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 8 -name n1_8 -include dft/scalar/n.h */ Chris@82: Chris@82: /* Chris@82: * This function contains 52 FP additions, 8 FP multiplications, Chris@82: * (or, 44 additions, 0 multiplications, 8 fused multiply/add), Chris@82: * 28 stack variables, 1 constants, and 32 memory accesses Chris@82: */ Chris@82: #include "dft/scalar/n.h" Chris@82: Chris@82: static void n1_8(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(KP707106781, +0.707106781186547524400844362104849039284835938); 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(32, is), MAKE_VOLATILE_STRIDE(32, os)) { Chris@82: E T3, Tn, Ti, TC, T6, TB, Tl, To, Td, TN, Tz, TH, Ta, TM, Tu; Chris@82: E TG; Chris@82: { Chris@82: E T1, T2, Tj, Tk; Chris@82: T1 = ri[0]; Chris@82: T2 = ri[WS(is, 4)]; Chris@82: T3 = T1 + T2; Chris@82: Tn = T1 - T2; Chris@82: { Chris@82: E Tg, Th, T4, T5; Chris@82: Tg = ii[0]; Chris@82: Th = ii[WS(is, 4)]; Chris@82: Ti = Tg + Th; Chris@82: TC = Tg - Th; Chris@82: T4 = ri[WS(is, 2)]; Chris@82: T5 = ri[WS(is, 6)]; Chris@82: T6 = T4 + T5; Chris@82: TB = T4 - T5; Chris@82: } Chris@82: Tj = ii[WS(is, 2)]; Chris@82: Tk = ii[WS(is, 6)]; Chris@82: Tl = Tj + Tk; Chris@82: To = Tj - Tk; Chris@82: { Chris@82: E Tb, Tc, Tv, Tw, Tx, Ty; Chris@82: Tb = ri[WS(is, 7)]; Chris@82: Tc = ri[WS(is, 3)]; Chris@82: Tv = Tb - Tc; Chris@82: Tw = ii[WS(is, 7)]; Chris@82: Tx = ii[WS(is, 3)]; Chris@82: Ty = Tw - Tx; Chris@82: Td = Tb + Tc; Chris@82: TN = Tw + Tx; Chris@82: Tz = Tv - Ty; Chris@82: TH = Tv + Ty; Chris@82: } Chris@82: { Chris@82: E T8, T9, Tq, Tr, Ts, Tt; Chris@82: T8 = ri[WS(is, 1)]; Chris@82: T9 = ri[WS(is, 5)]; Chris@82: Tq = T8 - T9; Chris@82: Tr = ii[WS(is, 1)]; Chris@82: Ts = ii[WS(is, 5)]; Chris@82: Tt = Tr - Ts; Chris@82: Ta = T8 + T9; Chris@82: TM = Tr + Ts; Chris@82: Tu = Tq + Tt; Chris@82: TG = Tt - Tq; Chris@82: } Chris@82: } Chris@82: { Chris@82: E T7, Te, TP, TQ; Chris@82: T7 = T3 + T6; Chris@82: Te = Ta + Td; Chris@82: ro[WS(os, 4)] = T7 - Te; Chris@82: ro[0] = T7 + Te; Chris@82: TP = Ti + Tl; Chris@82: TQ = TM + TN; Chris@82: io[WS(os, 4)] = TP - TQ; Chris@82: io[0] = TP + TQ; Chris@82: } Chris@82: { Chris@82: E Tf, Tm, TL, TO; Chris@82: Tf = Td - Ta; Chris@82: Tm = Ti - Tl; Chris@82: io[WS(os, 2)] = Tf + Tm; Chris@82: io[WS(os, 6)] = Tm - Tf; Chris@82: TL = T3 - T6; Chris@82: TO = TM - TN; Chris@82: ro[WS(os, 6)] = TL - TO; Chris@82: ro[WS(os, 2)] = TL + TO; Chris@82: } Chris@82: { Chris@82: E Tp, TA, TJ, TK; Chris@82: Tp = Tn + To; Chris@82: TA = Tu + Tz; Chris@82: ro[WS(os, 5)] = FNMS(KP707106781, TA, Tp); Chris@82: ro[WS(os, 1)] = FMA(KP707106781, TA, Tp); Chris@82: TJ = TC - TB; Chris@82: TK = TG + TH; Chris@82: io[WS(os, 5)] = FNMS(KP707106781, TK, TJ); Chris@82: io[WS(os, 1)] = FMA(KP707106781, TK, TJ); Chris@82: } Chris@82: { Chris@82: E TD, TE, TF, TI; Chris@82: TD = TB + TC; Chris@82: TE = Tz - Tu; Chris@82: io[WS(os, 7)] = FNMS(KP707106781, TE, TD); Chris@82: io[WS(os, 3)] = FMA(KP707106781, TE, TD); Chris@82: TF = Tn - To; Chris@82: TI = TG - TH; Chris@82: ro[WS(os, 7)] = FNMS(KP707106781, TI, TF); Chris@82: ro[WS(os, 3)] = FMA(KP707106781, TI, TF); Chris@82: } Chris@82: } Chris@82: } Chris@82: } Chris@82: Chris@82: static const kdft_desc desc = { 8, "n1_8", {44, 0, 8, 0}, &GENUS, 0, 0, 0, 0 }; Chris@82: Chris@82: void X(codelet_n1_8) (planner *p) { Chris@82: