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 6 -name n1_6 -include n.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 36 FP additions, 12 FP multiplications, cannam@95: * (or, 24 additions, 0 multiplications, 12 fused multiply/add), cannam@95: * 30 stack variables, 2 constants, and 24 memory accesses cannam@95: */ cannam@95: #include "n.h" cannam@95: cannam@95: static void n1_6(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(KP866025403, +0.866025403784438646763723170752936183471402627); cannam@95: DK(KP500000000, +0.500000000000000000000000000000000000000000000); 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(24, is), MAKE_VOLATILE_STRIDE(24, os)) { cannam@95: E TA, Tz; cannam@95: { cannam@95: E Tb, T3, Tx, Tp, Tj, Te, Ts, Ta, Tu, Ti, Tk; cannam@95: { cannam@95: E T1, T2, Tn, To; cannam@95: T1 = ri[0]; cannam@95: T2 = ri[WS(is, 3)]; cannam@95: Tn = ii[0]; cannam@95: To = ii[WS(is, 3)]; cannam@95: { cannam@95: E T4, T5, T7, T8; cannam@95: T4 = ri[WS(is, 2)]; cannam@95: Tb = T1 + T2; cannam@95: T3 = T1 - T2; cannam@95: Tx = Tn + To; cannam@95: Tp = Tn - To; cannam@95: T5 = ri[WS(is, 5)]; cannam@95: T7 = ri[WS(is, 4)]; cannam@95: T8 = ri[WS(is, 1)]; cannam@95: { cannam@95: E Tg, Tc, T6, Td, T9, Th; cannam@95: Tg = ii[WS(is, 2)]; cannam@95: Tc = T4 + T5; cannam@95: T6 = T4 - T5; cannam@95: Td = T7 + T8; cannam@95: T9 = T7 - T8; cannam@95: Th = ii[WS(is, 5)]; cannam@95: Tj = ii[WS(is, 4)]; cannam@95: Te = Tc + Td; cannam@95: TA = Td - Tc; cannam@95: Ts = T9 - T6; cannam@95: Ta = T6 + T9; cannam@95: Tu = Tg + Th; cannam@95: Ti = Tg - Th; cannam@95: Tk = ii[WS(is, 1)]; cannam@95: } cannam@95: } cannam@95: } cannam@95: ro[WS(os, 3)] = T3 + Ta; cannam@95: ro[0] = Tb + Te; cannam@95: { cannam@95: E Tf, Tv, Tl, Ty, Tr; cannam@95: Tf = FNMS(KP500000000, Ta, T3); cannam@95: Tv = Tj + Tk; cannam@95: Tl = Tj - Tk; cannam@95: { cannam@95: E Tt, Tw, Tq, Tm; cannam@95: Tt = FNMS(KP500000000, Te, Tb); cannam@95: Ty = Tu + Tv; cannam@95: Tw = Tu - Tv; cannam@95: Tq = Ti + Tl; cannam@95: Tm = Ti - Tl; cannam@95: io[0] = Tx + Ty; cannam@95: ro[WS(os, 1)] = FMA(KP866025403, Tm, Tf); cannam@95: ro[WS(os, 5)] = FNMS(KP866025403, Tm, Tf); cannam@95: Tr = FNMS(KP500000000, Tq, Tp); cannam@95: io[WS(os, 3)] = Tp + Tq; cannam@95: ro[WS(os, 2)] = FNMS(KP866025403, Tw, Tt); cannam@95: ro[WS(os, 4)] = FMA(KP866025403, Tw, Tt); cannam@95: } cannam@95: io[WS(os, 5)] = FNMS(KP866025403, Ts, Tr); cannam@95: io[WS(os, 1)] = FMA(KP866025403, Ts, Tr); cannam@95: Tz = FNMS(KP500000000, Ty, Tx); cannam@95: } cannam@95: } cannam@95: io[WS(os, 4)] = FMA(KP866025403, TA, Tz); cannam@95: io[WS(os, 2)] = FNMS(KP866025403, TA, Tz); cannam@95: } cannam@95: } cannam@95: } cannam@95: cannam@95: static const kdft_desc desc = { 6, "n1_6", {24, 0, 12, 0}, &GENUS, 0, 0, 0, 0 }; cannam@95: cannam@95: void X(codelet_n1_6) (planner *p) { cannam@95: X(kdft_register) (p, n1_6, &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 6 -name n1_6 -include n.