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 6 -name n1_6 -include dft/scalar/n.h */ Chris@82: Chris@82: /* Chris@82: * This function contains 36 FP additions, 12 FP multiplications, Chris@82: * (or, 24 additions, 0 multiplications, 12 fused multiply/add), Chris@82: * 23 stack variables, 2 constants, and 24 memory accesses Chris@82: */ Chris@82: #include "dft/scalar/n.h" Chris@82: Chris@82: static void n1_6(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(KP866025403, +0.866025403784438646763723170752936183471402627); Chris@82: DK(KP500000000, +0.500000000000000000000000000000000000000000000); 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(24, is), MAKE_VOLATILE_STRIDE(24, os)) { Chris@82: E T3, Tb, Tp, Tx, T6, Tc, T9, Td, Ta, Te, Ti, Tu, Tl, Tv, Tq; Chris@82: E Ty; Chris@82: { Chris@82: E T1, T2, Tn, To; Chris@82: T1 = ri[0]; Chris@82: T2 = ri[WS(is, 3)]; Chris@82: T3 = T1 - T2; Chris@82: Tb = T1 + T2; Chris@82: Tn = ii[0]; Chris@82: To = ii[WS(is, 3)]; Chris@82: Tp = Tn - To; Chris@82: Tx = Tn + To; Chris@82: } Chris@82: { Chris@82: E T4, T5, T7, T8; Chris@82: T4 = ri[WS(is, 2)]; Chris@82: T5 = ri[WS(is, 5)]; Chris@82: T6 = T4 - T5; Chris@82: Tc = T4 + T5; Chris@82: T7 = ri[WS(is, 4)]; Chris@82: T8 = ri[WS(is, 1)]; Chris@82: T9 = T7 - T8; Chris@82: Td = T7 + T8; Chris@82: } Chris@82: Ta = T6 + T9; Chris@82: Te = Tc + Td; Chris@82: { Chris@82: E Tg, Th, Tj, Tk; Chris@82: Tg = ii[WS(is, 2)]; Chris@82: Th = ii[WS(is, 5)]; Chris@82: Ti = Tg - Th; Chris@82: Tu = Tg + Th; Chris@82: Tj = ii[WS(is, 4)]; Chris@82: Tk = ii[WS(is, 1)]; Chris@82: Tl = Tj - Tk; Chris@82: Tv = Tj + Tk; Chris@82: } Chris@82: Tq = Ti + Tl; Chris@82: Ty = Tu + Tv; Chris@82: ro[WS(os, 3)] = T3 + Ta; Chris@82: io[WS(os, 3)] = Tp + Tq; Chris@82: ro[0] = Tb + Te; Chris@82: io[0] = Tx + Ty; Chris@82: { Chris@82: E Tf, Tm, Tr, Ts; Chris@82: Tf = FNMS(KP500000000, Ta, T3); Chris@82: Tm = Ti - Tl; Chris@82: ro[WS(os, 5)] = FNMS(KP866025403, Tm, Tf); Chris@82: ro[WS(os, 1)] = FMA(KP866025403, Tm, Tf); Chris@82: Tr = FNMS(KP500000000, Tq, Tp); Chris@82: Ts = T9 - T6; Chris@82: io[WS(os, 1)] = FMA(KP866025403, Ts, Tr); Chris@82: io[WS(os, 5)] = FNMS(KP866025403, Ts, Tr); Chris@82: } Chris@82: { Chris@82: E Tt, Tw, Tz, TA; Chris@82: Tt = FNMS(KP500000000, Te, Tb); Chris@82: Tw = Tu - Tv; Chris@82: ro[WS(os, 2)] = FNMS(KP866025403, Tw, Tt); Chris@82: ro[WS(os, 4)] = FMA(KP866025403, Tw, Tt); Chris@82: Tz = FNMS(KP500000000, Ty, Tx); Chris@82: TA = Td - Tc; Chris@82: io[WS(os, 2)] = FNMS(KP866025403, TA, Tz); Chris@82: io[WS(os, 4)] = FMA(KP866025403, TA, Tz); Chris@82: } Chris@82: } Chris@82: } Chris@82: } Chris@82: Chris@82: static const kdft_desc desc = { 6, "n1_6", {24, 0, 12, 0}, &GENUS, 0, 0, 0, 0 }; Chris@82: Chris@82: void X(codelet_n1_6) (planner *p) { Chris@82: