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 8 -name n1_8 -include n.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 52 FP additions, 8 FP multiplications, cannam@95: * (or, 44 additions, 0 multiplications, 8 fused multiply/add), cannam@95: * 36 stack variables, 1 constants, and 32 memory accesses cannam@95: */ cannam@95: #include "n.h" cannam@95: cannam@95: static void n1_8(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(KP707106781, +0.707106781186547524400844362104849039284835938); 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(32, is), MAKE_VOLATILE_STRIDE(32, os)) { cannam@95: E TF, TE, TD, TI; cannam@95: { cannam@95: E Tn, T3, TC, Ti, TB, T6, To, Tl, Td, TN, Tz, TH, Ta, Tq, Tt; cannam@95: E TM; cannam@95: { cannam@95: E T4, T5, Tj, Tk; cannam@95: { cannam@95: E T1, T2, Tg, Th; cannam@95: T1 = ri[0]; cannam@95: T2 = ri[WS(is, 4)]; cannam@95: Tg = ii[0]; cannam@95: Th = ii[WS(is, 4)]; cannam@95: T4 = ri[WS(is, 2)]; cannam@95: Tn = T1 - T2; cannam@95: T3 = T1 + T2; cannam@95: TC = Tg - Th; cannam@95: Ti = Tg + Th; cannam@95: T5 = ri[WS(is, 6)]; cannam@95: } cannam@95: Tj = ii[WS(is, 2)]; cannam@95: Tk = ii[WS(is, 6)]; cannam@95: { cannam@95: E Tb, Tc, Tw, Tx; cannam@95: Tb = ri[WS(is, 7)]; cannam@95: TB = T4 - T5; cannam@95: T6 = T4 + T5; cannam@95: To = Tj - Tk; cannam@95: Tl = Tj + Tk; cannam@95: Tc = ri[WS(is, 3)]; cannam@95: Tw = ii[WS(is, 7)]; cannam@95: Tx = ii[WS(is, 3)]; cannam@95: { cannam@95: E T8, Tv, Ty, T9, Tr, Ts; cannam@95: T8 = ri[WS(is, 1)]; cannam@95: Td = Tb + Tc; cannam@95: Tv = Tb - Tc; cannam@95: TN = Tw + Tx; cannam@95: Ty = Tw - Tx; cannam@95: T9 = ri[WS(is, 5)]; cannam@95: Tr = ii[WS(is, 1)]; cannam@95: Ts = ii[WS(is, 5)]; cannam@95: Tz = Tv - Ty; cannam@95: TH = Tv + Ty; cannam@95: Ta = T8 + T9; cannam@95: Tq = T8 - T9; cannam@95: Tt = Tr - Ts; cannam@95: TM = Tr + Ts; cannam@95: } cannam@95: } cannam@95: } cannam@95: { cannam@95: E TL, TG, Tu, Tf, Tm, TO; cannam@95: { cannam@95: E T7, Te, TP, TQ; cannam@95: TL = T3 - T6; cannam@95: T7 = T3 + T6; cannam@95: TG = Tt - Tq; cannam@95: Tu = Tq + Tt; cannam@95: Te = Ta + Td; cannam@95: Tf = Td - Ta; cannam@95: Tm = Ti - Tl; cannam@95: TP = Ti + Tl; cannam@95: TQ = TM + TN; cannam@95: TO = TM - TN; cannam@95: ro[0] = T7 + Te; cannam@95: ro[WS(os, 4)] = T7 - Te; cannam@95: io[0] = TP + TQ; cannam@95: io[WS(os, 4)] = TP - TQ; cannam@95: } cannam@95: { cannam@95: E Tp, TA, TJ, TK; cannam@95: TF = Tn - To; cannam@95: Tp = Tn + To; cannam@95: io[WS(os, 6)] = Tm - Tf; cannam@95: io[WS(os, 2)] = Tf + Tm; cannam@95: ro[WS(os, 2)] = TL + TO; cannam@95: ro[WS(os, 6)] = TL - TO; cannam@95: TA = Tu + Tz; cannam@95: TE = Tz - Tu; cannam@95: TD = TB + TC; cannam@95: TJ = TC - TB; cannam@95: TK = TG + TH; cannam@95: TI = TG - TH; cannam@95: ro[WS(os, 1)] = FMA(KP707106781, TA, Tp); cannam@95: ro[WS(os, 5)] = FNMS(KP707106781, TA, Tp); cannam@95: io[WS(os, 1)] = FMA(KP707106781, TK, TJ); cannam@95: io[WS(os, 5)] = FNMS(KP707106781, TK, TJ); cannam@95: } cannam@95: } cannam@95: } cannam@95: io[WS(os, 3)] = FMA(KP707106781, TE, TD); cannam@95: io[WS(os, 7)] = FNMS(KP707106781, TE, TD); cannam@95: ro[WS(os, 3)] = FMA(KP707106781, TI, TF); cannam@95: ro[WS(os, 7)] = FNMS(KP707106781, TI, TF); cannam@95: } cannam@95: } cannam@95: } cannam@95: cannam@95: static const kdft_desc desc = { 8, "n1_8", {44, 0, 8, 0}, &GENUS, 0, 0, 0, 0 }; cannam@95: cannam@95: void X(codelet_n1_8) (planner *p) { cannam@95: X(kdft_register) (p, n1_8, &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 8 -name n1_8 -include n.