Chris@82: /* Chris@82: * Copyright (c) 2003, 2007-14 Matteo Frigo Chris@82: * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology Chris@82: * Chris@82: * This program is free software; you can redistribute it and/or modify Chris@82: * it under the terms of the GNU General Public License as published by Chris@82: * the Free Software Foundation; either version 2 of the License, or Chris@82: * (at your option) any later version. Chris@82: * Chris@82: * This program is distributed in the hope that it will be useful, Chris@82: * but WITHOUT ANY WARRANTY; without even the implied warranty of Chris@82: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the Chris@82: * GNU General Public License for more details. Chris@82: * Chris@82: * You should have received a copy of the GNU General Public License Chris@82: * along with this program; if not, write to the Free Software Chris@82: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA Chris@82: * Chris@82: */ Chris@82: Chris@82: /* This file was automatically generated --- DO NOT EDIT */ Chris@82: /* Generated on Thu May 24 08:04:10 EDT 2018 */ Chris@82: Chris@82: #include "dft/codelet-dft.h" Chris@82: Chris@82: #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA) Chris@82: Chris@82: /* Generated by: ../../../genfft/gen_notw.native -fma -compact -variables 4 -pipeline-latency 4 -n 7 -name n1_7 -include dft/scalar/n.h */ Chris@82: Chris@82: /* Chris@82: * This function contains 60 FP additions, 42 FP multiplications, Chris@82: * (or, 18 additions, 0 multiplications, 42 fused multiply/add), Chris@82: * 41 stack variables, 6 constants, and 28 memory accesses Chris@82: */ Chris@82: #include "dft/scalar/n.h" Chris@82: Chris@82: static void n1_7(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) Chris@82: { Chris@82: DK(KP974927912, +0.974927912181823607018131682993931217232785801); Chris@82: DK(KP900968867, +0.900968867902419126236102319507445051165919162); Chris@82: DK(KP692021471, +0.692021471630095869627814897002069140197260599); Chris@82: DK(KP801937735, +0.801937735804838252472204639014890102331838324); Chris@82: DK(KP554958132, +0.554958132087371191422194871006410481067288862); Chris@82: DK(KP356895867, +0.356895867892209443894399510021300583399127187); Chris@82: { Chris@82: INT i; Chris@82: 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)) { Chris@82: E T1, Tz, T4, TI, Ta, TG, T7, TH, Tb, Tp, TT, TO, TJ, Tu, Tg; Chris@82: E TB, Tm, TC, Tj, TA, Tn, Ts, TQ, TL, TD, Tx; Chris@82: T1 = ri[0]; Chris@82: Tz = ii[0]; Chris@82: { Chris@82: E T2, T3, Te, Tf; Chris@82: T2 = ri[WS(is, 1)]; Chris@82: T3 = ri[WS(is, 6)]; Chris@82: T4 = T2 + T3; Chris@82: TI = T3 - T2; Chris@82: { Chris@82: E T8, T9, T5, T6; Chris@82: T8 = ri[WS(is, 3)]; Chris@82: T9 = ri[WS(is, 4)]; Chris@82: Ta = T8 + T9; Chris@82: TG = T9 - T8; Chris@82: T5 = ri[WS(is, 2)]; Chris@82: T6 = ri[WS(is, 5)]; Chris@82: T7 = T5 + T6; Chris@82: TH = T6 - T5; Chris@82: } Chris@82: Tb = FNMS(KP356895867, T7, T4); Chris@82: Tp = FNMS(KP356895867, T4, Ta); Chris@82: TT = FMA(KP554958132, TG, TI); Chris@82: TO = FMA(KP554958132, TH, TG); Chris@82: TJ = FNMS(KP554958132, TI, TH); Chris@82: Tu = FNMS(KP356895867, Ta, T7); Chris@82: Te = ii[WS(is, 