/* * Copyright (c) 2003 Matteo Frigo * Copyright (c) 2003 Massachusetts Institute of Technology * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by * the Free Software Foundation; either version 2 of the License, or * (at your option) any later version. * * This program is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * GNU General Public License for more details. * * You should have received a copy of the GNU General Public License * along with this program; if not, write to the Free Software * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA * */ /* This file was automatically generated --- DO NOT EDIT */ /* Generated on Sat Jul 5 21:39:14 EDT 2003 */ #include "codelet-dft.h" /* Generated by: /homee/stevenj/cvs/fftw3.0.1/genfft/gen_twidsq -compact -variables 4 -reload-twiddle -dif -n 4 -name q1_4 -include q.h */ /* * This function contains 88 FP additions, 48 FP multiplications, * (or, 64 additions, 24 multiplications, 24 fused multiply/add), * 37 stack variables, and 64 memory accesses */ /* * Generator Id's : * $Id: q1_4.c,v 1.1 2008/10/17 06:11:08 scuri Exp $ * $Id: q1_4.c,v 1.1 2008/10/17 06:11:08 scuri Exp $ * $Id: q1_4.c,v 1.1 2008/10/17 06:11:08 scuri Exp $ */ #include "q.h" static const R *q1_4(R *rio, R *iio, const R *W, stride is, stride vs, int m, int dist) { int i; for (i = m; i > 0; i = i - 1, rio = rio + dist, iio = iio + dist, W = W + 6) { E T3, Te, Tb, Tq, T6, T8, Th, Tr, Tv, TG, TD, TS, Ty, TA, TJ; E TT, TX, T18, T15, T1k, T10, T12, T1b, T1l, T1p, T1A, T1x, T1M, T1s, T1u; E T1D, T1N; { E T1, T2, T9, Ta; T1 = rio[0]; T2 = rio[WS(is, 2)]; T3 = T1 + T2; Te = T1 - T2; T9 = iio[0]; Ta = iio[WS(is, 2)]; Tb = T9 - Ta; Tq = T9 + Ta; } { E T4, T5, Tf, Tg; T4 = rio[WS(is, 1)]; T5 = rio[WS(is, 3)]; T6 = T4 + T5; T8 = T4 - T5; Tf = iio[WS(is, 1)]; Tg = iio[WS(is, 3)]; Th = Tf - Tg; Tr = Tf + Tg; } { E Tt, Tu, TB, TC; Tt = rio[WS(vs, 1)]; Tu = rio[WS(vs, 1) + WS(is, 2)]; Tv = Tt + Tu; TG = Tt - Tu; TB = iio[WS(vs, 1)]; TC = iio[WS(vs, 1) + WS(is, 2)]; TD = TB - TC; TS = TB + TC; } { E Tw, Tx, TH, TI; Tw = rio[WS(vs, 1) + WS(is, 1)]; Tx = rio[WS(vs, 1) + WS(is, 3)]; Ty = Tw + Tx; TA = Tw - Tx; TH = iio[WS(vs, 1) + WS(is, 1)]; TI = iio[WS(vs, 1) + WS(is, 3)]; TJ = TH - TI; TT = TH + TI; } { E TV, TW, T13, T14; TV = rio[WS(vs, 2)]; TW = rio[WS(vs, 2) + WS(is, 2)]; TX = TV + TW; T18 = TV - TW; T13 = iio[WS(vs, 2)]; T14 = iio[WS(vs, 2) + WS(is, 2)]; T15 = T13 - T14; T1k = T13 + T14; } { E TY, TZ, T19, T1a; TY = rio[WS(vs, 2) + WS(is, 1)]; TZ = rio[WS(vs, 2) + WS(is, 3)]; T10 = TY + TZ; T12 = TY - TZ; T19 = iio[WS(vs, 2) + WS(is, 1)]; T1a = iio[WS(vs, 2) + WS(is, 3)]; T1b = T19 - T1a; T1l = T19 + T1a; } { E T1n, T1o, T1v, T1w; T1n = rio[WS(vs, 3)]; T1o = rio[WS(vs, 3) + WS(is, 2)]; T1p = T1n + T1o; T1A = T1n - T1o; T1v = iio[WS(vs, 3)]; T1w = iio[WS(vs, 3) + WS(is, 2)]; T1x = T1v - T1w; T1M = T1v + T1w; } { E T1q, T1r, T1B, T1C; T1q = rio[WS(vs, 3) + WS(is, 1)]; T1r = rio[WS(vs, 3) + WS(is, 3)]; T1s = T1q + T1r; T1u = T1q - T1r; T1B = iio[WS(vs, 3) + WS(is, 1)]; T1C = iio[WS(vs, 3) + WS(is, 3)]; T1D = T1B - T1C; T1N = T1B + T1C; } rio[0] = T3 + T6; iio[0] = Tq + Tr; rio[WS(is, 1)] = Tv + Ty; iio[WS(is, 1)] = TS + TT; rio[WS(is, 2)] = TX + T10; iio[WS(is, 2)] = T1k + T1l; iio[WS(is, 3)] = T1M + T1N; rio[WS(is, 3)] = T1p + T1s; { E Tc, Ti, T7, Td; Tc = T8 + Tb; Ti = Te - Th; T7 = W[4]; Td = W[5]; iio[WS(vs, 3)] = FNMS(Td, Ti, T7 * Tc); rio[WS(vs, 3)] = FMA(Td, Tc, T7 * Ti); } { E T1K, T1O, T1J, T1L; T1K = T1p - T1s; T1O = T1M - T1N; T1J = W[2]; T1L = W[3]; rio[WS(vs, 2) + WS(is, 3)] = FMA(T1J, T1K, T1L * T1O); iio[WS(vs, 2) + WS(is, 3)] = FNMS(T1L, T1K, T1J * T1O); } { E Tk, Tm, Tj, Tl; Tk = Tb - T8; Tm = Te + Th; Tj = W[0]; Tl = W[1]; iio[WS(vs, 1)] = FNMS(Tl, Tm, Tj * Tk); rio[WS(vs, 1)] = FMA(Tl, Tk, Tj * Tm); } { E To, Ts, Tn, Tp; To = T3 - T6; Ts = Tq - Tr; Tn = W[2]; Tp = W[3]; rio[WS(vs, 2)] = FMA(Tn, To, Tp * Ts); iio[WS(vs, 2)] = FNMS(Tp, To, Tn * Ts); } { E T16, T1c, T11, T17; T16 = T12 + T15; T1c = T18 - T1b; T11 = W[4]; T17 = W[5]; iio[WS(vs, 3) + WS(is, 2)] = FNMS(T17, T1c, T11 * T16); rio[WS(vs, 3) + WS(is, 2)] = FMA(T17, T16, T11 * T1c); } { E T1G, T1I, T1F, T1H; T1G = T1x - T1u; T1I = T1A + T1D; T1F = W[0]; T1H = W[1]; iio[WS(vs, 1) + WS(is, 3)] = FNMS(T1H, T1I, T1F * T1G); rio[WS(vs, 1) + WS(is, 3)] = FMA(T1H, T1G, T1F * T1I); } { E TQ, TU, TP, TR; TQ = Tv - Ty; TU = TS - TT; TP = W[2]; TR = W[3]; rio[WS(vs, 2) + WS(is, 1)] = FMA(TP, TQ, TR * TU); iio[WS(vs, 2) + WS(is, 1)] = FNMS(TR, TQ, TP * TU); } { E T1e, T1g, T1d, T1f; T1e = T15 - T12; T1g = T18 + T1b; T1d = W[0]; T1f = W[1]; iio[WS(vs, 1) + WS(is, 2)] = FNMS(T1f, T1g, T1d * T1e); rio[WS(vs, 1) + WS(is, 2)] = FMA(T1f, T1e, T1d * T1g); } { E T1i, T1m, T1h, T1j; T1i = TX - T10; T1m = T1k - T1l; T1h = W[2]; T1j = W[3]; rio[WS(vs, 2) + WS(is, 2)] = FMA(T1h, T1i, T1j * T1m); iio[WS(vs, 2) + WS(is, 2)] = FNMS(T1j, T1i, T1h * T1m); } { E T1y, T1E, T1t, T1z; T1y = T1u + T1x; T1E = T1A - T1D; T1t = W[4]; T1z = W[5]; iio[WS(vs, 3) + WS(is, 3)] = FNMS(T1z, T1E, T1t * T1y); rio[WS(vs, 3) + WS(is, 3)] = FMA(T1z, T1y, T1t * T1E); } { E TM, TO, TL, TN; TM = TD - TA; TO = TG + TJ; TL = W[0]; TN = W[1]; iio[WS(vs, 1) + WS(is, 1)] = FNMS(TN, TO, TL * TM); rio[WS(vs, 1) + WS(is, 1)] = FMA(TN, TM, TL * TO); } { E TE, TK, Tz, TF; TE = TA + TD; TK = TG - TJ; Tz = W[4]; TF = W[5]; iio[WS(vs, 3) + WS(is, 1)] = FNMS(TF, TK, Tz * TE); rio[WS(vs, 3) + WS(is, 1)] = FMA(TF, TE, Tz * TK); } } return W; } static const tw_instr twinstr[] = { {TW_FULL, 0, 4}, {TW_NEXT, 1, 0} }; static const ct_desc desc = { 4, "q1_4", twinstr, {64, 24, 24, 0}, &GENUS, 0, 0, 0 }; void X(codelet_q1_4) (planner *p) { X(kdft_difsq_register) (p, q1_4, &desc); }