cannam@167: /* cannam@167: * Copyright (c) 2003, 2007-14 Matteo Frigo cannam@167: * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology cannam@167: * cannam@167: * This program is free software; you can redistribute it and/or modify cannam@167: * it under the terms of the GNU General Public License as published by cannam@167: * the Free Software Foundation; either version 2 of the License, or cannam@167: * (at your option) any later version. cannam@167: * cannam@167: * This program is distributed in the hope that it will be useful, cannam@167: * but WITHOUT ANY WARRANTY; without even the implied warranty of cannam@167: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the cannam@167: * GNU General Public License for more details. cannam@167: * cannam@167: * You should have received a copy of the GNU General Public License cannam@167: * along with this program; if not, write to the Free Software cannam@167: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA cannam@167: * cannam@167: */ cannam@167: cannam@167: /* This file was automatically generated --- DO NOT EDIT */ cannam@167: /* Generated on Thu May 24 08:04:10 EDT 2018 */ cannam@167: cannam@167: #include "dft/codelet-dft.h" cannam@167: cannam@167: #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA) cannam@167: cannam@167: /* Generated by: ../../../genfft/gen_notw.native -fma -compact -variables 4 -pipeline-latency 4 -n 9 -name n1_9 -include dft/scalar/n.h */ cannam@167: cannam@167: /* cannam@167: * This function contains 80 FP additions, 56 FP multiplications, cannam@167: * (or, 24 additions, 0 multiplications, 56 fused multiply/add), cannam@167: * 41 stack variables, 10 constants, and 36 memory accesses cannam@167: */ cannam@167: #include "dft/scalar/n.h" cannam@167: cannam@167: static void n1_9(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) cannam@167: { cannam@167: DK(KP954188894, +0.954188894138671133499268364187245676532219158); cannam@167: DK(KP363970234, +0.363970234266202361351047882776834043890471784); cannam@167: DK(KP852868531, +0.852868531952443209628250963940074071936020296); cannam@167: DK(KP492403876, +0.492403876506104029683371512294761506835321626); cannam@167: DK(KP984807753, +0.984807753012208059366743024589523013670643252); cannam@167: DK(KP777861913, +0.777861913430206160028177977318626690410586096); cannam@167: DK(KP839099631, +0.839099631177280011763127298123181364687434283); cannam@167: DK(KP176326980, +0.176326980708464973471090386868618986121633062); cannam@167: DK(KP866025403, +0.866025403784438646763723170752936183471402627); cannam@167: DK(KP500000000, +0.500000000000000000000000000000000000000000000); cannam@167: { cannam@167: INT i; cannam@167: for (i = v; i > 0; i = i - 1, ri = ri + ivs, ii = ii + ivs, ro = ro + ovs, io = io + ovs, MAKE_VOLATILE_STRIDE(36, is), MAKE_VOLATILE_STRIDE(36, os)) { cannam@167: E T5, TL, Tm, Tl, T1f, TM, Ta, T1c, TF, TW, TI, TX, Tf, T1d, Ts; cannam@167: E TZ, Tx, T10; cannam@167: { cannam@167: E T1, T2, T3, T4; cannam@167: T1 = ri[0]; cannam@167: T2 = ri[WS(is, 3)]; cannam@167: T3 = ri[WS(is, 6)]; cannam@167: T4 = T2 + T3; cannam@167: T5 = T1 + T4; cannam@167: TL = FNMS(KP500000000, T4, T1); cannam@167: Tm = T3 - T2; cannam@167: } cannam@167: { cannam@167: E Th, Ti, Tj, Tk; cannam@167: Th = ii[0]; cannam@167: Ti = ii[WS(is, 3)]; cannam@167: Tj = ii[WS(is, 6)]; cannam@167: Tk = Ti + Tj; cannam@167: Tl = FNMS(KP500000000, Tk, Th); cannam@167: T1f = Th + Tk; cannam@167: TM = Ti - Tj; cannam@167: } cannam@167: { cannam@167: E T6, Tz, T9, TE, TC, TH, TD, TG; cannam@167: T6 = ri[WS(is, 1)]; cannam@167: Tz = ii[WS(is, 1)]; cannam@167: { cannam@167: E T7, T8, TA, TB; cannam@167: T7 = ri[WS(is, 4)]; cannam@167: T8 = ri[WS(is, 7)]; cannam@167: T9 = T7 + T8; cannam@167: TE = T7 - T8; cannam@167: TA = ii[WS(is, 4)]; cannam@167: TB = ii[WS(is, 7)]; cannam@167: TC = TA + TB; cannam@167: TH = TB - TA; cannam@167: } cannam@167: Ta = T6 + T9; cannam@167: T1c = Tz + TC; cannam@167: TD = FNMS(KP500000000, TC, Tz); cannam@167: TF = FNMS(KP866025403, TE, TD); cannam@167: TW = FMA(KP866025403, TE, TD); cannam@167: TG = FNMS(KP500000000, T9, T6); cannam@167: TI = FNMS(KP866025403, TH, TG); cannam@167: TX = FMA(KP866025403, TH, TG); cannam@167: } cannam@167: { cannam@167: E Tb, Tt, Te, Tw, Tr, Tu, To, Tv; cannam@167: Tb = ri[WS(is, 2)]; cannam@167: Tt = ii[WS(is, 2)]; cannam@167: { cannam@167: E Tc, Td, Tp, Tq; cannam@167: Tc = ri[WS(is, 5)]; cannam@167: Td = ri[WS(is, 8)]; cannam@167: Te = Tc + Td; cannam@167: Tw = Td - Tc; cannam@167: Tp = ii[WS(is, 5)]; cannam@167: Tq = ii[WS(is, 8)]; cannam@167: Tr = Tp - Tq; cannam@167: Tu = Tp + Tq; cannam@167: } cannam@167: Tf = Tb + Te; cannam@167: T1d = Tt + Tu; cannam@167: To = FNMS(KP500000000, Te, Tb); cannam@167: Ts = FMA(KP866025403, Tr, To); cannam@167: TZ = FNMS(KP866025403, Tr, To); cannam@167: Tv = FNMS(KP500000000, Tu, Tt); cannam@167: Tx = FMA(KP866025403, Tw, Tv); cannam@167: T10 = FNMS(KP866025403, Tw, Tv); cannam@167: } cannam@167: { cannam@167: E T1e, Tg, T1b, T1i, T1g, T1h; cannam@167: T1e = T1c - T1d; cannam@167: Tg = Ta + Tf; cannam@167: T1b = FNMS(KP500000000, Tg, T5); cannam@167: ro[0] = T5 + Tg; cannam@167: ro[WS(os, 3)] = FMA(KP866025403, T1e, T1b); cannam@167: ro[WS(os, 6)] = FNMS(KP866025403, T1e, T1b); cannam@167: T1i = Tf - Ta; cannam@167: T1g = T1c + T1d; cannam@167: T1h = FNMS(KP500000000, T1g, T1f); cannam@167: io[WS(os, 3)] = FMA(KP866025403, T1i, T1h); cannam@167: io[0] = T1f + T1g; cannam@167: io[WS(os, 6)] = FNMS(KP866025403, T1i, T1h); cannam@167: } cannam@167: { cannam@167: E Tn, TN, TK, TS, TQ, TU, TR, TT; cannam@167: Tn = FMA(KP866025403, Tm, Tl); cannam@167: TN = FMA(KP866025403, TM, TL); cannam@167: { cannam@167: E Ty, TJ, TO, TP; cannam@167: Ty = FNMS(KP176326980, Tx, Ts); cannam@167: TJ = FNMS(KP839099631, TI, TF); cannam@167: TK = FNMS(KP777861913, TJ, Ty); cannam@167: TS = FMA(KP777861913, TJ, Ty); cannam@167: TO = FMA(KP176326980, Ts, Tx); cannam@167: TP = FMA(KP839099631, TF, TI); cannam@167: TQ = FMA(KP777861913, TP, TO); cannam@167: TU = FNMS(KP777861913, TP, TO); cannam@167: } cannam@167: io[WS(os, 1)] = FNMS(KP984807753, TK, Tn); cannam@167: ro[WS(os, 1)] = FMA(KP984807753, TQ, TN); cannam@167: TR = FNMS(KP492403876, TQ, TN); cannam@167: ro[WS(os, 4)] = FMA(KP852868531, TS, TR); cannam@167: ro[WS(os, 7)] = FNMS(KP852868531, TS, TR); cannam@167: TT = FMA(KP492403876, TK, Tn); cannam@167: io[WS(os, 7)] = FNMS(KP852868531, TU, TT); cannam@167: io[WS(os, 4)] = FMA(KP852868531, TU, TT); cannam@167: } cannam@167: { cannam@167: E TV, T17, T12, T1a, T16, T18, T13, T19; cannam@167: TV = FNMS(KP866025403, TM, TL); cannam@167: T17 = FNMS(KP866025403, Tm, Tl); cannam@167: { cannam@167: E TY, T11, T14, T15; cannam@167: TY = FMA(KP176326980, TX, TW); cannam@167: T11 = FNMS(KP363970234, T10, TZ); cannam@167: T12 = FNMS(KP954188894, T11, TY); cannam@167: T1a = FMA(KP954188894, T11, TY); cannam@167: T14 = FNMS(KP176326980, TW, TX); cannam@167: T15 = FMA(KP363970234, TZ, T10); cannam@167: T16 = FNMS(KP954188894, T15, T14); cannam@167: T18 = FMA(KP954188894, T15, T14); cannam@167: } cannam@167: ro[WS(os, 2)] = FMA(KP984807753, T12, TV); cannam@167: io[WS(os, 2)] = FNMS(KP984807753, T18, T17); cannam@167: T13 = FNMS(KP492403876, T12, TV); cannam@167: ro[WS(os, 5)] = FNMS(KP852868531, T16, T13); cannam@167: ro[WS(os, 8)] = FMA(KP852868531, T16, T13); cannam@167: T19 = FMA(KP492403876, T18, T17); cannam@167: io[WS(os, 5)] = FNMS(KP852868531, T1a, T19); cannam@167: io[WS(os, 8)] = FMA(KP852868531, T1a, T19); cannam@167: } cannam@167: } cannam@167: } cannam@167: } cannam@167: cannam@167: static const kdft_desc desc = { 9, "n1_9", {24, 0, 56, 0}, &GENUS, 0, 0, 0, 0 }; cannam@167: cannam@167: void X(codelet_n1_9) (planner *p) { cannam@167: X(kdft_register) (p, n1_9, &desc); cannam@167: } cannam@167: cannam@167: #else cannam@167: cannam@167: /* Generated by: ../../../genfft/gen_notw.native -compact -variables 4 -pipeline-latency 4 -n 9 -name n1_9 -include dft/scalar/n.h */ cannam@167: cannam@167: /* cannam@167: * This function contains 80 FP additions, 40 FP multiplications, cannam@167: * (or, 60 additions, 20 multiplications, 20 fused multiply/add), cannam@167: * 39 stack variables, 8 constants, and 36 memory accesses cannam@167: */ cannam@167: #include "dft/scalar/n.h" cannam@167: cannam@167: static void