cannam@127: /* cannam@127: * Copyright (c) 2003, 2007-14 Matteo Frigo cannam@127: * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology cannam@127: * cannam@127: * This program is free software; you can redistribute it and/or modify cannam@127: * it under the terms of the GNU General Public License as published by cannam@127: * the Free Software Foundation; either version 2 of the License, or cannam@127: * (at your option) any later version. cannam@127: * cannam@127: * This program is distributed in the hope that it will be useful, cannam@127: * but WITHOUT ANY WARRANTY; without even the implied warranty of cannam@127: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the cannam@127: * GNU General Public License for more details. cannam@127: * cannam@127: * You should have received a copy of the GNU General Public License cannam@127: * along with this program; if not, write to the Free Software cannam@127: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA cannam@127: * cannam@127: */ cannam@127: cannam@127: /* This file was automatically generated --- DO NOT EDIT */ cannam@127: /* Generated on Sat Jul 30 16:35:50 EDT 2016 */ cannam@127: cannam@127: #include "codelet-dft.h" cannam@127: cannam@127: #ifdef HAVE_FMA cannam@127: cannam@127: /* 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@127: cannam@127: /* cannam@127: * This function contains 52 FP additions, 8 FP multiplications, cannam@127: * (or, 44 additions, 0 multiplications, 8 fused multiply/add), cannam@127: * 36 stack variables, 1 constants, and 32 memory accesses cannam@127: */ cannam@127: #include "n.h" cannam@127: cannam@127: 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@127: { cannam@127: DK(KP707106781, +0.707106781186547524400844362104849039284835938); cannam@127: { cannam@127: INT i; cannam@127: 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@127: E TF, TE, TD, TI; cannam@127: { cannam@127: E Tn, T3, TC, Ti, TB, T6, To, Tl, Td, TN, Tz, TH, Ta, Tq, Tt; cannam@127: E TM; cannam@127: { cannam@127: E T4, T5, Tj, Tk; cannam@127: { cannam@127: E T1, T2, Tg, Th; cannam@127: T1 = ri[0]; cannam@127: T2 = ri[WS(is, 4)]; cannam@127: Tg = ii[0]; cannam@127: Th = ii[WS(is, 4)]; cannam@127: T4 = ri[WS(is, 2)]; cannam@127: Tn = T1 - T2; cannam@127: T3 = T1 + T2; cannam@127: TC = Tg - Th; cannam@127: Ti = Tg + Th; cannam@127: T5 = ri[WS(is, 6)]; cannam@127: } cannam@127: Tj = ii[WS(is, 2)]; cannam@127: Tk = ii[WS(is, 6)]; cannam@127: { cannam@127: E Tb, Tc, Tw, Tx; cannam@127: Tb = ri[WS(is, 7)]; cannam@127: TB = T4 - T5; cannam@127: T6 = T4 + T5; cannam@127: To = Tj - Tk; cannam@127: Tl = Tj + Tk; cannam@127: Tc = ri[WS(is, 3)]; cannam@127: Tw = ii[WS(is, 7)]; cannam@127: Tx = ii[WS(is, 3)]; cannam@127: { cannam@127: E T8, Tv, Ty, T9, Tr, Ts; cannam@127: T8 = ri[WS(is, 1)]; cannam@127: Td = Tb + Tc; cannam@127: Tv = Tb - Tc; cannam@127: TN = Tw + Tx; cannam@127: Ty = Tw - Tx; cannam@127: T9 = ri[WS(is, 5)]; cannam@127: Tr = ii[WS(is, 1)]; cannam@127: Ts = ii[WS(is, 5)]; cannam@127: Tz = Tv - Ty; cannam@127: TH = Tv + Ty; cannam@127: Ta = T8 + T9; cannam@127: Tq = T8 - T9; cannam@127: Tt = Tr - Ts; cannam@127: TM = Tr + Ts; cannam@127: } cannam@127: } cannam@127: } cannam@127: { cannam@127: E TL, TG, Tu, Tf, Tm, TO; cannam@127: { cannam@127: E T7, Te, TP, TQ; cannam@127: TL = T3 - T6; cannam@127: T7 = T3 + T6; cannam@127: TG = Tt - Tq; cannam@127: Tu = Tq + Tt; cannam@127: Te = Ta + Td; cannam@127: Tf = Td - Ta; cannam@127: Tm = Ti - Tl; cannam@127: TP = Ti + Tl; cannam@127: TQ = TM + TN; cannam@127: TO = TM - TN; cannam@127: ro[0] = T7 + Te; cannam@127: ro[WS(os, 4)] = T7 - Te; cannam@127: io[0] = TP + TQ; cannam@127: io[WS(os, 4)] = TP - TQ; cannam@127: } cannam@127: { cannam@127: E Tp, TA, TJ, TK; cannam@127: TF = Tn - To; cannam@127: Tp = Tn + To; cannam@127: io[WS(os, 6)] = Tm - Tf; cannam@127: io[WS(os, 2)] = Tf + Tm; cannam@127: ro[WS(os, 2)] = TL + TO; cannam@127: ro[WS(os, 6)] = TL - TO; cannam@127: TA = Tu + Tz; cannam@127: TE = Tz - Tu; cannam@127: TD = TB + TC; cannam@127: TJ = TC - TB; cannam@127: TK = TG + TH; cannam@127: TI = TG - TH; cannam@127: ro[WS(os, 1)] = FMA(KP707106781, TA, Tp); cannam@127: ro[WS(os, 5)] = FNMS(KP707106781, TA, Tp); cannam@127: io[WS(os, 1)] = FMA(KP707106781, TK, TJ); cannam@127: io[WS(os, 5)] = FNMS(KP707106781, TK, TJ); cannam@127: } cannam@127: } cannam@127: } cannam@127: io[WS(os, 3)] = FMA(KP707106781, TE, TD); cannam@127: io[WS(os, 7)] = FNMS(KP707106781, TE, TD); cannam@127: ro[WS(os, 3)] = FMA(KP707106781, TI, TF); cannam@127: ro[WS(os, 7)] = FNMS(KP707106781, TI, TF); cannam@127: } cannam@127: } cannam@127: } cannam@127: cannam@127: static const kdft_desc desc = { 8, "n1_8", {44, 0, 8, 0}, &GENUS, 0, 0, 0, 0 }; cannam@127: cannam@127: void X(codelet_n1_8) (planner *p) { cannam@127: X(kdft_register) (p, n1_8, &desc); cannam@127: } cannam@127: cannam@127: #else /* HAVE_FMA */ cannam@127: cannam@127: /* Generated by: ../../../genfft/gen_notw.native -compact -variables 4 -pipeline-latency 4 -n 8 -name n1_8 -include n.h */ cannam@127: cannam@127: /* cannam@127: * This function contains 52 FP additions, 4 FP multiplications, cannam@127: * (or, 52 additions, 4 multiplications, 0 fused multiply/add), cannam@127: * 28 stack variables, 1 constants, and 32 memory accesses cannam@127: */ cannam@127: #include "n.h" cannam@127: cannam@127: 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@127: { cannam@127: DK(KP707106781, +0.707106781186547524400844362104849039284835938); cannam@127: { cannam@127: INT i; cannam@127: 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@127: E T3, Tn, Ti, TC, T6, TB, Tl, To, Td, TN, Tz, TH, Ta, TM, Tu; cannam@127: E TG; cannam@127: { cannam@127: E T1, T2, Tj, Tk; cannam@127: T1 = ri[0]; cannam@127: T2 = ri[WS(is, 4)]; cannam@127: T3 = T1 + T2; cannam@127: Tn = T1 - T2; cannam@127: { cannam@127: E Tg, Th, T4, T5; cannam@127: Tg = ii[0]; cannam@127: Th = ii[WS(is, 4)]; cannam@127: Ti = Tg + Th; cannam@127: TC = Tg - Th; cannam@127: T4 = ri[WS(is, 2)]; cannam@127: T5 = ri[WS(is, 6)]; cannam@127: T6 = T4 + T5; cannam@127: TB = T4 - T5; cannam@127: } cannam@127: Tj = ii[WS(is, 2)]; cannam@127: Tk = ii[WS(is, 6)]; cannam@127: Tl = Tj + Tk; cannam@127: To = Tj - Tk; cannam@127: { cannam@127: E Tb, Tc, Tv, Tw, Tx, Ty; cannam@127: Tb = ri[WS(is, 7)]; cannam@127: Tc = ri[WS(is, 3)]; cannam@127: Tv = Tb - Tc; cannam@127: Tw = ii[WS(is, 7)]; cannam@127: Tx = ii[WS(is, 3)]; cannam@127: Ty = Tw - Tx; cannam@127: Td = Tb + Tc; cannam@127: TN = Tw + Tx; cannam@127: Tz = Tv - Ty; cannam@127: TH = Tv + Ty; cannam@127: } cannam@127: { cannam@127: E T8, T9, Tq, Tr, Ts, Tt; cannam@127: T8 = ri[WS(is, 1)]; cannam@127: T9 = ri[WS(is, 5)]; cannam@127: Tq = T8 - T9; cannam@127: Tr = ii[WS(is, 1)]; cannam@127: Ts = ii[WS(is, 5)]; cannam@127: Tt = Tr - Ts; cannam@127: Ta = T8 + T9; cannam@127: TM = Tr + Ts; cannam@127: Tu = Tq + Tt; cannam@127: TG = Tt - Tq; cannam@127: } cannam@127: } cannam@127: { cannam@127: E T7, Te, TP, TQ; cannam@127: T7 = T3 + T6; cannam@127: Te = Ta + Td; cannam@127: ro[WS(os, 4)] = T7 - Te; cannam@127: ro[0] = T7 + Te; cannam@127: TP = Ti + Tl; cannam@127: TQ = TM + TN; cannam@127: io[WS(os, 4)] = TP - TQ; cannam@127: io[0] = TP + TQ; cannam@127: } cannam@127: { cannam@127: E Tf, Tm, TL, TO; cannam@127: Tf = Td - Ta; cannam@127: Tm = Ti - Tl; cannam@127: io[WS(os, 2)] = Tf + Tm; cannam@127: io[WS(os, 6)] = Tm - Tf; cannam@127: TL = T3 - T6; cannam@127: TO = TM - TN; cannam@127: ro[WS(os, 6)] = TL - TO; cannam@127: ro[WS(os, 2)] = TL + TO; cannam@127: } cannam@127: { cannam@127: E Tp, TA, TJ, TK; cannam@127: Tp = Tn + To; cannam@127: TA = KP707106781 * (Tu + Tz); cannam@127: ro[WS(os, 5)] = Tp - TA; cannam@127: ro[WS(os, 1)] = Tp + TA; cannam@127: TJ = TC - TB; cannam@127: TK = KP707106781 * (TG + TH); cannam@127: io[WS(os, 5)] = TJ - TK; cannam@127: io[WS(os, 1)] = TJ + TK; cannam@127: } cannam@127: { cannam@127: E TD, TE, TF, TI; cannam@127: TD = TB + TC; cannam@127: TE = KP707106781 * (Tz - Tu); cannam@127: io[WS(os, 7)] = TD - TE; cannam@127: io[WS(os, 3)] = TD + TE; cannam@127: TF = Tn - To; cannam@127: TI = KP707106781 * (TG - TH); cannam@127: ro[WS(os, 7)] = TF - TI; cannam@127: ro[WS(os, 3)] = TF + TI; cannam@127: } cannam@127: } cannam@127: } cannam@127: } cannam@127: cannam@127: static const kdft_desc desc = { 8, "n1_8", {52, 4, 0, 0}, &GENUS, 0, 0, 0, 0 }; cannam@127: cannam@127: void X(codelet_n1_8) (planner *p) { cannam@127: X(kdft_register) (p, n1_8, &desc); cannam@127: } cannam@127: cannam@127: #endif /* HAVE_FMA */