Chris@10: /* Chris@10: * Copyright (c) 2003, 2007-11 Matteo Frigo Chris@10: * Copyright (c) 2003, 2007-11 Massachusetts Institute of Technology Chris@10: * Chris@10: * This program is free software; you can redistribute it and/or modify Chris@10: * it under the terms of the GNU General Public License as published by Chris@10: * the Free Software Foundation; either version 2 of the License, or Chris@10: * (at your option) any later version. Chris@10: * Chris@10: * This program is distributed in the hope that it will be useful, Chris@10: * but WITHOUT ANY WARRANTY; without even the implied warranty of Chris@10: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the Chris@10: * GNU General Public License for more details. Chris@10: * Chris@10: * You should have received a copy of the GNU General Public License Chris@10: * along with this program; if not, write to the Free Software Chris@10: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA Chris@10: * Chris@10: */ Chris@10: Chris@10: /* This file was automatically generated --- DO NOT EDIT */ Chris@10: /* Generated on Sun Nov 25 07:35:42 EST 2012 */ Chris@10: Chris@10: #include "codelet-dft.h" Chris@10: Chris@10: #ifdef HAVE_FMA Chris@10: Chris@10: /* Generated by: ../../../genfft/gen_notw.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 6 -name n1_6 -include n.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 36 FP additions, 12 FP multiplications, Chris@10: * (or, 24 additions, 0 multiplications, 12 fused multiply/add), Chris@10: * 30 stack variables, 2 constants, and 24 memory accesses Chris@10: */ Chris@10: #include "n.h" Chris@10: Chris@10: 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@10: { Chris@10: DK(KP866025403, +0.866025403784438646763723170752936183471402627); Chris@10: DK(KP500000000, +0.500000000000000000000000000000000000000000000); Chris@10: { Chris@10: INT i; Chris@10: 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@10: E TA, Tz; Chris@10: { Chris@10: E Tb, T3, Tx, Tp, Tj, Te, Ts, Ta, Tu, Ti, Tk; Chris@10: { Chris@10: E T1, T2, Tn, To; Chris@10: T1 = ri[0]; Chris@10: T2 = ri[WS(is, 3)]; Chris@10: Tn = ii[0]; Chris@10: To = ii[WS(is, 3)]; Chris@10: { Chris@10: E T4, T5, T7, T8; Chris@10: T4 = ri[WS(is, 2)]; Chris@10: Tb = T1 + T2; Chris@10: T3 = T1 - T2; Chris@10: Tx = Tn + To; Chris@10: Tp = Tn - To; Chris@10: T5 = ri[WS(is, 5)]; Chris@10: T7 = ri[WS(is, 4)]; Chris@10: T8 = ri[WS(is, 1)]; Chris@10: { Chris@10: E Tg, Tc, T6, Td, T9, Th; Chris@10: Tg = ii[WS(is, 2)]; Chris@10: Tc = T4 + T5; Chris@10: T6 = T4 - T5; Chris@10: Td = T7 + T8; Chris@10: T9 = T7 - T8; Chris@10: Th = ii[WS(is, 5)]; Chris@10: Tj = ii[WS(is, 4)]; Chris@10: Te = Tc + Td; Chris@10: TA = Td - Tc; Chris@10: Ts = T9 - T6; Chris@10: Ta = T6 + T9; Chris@10: Tu = Tg + Th; Chris@10: Ti = Tg - Th; Chris@10: Tk = ii[WS(is, 1)]; Chris@10: } Chris@10: } Chris@10: } Chris@10: ro[WS(os, 3)] = T3 + Ta; Chris@10: ro[0] = Tb + Te; Chris@10: { Chris@10: E Tf, Tv, Tl, Ty, Tr; Chris@10: Tf = FNMS(KP500000000, Ta, T3); Chris@10: Tv = Tj + Tk; Chris@10: Tl = Tj - Tk; Chris@10: { Chris@10: E Tt, Tw, Tq, Tm; Chris@10: Tt = FNMS(KP500000000, Te, Tb); Chris@10: Ty = Tu + Tv; Chris@10: Tw = Tu - Tv; Chris@10: Tq = Ti + Tl; Chris@10: Tm = Ti - Tl; Chris@10: io[0] = Tx + Ty; Chris@10: ro[WS(os, 1)] = FMA(KP866025403, Tm, Tf); Chris@10: ro[WS(os, 5)] = FNMS(KP866025403, Tm, Tf); Chris@10: Tr = FNMS(KP500000000, Tq, Tp); Chris@10: io[WS(os, 3)] = Tp + Tq; Chris@10: ro[WS(os, 2)] = FNMS(KP866025403, Tw, Tt); Chris@10: ro[WS(os, 4)] = FMA(KP866025403, Tw, Tt); Chris@10: } Chris@10: io[WS(os, 5)] = FNMS(KP866025403, Ts, Tr); Chris@10: io[WS(os, 1)] = FMA(KP866025403, Ts, Tr); Chris@10: Tz = FNMS(KP500000000, Ty, Tx); Chris@10: } Chris@10: } Chris@10: io[WS(os, 4)] = FMA(KP866025403, TA, Tz); Chris@10: io[WS(os, 2)] = FNMS(KP866025403, TA, Tz); Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kdft_desc desc = { 6, "n1_6", {24, 0, 12, 0}, &GENUS, 0, 0, 0, 0 }; Chris@10: Chris@10: void X(codelet_n1_6) (planner *p) { Chris@10: X(kdft_register) (p, n1_6, &desc); Chris@10: } Chris@10: Chris@10: #else /* HAVE_FMA */ Chris@10: Chris@10: /* Generated by: ../../../genfft/gen_notw.native -compact -variables 4 -pipeline-latency 4 -n 6 -name n1_6 -include n.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 36 FP additions, 8 FP multiplications, Chris@10: * (or, 32 additions, 4 multiplications, 4 fused multiply/add), Chris@10: * 23 stack variables, 2 constants, and 24 memory accesses Chris@10: */ Chris@10: #include "n.h" Chris@10: Chris@10: 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@10: { Chris@10: DK(KP866025403, +0.866025403784438646763723170752936183471402627); Chris@10: DK(KP500000000, +0.500000000000000000000000000000000000000000000); Chris@10: { Chris@10: INT i; Chris@10: 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@10: E T3, Tb, Tq, Tx, T6, Tc, T9, Td, Ta, Te, Ti, Tu, Tl, Tv, Tr; Chris@10: E Ty; Chris@10: { Chris@10: E T1, T2, To, Tp; Chris@10: T1 = ri[0]; Chris@10: T2 = ri[WS(is, 3)]; Chris@10: T3 = T1 - T2; Chris@10: Tb = T1 + T2; Chris@10: To = ii[0]; Chris@10: Tp = ii[WS(is, 3)]; Chris@10: Tq = To - Tp; Chris@10: Tx = To + Tp; Chris@10: } Chris@10: { Chris@10: E T4, T5, T7, T8; Chris@10: T4 = ri[WS(is, 2)]; Chris@10: T5 = ri[WS(is, 5)]; Chris@10: T6 = T4 - T5; Chris@10: Tc = T4 + T5; Chris@10: T7 = ri[WS(is, 4)]; Chris@10: T8 = ri[WS(is, 1)]; Chris@10: T9 = T7 - T8; Chris@10: Td = T7 + T8; Chris@10: } Chris@10: Ta = T6 + T9; Chris@10: Te = Tc + Td; Chris@10: { Chris@10: E Tg, Th, Tj, Tk; Chris@10: Tg = ii[WS(is, 2)]; Chris@10: Th = ii[WS(is, 5)]; Chris@10: Ti = Tg - Th; Chris@10: Tu = Tg + Th; Chris@10: Tj = ii[WS(is, 4)]; Chris@10: Tk = ii[WS(is, 1)]; Chris@10: Tl = Tj - Tk; Chris@10: Tv = Tj + Tk; Chris@10: } Chris@10: Tr = Ti + Tl; Chris@10: Ty = Tu + Tv; Chris@10: ro[WS(os, 3)] = T3 + Ta; Chris@10: io[WS(os, 3)] = Tq + Tr; Chris@10: ro[0] = Tb + Te; Chris@10: io[0] = Tx + Ty; Chris@10: { Chris@10: E Tf, Tm, Tn, Ts; Chris@10: Tf = FNMS(KP500000000, Ta, T3); Chris@10: Tm = KP866025403 * (Ti - Tl); Chris@10: ro[WS(os, 5)] = Tf - Tm; Chris@10: ro[WS(os, 1)] = Tf + Tm; Chris@10: Tn = KP866025403 * (T9 - T6); Chris@10: Ts = FNMS(KP500000000, Tr, Tq); Chris@10: io[WS(os, 1)] = Tn + Ts; Chris@10: io[WS(os, 5)] = Ts - Tn; Chris@10: } Chris@10: { Chris@10: E Tt, Tw, Tz, TA; Chris@10: Tt = FNMS(KP500000000, Te, Tb); Chris@10: Tw = KP866025403 * (Tu - Tv); Chris@10: ro[WS(os, 2)] = Tt - Tw; Chris@10: ro[WS(os, 4)] = Tt + Tw; Chris@10: Tz = FNMS(KP500000000, Ty, Tx); Chris@10: TA = KP866025403 * (Td - Tc); Chris@10: io[WS(os, 2)] = Tz - TA; Chris@10: io[WS(os, 4)] = TA + Tz; Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kdft_desc desc = { 6, "n1_6", {32, 4, 4, 0}, &GENUS, 0, 0, 0, 0 }; Chris@10: Chris@10: void X(codelet_n1_6) (planner *p) { Chris@10: X(kdft_register) (p, n1_6, &desc); Chris@10: } Chris@10: Chris@10: #endif /* HAVE_FMA */