X(kdft_register) (p, n1_8, &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 8 -name n1_8 -include dft/scalar/n.h */ Chris@82: Chris@82: /* Chris@82: * This function contains 52 FP additions, 4 FP multiplications, Chris@82: * (or, 52 additions, 4 multiplications, 0 fused multiply/add), Chris@82: * 28 stack variables, 1 constants, and 32 memory accesses Chris@82: */ Chris@82: #include "dft/scalar/n.h" Chris@82: Chris@82: static void n1_8(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(KP707106781, +0.707106781186547524400844362104849039284835938); 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(32, is), MAKE_VOLATILE_STRIDE(32, os)) { Chris@82: E T3, Tn, Ti, TC, T6, TB, Tl, To, Td, TN, Tz, TH, Ta, TM, Tu; Chris@82: E TG; Chris@82: { Chris@82: E T1, T2, Tj, Tk; Chris@82: T1 = ri[0]; Chris@82: T2 = ri[WS(is, 4)]; Chris@82: T3 = T1 + T2; Chris@82: Tn = T1 - T2; Chris@82: { Chris@82: E Tg, Th, T4, T5; Chris@82: Tg = ii[0]; Chris@82: Th = ii[WS(is, 4)]; Chris@82: Ti = Tg + Th; Chris@82: TC = Tg - Th; Chris@82: T4 = ri[WS(is, 2)]; Chris@82: T5 = ri[WS(is, 6)]; Chris@82: T6 = T4 + T5; Chris@82: TB = T4 - T5; Chris@82: } Chris@82: Tj = ii[WS(is, 2)]; Chris@82: Tk = ii[WS(is, 6)]; Chris@82: Tl = Tj + Tk; Chris@82: To = Tj - Tk; Chris@82: { Chris@82: E Tb, Tc, Tv, Tw, Tx, Ty; Chris@82: Tb = ri[WS(is, 7)]; Chris@82: Tc = ri[WS(is, 3)]; Chris@82: Tv = Tb - Tc; Chris@82: Tw = ii[WS(is, 7)]; Chris@82: Tx = ii[WS(is, 3)]; Chris@82: Ty = Tw - Tx; Chris@82: Td = Tb + Tc; Chris@82: TN = Tw + Tx; Chris@82: Tz = Tv - Ty; Chris@82: TH = Tv + Ty; Chris@82: } Chris@82: { Chris@82: E T8, T9, Tq, Tr, Ts, Tt; Chris@82: T8 = ri[WS(is, 1)]; Chris@82: T9 = ri[WS(is, 5)]; Chris@82: Tq = T8 - T9; Chris@82: Tr = ii[WS(is, 1)]; Chris@82: Ts = ii[WS(is, 5)]; Chris@82: Tt = Tr - Ts; Chris@82: Ta = T8 + T9; Chris@82: TM = Tr + Ts; Chris@82: Tu = Tq + Tt; Chris@82: TG = Tt - Tq; Chris@82: } Chris@82: } Chris@82: { Chris@82: E T7, Te, TP, TQ; Chris@82: T7 = T3 + T6; Chris@82: Te = Ta + Td; Chris@82: ro[WS(os, 4)] = T7 - Te; Chris@82: ro[0] = T7 + Te; Chris@82: TP = Ti + Tl; Chris@82: TQ = TM + TN; Chris@82: io[WS(os, 4)] = TP - TQ; Chris@82: io[0] = TP + TQ; Chris@82: } Chris@82: { Chris@82: E Tf, Tm, TL, TO; Chris@82: Tf = Td - Ta; Chris@82: Tm = Ti - Tl; Chris@82: io[WS(os, 2)] = Tf + Tm; Chris@82: io[WS(os, 6)] = Tm - Tf; Chris@82: TL = T3 - T6; Chris@82: TO = TM - TN; Chris@82: ro[WS(os, 6)] = TL - TO; Chris@82: ro[WS(os, 2)] = TL + TO; Chris@82: } Chris@82: { Chris@82: E Tp, TA, TJ, TK; Chris@82: Tp = Tn + To; Chris@82: TA = KP707106781 * (Tu + Tz); Chris@82: ro[WS(os, 5)] = Tp - TA; Chris@82: ro[WS(os, 1)] = Tp + TA; Chris@82: TJ = TC - TB; Chris@82: TK = KP707106781 * (TG + TH); Chris@82: io[WS(os, 5)] = TJ - TK; Chris@82: io[WS(os, 1)] = TJ + TK; Chris@82: } Chris@82: { Chris@82: E TD, TE, TF, TI; Chris@82: TD = TB + TC; Chris@82: TE = KP707106781 * (Tz - Tu); Chris@82: io[WS(os, 7)] = TD - TE; Chris@82: io[WS(os, 3)] = TD + TE; Chris@82: TF = Tn - To; Chris@82: TI = KP707106781 * (TG - TH); Chris@82: ro[WS(os, 7)] = TF - TI; Chris@82: ro[WS(os, 3)] = TF + TI; Chris@82: } Chris@82: } Chris@82: } Chris@82: } Chris@82: Chris@82: static const kdft_desc desc = { 8, "n1_8", {52, 4, 0, 0}, &GENUS, 0, 0, 0, 0 }; Chris@82: Chris@82: void X(codelet_n1_8) (planner *p) { Chris@82: X(kdft_register) (p, n1_8, &desc); Chris@82: } Chris@82: Chris@82: #endif