X(kdft_register) (p, n1_6, &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 6 -name n1_6 -include dft/scalar/n.h */ Chris@82: Chris@82: /* Chris@82: * This function contains 36 FP additions, 8 FP multiplications, Chris@82: * (or, 32 additions, 4 multiplications, 4 fused multiply/add), Chris@82: * 23 stack variables, 2 constants, and 24 memory accesses Chris@82: */ Chris@82: #include "dft/scalar/n.h" Chris@82: Chris@82: static void n1_6(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(KP866025403, +0.866025403784438646763723170752936183471402627); Chris@82: DK(KP500000000, +0.500000000000000000000000000000000000000000000); 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(24, is), MAKE_VOLATILE_STRIDE(24, os)) { Chris@82: E T3, Tb, Tq, Tx, T6, Tc, T9, Td, Ta, Te, Ti, Tu, Tl, Tv, Tr; Chris@82: E Ty; Chris@82: { Chris@82: E T1, T2, To, Tp; Chris@82: T1 = ri[0]; Chris@82: T2 = ri[WS(is, 3)]; Chris@82: T3 = T1 - T2; Chris@82: Tb = T1 + T2; Chris@82: To = ii[0]; Chris@82: Tp = ii[WS(is, 3)]; Chris@82: Tq = To - Tp; Chris@82: Tx = To + Tp; Chris@82: } Chris@82: { Chris@82: E T4, T5, T7, T8; Chris@82: T4 = ri[WS(is, 2)]; Chris@82: T5 = ri[WS(is, 5)]; Chris@82: T6 = T4 - T5; Chris@82: Tc = T4 + T5; Chris@82: T7 = ri[WS(is, 4)]; Chris@82: T8 = ri[WS(is, 1)]; Chris@82: T9 = T7 - T8; Chris@82: Td = T7 + T8; Chris@82: } Chris@82: Ta = T6 + T9; Chris@82: Te = Tc + Td; Chris@82: { Chris@82: E Tg, Th, Tj, Tk; Chris@82: Tg = ii[WS(is, 2)]; Chris@82: Th = ii[WS(is, 5)]; Chris@82: Ti = Tg - Th; Chris@82: Tu = Tg + Th; Chris@82: Tj = ii[WS(is, 4)]; Chris@82: Tk = ii[WS(is, 1)]; Chris@82: Tl = Tj - Tk; Chris@82: Tv = Tj + Tk; Chris@82: } Chris@82: Tr = Ti + Tl; Chris@82: Ty = Tu + Tv; Chris@82: ro[WS(os, 3)] = T3 + Ta; Chris@82: io[WS(os, 3)] = Tq + Tr; Chris@82: ro[0] = Tb + Te; Chris@82: io[0] = Tx + Ty; Chris@82: { Chris@82: E Tf, Tm, Tn, Ts; Chris@82: Tf = FNMS(KP500000000, Ta, T3); Chris@82: Tm = KP866025403 * (Ti - Tl); Chris@82: ro[WS(os, 5)] = Tf - Tm; Chris@82: ro[WS(os, 1)] = Tf + Tm; Chris@82: Tn = KP866025403 * (T9 - T6); Chris@82: Ts = FNMS(KP500000000, Tr, Tq); Chris@82: io[WS(os, 1)] = Tn + Ts; Chris@82: io[WS(os, 5)] = Ts - Tn; Chris@82: } Chris@82: { Chris@82: E Tt, Tw, Tz, TA; Chris@82: Tt = FNMS(KP500000000, Te, Tb); Chris@82: Tw = KP866025403 * (Tu - Tv); Chris@82: ro[WS(os, 2)] = Tt - Tw; Chris@82: ro[WS(os, 4)] = Tt + Tw; Chris@82: Tz = FNMS(KP500000000, Ty, Tx); Chris@82: TA = KP866025403 * (Td - Tc); Chris@82: io[WS(os, 2)] = Tz - TA; Chris@82: io[WS(os, 4)] = TA + Tz; Chris@82: } Chris@82: } Chris@82: } Chris@82: } Chris@82: Chris@82: static const kdft_desc desc = { 6, "n1_6", {32, 4, 4, 0}, &GENUS, 0, 0, 0, 0 }; Chris@82: Chris@82: void X(codelet_n1_6) (planner *p) { Chris@82: X(kdft_register) (p, n1_6, &desc); Chris@82: } Chris@82: Chris@82: #endif