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 52 FP additions, 4 FP multiplications, cannam@95: * (or, 52 additions, 4 multiplications, 0 fused multiply/add), cannam@95: * 28 stack variables, 1 constants, and 32 memory accesses cannam@95: */ cannam@95: #include "n.h" cannam@95: cannam@95: static void n1_8(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(KP707106781, +0.707106781186547524400844362104849039284835938); 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(32, is), MAKE_VOLATILE_STRIDE(32, os)) { cannam@95: E T3, Tn, Ti, TC, T6, TB, Tl, To, Td, TN, Tz, TH, Ta, TM, Tu; cannam@95: E TG; cannam@95: { cannam@95: E T1, T2, Tj, Tk; cannam@95: T1 = ri[0]; cannam@95: T2 = ri[WS(is, 4)]; cannam@95: T3 = T1 + T2; cannam@95: Tn = T1 - T2; cannam@95: { cannam@95: E Tg, Th, T4, T5; cannam@95: Tg = ii[0]; cannam@95: Th = ii[WS(is, 4)]; cannam@95: Ti = Tg + Th; cannam@95: TC = Tg - Th; cannam@95: T4 = ri[WS(is, 2)]; cannam@95: T5 = ri[WS(is, 6)]; cannam@95: T6 = T4 + T5; cannam@95: TB = T4 - T5; cannam@95: } cannam@95: Tj = ii[WS(is, 2)]; cannam@95: Tk = ii[WS(is, 6)]; cannam@95: Tl = Tj + Tk; cannam@95: To = Tj - Tk; cannam@95: { cannam@95: E Tb, Tc, Tv, Tw, Tx, Ty; cannam@95: Tb = ri[WS(is, 7)]; cannam@95: Tc = ri[WS(is, 3)]; cannam@95: Tv = Tb - Tc; cannam@95: Tw = ii[WS(is, 7)]; cannam@95: Tx = ii[WS(is, 3)]; cannam@95: Ty = Tw - Tx; cannam@95: Td = Tb + Tc; cannam@95: TN = Tw + Tx; cannam@95: Tz = Tv - Ty; cannam@95: TH = Tv + Ty; cannam@95: } cannam@95: { cannam@95: E T8, T9, Tq, Tr, Ts, Tt; cannam@95: T8 = ri[WS(is, 1)]; cannam@95: T9 = ri[WS(is, 5)]; cannam@95: Tq = T8 - T9; cannam@95: Tr = ii[WS(is, 1)]; cannam@95: Ts = ii[WS(is, 5)]; cannam@95: Tt = Tr - Ts; cannam@95: Ta = T8 + T9; cannam@95: TM = Tr + Ts; cannam@95: Tu = Tq + Tt; cannam@95: TG = Tt - Tq; cannam@95: } cannam@95: } cannam@95: { cannam@95: E T7, Te, TP, TQ; cannam@95: T7 = T3 + T6; cannam@95: Te = Ta + Td; cannam@95: ro[WS(os, 4)] = T7 - Te; cannam@95: ro[0] = T7 + Te; cannam@95: TP = Ti + Tl; cannam@95: TQ = TM + TN; cannam@95: io[WS(os, 4)] = TP - TQ; cannam@95: io[0] = TP + TQ; cannam@95: } cannam@95: { cannam@95: E Tf, Tm, TL, TO; cannam@95: Tf = Td - Ta; cannam@95: Tm = Ti - Tl; cannam@95: io[WS(os, 2)] = Tf + Tm; cannam@95: io[WS(os, 6)] = Tm - Tf; cannam@95: TL = T3 - T6; cannam@95: TO = TM - TN; cannam@95: ro[WS(os, 6)] = TL - TO; cannam@95: ro[WS(os, 2)] = TL + TO; cannam@95: } cannam@95: { cannam@95: E Tp, TA, TJ, TK; cannam@95: Tp = Tn + To; cannam@95: TA = KP707106781 * (Tu + Tz); cannam@95: ro[WS(os, 5)] = Tp - TA; cannam@95: ro[WS(os, 1)] = Tp + TA; cannam@95: TJ = TC - TB; cannam@95: TK = KP707106781 * (TG + TH); cannam@95: io[WS(os, 5)] = TJ - TK; cannam@95: io[WS(os, 1)] = TJ + TK; cannam@95: } cannam@95: { cannam@95: E TD, TE, TF, TI; cannam@95: TD = TB + TC; cannam@95: TE = KP707106781 * (Tz - Tu); cannam@95: io[WS(os, 7)] = TD - TE; cannam@95: io[WS(os, 3)] = TD + TE; cannam@95: TF = Tn - To; cannam@95: TI = KP707106781 * (TG - TH); cannam@95: ro[WS(os, 7)] = TF - TI; cannam@95: ro[WS(os, 3)] = TF + TI; cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: cannam@95: static const kdft_desc desc = { 8, "n1_8", {52, 4, 0, 0}, &GENUS, 0, 0, 0, 0 }; cannam@95: cannam@95: void X(codelet_n1_8) (planner *p) { cannam@95: X(kdft_register) (p, n1_8, &desc); cannam@95: } cannam@95: cannam@95: #endif /* HAVE_FMA */