cannam@95: /* cannam@95: * Copyright (c) 2003, 2007-11 Matteo Frigo cannam@95: * Copyright (c) 2003, 2007-11 Massachusetts Institute of Technology cannam@95: * cannam@95: * This program is free software; you can redistribute it and/or modify cannam@95: * it under the terms of the GNU General Public License as published by cannam@95: * the Free Software Foundation; either version 2 of the License, or cannam@95: * (at your option) any later version. cannam@95: * cannam@95: * This program is distributed in the hope that it will be useful, cannam@95: * but WITHOUT ANY WARRANTY; without even the implied warranty of cannam@95: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the cannam@95: * GNU General Public License for more details. cannam@95: * cannam@95: * You should have received a copy of the GNU General Public License cannam@95: * along with this program; if not, write to the Free Software cannam@95: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA cannam@95: * cannam@95: */ cannam@95: cannam@95: /* This file was automatically generated --- DO NOT EDIT */ cannam@95: /* Generated on Sun Nov 25 07:35:42 EST 2012 */ cannam@95: cannam@95: #include "codelet-dft.h" cannam@95: cannam@95: #ifdef HAVE_FMA cannam@95: cannam@95: /* Generated by: ../../../genfft/gen_notw.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 7 -name n1_7 -include n.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 60 FP additions, 42 FP multiplications, cannam@95: * (or, 18 additions, 0 multiplications, 42 fused multiply/add), cannam@95: * 51 stack variables, 6 constants, and 28 memory accesses cannam@95: */ cannam@95: #include "n.h" cannam@95: cannam@95: static void n1_7(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) cannam@95: { cannam@95: DK(KP974927912, +0.974927912181823607018131682993931217232785801); cannam@95: DK(KP900968867, +0.900968867902419126236102319507445051165919162); cannam@95: DK(KP801937735, +0.801937735804838252472204639014890102331838324); cannam@95: DK(KP692021471, +0.692021471630095869627814897002069140197260599); cannam@95: DK(KP356895867, +0.356895867892209443894399510021300583399127187); cannam@95: DK(KP554958132, +0.554958132087371191422194871006410481067288862); cannam@95: { cannam@95: INT i; cannam@95: for (i = v; i > 0; i = i - 1, ri = ri + ivs, ii = ii + ivs, ro = ro + ovs, io = io + ovs, MAKE_VOLATILE_STRIDE(28, is), MAKE_VOLATILE_STRIDE(28, os)) { cannam@95: E Tz, TP, Ty, TK, TN, TE, Tw, TF; cannam@95: { cannam@95: E T1, TI, T4, TG, Ta, TT, Tp, TH, T7, Tk, TJ, TO, Tu, Tb, TB; cannam@95: E Tg, Tl, Th, Ti; cannam@95: T1 = ri[0]; cannam@95: Tz = ii[0]; cannam@95: { cannam@95: E T5, T6, Te, Tf; cannam@95: { cannam@95: E T2, T3, T8, T9; cannam@95: T2 = ri[WS(is, 1)]; cannam@95: T3 = ri[WS(is, 6)]; cannam@95: T8 = ri[WS(is, 3)]; cannam@95: T9 = ri[WS(is, 4)]; cannam@95: T5 = ri[WS(is, 2)]; cannam@95: TI = T3 - T2; cannam@95: T4 = T2 + T3; cannam@95: TG = T9 - T8; cannam@95: Ta = T8 + T9; cannam@95: T6 = ri[WS(is, 5)]; cannam@95: } cannam@95: Te = ii[WS(is, 2)]; cannam@95: TT = FMA(KP554958132, TG, TI); cannam@95: Tp = FNMS(KP356895867, T4, Ta); cannam@95: TH = T6 - T5; cannam@95: T7 = T5 + T6; cannam@95: Tf = ii[WS(is, 5)]; cannam@95: Tk = ii[WS(is, 3)]; cannam@95: TJ = FNMS(KP554958132, TI, TH); cannam@95: TO = FMA(KP554958132, TH, TG); cannam@95: Tu = FNMS(KP356895867, Ta, T7); cannam@95: Tb = FNMS(KP356895867, T7, T4); cannam@95: TB = Te + Tf; cannam@95: Tg = Te - Tf; cannam@95: Tl = ii[WS(is, 4)]; cannam@95: Th = ii[WS(is, 1)]; cannam@95: Ti = ii[WS(is, 6)]; cannam@95: } cannam@95: { cannam@95: E Tm, TA, Tj, TD, Ts, TL, Tx, TU, To, TR, Td, TM, Tv; cannam@95: { cannam@95: E TC, TQ, Tn, Tc; cannam@95: ro[0] = T1 + T4 + T7 + Ta; cannam@95: TC = Tk + Tl; cannam@95: Tm = Tk - Tl; cannam@95: TA = Th + Ti; cannam@95: Tj = Th - Ti; cannam@95: TD = FNMS(KP356895867, TC, TB); cannam@95: Ts = FMA(KP554958132, Tg, Tm); cannam@95: TL = FNMS(KP356895867, TA, TC); cannam@95: TQ = FNMS(KP356895867, TB, TA); cannam@95: Tx = FNMS(KP554958132, Tj, Tg); cannam@95: Tn = FMA(KP554958132, Tm, Tj); cannam@95: io[0] = Tz + TA + TB + TC; cannam@95: Tc = FNMS(KP692021471, Tb, Ta); cannam@95: TU = FMA(KP801937735, TT, TH); cannam@95: To = FMA(KP801937735, Tn, Tg); cannam@95: TR = FNMS(KP692021471, TQ, TC); cannam@95: Td = FNMS(KP900968867, Tc, T1); cannam@95: } cannam@95: { cannam@95: E Tt, Tr, TS, Tq; cannam@95: Tt = FNMS(KP801937735, Ts, Tj); cannam@95: Tq = FNMS(KP692021471, Tp, T7); cannam@95: TS = FNMS(KP900968867, TR, Tz); cannam@95: ro[WS(os, 1)] = FMA(KP974927912, To, Td); cannam@95: ro[WS(os, 6)] = FNMS(KP974927912, To, Td); cannam@95: Tr = FNMS(KP900968867, Tq, T1); cannam@95: io[WS(os, 6)] = FNMS(KP974927912, TU, TS); cannam@95: io[WS(os, 1)] = FMA(KP974927912, TU, TS); cannam@95: TP = FNMS(KP801937735, TO, TI); cannam@95: ro[WS(os, 2)] = FMA(KP974927912, Tt, Tr); cannam@95: ro[WS(os, 5)] = FNMS(KP974927912, Tt, Tr); cannam@95: TM = FNMS(KP692021471, TL, TB); cannam@95: } cannam@95: Ty = FNMS(KP801937735, Tx, Tm); cannam@95: Tv = FNMS(KP692021471, Tu, T4); cannam@95: TK = FNMS(KP801937735, TJ, TG); cannam@95: TN = FNMS(KP900968867, TM, Tz); cannam@95: TE = FNMS(KP692021471, TD, TA); cannam@95: Tw = FNMS(KP900968867, Tv, T1); cannam@95: } cannam@95: } cannam@95: io[WS(os, 5)] = FNMS(KP974927912, TP, TN); cannam@95: io[WS(os, 2)] = FMA(KP974927912, TP, TN); cannam@95: TF = FNMS(KP900968867, TE, Tz); cannam@95: ro[WS(os, 3)] = FMA(KP974927912, Ty, Tw); cannam@95: ro[WS(os, 4)] = FNMS(KP974927912, Ty, Tw); cannam@95: io[WS(os, 4)] = FNMS(KP974927912, TK, TF); cannam@95: io[WS(os, 3)] = FMA(KP974927912, TK, TF); cannam@95: } cannam@95: } cannam@95: } cannam@95: cannam@95: static const kdft_desc desc = { 7, "n1_7", {18, 0, 42, 0}, &GENUS, 0, 0, 0, 0 }; cannam@95: cannam@95: void X(codelet_n1_7) (planner *p) { cannam@95: X(kdft_register) (p, n1_7, &desc); cannam@95: } cannam@95: cannam@95: #else /* HAVE_FMA */ cannam@95: cannam@95: /* Generated by: ../../../genfft/gen_notw.native -compact -variables 4 -pipeline-latency 4 -n 7 -name n1_7 -include n.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 60 FP additions, 36 FP multiplications, cannam@95: * (or, 36 additions, 12 multiplications, 24 fused multiply/add), cannam@95: * 25 stack variables, 6 constants, and 28 memory accesses cannam@95: */ cannam@95: #include "n.h" cannam@95: cannam@95: static void n1_7(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) cannam@95: { cannam@95: DK(KP222520933, +0.222520933956314404288902564496794759466355569); cannam@95: DK(KP900968867, +0.900968867902419126236102319507445051165919162); cannam@95: DK(KP623489801, +0.623489801858733530525004884004239810632274731); cannam@95: DK(KP433883739, +0.433883739117558120475768332848358754609990728); cannam@95: DK(KP781831482, +0.781831482468029808708444526674057750232334519); cannam@95: DK(KP974927912, +0.974927912181823607018131682993931217232785801); cannam@95: { cannam@95: INT i; cannam@95: for (i = v; i > 0; i = i - 1, ri = ri + ivs, ii = ii + ivs, ro = ro + ovs, io = io + ovs, MAKE_VOLATILE_STRIDE(28, is), MAKE_VOLATILE_STRIDE(28, os)) { cannam@95: E T1, Tu, T4, Tq, Te, Tx, T7, Ts, Tk, Tv, Ta, Tr, Th, Tw; cannam@95: T1 = ri[0]; cannam@95: Tu = ii[0]; cannam@95: { cannam@95: E T2, T3, Tc, Td; cannam@95: T2 = ri[WS(is, 1)]; cannam@95: T3 = ri[WS(is, 6)]; cannam@95: T4 = T2 + T3; cannam@95: Tq = T3 - T2; cannam@95: Tc = ii[WS(is, 1)]; cannam@95: Td = ii[WS(is, 6)]; cannam@95: Te = Tc - Td; cannam@95: Tx = Tc + Td; cannam@95: } cannam@95: { cannam@95: E T5, T6, Ti, Tj; cannam@95: T5 = ri[WS(is, 2)]; cannam@95: T6 = ri[WS(is, 5)]; cannam@95: T7 = T5 + T6; cannam@95: Ts = T6 - T5; cannam@95: Ti = ii[WS(is, 2)]; cannam@95: Tj = ii[WS(is, 5)]; cannam@95: Tk = Ti - Tj; cannam@95: Tv = Ti + Tj; cannam@95: } cannam@95: { cannam@95: E T8, T9, Tf, Tg; cannam@95: T8 = ri[WS(is, 3)]; cannam@95: T9 = ri[WS(is, 4)]; cannam@95: Ta = T8 + T9; cannam@95: Tr = T9 - T8; cannam@95: Tf = ii[WS(is, 3)]; cannam@95: Tg = ii[WS(is, 4)]; cannam@95: Th = Tf - Tg; cannam@95: Tw = Tf + Tg; cannam@95: } cannam@95: ro[0] = T1 + T4 + T7 + Ta; cannam@95: io[0] = Tu + Tx + Tv + Tw; cannam@95: { cannam@95: E Tl, Tb, TB, TC; cannam@95: Tl = FNMS(KP781831482, Th, KP974927912 * Te) - (KP433883739 * Tk); cannam@95: Tb = FMA(KP623489801, Ta, T1) + FNMA(KP900968867, T7, KP222520933 * T4); cannam@95: ro[WS(os, 5)] = Tb - Tl; cannam@95: ro[WS(os, 2)] = Tb + Tl; cannam@95: TB = FNMS(KP781831482, Tr, KP974927912 * Tq) - (KP433883739 * Ts); cannam@95: TC = FMA(KP623489801, Tw, Tu) + FNMA(KP900968867, Tv, KP222520933 * Tx); cannam@95: io[WS(os, 2)] = TB + TC; cannam@95: io[WS(os, 5)] = TC - TB; cannam@95: } cannam@95: { cannam@95: E Tn, Tm, Tz, TA; cannam@95: Tn = FMA(KP781831482, Te, KP974927912 * Tk) + (KP433883739 * Th); cannam@95: Tm = FMA(KP623489801, T4, T1) + FNMA(KP900968867, Ta, KP222520933 * T7); cannam@95: ro[WS(os, 6)] = Tm - Tn; cannam@95: ro[WS(os, 1)] = Tm + Tn; cannam@95: Tz = FMA(KP781831482, Tq, KP974927912 * Ts) + (KP433883739 * Tr); cannam@95: TA = FMA(KP623489801, Tx, Tu) + FNMA(KP900968867, Tw, KP222520933 * Tv); cannam@95: io[WS(os, 1)] = Tz + TA; cannam@95: io[WS(os, 6)] = TA - Tz; cannam@95: } cannam@95: { cannam@95: E Tp, To, Tt, Ty; cannam@95: Tp = FMA(KP433883739, Te, KP974927912 * Th) - (KP781831482 * Tk); cannam@95: To = FMA(KP623489801, T7, T1) + FNMA(KP222520933, Ta, KP900968867 * T4); cannam@95: ro[WS(os, 4)] = To - Tp; cannam@95: ro[WS(os, 3)] = To + Tp; cannam@95: Tt = FMA(KP433883739, Tq, KP974927912 * Tr) - (KP781831482 * Ts); cannam@95: Ty = FMA(KP623489801, Tv, Tu) + FNMA(KP222520933, Tw, KP900968867 * Tx); cannam@95: io[WS(os, 3)] = Tt + Ty; cannam@95: io[WS(os, 4)] = Ty - Tt; cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: cannam@95: static const kdft_desc desc = { 7, "n1_7", {36, 12, 24, 0}, &GENUS, 0, 0, 0, 0 }; cannam@95: cannam@95: void X(codelet_n1_7) (planner *p) { cannam@95: X(kdft_register) (p, n1_7, &desc); cannam@95: } cannam@95: cannam@95: #endif /* HAVE_FMA */