/* * Copyright (c) 2003, 2007-8 Matteo Frigo * Copyright (c) 2003, 2007-8 Massachusetts Institute of Technology * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by * the Free Software Foundation; either version 2 of the License, or * (at your option) any later version. * * This program is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * GNU General Public License for more details. * * You should have received a copy of the GNU General Public License * along with this program; if not, write to the Free Software * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA * */ /* This file was automatically generated --- DO NOT EDIT */ /* Generated on Sun Jul 12 06:43:48 EDT 2009 */ #include "codelet-rdft.h" #ifdef HAVE_FMA /* Generated by: ../../../genfft/gen_hc2hc -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 12 -dit -name hf_12 -include hf.h */ /* * This function contains 118 FP additions, 68 FP multiplications, * (or, 72 additions, 22 multiplications, 46 fused multiply/add), * 84 stack variables, 2 constants, and 48 memory accesses */ #include "hf.h" static void hf_12(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms) { DK(KP866025403, +0.866025403784438646763723170752936183471402627); DK(KP500000000, +0.500000000000000000000000000000000000000000000); INT m; for (m = mb, W = W + ((mb - 1) * 22); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 22, MAKE_VOLATILE_STRIDE(rs)) { E T2u, T2n; { E T1, T2i, T2e, Tl, T1Y, T10, T1S, TG, T2f, T1s, T2s, Ty, T1Z, T1H, T21; E T1d, TI, TL, T2h, T1l, T2p, Te, TJ, T1w, TO, TR, TN, TK, TQ; { E TW, TZ, TY, T1X, TX; T1 = cr[0]; T2i = ci[0]; { E Th, Tk, Tg, Tj, T2d, Ti, TV; Th = cr[WS(rs, 6)]; Tk = ci[WS(rs, 6)]; Tg = W[10]; Tj = W[11]; TW = cr[WS(rs, 9)]; TZ = ci[WS(rs, 9)]; T2d = Tg * Tk; Ti = Tg * Th; TV = W[16]; TY = W[17]; T2e = FNMS(Tj, Th, T2d); Tl = FMA(Tj, Tk, Ti); T1X = TV * TZ; TX = TV * TW; } { E Tn, Tq, Tt, T1o, To, Tw, Ts, Tp, Tv; { E TC, TF, TB, TE, T1R, TD, Tm; TC = cr[WS(rs, 3)]; TF = ci[WS(rs, 3)]; T1Y = FNMS(TY, TW, T1X); T10 = FMA(TY, TZ, TX); TB = W[4]; TE = W[5]; Tn = cr[WS(rs, 10)]; Tq = ci[WS(rs, 10)]; T1R = TB * TF; TD = TB * TC; Tm = W[18]; Tt = cr[WS(rs, 2)]; T1S = FNMS(TE, TC, T1R); TG = FMA(TE, TF, TD); T1o = Tm * Tq; To = Tm * Tn; Tw = ci[WS(rs, 2)]; Ts = W[2]; Tp = W[19]; Tv = W[3]; } { E T12, T15, T13, T1D, T18, T1b, T17, T14, T1a; { E T1p, Tr, T1r, Tx, T1q, Tu, T11; T12 = cr[WS(rs, 1)]; T1q = Ts * Tw; Tu = Ts * Tt; T1p = FNMS(Tp, Tn, T1o); Tr = FMA(Tp, Tq, To); T1r = FNMS(Tv, Tt, T1q); Tx = FMA(Tv, Tw, Tu); T15 = ci[WS(rs, 1)]; T11 = W[0]; T2f = T1p + T1r; T1s = T1p - T1r; T2s = Tx - Tr; Ty = Tr + Tx; T13 = T11 * T12; T1D = T11 * T15; } T18 = cr[WS(rs, 5)]; T1b = ci[WS(rs, 5)]; T17 = W[8]; T14 = W[1]; T1a = W[9]; { E T3, T6, T4, T1h, T9, Tc, T8, T5, Tb; { E T1E, T16, T1G, T1c, T1F, T19, T2; T3 = cr[WS(rs, 4)]; T1F = T17 * T1b; T19 = T17 * T18; T1E = FNMS(T14, T12, T1D); T16 = FMA(T14, T15, T13); T1G = FNMS(T1a, T18, T1F); T1c = FMA(T1a, T1b, T19); T6 = ci[WS(rs, 4)]; T2 = W[6]; T1Z = T1E + T1G; T1H = T1E - T1G; T21 = T1c - T16; T1d = T16 + T1c; T4 = T2 * T3; T1h = T2 * T6; } T9 = cr[WS(rs, 8)]; Tc = ci[WS(rs, 8)]; T8 = W[14]; T5 = W[7]; Tb = W[15]; { E T1i, T7, T1k, Td, T1j, Ta, TH; TI = cr[WS(rs, 7)]; T1j = T8 * Tc; Ta = T8 * T9; T1i = FNMS(T5, T3, T1h); T7 = FMA(T5, T6, T4); T1k = FNMS(Tb, T9, T1j); Td = FMA(Tb, Tc, Ta); TL = ci[WS(rs, 7)]; TH = W[12]; T2h = T1i + T1k; T1l = T1i - T1k; T2p = Td - T7; Te = T7 + Td; TJ = TH * TI; T1w = TH * TL; } TO = cr[WS(rs, 11)]; TR = ci[WS(rs, 11)]; TN = W[20]; TK = W[13]; TQ = W[21]; } } } } { E T1g, T1n, T2r, T1A, T1V, T28, TA, T2o, T1v, T1C, T1U, T29, T2m, T2k, T2l; E T1f, T2a, T20; { E T2g, T1T, TT, T2j, TU, T1e; { E Tf, T1x, TM, T1z, TS, Tz, T1y, TP; T1g = FNMS(KP500000000, Te, T1); Tf = T1 + Te; T1y = TN * TR; TP = TN * TO; T1x = FNMS(TK, TI, T1w); TM = FMA(TK, TL, TJ); T1z = FNMS(TQ, TO, T1y); TS = FMA(TQ, TR, TP); Tz = Tl + Ty; T1n = FNMS(KP500000000, Ty, Tl); T2r = FNMS(KP500000000, T2f, T2e); T2g = T2e + T2f; T1T = T1x + T1z; T1A = T1x - T1z; T1V = TS - TM; TT = TM + TS; T28 = Tf - Tz; TA = Tf + Tz; T2j = T2h + T2i; T2o = FNMS(KP500000000, T2h, T2i); } T1v = FNMS(KP500000000, TT, TG); TU = TG + TT; T1e = T10 + T1d; T1C = FNMS(KP500000000, T1d, T10); T1U = FNMS(KP500000000, T1T, T1S); T29 = T1S + T1T; T2m = T2j - T2g; T2k = T2g + T2j; T2l = TU - T1e; T1f = TU + T1e; T2a = T1Y + T1Z; T20 = FNMS(KP500000000, T1Z, T1Y); } { E T1m, T1K, T2y, T2q, T2z, T2t, T1L, T1t, T1B, T1N, T2c, T2b; ci[WS(rs, 8)] = T2l + T2m; cr[WS(rs, 9)] = T2l - T2m; cr[0] = TA + T1f; ci[WS(rs, 5)] = TA - T1f; T2c = T29 + T2a; T2b = T29 - T2a; T1m = FNMS(KP866025403, T1l, T1g); T1K = FMA(KP866025403, T1l, T1g); ci[WS(rs, 11)] = T2c + T2k; cr[WS(rs, 6)] = T2c - T2k; ci[WS(rs, 2)] = T28 + T2b; cr[WS(rs, 3)] = T28 - T2b; T2y = FMA(KP866025403, T2p, T2o); T2q = FNMS(KP866025403, T2p, T2o); T2z = FMA(KP866025403, T2s, T2r); T2t = FNMS(KP866025403, T2s, T2r); T1L = FMA(KP866025403, T1s, T1n); T1t = FNMS(KP866025403, T1s, T1n); T1B = FNMS(KP866025403, T1A, T1v); T1N = FMA(KP866025403, T1A, T1v); { E T1Q, T23, T27, T2A, T1P, T2x, T24, T1M; { E T1u, T25, T26, T1O, T1I, T2w, T2v, T1W, T22, T2B, T1J, T2C; T1Q = T1m - T1t; T1u = T1m + T1t; T25 = FMA(KP866025403, T1V, T1U); T1W = FNMS(KP866025403, T1V, T1U); T26 = FMA(KP866025403, T21, T20); T22 = FNMS(KP866025403, T21, T20); T1O = FMA(KP866025403, T1H, T1C); T1I = FNMS(KP866025403, T1H, T1C); T2w = T2t + T2q; T2u = T2q - T2t; T23 = T1W - T22; T2v = T1W + T22; T2B = T25 + T26; T27 = T25 - T26; T2n = T1I - T1B; T1J = T1B + T1I; T2C = T2z + T2y; T2A = T2y - T2z; ci[WS(rs, 9)] = T2w - T2v; cr[WS(rs, 8)] = -(T2v + T2w); ci[WS(rs, 3)] = T1u + T1J; cr[WS(rs, 2)] = T1u - T1J; cr[WS(rs, 10)] = T2B - T2C; ci[WS(rs, 7)] = T2B + T2C; T1P = T1N + T1O; T2x = T1O - T1N; } T24 = T1K - T1L; T1M = T1K + T1L; ci[WS(rs, 10)] = T2x + T2A; cr[WS(rs, 7)] = T2x - T2A; cr[WS(rs, 4)] = T1M + T1P; ci[WS(rs, 1)] = T1M - T1P; cr[WS(rs, 1)] = T24 + T27; ci[WS(rs, 4)] = T24 - T27; cr[WS(rs, 5)] = T1Q + T23; ci[0] = T1Q - T23; } } } } ci[WS(rs, 6)] = T2n + T2u; cr[WS(rs, 11)] = T2n - T2u; } } static const tw_instr twinstr[] = { {TW_FULL, 1, 12}, {TW_NEXT, 1, 0} }; static const hc2hc_desc desc = { 12, "hf_12", twinstr, &GENUS, {72, 22, 46, 0} }; void X(codelet_hf_12) (planner *p) { X(khc2hc_register) (p, hf_12, &desc); } #else /* HAVE_FMA */ /* Generated by: ../../../genfft/gen_hc2hc -compact -variables 4 -pipeline-latency 4 -n 12 -dit -name hf_12 -include hf.h */ /* * This function contains 118 FP additions, 60 FP multiplications, * (or, 88 additions, 30 multiplications, 30 fused multiply/add), * 47 stack variables, 2 constants, and 48 memory accesses */ #include "hf.h" static void hf_12(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms) { DK(KP500000000, +0.500000000000000000000000000000000000000000000); DK(KP866025403, +0.866025403784438646763723170752936183471402627); INT