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 36 FP additions, 8 FP multiplications, cannam@95: * (or, 32 additions, 4 multiplications, 4 fused multiply/add), cannam@95: * 23 stack variables, 2 constants, and 24 memory accesses cannam@95: */ cannam@95: #include "n.h" cannam@95: cannam@95: static void n1_6(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(KP866025403, +0.866025403784438646763723170752936183471402627); cannam@95: DK(KP500000000, +0.500000000000000000000000000000000000000000000); 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(24, is), MAKE_VOLATILE_STRIDE(24, os)) { cannam@95: E T3, Tb, Tq, Tx, T6, Tc, T9, Td, Ta, Te, Ti, Tu, Tl, Tv, Tr; cannam@95: E Ty; cannam@95: { cannam@95: E T1, T2, To, Tp; cannam@95: T1 = ri[0]; cannam@95: T2 = ri[WS(is, 3)]; cannam@95: T3 = T1 - T2; cannam@95: Tb = T1 + T2; cannam@95: To = ii[0]; cannam@95: Tp = ii[WS(is, 3)]; cannam@95: Tq = To - Tp; cannam@95: Tx = To + Tp; cannam@95: } cannam@95: { cannam@95: E T4, T5, T7, T8; cannam@95: T4 = ri[WS(is, 2)]; cannam@95: T5 = ri[WS(is, 5)]; cannam@95: T6 = T4 - T5; cannam@95: Tc = T4 + T5; cannam@95: T7 = ri[WS(is, 4)]; cannam@95: T8 = ri[WS(is, 1)]; cannam@95: T9 = T7 - T8; cannam@95: Td = T7 + T8; cannam@95: } cannam@95: Ta = T6 + T9; cannam@95: Te = Tc + Td; cannam@95: { cannam@95: E Tg, Th, Tj, Tk; cannam@95: Tg = ii[WS(is, 2)]; cannam@95: Th = ii[WS(is, 5)]; cannam@95: Ti = Tg - Th; cannam@95: Tu = Tg + Th; cannam@95: Tj = ii[WS(is, 4)]; cannam@95: Tk = ii[WS(is, 1)]; cannam@95: Tl = Tj - Tk; cannam@95: Tv = Tj + Tk; cannam@95: } cannam@95: Tr = Ti + Tl; cannam@95: Ty = Tu + Tv; cannam@95: ro[WS(os, 3)] = T3 + Ta; cannam@95: io[WS(os, 3)] = Tq + Tr; cannam@95: ro[0] = Tb + Te; cannam@95: io[0] = Tx + Ty; cannam@95: { cannam@95: E Tf, Tm, Tn, Ts; cannam@95: Tf = FNMS(KP500000000, Ta, T3); cannam@95: Tm = KP866025403 * (Ti - Tl); cannam@95: ro[WS(os, 5)] = Tf - Tm; cannam@95: ro[WS(os, 1)] = Tf + Tm; cannam@95: Tn = KP866025403 * (T9 - T6); cannam@95: Ts = FNMS(KP500000000, Tr, Tq); cannam@95: io[WS(os, 1)] = Tn + Ts; cannam@95: io[WS(os, 5)] = Ts - Tn; cannam@95: } cannam@95: { cannam@95: E Tt, Tw, Tz, TA; cannam@95: Tt = FNMS(KP500000000, Te, Tb); cannam@95: Tw = KP866025403 * (Tu - Tv); cannam@95: ro[WS(os, 2)] = Tt - Tw; cannam@95: ro[WS(os, 4)] = Tt + Tw; cannam@95: Tz = FNMS(KP500000000, Ty, Tx); cannam@95: TA = KP866025403 * (Td - Tc); cannam@95: io[WS(os, 2)] = Tz - TA; cannam@95: io[WS(os, 4)] = TA + Tz; cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: cannam@95: static const kdft_desc desc = { 6, "n1_6", {32, 4, 4, 0}, &GENUS, 0, 0, 0, 0 }; cannam@95: cannam@95: void X(codelet_n1_6) (planner *p) { cannam@95: X(kdft_register) (p, n1_6, &desc); cannam@95: } cannam@95: cannam@95: #endif /* HAVE_FMA */