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 07:36:09 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_twiddle.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -twiddle-log3 -precompute-twiddles -n 10 -name t2_10 -include t.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 114 FP additions, 94 FP multiplications, cannam@95: * (or, 48 additions, 28 multiplications, 66 fused multiply/add), cannam@95: * 85 stack variables, 4 constants, and 40 memory accesses cannam@95: */ cannam@95: #include "t.h" cannam@95: cannam@95: static void t2_10(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) cannam@95: { cannam@95: DK(KP951056516, +0.951056516295153572116439333379382143405698634); cannam@95: DK(KP559016994, +0.559016994374947424102293417182819058860154590); cannam@95: DK(KP250000000, +0.250000000000000000000000000000000000000000000); cannam@95: DK(KP618033988, +0.618033988749894848204586834365638117720309180); cannam@95: { cannam@95: INT m; cannam@95: for (m = mb, W = W + (mb * 6); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 6, MAKE_VOLATILE_STRIDE(20, rs)) { cannam@95: E T27, T2b, T2a, T2c; cannam@95: { cannam@95: E T2, T3, T8, Tc, T5, T4, TX, T11, TE, T6, TB, TA; cannam@95: T2 = W[0]; cannam@95: T3 = W[2]; cannam@95: T8 = W[4]; cannam@95: Tc = W[5]; cannam@95: T5 = W[1]; cannam@95: T4 = T2 * T3; cannam@95: TX = T3 * T8; cannam@95: TA = T2 * T8; cannam@95: T11 = T3 * Tc; cannam@95: TE = T2 * Tc; cannam@95: T6 = W[3]; cannam@95: TB = FMA(T5, Tc, TA); cannam@95: { cannam@95: E T2d, T24, T1c, Tk, T1i, T28, T2l, T1a, T2f, T1I, T1R, T1Z, TL, T1v, T1d; cannam@95: E Tz, T1S, T1r, TH, T1t; cannam@95: { cannam@95: E T1, TF, TY, T12, Tl, T7, T23, To, Tb, Te, Ti, Th, Td, Tw, Ts; cannam@95: E Ta; cannam@95: T1 = ri[0]; cannam@95: TF = FNMS(T5, T8, TE); cannam@95: TY = FMA(T6, Tc, TX); cannam@95: T12 = FNMS(T6, T8, T11); cannam@95: Tl = FMA(T5, T6, T4); cannam@95: T7 = FNMS(T5, T6, T4); cannam@95: Ta = T2 * T6; cannam@95: T23 = ii[0]; cannam@95: { cannam@95: E Tg, T9, Tv, Tr; cannam@95: Tg = T7 * Tc; cannam@95: T9 = T7 * T8; cannam@95: Tv = Tl * Tc; cannam@95: Tr = Tl * T8; cannam@95: To = FNMS(T5, T3, Ta); cannam@95: Tb = FMA(T5, T3, Ta); cannam@95: Te = ri[WS(rs, 5)]; cannam@95: Ti = ii[WS(rs, 5)]; cannam@95: Th = FNMS(Tb, T8, Tg); cannam@95: Td = FMA(Tb, Tc, T9); cannam@95: Tw = FNMS(To, T8, Tv); cannam@95: Ts = FMA(To, Tc, Tr); cannam@95: } cannam@95: { cannam@95: E T18, T1G, T1g, TW, T1P, T1C, T14, T1E; cannam@95: { cannam@95: E TR, T1z, TV, T1B, TZ, T13, T15, T17, T10, T1D; cannam@95: { cannam@95: E TO, TQ, TP, T22, Tj, T1y, T21, Tf; cannam@95: TO = ri[WS(rs, 4)]; cannam@95: T21 = Td * Ti; cannam@95: Tf = Td * Te; cannam@95: TQ = ii[WS(rs, 4)]; cannam@95: TP = T7 * TO; cannam@95: T22 = FNMS(Th, Te, T21); cannam@95: Tj = FMA(Th, Ti, Tf); cannam@95: T1y = T7 * TQ; cannam@95: TR = FMA(Tb, TQ, TP); cannam@95: T2d = T23 - T22; cannam@95: T24 = T22 + T23; cannam@95: T1c = T1 + Tj; cannam@95: Tk = T1 - Tj; cannam@95: T1z = FNMS(Tb, TO, T1y); cannam@95: } cannam@95: T15 = ri[WS(rs, 1)]; cannam@95: T17 = ii[WS(rs, 1)]; cannam@95: { cannam@95: E TS, TU, T16, T1F, TT, T1A; cannam@95: TS = ri[WS(rs, 9)]; cannam@95: TU = ii[WS(rs, 9)]; cannam@95: T16 = T2 * T15; cannam@95: T1F = T2 * T17; cannam@95: TT = T8 * TS; cannam@95: T1A = T8 * TU; cannam@95: T18 = FMA(T5, T17, T16); cannam@95: T1G = FNMS(T5, T15, T1F); cannam@95: TV = FMA(Tc, TU, TT); cannam@95: T1B = FNMS(Tc, TS, T1A); cannam@95: } cannam@95: TZ = ri[WS(rs, 6)]; cannam@95: T13 = ii[WS(rs, 6)]; cannam@95: T1g = TR + TV; cannam@95: TW = TR - TV; cannam@95: T1P = T1z + T1B; cannam@95: T1C = T1z - T1B; cannam@95: T10 = TY * TZ; cannam@95: T1D = TY * T13; cannam@95: T14 = FMA(T12, T13, T10); cannam@95: T1E = FNMS(T12, TZ, T1D); cannam@95: } cannam@95: { cannam@95: E Tq, T1o, Ty, TC, TG, T1q, TD, T1s; cannam@95: { cannam@95: E TI, TK, Tt, T1p; cannam@95: { cannam@95: E Tm, T1n, Tp, Tn; cannam@95: Tm = ri[WS(rs, 2)]; cannam@95: Tp = ii[WS(rs, 2)]; cannam@95: { cannam@95: E T19, T1h, T1Q, T1H; cannam@95: T19 = T14 - T18; cannam@95: T1h = T14 + T18; cannam@95: T1Q = T1E + T1G; cannam@95: T1H = T1E - T1G; cannam@95: Tn = Tl * Tm; cannam@95: T1i = T1g + T1h; cannam@95: T28 = T1g - T1h; cannam@95: T2l = TW - T19; cannam@95: T1a = TW + T19; cannam@95: T2f = T1C + T1H; cannam@95: T1I = T1C - T1H; cannam@95: T1R = T1P - T1Q; cannam@95: T1Z = T1P + T1Q; cannam@95: T1n = Tl * Tp; cannam@95: } cannam@95: Tq = FMA(To, Tp, Tn); cannam@95: TI = ri[WS(rs, 3)]; cannam@95: TK = ii[WS(rs, 3)]; cannam@95: T1o = FNMS(To, Tm, T1n); cannam@95: } cannam@95: { cannam@95: E Tx, Tu, TJ, T1u; cannam@95: Tt = ri[WS(rs, 7)]; cannam@95: TJ = T3 * TI; cannam@95: T1u = T3 * TK; cannam@95: Tx = ii[WS(rs, 7)]; cannam@95: Tu = Ts * Tt; cannam@95: TL = FMA(T6, TK, TJ); cannam@95: T1v = FNMS(T6, TI, T1u); cannam@95: T1p = Ts * Tx; cannam@95: Ty = FMA(Tw, Tx, Tu); cannam@95: } cannam@95: TC = ri[WS(rs, 8)]; cannam@95: TG = ii[WS(rs, 8)]; cannam@95: T1q = FNMS(Tw, Tt, T1p); cannam@95: } cannam@95: T1d = Tq + Ty; cannam@95: Tz = Tq - Ty; cannam@95: TD = TB * TC; cannam@95: T1s = TB * TG; cannam@95: T1S = T1o + T1q; cannam@95: T1r = T1o - T1q; cannam@95: TH = FMA(TF, TG, TD); cannam@95: T1t = FNMS(TF, TC, T1s); cannam@95: } cannam@95: } cannam@95: } cannam@95: { cannam@95: E T1f, T29, T1Y, T1U, T2j, T2n, T2m, T2o; cannam@95: { cannam@95: E T2k, T2e, T1l, T1L, T1J, T1k, T1b, T1e, TM; cannam@95: T1e = TH + TL; cannam@95: TM = TH - TL; cannam@95: { cannam@95: E T1w, T1T, TN, T1x; cannam@95: T1w = T1t - T1v; cannam@95: T1T = T1t + T1v; cannam@95: T1f = T1d + T1e; cannam@95: T29 = T1d - T1e; cannam@95: T2k = Tz - TM; cannam@95: TN = Tz + TM; cannam@95: T1x = T1r - T1w; cannam@95: T2e = T1r + T1w; cannam@95: T1Y = T1S + T1T; cannam@95: T1U = T1S - T1T; cannam@95: T1l = TN - T1a; cannam@95: T1b = TN + T1a; cannam@95: T1L = FNMS(KP618033988, T1x, T1I); cannam@95: T1J = FMA(KP618033988, T1I, T1x); cannam@95: } cannam@95: T1k = FNMS(KP250000000, T1b, Tk); cannam@95: ri[WS(rs, 5)] = Tk + T1b; cannam@95: { cannam@95: E T2g, T2i, T2h, T1K, T1m; cannam@95: T2g = T2e + T2f; cannam@95: T2i = T2e - T2f; cannam@95: T1K = FNMS(KP559016994, T1l, T1k); cannam@95: T1m = FMA(KP559016994, T1l, T1k); cannam@95: T2h = FNMS(KP250000000, T2g, T2d); cannam@95: ri[WS(rs, 1)] = FMA(KP951056516, T1J, T1m); cannam@95: ri[WS(rs, 9)] = FNMS(KP951056516, T1J, T1m); cannam@95: ri[WS(rs, 3)] = FMA(KP951056516, T1L, T1K); cannam@95: ri[WS(rs, 7)] = FNMS(KP951056516, T1L, T1K); cannam@95: ii[WS(rs, 5)] = T2g + T2d; cannam@95: T2j = FMA(KP559016994, T2i, T2h); cannam@95: T2n = FNMS(KP559016994, T2i, T2h); cannam@95: T2m = FMA(KP618033988, T2l, T2k); cannam@95: T2o = FNMS(KP618033988, T2k, T2l); cannam@95: } cannam@95: } cannam@95: { cannam@95: E T1O, T1W, T1V, T1X, T1j, T1N, T1M, T20, T26, T25; cannam@95: T1j = T1f + T1i; cannam@95: T1N = T1f - T1i; cannam@95: ii[WS(rs, 7)] = FMA(KP951056516, T2o, T2n); cannam@95: ii[WS(rs, 3)] = FNMS(KP951056516, T2o, T2n); cannam@95: ii[WS(rs, 9)] = FMA(KP951056516, T2m, T2j); cannam@95: ii[WS(rs, 1)] = FNMS(KP951056516, T2m, T2j); cannam@95: T1M = FNMS(KP250000000, T1j, T1c); cannam@95: ri[0] = T1c + T1j; cannam@95: T1O = FNMS(KP559016994, T1N, T1M); cannam@95: T1W = FMA(KP559016994, T1N, T1M); cannam@95: T1V = FNMS(KP618033988, T1U, T1R); cannam@95: T1X = FMA(KP618033988, T1R, T1U); cannam@95: T20 = T1Y + T1Z; cannam@95: T26 = T1Y - T1Z; cannam@95: ri[WS(rs, 6)] = FMA(KP951056516, T1X, T1W); cannam@95: ri[WS(rs, 4)] = FNMS(KP951056516, T1X, T1W); cannam@95: ri[WS(rs, 8)] = FMA(KP951056516, T1V, T1O); cannam@95: ri[WS(rs, 2)] = FNMS(KP951056516, T1V, T1O); cannam@95: T25 = FNMS(KP250000000, T20, T24); cannam@95: ii[0] = T20 + T24; cannam@95: T27 = FNMS(KP559016994, T26, T25); cannam@95: T2b = FMA(KP559016994, T26, T25); cannam@95: T2a = FNMS(KP618033988, T29, T28); cannam@95: T2c = FMA(KP618033988, T28, T29); cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: ii[WS(rs, 6)] = FNMS(KP951056516, T2c, T2b); cannam@95: ii[WS(rs, 4)] = FMA(KP951056516, T2c, T2b); cannam@95: ii[WS(rs, 8)] = FNMS(KP951056516, T2a, T27); cannam@95: ii[WS(rs, 2)] = FMA(KP951056516, T2a, T27); cannam@95: } cannam@95: } cannam@95: } cannam@95: cannam@95: static const tw_instr twinstr[] = { cannam@95: {TW_CEXP, 0, 1}, cannam@95: {TW_CEXP, 0, 3}, cannam@95: {TW_CEXP, 0, 9}, cannam@95: {TW_NEXT, 1, 0} cannam@95: }; cannam@95: cannam@95: static const ct_desc desc = { 10, "t2_10", twinstr, &GENUS, {48, 28, 66, 0}, 0, 0, 0 }; cannam@95: cannam@95: void X(codelet_t2_10) (planner *p) { cannam@95: X(kdft_dit_register) (p, t2_10, &desc); cannam@95: } cannam@95: #else /* HAVE_FMA */ cannam@95: cannam@95: /* Generated by: ../../../genfft/gen_twiddle.native -compact -variables 4 -pipeline-latency 4 -twiddle-log3 -precompute-twiddles -n 10 -name t2_10 -include t.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 114 FP additions, 80 FP multiplications, cannam@95: * (or, 76 additions, 42 multiplications, 38 fused multiply/add), cannam@95: * 63 stack variables, 4 constants, and 40 memory accesses cannam@95: */ cannam@95: #include "t.h" cannam@95: cannam@95: static void t2_10(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) cannam@95: { cannam@95: DK(KP587785252, +0.587785252292473129168705954639072768597652438); cannam@95: DK(KP951056516, +0.951056516295153572116439333379382143405698634); cannam@95: DK(KP250000000, +0.250000000000000000000000000000000000000000000); cannam@95: DK(KP559016994, +0.559016994374947424102293417182819058860154590); cannam@95: { cannam@95: INT m; cannam@95: for (m = mb, W = W + (mb * 6); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 6, MAKE_VOLATILE_STRIDE(20, rs)) { cannam@95: E T2, T5, T3, T6, T8, Tm, Tc, Tk, T9, Td, Te, TM, TO, Tg, Tp; cannam@95: E Tv, Tx, Tr; cannam@95: { cannam@95: E T4, Tb, T7, Ta; cannam@95: T2 = W[0]; cannam@95: T5 = W[1]; cannam@95: T3 = W[2]; cannam@95: T6 = W[3]; cannam@95: T4 = T2 * T3; cannam@95: Tb = T5 * T3; cannam@95: T7 = T5 * T6; cannam@95: Ta = T2 * T6; cannam@95: T8 = T4 - T7; cannam@95: Tm = Ta - Tb; cannam@95: Tc = Ta + Tb; cannam@95: Tk = T4 + T7; cannam@95: T9 = W[4]; cannam@95: Td = W[5]; cannam@95: Te = FMA(T8, T9, Tc * Td); cannam@95: TM = FMA(T3, T9, T6 * Td); cannam@95: TO = FNMS(T6, T9, T3 * Td); cannam@95: Tg = FNMS(Tc, T9, T8 * Td); cannam@95: Tp = FMA(Tk, T9, Tm * Td); cannam@95: Tv = FMA(T2, T9, T5 * Td); cannam@95: Tx = FNMS(T5, T9, T2 * Td); cannam@95: Tr = FNMS(Tm, T9, Tk * Td); cannam@95: } cannam@95: { cannam@95: E Tj, T1S, TX, T1G, TL, TU, TV, T1s, T1t, T1C, T11, T12, T13, T1h, T1k; cannam@95: E T1Q, Tu, TD, TE, T1v, T1w, T1B, TY, TZ, T10, T1a, T1d, T1P; cannam@95: { cannam@95: E T1, T1F, Ti, T1E, Tf, Th; cannam@95: T1 = ri[0]; cannam@95: T1F = ii[0]; cannam@95: Tf = ri[WS(rs, 5)]; cannam@95: Th = ii[WS(rs, 5)]; cannam@95: Ti = FMA(Te, Tf, Tg * Th); cannam@95: T1E = FNMS(Tg, Tf, Te * Th); cannam@95: Tj = T1 - Ti; cannam@95: T1S = T1F - T1E; cannam@95: TX = T1 + Ti; cannam@95: T1G = T1E + T1F; cannam@95: } cannam@95: { cannam@95: E TH, T1f, TT, T1j, TK, T1g, TQ, T1i; cannam@95: { cannam@95: E TF, TG, TR, TS; cannam@95: TF = ri[WS(rs, 4)]; cannam@95: TG = ii[WS(rs, 4)]; cannam@95: TH = FMA(T8, TF, Tc * TG); cannam@95: T1f = FNMS(Tc, TF, T8 * TG); cannam@95: TR = ri[WS(rs, 1)]; cannam@95: TS = ii[WS(rs, 1)]; cannam@95: TT = FMA(T2, TR, T5 * TS); cannam@95: T1j = FNMS(T5, TR, T2 * TS); cannam@95: } cannam@95: { cannam@95: E TI, TJ, TN, TP; cannam@95: TI = ri[WS(rs, 9)]; cannam@95: TJ = ii[WS(rs, 9)]; cannam@95: TK = FMA(T9, TI, Td * TJ); cannam@95: T1g = FNMS(Td, TI, T9 * TJ); cannam@95: TN = ri[WS(rs, 6)]; cannam@95: TP = ii[WS(rs, 6)]; cannam@95: TQ = FMA(TM, TN, TO * TP); cannam@95: T1i = FNMS(TO, TN, TM * TP); cannam@95: } cannam@95: TL = TH - TK; cannam@95: TU = TQ - TT; cannam@95: TV = TL + TU; cannam@95: T1s = T1f + T1g; cannam@95: T1t = T1i + T1j; cannam@95: T1C = T1s + T1t; cannam@95: T11 = TH + TK; cannam@95: T12 = TQ + TT; cannam@95: T13 = T11 + T12; cannam@95: T1h = T1f - T1g; cannam@95: T1k = T1i - T1j; cannam@95: T1Q = T1h + T1k; cannam@95: } cannam@95: { cannam@95: E To, T18, TC, T1c, Tt, T19, Tz, T1b; cannam@95: { cannam@95: E Tl, Tn, TA, TB; cannam@95: Tl = ri[WS(rs, 2)]; cannam@95: Tn = ii[WS(rs, 2)]; cannam@95: To = FMA(Tk, Tl, Tm * Tn); cannam@95: T18 = FNMS(Tm, Tl, Tk * Tn); cannam@95: TA = ri[WS(rs, 3)]; cannam@95: TB = ii[WS(rs, 3)]; cannam@95: TC = FMA(T3, TA, T6 * TB); cannam@95: T1c = FNMS(T6, TA, T3 * TB); cannam@95: } cannam@95: { cannam@95: E Tq, Ts, Tw, Ty; cannam@95: Tq = ri[WS(rs, 7)]; cannam@95: Ts = ii[WS(rs, 7)]; cannam@95: Tt = FMA(Tp, Tq, Tr * Ts); cannam@95: T19 = FNMS(Tr, Tq, Tp * Ts); cannam@95: Tw = ri[WS(rs, 8)]; cannam@95: Ty = ii[WS(rs, 8)]; cannam@95: Tz = FMA(Tv, Tw, Tx * Ty); cannam@95: T1b = FNMS(Tx, Tw, Tv * Ty); cannam@95: } cannam@95: Tu = To - Tt; cannam@95: TD = Tz - TC; cannam@95: TE = Tu + TD; cannam@95: T1v = T18 + T19; cannam@95: T1w = T1b + T1c; cannam@95: T1B = T1v + T1w; cannam@95: TY = To + Tt; cannam@95: TZ = Tz + TC; cannam@95: T10 = TY + TZ; cannam@95: T1a = T18 - T19; cannam@95: T1d = T1b - T1c; cannam@95: T1P = T1a + T1d; cannam@95: } cannam@95: { cannam@95: E T15, TW, T16, T1m, T1o, T1e, T1l, T1n, T17; cannam@95: T15 = KP559016994 * (TE - TV); cannam@95: TW = TE + TV; cannam@95: T16 = FNMS(KP250000000, TW, Tj); cannam@95: T1e = T1a - T1d; cannam@95: T1l = T1h - T1k; cannam@95: T1m = FMA(KP951056516, T1e, KP587785252 * T1l); cannam@95: T1o = FNMS(KP587785252, T1e, KP951056516 * T1l); cannam@95: ri[WS(rs, 5)] = Tj + TW; cannam@95: T1n = T16 - T15; cannam@95: ri[WS(rs, 7)] = T1n - T1o; cannam@95: ri[WS(rs, 3)] = T1n + T1o; cannam@95: T17 = T15 + T16; cannam@95: ri[WS(rs, 9)] = T17 - T1m; cannam@95: ri[WS(rs, 1)] = T17 + T1m; cannam@95: } cannam@95: { cannam@95: E T1R, T1T, T1U, T1Y, T20, T1W, T1X, T1Z, T1V; cannam@95: T1R = KP559016994 * (T1P - T1Q); cannam@95: T1T = T1P + T1Q; cannam@95: T1U = FNMS(KP250000000, T1T, T1S); cannam@95: T1W = Tu - TD; cannam@95: T1X = TL - TU; cannam@95: T1Y = FMA(KP951056516, T1W, KP587785252 * T1X); cannam@95: T20 = FNMS(KP587785252, T1W, KP951056516 * T1X); cannam@95: ii[WS(rs, 5)] = T1T + T1S; cannam@95: T1Z = T1U - T1R; cannam@95: ii[WS(rs, 3)] = T1Z - T20; cannam@95: ii[WS(rs, 7)] = T20 + T1Z; cannam@95: T1V = T1R + T1U; cannam@95: ii[WS(rs, 1)] = T1V - T1Y; cannam@95: ii[WS(rs, 9)] = T1Y + T1V; cannam@95: } cannam@95: { cannam@95: E T1q, T14, T1p, T1y, T1A, T1u, T1x, T1z, T1r; cannam@95: T1q = KP559016994 * (T10 - T13); cannam@95: T14 = T10 + T13; cannam@95: T1p = FNMS(KP250000000, T14, TX); cannam@95: T1u = T1s - T1t; cannam@95: T1x = T1v - T1w; cannam@95: T1y = FNMS(KP587785252, T1x, KP951056516 * T1u); cannam@95: T1A = FMA(KP951056516, T1x, KP587785252 * T1u); cannam@95: ri[0] = TX + T14; cannam@95: T1z = T1q + T1p; cannam@95: ri[WS(rs, 4)] = T1z - T1A; cannam@95: ri[WS(rs, 6)] = T1z + T1A; cannam@95: T1r = T1p - T1q; cannam@95: ri[WS(rs, 2)] = T1r - T1y; cannam@95: ri[WS(rs, 8)] = T1r + T1y; cannam@95: } cannam@95: { cannam@95: E T1L, T1D, T1K, T1J, T1N, T1H, T1I, T1O, T1M; cannam@95: T1L = KP559016994 * (T1B - T1C); cannam@95: T1D = T1B + T1C; cannam@95: T1K = FNMS(KP250000000, T1D, T1G); cannam@95: T1H = T11 - T12; cannam@95: T1I = TY - TZ; cannam@95: T1J = FNMS(KP587785252, T1I, KP951056516 * T1H); cannam@95: T1N = FMA(KP951056516, T1I, KP587785252 * T1H); cannam@95: ii[0] = T1D + T1G; cannam@95: T1O = T1L + T1K; cannam@95: ii[WS(rs, 4)] = T1N + T1O; cannam@95: ii[WS(rs, 6)] = T1O - T1N; cannam@95: T1M = T1K - T1L; cannam@95: ii[WS(rs, 2)] = T1J + T1M; cannam@95: ii[WS(rs, 8)] = T1M - T1J; cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: cannam@95: static const tw_instr twinstr[] = { cannam@95: {TW_CEXP, 0, 1}, cannam@95: {TW_CEXP, 0, 3}, cannam@95: {TW_CEXP, 0, 9}, cannam@95: {TW_NEXT, 1, 0} cannam@95: }; cannam@95: cannam@95: static const ct_desc desc = { 10, "t2_10", twinstr, &GENUS, {76, 42, 38, 0}, 0, 0, 0 }; cannam@95: cannam@95: void X(codelet_t2_10) (planner *p) { cannam@95: X(kdft_dit_register) (p, t2_10, &desc); cannam@95: } cannam@95: #endif /* HAVE_FMA */