m; for (m = mb, W = W + ((mb - 1) * 22); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 22, MAKE_VOLATILE_STRIDE(rs)) { E T1, T1W, T18, T23, Tc, T15, T1V, T22, TR, T1E, T1o, T1D, T12, T1l, T1F; E T1G, Ti, T1S, T1d, T26, Tt, T1a, T1T, T25, TA, T1y, T1j, T1B, TL, T1g; E T1z, T1A; { E T6, T16, Tb, T17; T1 = cr[0]; T1W = ci[0]; { E T3, T5, T2, T4; T3 = cr[WS(rs, 4)]; T5 = ci[WS(rs, 4)]; T2 = W[6]; T4 = W[7]; T6 = FMA(T2, T3, T4 * T5); T16 = FNMS(T4, T3, T2 * T5); } { E T8, Ta, T7, T9; T8 = cr[WS(rs, 8)]; Ta = ci[WS(rs, 8)]; T7 = W[14]; T9 = W[15]; Tb = FMA(T7, T8, T9 * Ta); T17 = FNMS(T9, T8, T7 * Ta); } T18 = KP866025403 * (T16 - T17); T23 = KP866025403 * (Tb - T6); Tc = T6 + Tb; T15 = FNMS(KP500000000, Tc, T1); T1V = T16 + T17; T22 = FNMS(KP500000000, T1V, T1W); } { E T11, T1n, TW, T1m; { E TO, TQ, TN, TP; TO = cr[WS(rs, 9)]; TQ = ci[WS(rs, 9)]; TN = W[16]; TP = W[17]; TR = FMA(TN, TO, TP * TQ); T1E = FNMS(TP, TO, TN * TQ); } { E TY, T10, TX, TZ; TY = cr[WS(rs, 5)]; T10 = ci[WS(rs, 5)]; TX = W[8]; TZ = W[9]; T11 = FMA(TX, TY, TZ * T10); T1n = FNMS(TZ, TY, TX * T10); } { E TT, TV, TS, TU; TT = cr[WS(rs, 1)]; TV = ci[WS(rs, 1)]; TS = W[0]; TU = W[1]; TW = FMA(TS, TT, TU * TV); T1m = FNMS(TU, TT, TS * TV); } T1o = KP866025403 * (T1m - T1n); T1D = KP866025403 * (T11 - TW); T12 = TW + T11; T1l = FNMS(KP500000000, T12, TR); T1F = T1m + T1n; T1G = FNMS(KP500000000, T1F, T1E); } { E Ts, T1c, Tn, T1b; { E Tf, Th, Te, Tg; Tf = cr[WS(rs, 6)]; Th = ci[WS(rs, 6)]; Te = W[10]; Tg = W[11]; Ti = FMA(Te, Tf, Tg * Th); T1S = FNMS(Tg, Tf, Te * Th); } { E Tp, Tr, To, Tq; Tp = cr[WS(rs, 2)]; Tr = ci[WS(rs, 2)]; To = W[2]; Tq = W[3]; Ts = FMA(To, Tp, Tq * Tr); T1c = FNMS(Tq, Tp, To * Tr); } { E Tk, Tm, Tj, Tl; Tk = cr[WS(rs, 10)]; Tm = ci[WS(rs, 10)]; Tj = W[18]; Tl = W[19]; Tn = FMA(Tj, Tk, Tl * Tm); T1b = FNMS(Tl, Tk, Tj * Tm); } T1d = KP866025403 * (T1b - T1c); T26 = KP866025403 * (Ts - Tn); Tt = Tn + Ts; T1a = FNMS(KP500000000, Tt, Ti); T1T = T1b + T1c; T25 = FNMS(KP500000000, T1T, T1S); } { E TK, T1i, TF, T1h; { E Tx, Tz, Tw, Ty; Tx = cr[WS(rs, 3)]; Tz = ci[WS(rs, 3)]; Tw = W[4]; Ty = W[5]; TA = FMA(Tw, Tx, Ty * Tz); T1y = FNMS(Ty, Tx, Tw * Tz); } { E TH, TJ, TG, TI; TH = cr[WS(rs, 11)]; TJ = ci[WS(rs, 11)]; TG = W[20]; TI = W[21]; TK = FMA(TG, TH, TI * TJ); T1i = FNMS(TI, TH, TG * TJ); } { E TC, TE, TB, TD; TC = cr[WS(rs, 7)]; TE = ci[WS(rs, 7)]; TB = W[12]; TD = W[13]; TF = FMA(TB, TC, TD * TE); T1h = FNMS(TD, TC, TB * TE); } T1j = KP866025403 * (T1h - T1i); T1B = KP866025403 * (TK - TF); TL = TF + TK; T1g = FNMS(KP500000000, TL, TA); T1z = T1h + T1i; T1A = FNMS(KP500000000, T1z, T1y); } { E Tv, T1N, T1Y, T20, T14, T1Z, T1Q, T1R; { E Td, Tu, T1U, T1X; Td = T1 + Tc; Tu = Ti + Tt; Tv = Td + Tu; T1N = Td - Tu; T1U = T1S + T1T; T1X = T1V + T1W; T1Y = T1U + T1X; T20 = T1X - T1U; } { E TM, T13, T1O, T1P; TM = TA + TL; T13 = TR + T12; T14 = TM + T13; T1Z = TM - T13; T1O = T1y + T1z; T1P = T1E + T1F; T1Q = T1O - T1P; T1R = T1O + T1P; } ci[WS(rs, 5)] = Tv - T14; cr[WS(rs, 9)] = T1Z - T20; ci[WS(rs, 8)] = T1Z + T20; cr[0] = Tv + T14; cr[WS(rs, 3)] = T1N - T1Q; cr[WS(rs, 6)] = T1R - T1Y; ci[WS(rs, 11)] = T1R + T1Y; ci[WS(rs, 2)] = T1N + T1Q; } { E T1f, T1x, T28, T2a, T1q, T21, T1I, T29; { E T19, T1e, T24, T27; T19 = T15 - T18; T1e = T1a - T1d; T1f = T19 + T1e; T1x = T19 - T1e; T24 = T22 - T23; T27 = T25 - T26; T28 = T24 - T27; T2a = T27 + T24; } { E T1k, T1p, T1C, T1H; T1k = T1g - T1j; T1p = T1l - T1o; T1q = T1k + T1p; T21 = T1p - T1k; T1C = T1A - T1B; T1H = T1D - T1G; T1I = T1C + T1H; T29 = T1H - T1C; } cr[WS(rs, 2)] = T1f - T1q; cr[WS(rs, 8)] = T29 - T2a; ci[WS(rs, 9)] = T29 + T2a; ci[WS(rs, 3)] = T1f + T1q; ci[0] = T1x - T1I; cr[WS(rs, 11)] = T21 - T28; ci[WS(rs, 6)] = T21 + T28; cr[WS(rs, 5)] = T1x + T1I; } { E T1t, T1J, T2e, T2g, T1w, T2b, T1M, T2f; { E T1r, T1s, T2c, T2d; T1r = T15 + T18; T1s = T1a + T1d; T1t = T1r + T1s; T1J = T1r - T1s; T2c = T23 + T22; T2d = T26 + T25; T2e = T2c - T2d; T2g = T2d + T2c; } { E T1u, T1v, T1K, T1L; T1u = T1g + T1j; T1v = T1l + T1o; T1w = T1u + T1v; T2b = T1v - T1u; T1K = T1B + T1A; T1L = T1D + T1G; T1M = T1K - T1L; T2f = T1K + T1L; } ci[WS(rs, 1)] = T1t - T1w; cr[WS(rs, 1)] = T1J + T1M; cr[WS(rs, 4)] = T1t + T1w; ci[WS(rs, 4)] = T1J - T1M; cr[WS(rs, 7)] = T2b - T2e; ci[WS(rs, 7)] = T2f + T2g; ci[WS(rs, 10)] = T2b + T2e; cr[WS(rs, 10)] = T2f - T2g; } } } static const tw_instr twinstr[] = { {TW_FULL, 1, 12}, {TW_NEXT, 1, 0} }; static const hc2hc_desc desc = { 12, "hf_12", twinstr, &GENUS, {88, 30, 30, 0} }; void X(codelet_hf_12) (planner *p) { X(khc2hc_register) (p, hf_12, &desc); } #endif /* HAVE_FMA */