2)]; Chris@82: Tf = ii[WS(is, 5)]; Chris@82: Tg = Te - Tf; Chris@82: TB = Te + Tf; Chris@82: { Chris@82: E Tk, Tl, Th, Ti; Chris@82: Tk = ii[WS(is, 3)]; Chris@82: Tl = ii[WS(is, 4)]; Chris@82: Tm = Tk - Tl; Chris@82: TC = Tk + Tl; Chris@82: Th = ii[WS(is, 1)]; Chris@82: Ti = ii[WS(is, 6)]; Chris@82: Tj = Th - Ti; Chris@82: TA = Th + Ti; Chris@82: } Chris@82: Tn = FMA(KP554958132, Tm, Tj); Chris@82: Ts = FMA(KP554958132, Tg, Tm); Chris@82: TQ = FNMS(KP356895867, TB, TA); Chris@82: TL = FNMS(KP356895867, TA, TC); Chris@82: TD = FNMS(KP356895867, TC, TB); Chris@82: Tx = FNMS(KP554958132, Tj, Tg); Chris@82: } Chris@82: ro[0] = T1 + T4 + T7 + Ta; Chris@82: io[0] = Tz + TA + TB + TC; Chris@82: { Chris@82: E To, Td, Tc, TU, TS, TR; Chris@82: To = FMA(KP801937735, Tn, Tg); Chris@82: Tc = FNMS(KP692021471, Tb, Ta); Chris@82: Td = FNMS(KP900968867, Tc, T1); Chris@82: ro[WS(os, 6)] = FNMS(KP974927912, To, Td); Chris@82: ro[WS(os, 1)] = FMA(KP974927912, To, Td); Chris@82: TU = FMA(KP801937735, TT, TH); Chris@82: TR = FNMS(KP692021471, TQ, TC); Chris@82: TS = FNMS(KP900968867, TR, Tz); Chris@82: io[WS(os, 1)] = FMA(KP974927912, TU, TS); Chris@82: io[WS(os, 6)] = FNMS(KP974927912, TU, TS); Chris@82: } Chris@82: { Chris@82: E Tt, Tr, Tq, TP, TN, TM; Chris@82: Tt = FNMS(KP801937735, Ts, Tj); Chris@82: Tq = FNMS(KP692021471, Tp, T7); Chris@82: Tr = FNMS(KP900968867, Tq, T1); Chris@82: ro[WS(os, 5)] = FNMS(KP974927912, Tt, Tr); Chris@82: ro[WS(os, 2)] = FMA(KP974927912, Tt, Tr); Chris@82: TP = FNMS(KP801937735, TO, TI); Chris@82: TM = FNMS(KP692021471, TL, TB); Chris@82: TN = FNMS(KP900968867, TM, Tz); Chris@82: io[WS(os, 2)] = FMA(KP974927912, TP, TN); Chris@82: io[WS(os, 5)] = FNMS(KP974927912, TP, TN); Chris@82: } Chris@82: { Chris@82: E Ty, Tw, Tv, TK, TF, TE; Chris@82: Ty = FNMS(KP801937735, Tx, Tm); Chris@82: Tv = FNMS(KP692021471, Tu, T4); Chris@82: Tw = FNMS(KP900968867, Tv, T1); Chris@82: ro[WS(os, 4)] = FNMS(KP974927912, Ty, Tw); Chris@82: ro[WS(os, 3)] = FMA(KP974927912, Ty, Tw); Chris@82: TK = FNMS(KP801937735, TJ, TG); Chris@82: TE = FNMS(KP692021471, TD, TA); Chris@82: TF = FNMS(KP900968867, TE, Tz); Chris@82: io[WS(os, 3)] = FMA(KP974927912, TK, TF); Chris@82: io[WS(os, 4)] = FNMS(KP974927912, TK, TF); Chris@82: } Chris@82: } Chris@82: } Chris@82: } Chris@82: Chris@82: static const kdft_desc desc = { 7, "n1_7", {18, 0, 42, 0}, &GENUS, 0, 0, 0, 0 }; Chris@82: Chris@82: void X(codelet_n1_7) (planner *p) { Chris@82: X(kdft_register) (p, n1_7, &desc); Chris@82: } Chris@82: Chris@82: #else Chris@82: Chris@82: /* Generated by: ../../../genfft/gen_notw.native -compact -variables 4 -pipeline-latency 4 -n 7 -name n1_7 -include dft/scalar/n.h */ Chris@82: Chris@82: /* Chris@82: * This function contains 60 FP additions, 36 FP multiplications, Chris@82: * (or, 36 additions, 12 multiplications, 24 fused multiply/add), Chris@82: * 25 stack variables, 6 constants, and 28 memory accesses Chris@82: */ Chris@82: #include "dft/scalar/n.h" Chris@82: Chris@82: static void n1_7(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) Chris@82: { Chris@82: DK(KP222520933, +0.222520933956314404288902564496794759466355569); Chris@82: DK(KP900968867, +0.900968867902419126236102319507445051165919162); Chris@82: DK(KP623489801, +0.623489801858733530525004884004239810632274731); Chris@82: DK(KP433883739, +0.433883739117558120475768332848358754609990728); Chris@82: DK(KP781831482, +0.781831482468029808708444526674057750232334519); Chris@82: DK(KP974927912, +0.974927912181823607018131682993931217232785801); Chris@82: { Chris@82: INT i; Chris@82: 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)) { Chris@82: E T1, Tu, T4, Tq, Te, Tx, T7, Ts, Tk, Tv, Ta, Tr, Th, Tw; Chris@82: T1 = ri[0]; Chris@82: Tu = ii[0]; Chris@82: { Chris@82: E T2, T3, Tc, Td; Chris@82: T2 = ri[WS(is, 1)]; Chris@82: T3 = ri[WS(is, 6)]; Chris@82: T4 = T2 + T3; Chris@82: Tq = T3 - T2; Chris@82: Tc = ii[WS(is, 1)]; Chris@82: Td = ii[WS(is, 6)]; Chris@82: Te = Tc - Td; Chris@82: Tx = Tc + Td; Chris@82: } Chris@82: { Chris@82: E T5, T6, Ti, Tj; Chris@82: T5 = ri[WS(is, 2)]; Chris@82: T6 = ri[WS(is, 5)]; Chris@82: T7 = T5 + T6; Chris@82: Ts = T6 - T5; Chris@82: Ti = ii[WS(is, 2)]; Chris@82: Tj = ii[WS(is, 5)]; Chris@82: Tk = Ti - Tj; Chris@82: Tv = Ti + Tj; Chris@82: } Chris@82: { Chris@82: E T8, T9, Tf, Tg; Chris@82: T8 = ri[WS(is, 3)]; Chris@82: T9 = ri[WS(is, 4)]; Chris@82: Ta = T8 + T9; Chris@82: Tr = T9 - T8; Chris@82: Tf = ii[WS(is, 3)]; Chris@82: Tg = ii[WS(is, 4)]; Chris@82: Th = Tf - Tg; Chris@82: Tw = Tf + Tg; Chris@82: } Chris@82: ro[0] = T1 + T4 + T7 + Ta; Chris@82: io[0] = Tu + Tx + Tv + Tw; Chris@82: { Chris@82: E Tl, Tb, TB, TC; Chris@82: Tl = FNMS(KP781831482, Th, KP974927912 * Te) - (KP433883739 * Tk); Chris@82: Tb = FMA(KP623489801, Ta, T1) + FNMA(KP900968867, T7, KP222520933 * T4); Chris@82: ro[WS(os, 5)] = Tb - Tl; Chris@82: ro[WS(os, 2)] = Tb + Tl; Chris@82: TB = FNMS(KP781831482, Tr, KP974927912 * Tq) - (KP433883739 * Ts); Chris@82: TC = FMA(KP623489801, Tw, Tu) + FNMA(KP900968867, Tv, KP222520933 * Tx); Chris@82: io[WS(os, 2)] = TB + TC; Chris@82: io[WS(os, 5)] = TC - TB; Chris@82: } Chris@82: { Chris@82: E Tn, Tm, Tz, TA; Chris@82: Tn = FMA(KP781831482, Te, KP974927912 * Tk) + (KP433883739 * Th); Chris@82: Tm = FMA(KP623489801, T4, T1) + FNMA(KP900968867, Ta, KP222520933 * T7); Chris@82: ro[WS(os, 6)] = Tm - Tn; Chris@82: ro[WS(os, 1)] = Tm + Tn; Chris@82: Tz = FMA(KP781831482, Tq, KP974927912 * Ts) + (KP433883739 * Tr); Chris@82: TA = FMA(KP623489801, Tx, Tu) + FNMA(KP900968867, Tw, KP222520933 * Tv); Chris@82: io[WS(os, 1)] = Tz + TA; Chris@82: io[WS(os, 6)] = TA - Tz; Chris@82: } Chris@82: { Chris@82: E Tp, To, Tt, Ty; Chris@82: Tp = FMA(KP433883739, Te, KP974927912 * Th) - (KP781831482 * Tk); Chris@82: To = FMA(KP623489801, T7, T1) + FNMA(KP222520933, Ta, KP900968867 * T4); Chris@82: ro[WS(os, 4)] = To - Tp; Chris@82: ro[WS(os, 3)] = To + Tp; Chris@82: Tt = FMA(KP433883739, Tq, KP974927912 * Tr) - (KP781831482 * Ts); Chris@82: Ty = FMA(KP623489801, Tv, Tu) + FNMA(KP222520933, Tw, KP900968867 * Tx); Chris@82: io[WS(os, 3)] = Tt + Ty; Chris@82: io[WS(os, 4)] = Ty - Tt; Chris@82: } Chris@82: } Chris@82: } Chris@82: } Chris@82: Chris@82: static const kdft_desc desc = { 7, "n1_7", {36, 12, 24, 0}, &GENUS, 0, 0, 0, 0 }; Chris@82: Chris@82: void X(codelet_n1_7) (planner *p) { Chris@82: X(kdft_register) (p, n1_7, &desc); Chris@82: } Chris@82: Chris@82: #endif