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 15 -name n1_15 -include n.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 156 FP additions, 84 FP multiplications, Chris@10: * (or, 72 additions, 0 multiplications, 84 fused multiply/add), Chris@10: * 75 stack variables, 6 constants, and 60 memory accesses Chris@10: */ Chris@10: #include "n.h" Chris@10: Chris@10: static void n1_15(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(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@10: DK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@10: DK(KP618033988, +0.618033988749894848204586834365638117720309180); Chris@10: DK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@10: DK(KP866025403, +0.866025403784438646763723170752936183471402627); Chris@10: DK(KP500000000, +0.500000000000000000000000000000000000000000000); 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(60, is), MAKE_VOLATILE_STRIDE(60, os)) { Chris@10: E T1r, T1g, T14, T13; Chris@10: { Chris@10: E T5, T2l, Tx, TV, T1z, T1X, T2s, Tr, T24, TT, T2e, T2n, T1Z, T1Q, T1B; Chris@10: E T11, T1H, TW, T2t, Tg, TX, T25, TI, T2h, T2m, T1Y, T1T, T1A; Chris@10: { Chris@10: E T1, T1v, T2, T3, Tu, Tv, TZ, T10; Chris@10: T1 = ri[0]; Chris@10: T1v = ii[0]; Chris@10: T2 = ri[WS(is, 5)]; Chris@10: T3 = ri[WS(is, 10)]; Chris@10: Tu = ii[WS(is, 5)]; Chris@10: Tv = ii[WS(is, 10)]; Chris@10: { Chris@10: E T1k, Tm, TM, TJ, Tl, T2c, T1j, T1m, TP, T1p, Tp, TQ; Chris@10: { Chris@10: E Th, T1h, TK, TL, Tk, Tn, To, T1i; Chris@10: { Chris@10: E Ti, Tj, T1y, T4; Chris@10: Th = ri[WS(is, 6)]; Chris@10: T1y = T3 - T2; Chris@10: T4 = T2 + T3; Chris@10: { Chris@10: E T1w, Tw, Tt, T1x; Chris@10: T1w = Tu + Tv; Chris@10: Tw = Tu - Tv; Chris@10: Ti = ri[WS(is, 11)]; Chris@10: T5 = T1 + T4; Chris@10: Tt = FNMS(KP500000000, T4, T1); Chris@10: T2l = T1v + T1w; Chris@10: T1x = FNMS(KP500000000, T1w, T1v); Chris@10: Tx = FNMS(KP866025403, Tw, Tt); Chris@10: TV = FMA(KP866025403, Tw, Tt); Chris@10: T1z = FMA(KP866025403, T1y, T1x); Chris@10: T1X = FNMS(KP866025403, T1y, T1x); Chris@10: Tj = ri[WS(is, 1)]; Chris@10: } Chris@10: T1h = ii[WS(is, 6)]; Chris@10: TK = ii[WS(is, 11)]; Chris@10: TL = ii[WS(is, 1)]; Chris@10: Tk = Ti + Tj; Chris@10: T1k = Tj - Ti; Chris@10: } Chris@10: Tm = ri[WS(is, 9)]; Chris@10: TM = TK - TL; Chris@10: T1i = TK + TL; Chris@10: TJ = FNMS(KP500000000, Tk, Th); Chris@10: Tl = Th + Tk; Chris@10: Tn = ri[WS(is, 14)]; Chris@10: To = ri[WS(is, 4)]; Chris@10: T2c = T1h + T1i; Chris@10: T1j = FNMS(KP500000000, T1i, T1h); Chris@10: T1m = ii[WS(is, 9)]; Chris@10: TP = ii[WS(is, 