, ms, &(x[WS(vs, 1)]));
Chris@19 54 ST(&(x[WS(vs, 1) + WS(rs, 1)]), T6, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
Chris@19 55 }
Chris@19 56 }
Chris@19 57 VLEAVE();
Chris@19 58 }
Chris@19 59
Chris@19 60 static const tw_instr twinstr[] = {
Chris@19 61 VTW(0, 1),
Chris@19 62 {TW_NEXT, VL, 0}
Chris@19 63 };
Chris@19 64
Chris@19 65 static const ct_desc desc = { 2, XSIMD_STRING("q1bv_2"), twinstr, &GENUS, {6, 4, 0, 0}, 0, 0, 0 };
Chris@19 66
Chris@19 67 void XSIMD(codelet_q1bv_2) (planner *p) {
Chris@19 68 X(kdft_difsq_register) (p, q1bv_2, &desc);
Chris@19 69 }
Chris@19 70 #else /* HAVE_FMA */
Chris@19 71
Chris@19 72 /* Generated by: ../../../genfft/gen_twidsq_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 2 -dif -name q1bv_2 -include q1b.h -sign 1 */
Chris@19 73
Chris@19 74 /*
Chris@19 75 * This function contains 6 FP additions, 4 FP multiplications,
Chris@19 76 * (or, 6 additions, 4 multiplications, 0 fused multiply/add),
Chris@19 77 * 8 stack variables, 0 constants, and 8 memory accesses
Chris@19 78 */
Chris@19 79 #include "q1b.h"
Chris@19 80
Chris@19 81 static void q1bv_2(R *ri, R *ii, const R *W, stride rs, stride vs, INT mb, INT me, INT ms)
Chris@19 82 {
Chris@19 83 {
Chris@19 84 INT m;
Chris@19 85 R *x;
Chris@19 86 x = ii;
Chris@19 87 for (m = mb, W = W + (mb * ((TWVL / VL) * 2)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 2), MAKE_VOLATILE_STRIDE(4, rs), MAKE_VOLATILE_STRIDE(4, vs)) {
Chris@19 88 V T1, T2, T3, T4, T5, T6;
Chris@19 89 T1 = LD(&(x[0]), ms, &(x[0]));
Chris@19 90 T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@19 91 T3 = BYTW(&(W[0]), VSUB(T1, T2));
Chris@19 92 T4 = LD(&(x[WS(vs, 1)]), ms, &(x[WS(vs, 1)]));
Chris@19 93 T5 = LD(&(x[WS(vs, 1) + WS(rs, 1)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
Chris@19 94 T6 = BYTW(&(W[0]), VSUB(T4, T5));
Chris@19 95 ST(&(x[WS(vs, 1)]), T3, ms, &(x[WS(vs, 1)]));
Chris@19 96 ST(&(x[WS(vs, 1) + WS(rs, 1)]), T6, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
Chris@19 97 ST(&(x[0]), VADD(T1, T2), ms, &(x[0]));
Chris@19 98 ST(&(x[WS(rs, 1)]), VADD(T4, T5), ms, &(x[WS(rs, 1)]));
Chris@19 99 }
Chris@19 100 }
Chris@19 101 VLEAVE();
Chris@19 102 }
Chris@19 103
Chris@19 104 static const tw_instr twinstr[] = {
Chris@19 105 VTW(0, 1),
Chris@19 106 {TW_NEXT, VL, 0}
Chris@19 107 };
Chris@19 108
Chris@19 109 static const ct_desc desc = { 2, XSIMD_STRING("q1bv_2"), twinstr, &GENUS, {6, 4, 0, 0}, 0, 0, 0 };
Chris@19 110
Chris@19 111 void XSIMD(codelet_q1bv_2) (planner *p) {
Chris@19 112 X(kdft_difsq_register) (p, q1bv_2, &desc);
Chris@19 113 }
Chris@19 114 #endif /* HAVE_FMA */