n1_9(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) cannam@167: { cannam@167: DK(KP939692620, +0.939692620785908384054109277324731469936208134); cannam@167: DK(KP342020143, +0.342020143325668733044099614682259580763083368); cannam@167: DK(KP984807753, +0.984807753012208059366743024589523013670643252); cannam@167: DK(KP173648177, +0.173648177666930348851716626769314796000375677); cannam@167: DK(KP642787609, +0.642787609686539326322643409907263432907559884); cannam@167: DK(KP766044443, +0.766044443118978035202392650555416673935832457); cannam@167: DK(KP500000000, +0.500000000000000000000000000000000000000000000); cannam@167: DK(KP866025403, +0.866025403784438646763723170752936183471402627); cannam@167: { cannam@167: INT i; cannam@167: for (i = v; i > 0; i = i - 1, ri = ri + ivs, ii = ii + ivs, ro = ro + ovs, io = io + ovs, MAKE_VOLATILE_STRIDE(36, is), MAKE_VOLATILE_STRIDE(36, os)) { cannam@167: E T5, TO, Th, Tk, T1g, TR, Ta, T1c, Tq, TW, Tv, TX, Tf, T1d, TB; cannam@167: E T10, TG, TZ; cannam@167: { cannam@167: E T1, T2, T3, T4; cannam@167: T1 = ri[0]; cannam@167: T2 = ri[WS(is, 3)]; cannam@167: T3 = ri[WS(is, 6)]; cannam@167: T4 = T2 + T3; cannam@167: T5 = T1 + T4; cannam@167: TO = KP866025403 * (T3 - T2); cannam@167: Th = FNMS(KP500000000, T4, T1); cannam@167: } cannam@167: { cannam@167: E TP, Ti, Tj, TQ; cannam@167: TP = ii[0]; cannam@167: Ti = ii[WS(is, 3)]; cannam@167: Tj = ii[WS(is, 6)]; cannam@167: TQ = Ti + Tj; cannam@167: Tk = KP866025403 * (Ti - Tj); cannam@167: T1g = TP + TQ; cannam@167: TR = FNMS(KP500000000, TQ, TP); cannam@167: } cannam@167: { cannam@167: E T6, Ts, T9, Tr, Tp, Tt, Tm, Tu; cannam@167: T6 = ri[WS(is, 1)]; cannam@167: Ts = ii[WS(is, 1)]; cannam@167: { cannam@167: E T7, T8, Tn, To; cannam@167: T7 = ri[WS(is, 4)]; cannam@167: T8 = ri[WS(is, 7)]; cannam@167: T9 = T7 + T8; cannam@167: Tr = KP866025403 * (T8 - T7); cannam@167: Tn = ii[WS(is, 4)]; cannam@167: To = ii[WS(is, 7)]; cannam@167: Tp = KP866025403 * (Tn - To); cannam@167: Tt = Tn + To; cannam@167: } cannam@167: Ta = T6 + T9; cannam@167: T1c = Ts + Tt; cannam@167: Tm = FNMS(KP500000000, T9, T6); cannam@167: Tq = Tm + Tp; cannam@167: TW = Tm - Tp; cannam@167: Tu = FNMS(KP500000000, Tt, Ts); cannam@167: Tv = Tr + Tu; cannam@167: TX = Tu - Tr; cannam@167: } cannam@167: { cannam@167: E Tb, TD, Te, TC, TA, TE, Tx, TF; cannam@167: Tb = ri[WS(is, 2)]; cannam@167: TD = ii[WS(is, 2)]; cannam@167: { cannam@167: E Tc, Td, Ty, Tz; cannam@167: Tc = ri[WS(is, 5)]; cannam@167: Td = ri[WS(is, 8)]; cannam@167: Te = Tc + Td; cannam@167: TC = KP866025403 * (Td - Tc); cannam@167: Ty = ii[WS(is, 5)]; cannam@167: Tz = ii[WS(is, 8)]; cannam@167: TA = KP866025403 * (Ty - Tz); cannam@167: TE = Ty + Tz; cannam@167: } cannam@167: Tf = Tb + Te; cannam@167: T1d = TD + TE; cannam@167: Tx = FNMS(KP500000000, Te, Tb); cannam@167: TB = Tx + TA; cannam@167: T10 = Tx - TA; cannam@167: TF = FNMS(KP500000000, TE, TD); cannam@167: TG = TC + TF; cannam@167: TZ = TF - TC; cannam@167: } cannam@167: { cannam@167: E T1e, Tg, T1b, T1f, T1h, T1i; cannam@167: T1e = KP866025403 * (T1c - T1d); cannam@167: Tg = Ta + Tf; cannam@167: T1b = FNMS(KP500000000, Tg, T5); cannam@167: ro[0] = T5 + Tg; cannam@167: ro[WS(os, 3)] = T1b + T1e; cannam@167: ro[WS(os, 6)] = T1b - T1e; cannam@167: T1f = KP866025403 * (Tf - Ta); cannam@167: T1h = T1c + T1d; cannam@167: T1i = FNMS(KP500000000, T1h, T1g); cannam@167: io[WS(os, 3)] = T1f + T1i; cannam@167: io[0] = T1g + T1h; cannam@167: io[WS(os, 6)] = T1i - T1f; cannam@167: } cannam@167: { cannam@167: E Tl, TS, TI, TN, TM, TT, TJ, TU; cannam@167: Tl = Th + Tk; cannam@167: TS = TO + TR; cannam@167: { cannam@167: E Tw, TH, TK, TL; cannam@167: Tw = FMA(KP766044443, Tq, KP642787609 * Tv); cannam@167: TH = FMA(KP173648177, TB, KP984807753 * TG); cannam@167: TI = Tw + TH; cannam@167: TN = KP866025403 * (TH - Tw); cannam@167: TK = FNMS(KP642787609, Tq, KP766044443 * Tv); cannam@167: TL = FNMS(KP984807753, TB, KP173648177 * TG); cannam@167: TM = KP866025403 * (TK - TL); cannam@167: TT = TK + TL; cannam@167: } cannam@167: ro[WS(os, 1)] = Tl + TI; cannam@167: io[WS(os, 1)] = TS + TT; cannam@167: TJ = FNMS(KP500000000, TI, Tl); cannam@167: ro[WS(os, 7)] = TJ - TM; cannam@167: ro[WS(os, 4)] = TJ + TM; cannam@167: TU = FNMS(KP500000000, TT, TS); cannam@167: io[WS(os, 4)] = TN + TU; cannam@167: io[WS(os, 7)] = TU - TN; cannam@167: } cannam@167: { cannam@167: E TV, T14, T12, T13, T17, T1a, T18, T19; cannam@167: TV = Th - Tk; cannam@167: T14 = TR - TO; cannam@167: { cannam@167: E TY, T11, T15, T16; cannam@167: TY = FMA(KP173648177, TW, KP984807753 * TX); cannam@167: T11 = FNMS(KP939692620, T10, KP342020143 * TZ); cannam@167: T12 = TY + T11; cannam@167: T13 = KP866025403 * (T11 - TY); cannam@167: T15 = FNMS(KP984807753, TW, KP173648177 * TX); cannam@167: T16 = FMA(KP342020143, T10, KP939692620 * TZ); cannam@167: T17 = T15 - T16; cannam@167: T1a = KP866025403 * (T15 + T16); cannam@167: } cannam@167: ro[WS(os, 2)] = TV + T12; cannam@167: io[WS(os, 2)] = T14 + T17; cannam@167: T18 = FNMS(KP500000000, T17, T14); cannam@167: io[WS(os, 5)] = T13 + T18; cannam@167: io[WS(os, 8)] = T18 - T13; cannam@167: T19 = FNMS(KP500000000, T12, TV); cannam@167: ro[WS(os, 8)] = T19 - T1a; cannam@167: ro[WS(os, 5)] = T19 + T1a; cannam@167: } cannam@167: } cannam@167: } cannam@167: } cannam@167: cannam@167: static const kdft_desc desc = { 9, "n1_9", {60, 20, 20, 0}, &GENUS, 0, 0, 0, 0 }; cannam@167: cannam@167: void X(codelet_n1_9) (planner *p) { cannam@167: X(kdft_register) (p, n1_9, &desc); cannam@167: } cannam@167: cannam@167: #endif