14)]; Chris@10: T1p = To - Tn; Chris@10: Tp = Tn + To; Chris@10: TQ = ii[WS(is, 4)]; Chris@10: } Chris@10: { Chris@10: E TN, TS, T1o, T2d; Chris@10: { Chris@10: E TO, T1n, TR, Tq; Chris@10: TN = FNMS(KP866025403, TM, TJ); Chris@10: TZ = FMA(KP866025403, TM, TJ); Chris@10: TO = FNMS(KP500000000, Tp, Tm); Chris@10: Tq = Tm + Tp; Chris@10: T1n = TP + TQ; Chris@10: TR = TP - TQ; Chris@10: T2s = Tl - Tq; Chris@10: Tr = Tl + Tq; Chris@10: T10 = FMA(KP866025403, TR, TO); Chris@10: TS = FNMS(KP866025403, TR, TO); Chris@10: T1o = FNMS(KP500000000, T1n, T1m); Chris@10: T2d = T1m + T1n; Chris@10: } Chris@10: { Chris@10: E T1O, T1l, T1P, T1q; Chris@10: T1O = FNMS(KP866025403, T1k, T1j); Chris@10: T1l = FMA(KP866025403, T1k, T1j); Chris@10: T24 = TN - TS; Chris@10: TT = TN + TS; Chris@10: T1P = FNMS(KP866025403, T1p, T1o); Chris@10: T1q = FMA(KP866025403, T1p, T1o); Chris@10: T2e = T2c - T2d; Chris@10: T2n = T2c + T2d; Chris@10: T1Z = T1O + T1P; Chris@10: T1Q = T1O - T1P; Chris@10: T1r = T1l - T1q; Chris@10: T1B = T1l + T1q; Chris@10: } Chris@10: } Chris@10: } Chris@10: { Chris@10: E T19, Tb, TB, Ty, Ta, T2f, T18, T1b, TE, T1e, Te, TF; Chris@10: { Chris@10: E T6, T16, Tz, TA, T9, T7, T8, Tc, Td, T17; Chris@10: T6 = ri[WS(is, 3)]; Chris@10: T7 = ri[WS(is, 8)]; Chris@10: T11 = TZ + T10; Chris@10: T1H = TZ - T10; Chris@10: T8 = ri[WS(is, 13)]; Chris@10: T16 = ii[WS(is, 3)]; Chris@10: Tz = ii[WS(is, 8)]; Chris@10: TA = ii[WS(is, 13)]; Chris@10: T9 = T7 + T8; Chris@10: T19 = T8 - T7; Chris@10: Tb = ri[WS(is, 12)]; Chris@10: TB = Tz - TA; Chris@10: T17 = Tz + TA; Chris@10: Ty = FNMS(KP500000000, T9, T6); Chris@10: Ta = T6 + T9; Chris@10: Tc = ri[WS(is, 2)]; Chris@10: Td = ri[WS(is, 7)]; Chris@10: T2f = T16 + T17; Chris@10: T18 = FNMS(KP500000000, T17, T16); Chris@10: T1b = ii[WS(is, 12)]; Chris@10: TE = ii[WS(is, 2)]; Chris@10: T1e = Td - Tc; Chris@10: Te = Tc + Td; Chris@10: TF = ii[WS(is, 7)]; Chris@10: } Chris@10: { Chris@10: E TC, TH, T1d, T2g; Chris@10: { Chris@10: E TD, T1c, TG, Tf; Chris@10: TC = FNMS(KP866025403, TB, Ty); Chris@10: TW = FMA(KP866025403, TB, Ty); Chris@10: TD = FNMS(KP500000000, Te, Tb); Chris@10: Tf = Tb + Te; Chris@10: T1c = TE + TF; Chris@10: TG = TE - TF; Chris@10: T2t = Ta - Tf; Chris@10: Tg = Ta + Tf; Chris@10: TX = FMA(KP866025403, TG, TD); Chris@10: TH = FNMS(KP866025403, TG, TD); Chris@10: T1d = FNMS(KP500000000, T1c, T1b); Chris@10: T2g = T1b + T1c; Chris@10: } Chris@10: { Chris@10: E T1R, T1a, T1S, T1f; Chris@10: T1R = FNMS(KP866025403, T19, T18); Chris@10: T1a = FMA(KP866025403, T19, T18); Chris@10: T25 = TC - TH; Chris@10: TI = TC + TH; Chris@10: T1S = FNMS(KP866025403, T1e, T1d); Chris@10: T1f = FMA(KP866025403, T1e, T1d); Chris@10: T2h = T2f - T2g; Chris@10: T2m = T2f + T2g; Chris@10: T1Y = T1R + T1S; Chris@10: T1T = T1R - T1S; Chris@10: T1g = T1a - T1f; Chris@10: T1A = T1a + T1f; Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: { Chris@10: E TY, T1G, T1M, T1L, T2a, T29, Ts, T22, T21, T20; Chris@10: T2a = Tg - Tr; Chris@10: Ts = Tg + Tr; Chris@10: TY = TW + TX; Chris@10: T1G = TW - TX; Chris@10: T29 = FNMS(KP250000000, Ts, T5); Chris@10: ro[0] = T5 + Ts; Chris@10: { Chris@10: E T2q, T2p, T2o, TU; Chris@10: T2o = T2m + T2n; Chris@10: T2q = T2m - T2n; Chris@10: { Chris@10: E T2k, T2i, T2b, T2j; Chris@10: T2k = FMA(KP618033988, T2e, T2h); Chris@10: T2i = FNMS(KP618033988, T2h, T2e); Chris@10: T2b = FNMS(KP559016994, T2a, T29); Chris@10: T2j = FMA(KP559016994, T2a, T29); Chris@10: ro[WS(os, 3)] = FMA(KP951056516, T2i, T2b); Chris@10: ro[WS(os, 12)] = FNMS(KP951056516, T2i, T2b); Chris@10: ro[WS(os, 6)] = FMA(KP951056516, T2k, T2j); Chris@10: ro[WS(os, 9)] = FNMS(KP951056516, T2k, T2j); Chris@10: T2p = FNMS(KP250000000, T2o, T2l); Chris@10: } Chris@10: io[0] = T2l + T2o; Chris@10: TU = TI + TT; Chris@10: T1M = TI - TT; Chris@10: { Chris@10: E T2r, T2v, T2w, T2u; Chris@10: T2r = FNMS(KP559016994, T2q, T2p); Chris@10: T2v = FMA(KP559016994, T2q, T2p); Chris@10: T2w = FMA(KP618033988, T2s, T2t); Chris@10: T2u = FNMS(KP618033988, T2t, T2s); Chris@10: io[WS(os, 9)] = FMA(KP951056516, T2w, T2v); Chris@10: io[WS(os, 6)] = FNMS(KP951056516, T2w, T2v); Chris@10: io[WS(os, 12)] = FMA(KP951056516, T2u, T2r); Chris@10: io[WS(os, 3)] = FNMS(KP951056516, T2u, T2r); Chris@10: T1L = FNMS(KP250000000, TU, Tx); Chris@10: } Chris@10: ro[WS(os, 5)] = Tx + TU; Chris@10: } Chris@10: T20 = T1Y + T1Z; Chris@10: T22 = T1Y - T1Z; Chris@10: { Chris@10: E T1N, T1V, T1W, T1U; Chris@10: T1N = FNMS(KP559016994, T1M, T1L); Chris@10: T1V = FMA(KP559016994, T1M, T1L); Chris@10: T1W = FMA(KP618033988, T1Q, T1T); Chris@10: T1U = FNMS(KP618033988, T1T, T1Q); Chris@10: ro[WS(os, 11)] = FMA(KP951056516, T1W, T1V); Chris@10: ro[WS(os, 14)] = FNMS(KP951056516, T1W, T1V); Chris@10: ro[WS(os, 8)] = FMA(KP951056516, T1U, T1N); Chris@10: ro[WS(os, 2)] = FNMS(KP951056516, T1U, T1N); Chris@10: T21 = FNMS(KP250000000, T20, T1X); Chris@10: } Chris@10: io[WS(os, 5)] = T1X + T20; Chris@10: { Chris@10: E T1E, T1D, T1C, T12; Chris@10: T1C = T1A + T1B; Chris@10: T1E = T1A - T1B; Chris@10: { Chris@10: E T23, T27, T28, T26; Chris@10: T23 = FNMS(KP559016994, T22, T21); Chris@10: T27 = FMA(KP559016994, T22, T21); Chris@10: T28 = FMA(KP618033988, T24, T25); Chris@10: T26 = FNMS(KP618033988, T25, T24); Chris@10: io[WS(os, 14)] = FMA(KP951056516, T28, T27); Chris@10: io[WS(os, 11)] = FNMS(KP951056516, T28, T27); Chris@10: io[WS(os, 8)] = FNMS(KP951056516, T26, T23); Chris@10: io[WS(os, 2)] = FMA(KP951056516, T26, T23); Chris@10: T1D = FNMS(KP250000000, T1C, T1z); Chris@10: } Chris@10: io[WS(os, 10)] = T1z + T1C; Chris@10: T12 = TY + T11; Chris@10: T14 = TY - T11; Chris@10: { Chris@10: E T1F, T1J, T1K, T1I; Chris@10: T1F = FMA(KP559016994, T1E, T1D); Chris@10: T1J = FNMS(KP559016994, T1E, T1D); Chris@10: T1K = FNMS(KP618033988, T1G, T1H); Chris@10: T1I = FMA(KP618033988, T1H, T1G); Chris@10: io[WS(os, 13)] = FNMS(KP951056516, T1K, T1J); Chris@10: io[WS(os, 7)] = FMA(KP951056516, T1K, T1J); Chris@10: io[WS(os, 4)] = FMA(KP951056516, T1I, T1F); Chris@10: io[WS(os, 1)] = FNMS(KP951056516, T1I, T1F); Chris@10: T13 = FNMS(KP250000000, T12, TV); Chris@10: } Chris@10: ro[WS(os, 10)] = TV + T12; Chris@10: } Chris@10: } Chris@10: } Chris@10: { Chris@10: E T1t, T15, T1s, T1u; Chris@10: T1t = FNMS(KP559016994, T14, T13); Chris@10: T15 = FMA(KP559016994, T14, T13); Chris@10: T1s = FMA(KP618033988, T1r, T1g); Chris@10: T1u = FNMS(KP618033988, T1g, T1r); Chris@10: ro[WS(os, 13)] = FMA(KP951056516, T1u, T1t); Chris@10: ro[WS(os, 7)] = FNMS(KP951056516, T1u, T1t); Chris@10: ro[WS(os, 1)] = FMA(KP951056516, T1s, T15); Chris@10: ro[WS(os, 4)] = FNMS(KP951056516, T1s, T15); Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kdft_desc desc = { 15, "n1_15", {72, 0, 84, 0}, &GENUS, 0, 0, 0, 0 }; Chris@10: Chris@10: void X(codelet_n1_15) (planner *p) { Chris@10: X(kdft_register) (p, n1_15, &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 15 -name n1_15 -include n.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 156 FP additions, 56 FP multiplications, Chris@10: * (or, 128 additions, 28 multiplications, 28 fused multiply/add), Chris@10: * 69 stack variables, 6 constants, and 60 memory accesses Chris@10: */ Chris@10: #include "n.h" Chris@10: Chris@10: static void n1_15(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(KP587785252, +0.587785252292473129168705954639072768597652438); Chris@10: DK(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@10: DK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@10: DK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@10: DK(KP500000000, +0.500000000000000000000000000000000000000000000); Chris@10: DK(KP866025403, +0.866025403784438646763723170752936183471402627); 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(60, is), MAKE_VOLATILE_STRIDE(60, os)) { Chris@10: E T5, T2l, Tx, TV, T1C, T20, Tl, Tq, Tr, TN, TS, TT, T2c, T2d, T2n; Chris@10: E T1O, T1P, T22, T1l, T1q, T1w, TZ, T10, T11, Ta, Tf, Tg, TC, TH, TI; Chris@10: E T2f, T2g, T2m, T1R, T1S, T21, T1a, T1f, T1v, TW, TX, TY; Chris@10: { Chris@10: E T1, T1z, T4, T1y, Tw, T1A, Tt, T1B; Chris@10: T1 = ri[0]; Chris@10: T1z = ii[0]; Chris@10: { Chris@10: E T2, T3, Tu, Tv; Chris@10: T2 = ri[WS(is, 5)]; Chris@10: T3 = ri[WS(is, 10)]; Chris@10: T4 = T2 + T3; Chris@10: T1y = KP866025403 * (T3 - T2); Chris@10: Tu = ii[WS(is, 5)]; Chris@10: Tv = ii[WS(is, 10)]; Chris@10: Tw = KP866025403 * (Tu - Tv); Chris@10: T1A = Tu + Tv; Chris@10: } Chris@10: T5 = T1 + T4; Chris@10: T2l = T1z + T1A; Chris@10: Tt = FNMS(KP500000000, T4, T1); Chris@10: Tx = Tt - Tw; Chris@10: TV = Tt + Tw; Chris@10: T1B = FNMS(KP500000000, T1A, T1z); Chris@10: T1C = T1y + T1B; Chris@10: T20 = T1B - T1y; Chris@10: } Chris@10: { Chris@10: E Th, Tk, TJ, T1h, T1i, T1j, TM, T1k, Tm, Tp, TO, T1m, T1n, T1o, TR; Chris@10: E T1p; Chris@10: { Chris@10: E Ti, Tj, TK, TL; Chris@10: Th = ri[WS(is, 6)]; Chris@10: Ti = ri[WS(is, 11)]; Chris@10: Tj = ri[WS(is, 1)]; Chris@10: Tk = Ti + Tj; Chris@10: TJ = FNMS(KP500000000, Tk, Th); Chris@10: T1h = KP866025403 * (Tj - Ti); Chris@10: T1i = ii[WS(is, 6)]; Chris@10: TK = ii[WS(is, 11)]; Chris@10: TL = ii[WS(is, 1)]; Chris@10: T1j = TK + TL; Chris@10: TM = KP866025403 * (TK - TL); Chris@10: T1k = FNMS(KP500000000, T1j, T1i); Chris@10: } Chris@10: { Chris@10: E Tn, To, TP, TQ; Chris@10: Tm = ri[WS(is, 9)]; Chris@10: Tn = ri[WS(is, 14)]; Chris@10: To = ri[WS(is, 4)]; Chris@10: Tp = Tn + To; Chris@10: TO = FNMS(KP500000000, Tp, Tm); Chris@10: T1m = KP866025403 * (To - Tn); Chris@10: T1n = ii[WS(is, 9)]; Chris@10: TP = ii[WS(is, 14)]; Chris@10: TQ = ii[WS(is, 4)]; Chris@10: T1o = TP + TQ; Chris@10: TR = KP866025403 * (TP - TQ); Chris@10: T1p = FNMS(KP500000000, T1o, T1n); Chris@10: } Chris@10: Tl = Th + Tk; Chris@10: Tq = Tm + Tp; Chris@10: Tr = Tl + Tq; Chris@10: TN = TJ - TM; Chris@10: TS = TO - TR; Chris@10: TT = TN + TS; Chris@10: T2c = T1i + T1j; Chris@10: T2d = T1n + T1o; Chris@10: T2n = T2c + T2d; Chris@10: T1O = T1k - T1h; Chris@10: T1P = T1p - T1m; Chris@10: T22 = T1O + T1P; Chris@10: T1l = T1h + T1k; Chris@10: T1q = T1m + T1p; Chris@10: T1w = T1l + T1q; Chris@10: TZ = TJ + TM; Chris@10: T10 = TO + TR; Chris@10: T11 = TZ + T10; Chris@10: } Chris@10: { Chris@10: E T6, T9, Ty, T16, T17, T18, TB, T19, Tb, Te, TD, T1b, T1c, T1d, TG; Chris@10: E T1e; Chris@10: { Chris@10: E T7, T8, Tz, TA; Chris@10: T6 = ri[WS(is, 