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 07:35:43 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 14 -name n1_14 -include n.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 148 FP additions, 84 FP multiplications, Chris@10: * (or, 64 additions, 0 multiplications, 84 fused multiply/add), Chris@10: * 80 stack variables, 6 constants, and 56 memory accesses Chris@10: */ Chris@10: #include "n.h" Chris@10: Chris@10: static void n1_14(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(KP974927912, +0.974927912181823607018131682993931217232785801); Chris@10: DK(KP801937735, +0.801937735804838252472204639014890102331838324); Chris@10: DK(KP900968867, +0.900968867902419126236102319507445051165919162); Chris@10: DK(KP554958132, +0.554958132087371191422194871006410481067288862); Chris@10: DK(KP692021471, +0.692021471630095869627814897002069140197260599); Chris@10: DK(KP356895867, +0.356895867892209443894399510021300583399127187); 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(56, is), MAKE_VOLATILE_STRIDE(56, os)) { Chris@10: E Tp, T1L, T24, T1W, T1X, T28, T2a, T1Y, T29, T2b; Chris@10: { Chris@10: E T3, T1x, T1b, To, T1i, T1M, Ts, Ta, T1k, Tv, Th, T1j, T1K, Ty, TZ; Chris@10: E T14, Tz, T1Z, T27, T2c, T1d, TI, T23, T1G, T1D, TW, T1e, T22, T1A, TP; Chris@10: E T1c, T1n, T1s, T1f, T1P; Chris@10: { Chris@10: E T1, T2, T19, T1a; Chris@10: T1 = ri[0]; Chris@10: T2 = ri[WS(is, 7)]; Chris@10: T19 = ii[0]; Chris@10: T1a = ii[WS(is, 7)]; Chris@10: { Chris@10: E Tq, T6, Tr, T9, Te, Tx, Tn, Tw, Tk, Tf, Tb, Tc; Chris@10: { Chris@10: E Tl, Tm, Ti, Tj; Chris@10: { Chris@10: E T4, T5, T7, T8; Chris@10: T4 = ri[WS(is, 2)]; Chris@10: Tp = T1 + T2; Chris@10: T3 = T1 - T2; Chris@10: T1x = T19 + T1a; Chris@10: T1b = T19 - T1a; Chris@10: T5 = ri[WS(is, 9)]; Chris@10: T7 = ri[WS(is, 12)]; Chris@10: T8 = ri[WS(is, 5)]; Chris@10: Tl = ri[WS(is, 8)]; Chris@10: Tq = T4 + T5; Chris@10: T6 = T4 - T5; Chris@10: Tr = T7 + T8; Chris@10: T9 = T7 - T8; Chris@10: Tm = ri[WS(is, 1)]; Chris@10: } Chris@10: Ti = ri[WS(is, 6)]; Chris@10: Tj = ri[WS(is, 13)]; Chris@10: Te = ri[WS(is, 10)]; Chris@10: Tx = Tl + Tm; Chris@10: Tn = Tl - Tm; Chris@10: Tw = Ti + Tj; Chris@10: Tk = Ti - Tj; Chris@10: Tf = ri[WS(is, 3)]; Chris@10: Tb = ri[WS(is, 4)]; Chris@10: Tc = ri[WS(is, 11)]; Chris@10: } Chris@10: { Chris@10: E Tu, Tg, Tt, Td; Chris@10: To = Tk + Tn; Chris@10: T1i = Tn - Tk; Chris@10: Tu = Te + Tf; Chris@10: Tg = Te - Tf; Chris@10: Tt = Tb + Tc; Chris@10: Td = Tb - Tc; Chris@10: T1M = Tr - Tq; Chris@10: Ts = Tq + Tr; Chris@10: Ta = T6 + T9; Chris@10: T1k = T9 - T6; Chris@10: T1L = Tt - Tu; Chris@10: Tv = Tt + Tu; Chris@10: Th = Td + Tg; Chris@10: T1j = Tg - Td; Chris@10: T1K = Tw - Tx; Chris@10: Ty = Tw + Tx; Chris@10: TZ = FNMS(KP356895867, Ta, To); Chris@10: T14 = FNMS(KP356895867, To, Th); Chris@10: Tz = FNMS(KP356895867, Th, Ta); Chris@10: T1Z = FNMS(KP356895867, Ts, Ty); Chris@10: } Chris@10: } Chris@10: { Chris@10: E T1B, TE, T1C, TH, T1F, TV, TJ, T1E, TS, T1z, TO, TK, T1y, TL; Chris@10: { Chris@10: E TF, TG, TT, TU, TC, TD; Chris@10: TC = ii[WS(is, 4)]; Chris@10: TD = ii[WS(is, 11)]; Chris@10: T27 = FNMS(KP356895867, Tv, Ts); Chris@10: T2c = FNMS(KP356895867, Ty, Tv); Chris@10: TF = ii[WS(is, 10)]; Chris@10: T1B = TC + TD; Chris@10: TE = TC - TD; Chris@10: TG = ii[WS(is, 3)]; Chris@10: TT = ii[WS(is, 8)]; Chris@10: TU = ii[WS(is, 1)]; Chris@10: { Chris@10: E TQ, TR, TM, TN; Chris@10: TQ = ii[WS(is, 6)]; Chris@10: T1C = TF + TG; Chris@10: TH = TF - TG; Chris@10: T1F = TT + TU; Chris@10: TV = TT - TU; Chris@10: TR = ii[WS(is, 13)]; Chris@10: TM = ii[WS(is, 12)]; Chris@10: TN = ii[WS(is, 5)]; Chris@10: TJ = ii[WS(is, 2)]; Chris@10: T1E = TQ + TR; Chris@10: TS = TQ - TR; Chris@10: T1z = TM + TN; Chris@10: TO = TM - TN; Chris@10: TK = ii[WS(is, 9)]; Chris@10: } Chris@10: } Chris@10: T1d = TE + TH; Chris@10: TI = TE - TH; Chris@10: T23 = T1F - T1E; Chris@10: T1G = T1E + T1F; Chris@10: T1D = T1B + T1C; Chris@10: T24 = T1C - T1B; Chris@10: T1y = TJ + TK; Chris@10: TL = TJ - TK; Chris@10: TW = TS - TV; Chris@10: T1e = TS + TV; Chris@10: T22 = T1y - T1z; Chris@10: T1A = T1y + T1z; Chris@10: TP = TL - TO; Chris@10: T1c = TL + TO; Chris@10: T1n = FNMS(KP356895867, T1c, T1e); Chris@10: T1s = FNMS(KP356895867, T1d, T1c); Chris@10: T1f = FNMS(KP356895867, T1e, T1d); Chris@10: T1P = FNMS(KP356895867, T1A, T1G); Chris@10: } Chris@10: } Chris@10: { Chris@10: E T1U, T1H, T11, T12, T1o, T1q; Chris@10: ro[WS(os, 7)] = T3 + Ta + Th + To; Chris@10: io[WS(os, 7)] = T1b + T1c + T1d + T1e; Chris@10: T1U = FNMS(KP356895867, T1D, T1A); Chris@10: T1H = FNMS(KP356895867, T1G, T1D); Chris@10: ro[0] = Tp + Ts + Tv + Ty; Chris@10: io[0] = T1x + T1A + T1D + T1G; Chris@10: { Chris@10: E TB, TY, T1u, T1w, T10; Chris@10: { Chris@10: E TA, TX, T1t, T1v; Chris@10: TA = FNMS(KP692021471, Tz, To); Chris@10: TX = FMA(KP554958132, TW, TP); Chris@10: T1t = FNMS(KP692021471, T1s, T1e); Chris@10: T1v = FMA(KP554958132, T1i, T1k); Chris@10: TB = FNMS(KP900968867, TA, T3); Chris@10: TY = FMA(KP801937735, TX, TI); Chris@10: T1u = FNMS(KP900968867, T1t, T1b); Chris@10: T1w = FMA(KP801937735, T1v, T1j); Chris@10: } Chris@10: T10 = FNMS(KP692021471, TZ, Th); Chris@10: ro[WS(os, 1)] = FMA(KP974927912, TY, TB); Chris@10: ro[WS(os, 13)] = FNMS(KP974927912, TY, TB); Chris@10: io[WS(os, 13)] = FNMS(KP974927912, T1w, T1u); Chris@10: io[WS(os, 1)] = FMA(KP974927912, T1w, T1u); Chris@10: T11 = FNMS(KP900968867, T10, T3); Chris@10: T12 = FMA(KP554958132, TI, TW); Chris@10: T1o = FNMS(KP692021471, T1n, T1d); Chris@10: T1q = FMA(KP554958132, T1j, T1i); Chris@10: } Chris@10: { Chris@10: E T1J, T1N, T2d, T2f; Chris@10: { Chris@10: E T16, T17, T1g, T1l; Chris@10: { Chris@10: E T13, T1p, T1r, T15; Chris@10: T15 = FNMS(KP692021471, T14, Ta); Chris@10: T13 = FNMS(KP801937735, T12, TP); Chris@10: T1p = FNMS(KP900968867, T1o, T1b); Chris@10: T1r = FNMS(KP801937735, T1q, T1k); Chris@10: T16 = FNMS(KP900968867, T15, T3); Chris@10: ro[WS(os, 9)] = FMA(KP974927912, T13, T11); Chris@10: ro[WS(os, 5)] = FNMS(KP974927912, T13, T11); Chris@10: io[WS(os, 9)] = FMA(KP974927912, T1r, T1p); Chris@10: io[WS(os, 5)] = FNMS(KP974927912, T1r, T1p); Chris@10: T17 = FNMS(KP554958132, TP, TI); Chris@10: } Chris@10: T1g = FNMS(KP692021471, T1f, T1c); Chris@10: T1l = FNMS(KP554958132, T1k, T1j); Chris@10: { Chris@10: E T18, T1h, T1m, T1I; Chris@10: T1I = FNMS(KP692021471, T1H, T1A); Chris@10: T18 = FNMS(KP801937735, T17, TW); Chris@10: T1h = FNMS(KP900968867, T1g, T1b); Chris@10: T1m = FNMS(KP801937735, T1l, T1i); Chris@10: T1J = FNMS(KP900968867, T1I, T1x); Chris@10: ro[WS(os, 3)] = FMA(KP974927912, T18, T16); Chris@10: ro[WS(os, 11)] = FNMS(KP974927912, T18, T16); Chris@10: io[WS(os, 11)] = FNMS(KP974927912, T1m, T1h); Chris@10: io[WS(os, 3)] = FMA(KP974927912, T1m, T1h); Chris@10: T1N = FMA(KP554958132, T1M, T1L); Chris@10: } Chris@10: T2d = FNMS(KP692021471, T2c, Ts); Chris@10: T2f = FMA(KP554958132, T22, T24); Chris@10: } Chris@10: { Chris@10: E T1R, T1S, T20, T25; Chris@10: { Chris@10: E T1O, T2e, T2g, T1Q; Chris@10: T1Q = FNMS(KP692021471, T1P, T1D); Chris@10: T1O = FNMS(KP801937735, T1N, T1K); Chris@10: T2e = FNMS(KP900968867, T2d, Tp); Chris@10: T2g = FNMS(KP801937735, T2f, T23); Chris@10: T1R = FNMS(KP900968867, T1Q, T1x); Chris@10: io[WS(os, 10)] = FNMS(KP974927912, T1O, T1J); Chris@10: io[WS(os, 4)] = FMA(KP974927912, T1O, T1J); Chris@10: ro[WS(os, 4)] = FMA(KP974927912, T2g, T2e); Chris@10: ro[WS(os, 10)] = FNMS(KP974927912, T2g, T2e); Chris@10: T1S = FMA(KP554958132, T1L, T1K); Chris@10: } Chris@10: T20 = FNMS(KP692021471, T1Z, Tv); Chris@10: T25 = FMA(KP554958132, T24, T23); Chris@10: { Chris@10: E T1T, T21, T26, T1V; Chris@10: T1V = FNMS(KP692021471, T1U, T1G); Chris@10: T1T = FMA(KP801937735, T1S, T1M); Chris@10: T21 = FNMS(KP900968867, T20, Tp); Chris@10: T26 = FMA(KP801937735, T25, T22); Chris@10: T1W = FNMS(KP900968867, T1V, T1x); Chris@10: io[WS(os, 12)] = FNMS(KP974927912, T1T, T1R); Chris@10: io[WS(os, 2)] = FMA(KP974927912, T1T, T1R); Chris@10: ro[WS(os, 2)] = FMA(KP974927912, T26, T21); Chris@10: ro[WS(os, 12)] = FNMS(KP974927912, T26, T21); Chris@10: T1X = FNMS(KP554958132, T1K, T1M); Chris@10: } Chris@10: T28 = FNMS(KP692021471, T27, Ty); Chris@10: T2a = FNMS(KP554958132, T23, T22); Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: T1Y = FNMS(KP801937735, T1X, T1L); Chris@10: T29 = FNMS(KP900968867, T28, Tp); Chris@10: T2b = FNMS(KP801937735, T2a, T24); Chris@10: io[WS(os, 8)] = FNMS(KP974927912, T1Y, T1W); Chris@10: io[WS(os, 6)] = FMA(KP974927912, T1Y, T1W); Chris@10: ro[WS(os, 6)] = FMA(KP974927912, T2b, T29); Chris@10: ro[WS(os, 8)] = FNMS(KP974927912, T2b, T29); Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kdft_desc desc = { 14, "n1_14", {64, 0, 84, 0}, &GENUS, 0, 0, 0, 0 }; Chris@10: Chris@10: void X(codelet_n1_14) (planner *p) { Chris@10: X(kdft_register) (p, n1_14, &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 14 -name n1_14 -include n.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 148 FP additions, 72 FP multiplications, Chris@10: * (or, 100 additions, 24 multiplications, 48 fused multiply/add), Chris@10: * 43 stack variables, 6 constants, and 56 memory accesses Chris@10: */ Chris@10: #include "n.h" Chris@10: Chris@10: static void n1_14(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(KP222520933, +0.222520933956314404288902564496794759466355569); Chris@10: DK(KP900968867, +0.900968867902419126236102319507445051165919162); Chris@10: DK(KP623489801, +0.623489801858733530525004884004239810632274731); Chris@10: DK(KP433883739, +0.433883739117558120475768332848358754609990728); Chris@10: DK(KP781831482, +0.781831482468029808708444526674057750232334519); Chris@10: DK(KP974927912, +0.974927912181823607018131682993931217232785801); 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(56, is), MAKE_VOLATILE_STRIDE(56, os)) { Chris@10: E T3, Tp, T16, T1f, Ta, T1q, Ts, T10, TG, T1z, T19, T1i, Th, T1s, Tv; Chris@10: E T12, TU, T1B, T17, T1o, To, T1r, Ty, T11, TN, T1A, T18, T1l; Chris@10: { Chris@10: E T1, T2, T14, T15; Chris@10: T1 = ri[0]; Chris@10: T2 = ri[WS(is, 7)]; Chris@10: T3 = T1 - T2; Chris@10: Tp = T1 + T2; Chris@10: T14 = ii[0]; Chris@10: T15 = ii[WS(is, 7)]; Chris@10: T16 = T14 - T15; Chris@10: T1f = T14 + T15; Chris@10: } Chris@10: { Chris@10: E T6, Tq, T9, Tr; Chris@10: { Chris@10: E T4, T5, T7, T8; Chris@10: T4 = ri[WS(is, 2)]; Chris@10: T5 = ri[WS(is, 9)]; Chris@10: T6 = T4 - T5; Chris@10: Tq = T4 + T5; Chris@10: T7 = ri[WS(is, 12)]; Chris@10: T8 = ri[WS(is, 5)]; Chris@10: T9 = T7 - T8; Chris@10: Tr = T7 + T8; Chris@10: } Chris@10: Ta = T6 + T9; Chris@10: T1q = Tr - Tq; Chris@10: Ts = Tq + Tr; Chris@10: T10 = T9 - T6; Chris@10: } Chris@10: { Chris@10: E TC, T1g, TF, T1h; Chris@10: { Chris@10: E TA, TB, TD, TE; Chris@10: TA = ii[WS(is, 2)]; Chris@10: TB = ii[WS(is, 9)]; Chris@10: TC = TA - TB; Chris@10: T1g = TA + TB; Chris@10: TD = ii[WS(is, 12)]; Chris@10: TE = ii[WS(is, 5)]; Chris@10: TF = TD - TE; Chris@10: T1h = TD + TE; Chris@10: } Chris@10: TG = TC - TF; Chris@10: T1z = T1g - T1h; Chris@10: T19 = TC + TF; Chris@10: T1i = T1g + T1h; Chris@10: } Chris@10: { Chris@10: E Td, Tt, Tg, Tu; Chris@10: { Chris@10: E Tb, Tc, Te, Tf; Chris@10: Tb = ri[WS(is, 4)]; Chris@10: Tc = ri[WS(is, 11)]; Chris@10: Td = Tb - Tc; Chris@10: Tt = Tb + Tc; Chris@10: Te = ri[WS(is, 10)]; Chris@10: Tf = ri[WS(is, 3)]; Chris@10: Tg = Te - Tf; Chris@10: Tu = Te + Tf; Chris@10: } Chris@10: Th = Td + Tg; Chris@10: T1s = Tt - Tu; Chris@10: Tv = Tt + Tu; Chris@10: T12 = Tg - Td; Chris@10: } Chris@10: { Chris@10: E TQ, T1m, TT, T1n; Chris@10: { Chris@10: E TO, TP, TR, TS; Chris@10: TO = ii[WS(is, 4)]; Chris@10: TP = ii[WS(is, 11)]; Chris@10: TQ = TO - TP; Chris@10: T1m = TO + TP; Chris@10: TR = ii[WS(is, 10)]; Chris@10: TS = ii[WS(is, 3)]; Chris@10: TT = TR - TS; Chris@10: T1n = TR + TS; Chris@10: } Chris@10: TU = TQ - TT; Chris@10: T1B = T1n - T1m; Chris@10: T17 = TQ + TT; Chris@10: T1o = T1m + T1n; Chris@10: } Chris@10: { Chris@10: E Tk, Tw, Tn, Tx; Chris@10: { Chris@10: E Ti, Tj, Tl, Tm; Chris@10: Ti = ri[WS(is, 6)]; Chris@10: Tj = ri[WS(is, 13)]; Chris@10: Tk = Ti - Tj; Chris@10: Tw = Ti + Tj; Chris@10: Tl = ri[WS(is, 8)]; Chris@10: Tm = ri[WS(is, 1)]; Chris@10: Tn = Tl - Tm; Chris@10: Tx = Tl + Tm; Chris@10: } Chris@10: To = Tk + Tn; Chris@10: T1r = Tw - Tx; Chris@10: Ty = Tw + Tx; Chris@10: T11 = Tn - Tk; Chris@10: } Chris@10: { Chris@10: E TJ, T1j, TM, T1k; Chris@10: { Chris@10: E TH, TI, TK, TL; Chris@10: TH = ii[WS(is, 6)]; Chris@10: TI = ii[WS(is, 13)]; Chris@10: TJ = TH - TI; Chris@10: T1j = TH + TI; Chris@10: TK = ii[WS(is, 8)]; Chris@10: TL = ii[WS(is, 1)]; Chris@10: TM = TK - TL; Chris@10: T1k = TK + TL; Chris@10: } Chris@10: TN = TJ - TM; Chris@10: T1A = T1k - T1j; Chris@10: T18 = TJ + TM; Chris@10: T1l = T1j + T1k; Chris@10: } Chris@10: ro[WS(os, 7)] = T3 + Ta + Th + To; Chris@10: io[WS(os, 7)] = T16 + T19 + T17 + T18; Chris@10: ro[0] = Tp + Ts + Tv + Ty; Chris@10: io[0] = T1f + T1i + T1o + T1l; Chris@10: { Chris@10: E TV, Tz, T1e, T1d; Chris@10: TV = FNMS(KP781831482, TN, KP974927912 * TG) - (KP433883739 * TU); Chris@10: Tz = FMA(KP623489801, To, T3) + FNMA(KP900968867, Th, KP222520933 * Ta); Chris@10: ro[WS(os, 5)] = Tz - TV; Chris@10: ro[WS(os, 9)] = Tz + TV; Chris@10: T1e = FNMS(KP781831482, T11, KP974927912 * T10) - (KP433883739 * T12); Chris@10: T1d = FMA(KP623489801, T18, T16) + FNMA(KP900968867, T17, KP222520933 * T19); Chris@10: io[WS(os, 5)] = T1d - T1e; Chris@10: io[WS(os, 9)] = T1e + T1d; Chris@10: } Chris@10: { Chris@10: E TX, TW, T1b, T1c; Chris@10: TX = FMA(KP781831482, TG, KP974927912 * TU) + (KP433883739 * TN); Chris@10: TW = FMA(KP623489801, Ta, T3) + FNMA(KP900968867, To, KP222520933 * Th); Chris@10: ro[WS(os, 13)] = TW - TX; Chris@10: ro[WS(os, 1)] = TW + TX; Chris@10: T1b = FMA(KP781831482, T10, KP974927912 * T12) + (KP433883739 * T11); Chris@10: T1c = FMA(KP623489801, T19, T16) + FNMA(KP900968867, T18, KP222520933 * T17); Chris@10: io[WS(os, 1)] = T1b + T1c; Chris@10: io[WS(os, 13)] = T1c - T1b; Chris@10: } Chris@10: { Chris@10: E TZ, TY, T13, T1a; Chris@10: TZ = FMA(KP433883739, TG, KP974927912 * TN) - (KP781831482 * TU); Chris@10: TY = FMA(KP623489801, Th, T3) + FNMA(KP222520933, To, KP900968867 * Ta); Chris@10: ro[WS(os, 11)] = TY - TZ; Chris@10: ro[WS(os, 3)] = TY + TZ; Chris@10: T13 = FMA(KP433883739, T10, KP974927912 * T11) - (KP781831482 * T12); Chris@10: T1a = FMA(KP623489801, T17, T16) + FNMA(KP222520933, T18, KP900968867 * T19); Chris@10: io[WS(os, 3)] = T13 + T1a; Chris@10: io[WS(os, 11)] = T1a - T13; Chris@10: } Chris@10: { Chris@10: E T1t, T1p, T1C, T1y; Chris@10: T1t = FNMS(KP433883739, T1r, KP781831482 * T1q) - (KP974927912 * T1s); Chris@10: T1p = FMA(KP623489801, T1i, T1f) + FNMA(KP900968867, T1l, KP222520933 * T1o); Chris@10: io[WS(os, 6)] = T1p - T1t; Chris@10: io[WS(os, 8)] = T1t + T1p; Chris@10: T1C = FNMS(KP433883739, T1A, KP781831482 * T1z) - (KP974927912 * T1B); Chris@10: T1y = FMA(KP623489801, Ts, Tp) + FNMA(KP900968867, Ty, KP222520933 * Tv); Chris@10: ro[WS(os, 6)] = T1y - T1C; Chris@10: ro[WS(os, 8)] = T1y + T1C; Chris@10: } Chris@10: { Chris@10: E T1v, T1u, T1E, T1D; Chris@10: T1v = FMA(KP433883739, T1q, KP781831482 * T1s) - (KP974927912 * T1r); Chris@10: T1u = FMA(KP623489801, T1o, T1f) + FNMA(KP222520933, T1l, KP900968867 * T1i); Chris@10: io[WS(os, 4)] = T1u - T1v; Chris@10: io[WS(os, 10)] = T1v + T1u; Chris@10: T1E = FMA(KP433883739, T1z, KP781831482 * T1B) - (KP974927912 * T1A); Chris@10: T1D = FMA(KP623489801, Tv, Tp) + FNMA(KP222520933, Ty, KP900968867 * Ts); Chris@10: ro[WS(os, 4)] = T1D - T1E; Chris@10: ro[WS(os, 10)] = T1D + T1E; Chris@10: } Chris@10: { Chris@10: E T1w, T1x, T1G, T1F; Chris@10: T1w = FMA(KP974927912, T1q, KP433883739 * T1s) + (KP781831482 * T1r); Chris@10: T1x = FMA(KP623489801, T1l, T1f) + FNMA(KP900968867, T1o, KP222520933 * T1i); Chris@10: io[WS(os, 2)] = T1w + T1x; Chris@10: io[WS(os, 12)] = T1x - T1w; Chris@10: T1G = FMA(KP974927912, T1z, KP433883739 * T1B) + (KP781831482 * T1A); Chris@10: T1F = FMA(KP623489801, Ty, Tp) + FNMA(KP900968867, Tv, KP222520933 * Ts); Chris@10: ro[WS(os, 12)] = T1F - T1G; Chris@10: ro[WS(os, 2)] = T1F + T1G; Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kdft_desc desc = { 14, "n1_14", {100, 24, 48, 0}, &GENUS, 0, 0, 0, 0 }; Chris@10: Chris@10: void X(codelet_n1_14) (planner *p) { Chris@10: X(kdft_register) (p, n1_14, &desc); Chris@10: } Chris@10: Chris@10: #endif /* HAVE_FMA */