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 07:35:50 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 -n 12 -name t1_12 -include t.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 118 FP additions, 68 FP multiplications, Chris@10: * (or, 72 additions, 22 multiplications, 46 fused multiply/add), Chris@10: * 84 stack variables, 2 constants, and 48 memory accesses Chris@10: */ Chris@10: #include "t.h" Chris@10: Chris@10: static void t1_12(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) Chris@10: { Chris@10: DK(KP866025403, +0.866025403784438646763723170752936183471402627); Chris@10: DK(KP500000000, +0.500000000000000000000000000000000000000000000); Chris@10: { Chris@10: INT m; Chris@10: for (m = mb, W = W + (mb * 22); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 22, MAKE_VOLATILE_STRIDE(24, rs)) { Chris@10: E T2B, T2C; Chris@10: { Chris@10: E T1, T2i, T2e, Tl, T1Y, T10, T1S, TG, T2f, T1s, T2r, Ty, T1Z, T1H, T21; Chris@10: E T1d, TI, TL, T2h, T1l, T2o, Te, TJ, T1w, TO, TR, TN, TK, TQ; Chris@10: { Chris@10: E TW, TZ, TY, T1X, TX; Chris@10: T1 = ri[0]; Chris@10: T2i = ii[0]; Chris@10: { Chris@10: E Th, Tk, Tg, Tj, T2d, Ti, TV; Chris@10: Th = ri[WS(rs, 6)]; Chris@10: Tk = ii[WS(rs, 6)]; Chris@10: Tg = W[10]; Chris@10: Tj = W[11]; Chris@10: TW = ri[WS(rs, 9)]; Chris@10: TZ = ii[WS(rs, 9)]; Chris@10: T2d = Tg * Tk; Chris@10: Ti = Tg * Th; Chris@10: TV = W[16]; Chris@10: TY = W[17]; Chris@10: T2e = FNMS(Tj, Th, T2d); Chris@10: Tl = FMA(Tj, Tk, Ti); Chris@10: T1X = TV * TZ; Chris@10: TX = TV * TW; Chris@10: } Chris@10: { Chris@10: E Tn, Tq, Tt, T1o, To, Tw, Ts, Tp, Tv; Chris@10: { Chris@10: E TC, TF, TB, TE, T1R, TD, Tm; Chris@10: TC = ri[WS(rs, 3)]; Chris@10: TF = ii[WS(rs, 3)]; Chris@10: T1Y = FNMS(TY, TW, T1X); Chris@10: T10 = FMA(TY, TZ, TX); Chris@10: TB = W[4]; Chris@10: TE = W[5]; Chris@10: Tn = ri[WS(rs, 10)]; Chris@10: Tq = ii[WS(rs, 10)]; Chris@10: T1R = TB * TF; Chris@10: TD = TB * TC; Chris@10: Tm = W[18]; Chris@10: Tt = ri[WS(rs, 2)]; Chris@10: T1S = FNMS(TE, TC, T1R); Chris@10: TG = FMA(TE, TF, TD); Chris@10: T1o = Tm * Tq; Chris@10: To = Tm * Tn; Chris@10: Tw = ii[WS(rs, 2)]; Chris@10: Ts = W[2]; Chris@10: Tp = W[19]; Chris@10: Tv = W[3]; Chris@10: } Chris@10: { Chris@10: E T12, T15, T13, T1D, T18, T1b, T17, T14, T1a; Chris@10: { Chris@10: E T1p, Tr, T1r, Tx, T1q, Tu, T11; Chris@10: T12 = ri[WS(rs, 1)]; Chris@10: T1q = Ts * Tw; Chris@10: Tu = Ts * Tt; Chris@10: T1p = FNMS(Tp, Tn, T1o); Chris@10: Tr = FMA(Tp, Tq, To); Chris@10: T1r = FNMS(Tv, Tt, T1q); Chris@10: Tx = FMA(Tv, Tw, Tu); Chris@10: T15 = ii[WS(rs, 1)]; Chris@10: T11 = W[0]; Chris@10: T2f = T1p + T1r; Chris@10: T1s = T1p - T1r; Chris@10: T2r = Tx - Tr; Chris@10: Ty = Tr + Tx; Chris@10: T13 = T11 * T12; Chris@10: T1D = T11 * T15; Chris@10: } Chris@10: T18 = ri[WS(rs, 5)]; Chris@10: T1b = ii[WS(rs, 5)]; Chris@10: T17 = W[8]; Chris@10: T14 = W[1]; Chris@10: T1a = W[9]; Chris@10: { Chris@10: E T3, T6, T4, T1h, T9, Tc, T8, T5, Tb; Chris@10: { Chris@10: E T1E, T16, T1G, T1c, T1F, T19, T2; Chris@10: T3 = ri[WS(rs, 4)]; Chris@10: T1F = T17 * T1b; Chris@10: T19 = T17 * T18; Chris@10: T1E = FNMS(T14, T12, T1D); Chris@10: T16 = FMA(T14, T15, T13); Chris@10: T1G = FNMS(T1a, T18, T1F); Chris@10: T1c = FMA(T1a, T1b, T19); Chris@10: T6 = ii[WS(rs, 4)]; Chris@10: T2 = W[6]; Chris@10: T1Z = T1E + T1G; Chris@10: T1H = T1E - T1G; Chris@10: T21 = T1c - T16; Chris@10: T1d = T16 + T1c; Chris@10: T4 = T2 * T3; Chris@10: T1h = T2 * T6; Chris@10: } Chris@10: T9 = ri[WS(rs, 8)]; Chris@10: Tc = ii[WS(rs, 8)]; Chris@10: T8 = W[14]; Chris@10: T5 = W[7]; Chris@10: Tb = W[15]; Chris@10: { Chris@10: E T1i, T7, T1k, Td, T1j, Ta, TH; Chris@10: TI = ri[WS(rs, 7)]; Chris@10: T1j = T8 * Tc; Chris@10: Ta = T8 * T9; Chris@10: T1i = FNMS(T5, T3, T1h); Chris@10: T7 = FMA(T5, T6, T4); Chris@10: T1k = FNMS(Tb, T9, T1j); Chris@10: Td = FMA(Tb, Tc, Ta); Chris@10: TL = ii[WS(rs, 7)]; Chris@10: TH = W[12]; Chris@10: T2h = T1i + T1k; Chris@10: T1l = T1i - T1k; Chris@10: T2o = Td - T7; Chris@10: Te = T7 + Td; Chris@10: TJ = TH * TI; Chris@10: T1w = TH * TL; Chris@10: } Chris@10: TO = ri[WS(rs, 11)]; Chris@10: TR = ii[WS(rs, 11)]; Chris@10: TN = W[20]; Chris@10: TK = W[13]; Chris@10: TQ = W[21]; Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: { Chris@10: E T1g, T1n, T2q, T1A, T1V, T28, TA, T2n, T1v, T1C, T1U, T29, T2m, T2k, T2l; Chris@10: E T1f, T2a, T20; Chris@10: { Chris@10: E T2g, T1T, TT, T2j, TU, T1e; Chris@10: { Chris@10: E Tf, T1x, TM, T1z, TS, Tz, T1y, TP; Chris@10: T1g = FNMS(KP500000000, Te, T1); Chris@10: Tf = T1 + Te; Chris@10: T1y = TN * TR; Chris@10: TP = TN * TO; Chris@10: T1x = FNMS(TK, TI, T1w); Chris@10: TM = FMA(TK, TL, TJ); Chris@10: T1z = FNMS(TQ, TO, T1y); Chris@10: TS = FMA(TQ, TR, TP); Chris@10: Tz = Tl + Ty; Chris@10: T1n = FNMS(KP500000000, Ty, Tl); Chris@10: T2q = FNMS(KP500000000, T2f, T2e); Chris@10: T2g = T2e + T2f; Chris@10: T1T = T1x + T1z; Chris@10: T1A = T1x - T1z; Chris@10: T1V = TS - TM; Chris@10: TT = TM + TS; Chris@10: T28 = Tf - Tz; Chris@10: TA = Tf + Tz; Chris@10: T2j = T2h + T2i; Chris@10: T2n = FNMS(KP500000000, T2h, T2i); Chris@10: } Chris@10: T1v = FNMS(KP500000000, TT, TG); Chris@10: TU = TG + TT; Chris@10: T1e = T10 + T1d; Chris@10: T1C = FNMS(KP500000000, T1d, T10); Chris@10: T1U = FNMS(KP500000000, T1T, T1S); Chris@10: T29 = T1S + T1T; Chris@10: T2m = T2j - T2g; Chris@10: T2k = T2g + T2j; Chris@10: T2l = TU - T1e; Chris@10: T1f = TU + T1e; Chris@10: T2a = T1Y + T1Z; Chris@10: T20 = FNMS(KP500000000, T1Z, T1Y); Chris@10: } Chris@10: { Chris@10: E T1m, T1K, T2y, T2p, T2x, T2s, T1L, T1t, T1B, T1N, T2c, T2b; Chris@10: ii[WS(rs, 9)] = T2m - T2l; Chris@10: ii[WS(rs, 3)] = T2l + T2m; Chris@10: ri[0] = TA + T1f; Chris@10: ri[WS(rs, 6)] = TA - T1f; Chris@10: T2c = T29 + T2a; Chris@10: T2b = T29 - T2a; Chris@10: T1m = FNMS(KP866025403, T1l, T1g); Chris@10: T1K = FMA(KP866025403, T1l, T1g); Chris@10: ii[0] = T2c + T2k; Chris@10: ii[WS(rs, 6)] = T2k - T2c; Chris@10: ri[WS(rs, 9)] = T28 + T2b; Chris@10: ri[WS(rs, 3)] = T28 - T2b; Chris@10: T2y = FNMS(KP866025403, T2o, T2n); Chris@10: T2p = FMA(KP866025403, T2o, T2n); Chris@10: T2x = FNMS(KP866025403, T2r, T2q); Chris@10: T2s = FMA(KP866025403, T2r, T2q); Chris@10: T1L = FMA(KP866025403, T1s, T1n); Chris@10: T1t = FNMS(KP866025403, T1s, T1n); Chris@10: T1B = FNMS(KP866025403, T1A, T1v); Chris@10: T1N = FMA(KP866025403, T1A, T1v); Chris@10: { Chris@10: E T24, T27, T1Q, T2u, T23, T2v, T2w, T2t; Chris@10: { Chris@10: E T1u, T1W, T22, T1O, T1I, T2z, T2A, T25, T26, T1M, T1J, T1P; Chris@10: T24 = T1m - T1t; Chris@10: T1u = T1m + T1t; Chris@10: T25 = FNMS(KP866025403, T1V, T1U); Chris@10: T1W = FMA(KP866025403, T1V, T1U); Chris@10: T26 = FNMS(KP866025403, T21, T20); Chris@10: T22 = FMA(KP866025403, T21, T20); Chris@10: T1O = FMA(KP866025403, T1H, T1C); Chris@10: T1I = FNMS(KP866025403, T1H, T1C); Chris@10: T2z = T2x + T2y; Chris@10: T2B = T2y - T2x; Chris@10: T27 = T25 - T26; Chris@10: T2A = T25 + T26; Chris@10: T1M = T1K + T1L; Chris@10: T1Q = T1K - T1L; Chris@10: T2C = T1B - T1I; Chris@10: T1J = T1B + T1I; Chris@10: T1P = T1N + T1O; Chris@10: T2u = T1N - T1O; Chris@10: ii[WS(rs, 8)] = T2A + T2z; Chris@10: ii[WS(rs, 2)] = T2z - T2A; Chris@10: ri[WS(rs, 8)] = T1u + T1J; Chris@10: ri[WS(rs, 2)] = T1u - T1J; Chris@10: ri[WS(rs, 10)] = T1M - T1P; Chris@10: ri[WS(rs, 4)] = T1M + T1P; Chris@10: T23 = T1W - T22; Chris@10: T2v = T1W + T22; Chris@10: T2w = T2s + T2p; Chris@10: T2t = T2p - T2s; Chris@10: } Chris@10: ii[WS(rs, 10)] = T2w - T2v; Chris@10: ii[WS(rs, 4)] = T2v + T2w; Chris@10: ri[WS(rs, 1)] = T1Q + T23; Chris@10: ri[WS(rs, 7)] = T1Q - T23; Chris@10: ii[WS(rs, 7)] = T2u + T2t; Chris@10: ii[WS(rs, 1)] = T2t - T2u; Chris@10: ri[WS(rs, 5)] = T24 + T27; Chris@10: ri[WS(rs, 11)] = T24 - T27; Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: ii[WS(rs, 11)] = T2C + T2B; Chris@10: ii[WS(rs, 5)] = T2B - T2C; Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const tw_instr twinstr[] = { Chris@10: {TW_FULL, 0, 12}, Chris@10: {TW_NEXT, 1, 0} Chris@10: }; Chris@10: Chris@10: static const ct_desc desc = { 12, "t1_12", twinstr, &GENUS, {72, 22, 46, 0}, 0, 0, 0 }; Chris@10: Chris@10: void X(codelet_t1_12) (planner *p) { Chris@10: X(kdft_dit_register) (p, t1_12, &desc); Chris@10: } Chris@10: #else /* HAVE_FMA */ Chris@10: Chris@10: /* Generated by: ../../../genfft/gen_twiddle.native -compact -variables 4 -pipeline-latency 4 -n 12 -name t1_12 -include t.