3)]; Chris@10: T7 = ri[WS(is, 8)]; Chris@10: T8 = ri[WS(is, 13)]; Chris@10: T9 = T7 + T8; Chris@10: Ty = FNMS(KP500000000, T9, T6); Chris@10: T16 = KP866025403 * (T8 - T7); Chris@10: T17 = ii[WS(is, 3)]; Chris@10: Tz = ii[WS(is, 8)]; Chris@10: TA = ii[WS(is, 13)]; Chris@10: T18 = Tz + TA; Chris@10: TB = KP866025403 * (Tz - TA); Chris@10: T19 = FNMS(KP500000000, T18, T17); Chris@10: } Chris@10: { Chris@10: E Tc, Td, TE, TF; Chris@10: Tb = ri[WS(is, 12)]; Chris@10: Tc = ri[WS(is, 2)]; Chris@10: Td = ri[WS(is, 7)]; Chris@10: Te = Tc + Td; Chris@10: TD = FNMS(KP500000000, Te, Tb); Chris@10: T1b = KP866025403 * (Td - Tc); Chris@10: T1c = ii[WS(is, 12)]; Chris@10: TE = ii[WS(is, 2)]; Chris@10: TF = ii[WS(is, 7)]; Chris@10: T1d = TE + TF; Chris@10: TG = KP866025403 * (TE - TF); Chris@10: T1e = FNMS(KP500000000, T1d, T1c); Chris@10: } Chris@10: Ta = T6 + T9; Chris@10: Tf = Tb + Te; Chris@10: Tg = Ta + Tf; Chris@10: TC = Ty - TB; Chris@10: TH = TD - TG; Chris@10: TI = TC + TH; Chris@10: T2f = T17 + T18; Chris@10: T2g = T1c + T1d; Chris@10: T2m = T2f + T2g; Chris@10: T1R = T19 - T16; Chris@10: T1S = T1e - T1b; Chris@10: T21 = T1R + T1S; Chris@10: T1a = T16 + T19; Chris@10: T1f = T1b + T1e; Chris@10: T1v = T1a + T1f; Chris@10: TW = Ty + TB; Chris@10: TX = TD + TG; Chris@10: TY = TW + TX; Chris@10: } Chris@10: { Chris@10: E T2a, Ts, T29, T2i, T2k, T2e, T2h, T2j, T2b; Chris@10: T2a = KP559016994 * (Tg - Tr); Chris@10: Ts = Tg + Tr; Chris@10: T29 = FNMS(KP250000000, Ts, T5); Chris@10: T2e = T2c - T2d; Chris@10: T2h = T2f - T2g; Chris@10: T2i = FNMS(KP587785252, T2h, KP951056516 * T2e); Chris@10: T2k = FMA(KP951056516, T2h, KP587785252 * T2e); Chris@10: ro[0] = T5 + Ts; Chris@10: T2j = T2a + T29; Chris@10: ro[WS(os, 9)] = T2j - T2k; Chris@10: ro[WS(os, 6)] = T2j + T2k; Chris@10: T2b = T29 - T2a; Chris@10: ro[WS(os, 12)] = T2b - T2i; Chris@10: ro[WS(os, 3)] = T2b + T2i; Chris@10: } Chris@10: { Chris@10: E T2q, T2o, T2p, T2u, T2w, T2s, T2t, T2v, T2r; Chris@10: T2q = KP559016994 * (T2m - T2n); Chris@10: T2o = T2m + T2n; Chris@10: T2p = FNMS(KP250000000, T2o, T2l); Chris@10: T2s = Tl - Tq; Chris@10: T2t = Ta - Tf; Chris@10: T2u = FNMS(KP587785252, T2t, KP951056516 * T2s); Chris@10: T2w = FMA(KP951056516, T2t, KP587785252 * T2s); Chris@10: io[0] = T2l + T2o; Chris@10: T2v = T2q + T2p; Chris@10: io[WS(os, 6)] = T2v - T2w; Chris@10: io[WS(os, 9)] = T2w + T2v; Chris@10: T2r = T2p - T2q; Chris@10: io[WS(os, 3)] = T2r - T2u; Chris@10: io[WS(os, 12)] = T2u + T2r; Chris@10: } Chris@10: { Chris@10: E T1M, TU, T1L, T1U, T1W, T1Q, T1T, T1V, T1N; Chris@10: T1M = KP559016994 * (TI - TT); Chris@10: TU = TI + TT; Chris@10: T1L = FNMS(KP250000000, TU, Tx); Chris@10: T1Q = T1O - T1P; Chris@10: T1T = T1R - T1S; Chris@10: T1U = FNMS(KP587785252, T1T, KP951056516 * T1Q); Chris@10: T1W = FMA(KP951056516, T1T, KP587785252 * T1Q); Chris@10: ro[WS(os, 5)] = Tx + TU; Chris@10: T1V = T1M + T1L; Chris@10: ro[WS(os, 14)] = T1V - T1W; Chris@10: ro[WS(os, 11)] = T1V + T1W; Chris@10: T1N = T1L - T1M; Chris@10: ro[WS(os, 2)] = T1N - T1U; Chris@10: ro[WS(os, 8)] = T1N + T1U; Chris@10: } Chris@10: { Chris@10: E T25, T23, T24, T1Z, T28, T1X, T1Y, T27, T26; Chris@10: T25 = KP559016994 * (T21 - T22); Chris@10: T23 = T21 + T22; Chris@10: T24 = FNMS(KP250000000, T23, T20); Chris@10: T1X = TN - TS; Chris@10: T1Y = TC - TH; Chris@10: T1Z = FNMS(KP587785252, T1Y, KP951056516 * T1X); Chris@10: T28 = FMA(KP951056516, T1Y, KP587785252 * T1X); Chris@10: io[WS(os, 5)] = T20 + T23; Chris@10: T27 = T25 + T24; Chris@10: io[WS(os, 11)] = T27 - T28; Chris@10: io[WS(os, 14)] = T28 + T27; Chris@10: T26 = T24 - T25; Chris@10: io[WS(os, 2)] = T1Z + T26; Chris@10: io[WS(os, 8)] = T26 - T1Z; Chris@10: } Chris@10: { Chris@10: E T1x, T1D, T1E, T1I, T1J, T1G, T1H, T1K, T1F; Chris@10: T1x = KP559016994 * (T1v - T1w); Chris@10: T1D = T1v + T1w; Chris@10: T1E = FNMS(KP250000000, T1D, T1C); Chris@10: T1G = TW - TX; Chris@10: T1H = TZ - T10; Chris@10: T1I = FMA(KP951056516, T1G, KP587785252 * T1H); Chris@10: T1J = FNMS(KP587785252, T1G, KP951056516 * T1H); Chris@10: io[WS(os, 10)] = T1C + T1D; Chris@10: T1K = T1E - T1x; Chris@10: io[WS(os, 7)] = T1J + T1K; Chris@10: io[WS(os, 13)] = T1K - T1J; Chris@10: T1F = T1x + T1E; Chris@10: io[WS(os, 1)] = T1F - T1I; Chris@10: io[WS(os, 4)] = T1I + T1F; Chris@10: } Chris@10: { Chris@10: E T13, T12, T14, T1s, T1u, T1g, T1r, T1t, T15; Chris@10: T13 = KP559016994 * (TY - T11); Chris@10: T12 = TY + T11; Chris@10: T14 = FNMS(KP250000000, T12, TV); Chris@10: T1g = T1a - T1f; Chris@10: T1r = T1l - T1q; Chris@10: T1s = FMA(KP951056516, T1g, KP587785252 * T1r); Chris@10: T1u = FNMS(KP587785252, T1g, KP951056516 * T1r); Chris@10: ro[WS(os, 10)] = TV + T12; Chris@10: T1t = T14 - T13; Chris@10: ro[WS(os, 7)] = T1t - T1u; Chris@10: ro[WS(os, 13)] = T1t + T1u; Chris@10: T15 = T13 + T14; Chris@10: ro[WS(os, 4)] = T15 - T1s; Chris@10: ro[WS(os, 1)] = T15 + T1s; Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kdft_desc desc = { 15, "n1_15", {128, 28, 28, 0}, &GENUS, 0, 0, 0, 0 }; Chris@10: Chris@10: void X(codelet_n1_15) (planner *p) { Chris@10: X(kdft_register) (p, n1_15, &desc); Chris@10: } Chris@10: Chris@10: #endif /* HAVE_FMA */