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