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 118 FP additions, 60 FP multiplications, Chris@10: * (or, 88 additions, 30 multiplications, 30 fused multiply/add), Chris@10: * 47 stack variables, 2 constants, and 48 memory accesses Chris@10: */ Chris@10: #include "t.h" Chris@10: Chris@10: static void t1_12(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) Chris@10: { Chris@10: DK(KP500000000, +0.500000000000000000000000000000000000000000000); Chris@10: DK(KP866025403, +0.866025403784438646763723170752936183471402627); Chris@10: { Chris@10: INT m; Chris@10: for (m = mb, W = W + (mb * 22); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 22, MAKE_VOLATILE_STRIDE(24, rs)) { Chris@10: E T1, T1W, T18, T21, Tc, T15, T1V, T22, TR, T1E, T1o, T1D, T12, T1l, T1F; Chris@10: E T1G, Ti, T1S, T1d, T24, Tt, T1a, T1T, T25, TA, T1z, T1j, T1y, TL, T1g; Chris@10: E T1A, T1B; Chris@10: { Chris@10: E T6, T16, Tb, T17; Chris@10: T1 = ri[0]; Chris@10: T1W = ii[0]; Chris@10: { Chris@10: E T3, T5, T2, T4; Chris@10: T3 = ri[WS(rs, 4)]; Chris@10: T5 = ii[WS(rs, 4)]; Chris@10: T2 = W[6]; Chris@10: T4 = W[7]; Chris@10: T6 = FMA(T2, T3, T4 * T5); Chris@10: T16 = FNMS(T4, T3, T2 * T5); Chris@10: } Chris@10: { Chris@10: E T8, Ta, T7, T9; Chris@10: T8 = ri[WS(rs, 8)]; Chris@10: Ta = ii[WS(rs, 8)]; Chris@10: T7 = W[14]; Chris@10: T9 = W[15]; Chris@10: Tb = FMA(T7, T8, T9 * Ta); Chris@10: T17 = FNMS(T9, T8, T7 * Ta); Chris@10: } Chris@10: T18 = KP866025403 * (T16 - T17); Chris@10: T21 = KP866025403 * (Tb - T6); Chris@10: Tc = T6 + Tb; Chris@10: T15 = FNMS(KP500000000, Tc, T1); Chris@10: T1V = T16 + T17; Chris@10: T22 = FNMS(KP500000000, T1V, T1W); Chris@10: } Chris@10: { Chris@10: E T11, T1n, TW, T1m; Chris@10: { Chris@10: E TO, TQ, TN, TP; Chris@10: TO = ri[WS(rs, 9)]; Chris@10: TQ = ii[WS(rs, 9)]; Chris@10: TN = W[16]; Chris@10: TP = W[17]; Chris@10: TR = FMA(TN, TO, TP * TQ); Chris@10: T1E = FNMS(TP, TO, TN * TQ); Chris@10: } Chris@10: { Chris@10: E TY, T10, TX, TZ; Chris@10: TY = ri[WS(rs, 5)]; Chris@10: T10 = ii[WS(rs, 5)]; Chris@10: TX = W[8]; Chris@10: TZ = W[9]; Chris@10: T11 = FMA(TX, TY, TZ * T10); Chris@10: T1n = FNMS(TZ, TY, TX * T10); Chris@10: } Chris@10: { Chris@10: E TT, TV, TS, TU; Chris@10: TT = ri[WS(rs, 1)]; Chris@10: TV = ii[WS(rs, 1)]; Chris@10: TS = W[0]; Chris@10: TU = W[1]; Chris@10: TW = FMA(TS, TT, TU * TV); Chris@10: T1m = FNMS(TU, TT, TS * TV); Chris@10: } Chris@10: T1o = KP866025403 * (T1m - T1n); Chris@10: T1D = KP866025403 * (T11 - TW); Chris@10: T12 = TW + T11; Chris@10: T1l = FNMS(KP500000000, T12, TR); Chris@10: T1F = T1m + T1n; Chris@10: T1G = FNMS(KP500000000, T1F, T1E); Chris@10: } Chris@10: { Chris@10: E Ts, T1c, Tn, T1b; Chris@10: { Chris@10: E Tf, Th, Te, Tg; Chris@10: Tf = ri[WS(rs, 6)]; Chris@10: Th = ii[WS(rs, 6)]; Chris@10: Te = W[10]; Chris@10: Tg = W[11]; Chris@10: Ti = FMA(Te, Tf, Tg * Th); Chris@10: T1S = FNMS(Tg, Tf, Te * Th); Chris@10: } Chris@10: { Chris@10: E Tp, Tr, To, Tq; Chris@10: Tp = ri[WS(rs, 2)]; Chris@10: Tr = ii[WS(rs, 2)]; Chris@10: To = W[2]; Chris@10: Tq = W[3]; Chris@10: Ts = FMA(To, Tp, Tq * Tr); Chris@10: T1c = FNMS(Tq, Tp, To * Tr); Chris@10: } Chris@10: { Chris@10: E Tk, Tm, Tj, Tl; Chris@10: Tk = ri[WS(rs, 10)]; Chris@10: Tm = ii[WS(rs, 10)]; Chris@10: Tj = W[18]; Chris@10: Tl = W[19]; Chris@10: Tn = FMA(Tj, Tk, Tl * Tm); Chris@10: T1b = FNMS(Tl, Tk, Tj * Tm); Chris@10: } Chris@10: T1d = KP866025403 * (T1b - T1c); Chris@10: T24 = KP866025403 * (Ts - Tn); Chris@10: Tt = Tn + Ts; Chris@10: T1a = FNMS(KP500000000, Tt, Ti); Chris@10: T1T = T1b + T1c; Chris@10: T25 = FNMS(KP500000000, T1T, T1S); Chris@10: } Chris@10: { Chris@10: E TK, T1i, TF, T1h; Chris@10: { Chris@10: E Tx, Tz, Tw, Ty; Chris@10: Tx = ri[WS(rs, 3)]; Chris@10: Tz = ii[WS(rs, 3)]; Chris@10: Tw = W[4]; Chris@10: Ty = W[5]; Chris@10: TA = FMA(Tw, Tx, Ty * Tz); Chris@10: T1z = FNMS(Ty, Tx, Tw * Tz); Chris@10: } Chris@10: { Chris@10: E TH, TJ, TG, TI; Chris@10: TH = ri[WS(rs, 11)]; Chris@10: TJ = ii[WS(rs, 11)]; Chris@10: TG = W[20]; Chris@10: TI = W[21]; Chris@10: TK = FMA(TG, TH, TI * TJ); Chris@10: T1i = FNMS(TI, TH, TG * TJ); Chris@10: } Chris@10: { Chris@10: E TC, TE, TB, TD; Chris@10: TC = ri[WS(rs, 7)]; Chris@10: TE = ii[WS(rs, 7)]; Chris@10: TB = W[12]; Chris@10: TD = W[13]; Chris@10: TF = FMA(TB, TC, TD * TE); Chris@10: T1h = FNMS(TD, TC, TB * TE); Chris@10: } Chris@10: T1j = KP866025403 * (T1h - T1i); Chris@10: T1y = KP866025403 * (TK - TF); Chris@10: TL = TF + TK; Chris@10: T1g = FNMS(KP500000000, TL, TA); Chris@10: T1A = T1h + T1i; Chris@10: T1B = FNMS(KP500000000, T1A, T1z); Chris@10: } Chris@10: { Chris@10: E Tv, T1N, T1Y, T20, T14, T1Z, T1Q, T1R; Chris@10: { Chris@10: E Td, Tu, T1U, T1X; Chris@10: Td = T1 + Tc; Chris@10: Tu = Ti + Tt; Chris@10: Tv = Td + Tu; Chris@10: T1N = Td - Tu; Chris@10: T1U = T1S + T1T; Chris@10: T1X = T1V + T1W; Chris@10: T1Y = T1U + T1X; Chris@10: T20 = T1X - T1U; Chris@10: } Chris@10: { Chris@10: E TM, T13, T1O, T1P; Chris@10: TM = TA + TL; Chris@10: T13 = TR + T12; Chris@10: T14 = TM + T13; Chris@10: T1Z = TM - T13; Chris@10: T1O = T1z + T1A; Chris@10: T1P = T1E + T1F; Chris@10: T1Q = T1O - T1P; Chris@10: T1R = T1O + T1P; Chris@10: } Chris@10: ri[WS(rs, 6)] = Tv - T14; Chris@10: ii[WS(rs, 6)] = T1Y - T1R; Chris@10: ri[0] = Tv + T14; Chris@10: ii[0] = T1R + T1Y; Chris@10: ri[WS(rs, 3)] = T1N - T1Q; Chris@10: ii[WS(rs, 3)] = T1Z + T20; Chris@10: ri[WS(rs, 9)] = T1N + T1Q; Chris@10: ii[WS(rs, 9)] = T20 - T1Z; Chris@10: } Chris@10: { Chris@10: E T1t, T1x, T27, T2a, T1w, T28, T1I, T29; Chris@10: { Chris@10: E T1r, T1s, T23, T26; Chris@10: T1r = T15 + T18; Chris@10: T1s = T1a + T1d; Chris@10: T1t = T1r + T1s; Chris@10: T1x = T1r - T1s; Chris@10: T23 = T21 + T22; Chris@10: T26 = T24 + T25; Chris@10: T27 = T23 - T26; Chris@10: T2a = T26 + T23; Chris@10: } Chris@10: { Chris@10: E T1u, T1v, T1C, T1H; Chris@10: T1u = T1g + T1j; Chris@10: T1v = T1l + T1o; Chris@10: T1w = T1u + T1v; Chris@10: T28 = T1u - T1v; Chris@10: T1C = T1y + T1B; Chris@10: T1H = T1D + T1G; Chris@10: T1I = T1C - T1H; Chris@10: T29 = T1C + T1H; Chris@10: } Chris@10: ri[WS(rs, 10)] = T1t - T1w; Chris@10: ii[WS(rs, 10)] = T2a - T29; Chris@10: ri[WS(rs, 4)] = T1t + T1w; Chris@10: ii[WS(rs, 4)] = T29 + T2a; Chris@10: ri[WS(rs, 7)] = T1x - T1I; Chris@10: ii[WS(rs, 7)] = T28 + T27; Chris@10: ri[WS(rs, 1)] = T1x + T1I; Chris@10: ii[WS(rs, 1)] = T27 - T28; Chris@10: } Chris@10: { Chris@10: E T1f, T1J, T2d, T2f, T1q, T2g, T1M, T2e; Chris@10: { Chris@10: E T19, T1e, T2b, T2c; Chris@10: T19 = T15 - T18; Chris@10: T1e = T1a - T1d; Chris@10: T1f = T19 + T1e; Chris@10: T1J = T19 - T1e; Chris@10: T2b = T25 - T24; Chris@10: T2c = T22 - T21; Chris@10: T2d = T2b + T2c; Chris@10: T2f = T2c - T2b; Chris@10: } Chris@10: { Chris@10: E T1k, T1p, T1K, T1L; Chris@10: T1k = T1g - T1j; Chris@10: T1p = T1l - T1o; Chris@10: T1q = T1k + T1p; Chris@10: T2g = T1k - T1p; Chris@10: T1K = T1B - T1y; Chris@10: T1L = T1G - T1D; Chris@10: T1M = T1K - T1L; Chris@10: T2e = T1K + T1L; Chris@10: } Chris@10: ri[WS(rs, 2)] = T1f - T1q; Chris@10: ii[WS(rs, 2)] = T2d - T2e; Chris@10: ri[WS(rs, 8)] = T1f + T1q; Chris@10: ii[WS(rs, 8)] = T2e + T2d; Chris@10: ri[WS(rs, 11)] = T1J - T1M; Chris@10: ii[WS(rs, 11)] = T2g + T2f; Chris@10: ri[WS(rs, 5)] = T1J + T1M; Chris@10: ii[WS(rs, 5)] = T2f - T2g; Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const tw_instr twinstr[] = { Chris@10: {TW_FULL, 0, 12}, Chris@10: {TW_NEXT, 1, 0} Chris@10: }; Chris@10: Chris@10: static const ct_desc desc = { 12, "t1_12", twinstr, &GENUS, {88, 30, 30, 0}, 0, 0, 0 }; Chris@10: Chris@10: void X(codelet_t1_12) (planner *p) { Chris@10: X(kdft_dit_register) (p, t1_12, &desc); Chris@10: } Chris@10: #endif /* HAVE_FMA */