/* * 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 10 -dit -name hf_10 -include hf.h */ /* * This function contains 102 FP additions, 72 FP multiplications, * (or, 48 additions, 18 multiplications, 54 fused multiply/add), * 72 stack variables, 4 constants, and 40 memory accesses */ #include "hf.h" static void hf_10(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms) { DK(KP951056516, +0.951056516295153572116439333379382143405698634); DK(KP559016994, +0.559016994374947424102293417182819058860154590); DK(KP250000000, +0.250000000000000000000000000000000000000000000); DK(KP618033988, +0.618033988749894848204586834365638117720309180); INT m; for (m = mb, W = W + ((mb - 1) * 18); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 18, MAKE_VOLATILE_STRIDE(rs)) { E T29, T2d, T2c, T2e; { E T23, T1U, T8, T12, T1y, T1P, T25, T1H, T2b, T18, T10, T1Y, T1I, Tl, T13; E T1J, Ty, T14, T1n, T1O, T24, T1K; { E T1, T1R, T3, T6, T2, T5; T1 = cr[0]; T1R = ci[0]; T3 = cr[WS(rs, 5)]; T6 = ci[WS(rs, 5)]; T2 = W[8]; T5 = W[9]; { E T1p, TY, T1x, T1F, TM, T16, T1r, TS; { E TF, T1w, TO, TR, T1u, TL, TN, TQ, T1q, TP; { E TU, TX, TT, TW; { E TB, TE, T1S, T4, TA, TD; TB = cr[WS(rs, 4)]; TE = ci[WS(rs, 4)]; T1S = T2 * T6; T4 = T2 * T3; TA = W[6]; TD = W[7]; { E T1T, T7, T1v, TC; T1T = FNMS(T5, T3, T1S); T7 = FMA(T5, T6, T4); T1v = TA * TE; TC = TA * TB; T23 = T1T + T1R; T1U = T1R - T1T; T8 = T1 - T7; T12 = T1 + T7; TF = FMA(TD, TE, TC); T1w = FNMS(TD, TB, T1v); } } TU = cr[WS(rs, 1)]; TX = ci[WS(rs, 1)]; TT = W[0]; TW = W[1]; { E TH, TK, TJ, T1t, TI, T1o, TV, TG; TH = cr[WS(rs, 9)]; TK = ci[WS(rs, 9)]; T1o = TT * TX; TV = TT * TU; TG = W[16]; TJ = W[17]; T1p = FNMS(TW, TU, T1o); TY = FMA(TW, TX, TV); T1t = TG * TK; TI = TG * TH; TO = cr[WS(rs, 6)]; TR = ci[WS(rs, 6)]; T1u = FNMS(TJ, TH, T1t); TL = FMA(TJ, TK, TI); TN = W[10]; TQ = W[11]; } } T1x = T1u - T1w; T1F = T1w + T1u; TM = TF - TL; T16 = TF + TL; T1q = TN * TR; TP = TN * TO; T1r = FNMS(TQ, TO, T1q); TS = FMA(TQ, TR, TP); } { E T1l, Te, T1e, Tx, Tn, Tq, Tp, T1j, Tk, T1f, To; { E Tt, Tw, Tv, T1d, Tu; { E Ta, Td, T9, Tc, T1k, Tb, Ts; Ta = cr[WS(rs, 2)]; Td = ci[WS(rs, 2)]; { E T1G, T1s, TZ, T17; T1G = T1r + T1p; T1s = T1p - T1r; TZ = TS - TY; T17 = TS + TY; T1y = T1s - T1x; T1P = T1x + T1s; T25 = T1F + T1G; T1H = T1F - T1G; T2b = T16 - T17; T18 = T16 + T17; T10 = TM + TZ; T1Y = TZ - TM; T9 = W[2]; } Tc = W[3]; Tt = cr[WS(rs, 3)]; Tw = ci[WS(rs, 3)]; T1k = T9 * Td; Tb = T9 * Ta; Ts = W[4]; Tv = W[5]; T1l = FNMS(Tc, Ta, T1k); Te = FMA(Tc, Td, Tb); T1d = Ts * Tw; Tu = Ts * Tt; } { E Tg, Tj, Tf, Ti, T1i, Th, Tm; Tg = cr[WS(rs, 7)]; Tj = ci[WS(rs, 7)]; T1e = FNMS(Tv, Tt, T1d); Tx = FMA(Tv, Tw, Tu); Tf = W[12]; Ti = W[13]; Tn = cr[WS(rs, 8)]; Tq = ci[WS(rs, 8)]; T1i = Tf * Tj; Th = Tf * Tg; Tm = W[14]; Tp = W[15]; T1j = FNMS(Ti, Tg, T1i); Tk = FMA(Ti, Tj, Th); T1f = Tm * Tq; To = Tm * Tn; } } { E T1m, T1g, Tr, T1h; T1m = T1j - T1l; T1I = T1l + T1j; Tl = Te - Tk; T13 = Te + Tk; T1g = FNMS(Tp, Tn, T1f); Tr = FMA(Tp, Tq, To); T1J = T1g + T1e; T1h = T1e - T1g; Ty = Tr - Tx; T14 = Tr + Tx; T1n = T1h - T1m; T1O = T1m + T1h; } } } } T24 = T1I + T1J; T1K = T1I - T1J; { E T2a, T15, Tz, T1Z; T2a = T13 - T14; T15 = T13 + T14; Tz = Tl + Ty; T1Z = Ty - Tl; { E T1L, T1N, T1E, T1M; { E T19, T1D, T1C, T11, T1b; T19 = T15 + T18; T1D = T15 - T18; T11 = Tz + T10; T1b = Tz - T10; { E T1B, T1z, T1a, T1A, T1c; T1B = FNMS(KP618033988, T1n, T1y); T1z = FMA(KP618033988, T1y, T1n); ci[WS(rs, 4)] = T8 + T11; T1a = FNMS(KP250000000, T11, T8); T1A = FNMS(KP559016994, T1b, T1a); T1c = FMA(KP559016994, T1b, T1a); T1C = FNMS(KP250000000, T19, T12); T1L = FNMS(KP618033988, T1K, T1H); T1N = FMA(KP618033988, T1H, T1K); cr[WS(rs, 1)] = FMA(KP951056516, T1z, T1c); ci[0] = FNMS(KP951056516, T1z, T1c); cr[WS(rs, 3)] = FMA(KP951056516, T1B, T1A); ci[WS(rs, 2)] = FNMS(KP951056516, T1B, T1A); } cr[0] = T12 + T19; T1E = FNMS(KP559016994, T1D, T1C); T1M = FMA(KP559016994, T1D, T1C); } { E T1X, T21, T20, T22, T1Q, T1W, T1V, T26, T28, T27; T1Q = T1O + T1P; T1W = T1P - T1O; ci[WS(rs, 3)] = FMA(KP951056516, T1N, T1M); cr[WS(rs, 4)] = FNMS(KP951056516, T1N, T1M); ci[WS(rs, 1)] = FMA(KP951056516, T1L, T1E); cr[WS(rs, 2)] = FNMS(KP951056516, T1L, T1E); T1V = FMA(KP250000000, T1Q, T1U); cr[WS(rs, 5)] = T1Q - T1U; T1X = FNMS(KP559016994, T1W, T1V); T21 = FMA(KP559016994, T1W, T1V); T20 = FNMS(KP618033988, T1Z, T1Y); T22 = FMA(KP618033988, T1Y, T1Z); T26 = T24 + T25; T28 = T24 - T25; ci[WS(rs, 8)] = FMA(KP951056516, T22, T21); cr[WS(rs, 9)] = FMS(KP951056516, T22, T21); ci[WS(rs, 6)] = FMA(KP951056516, T20, T1X); cr[WS(rs, 7)] = FMS(KP951056516, T20, T1X); T27 = FNMS(KP250000000, T26, T23); ci[WS(rs, 9)] = T26 + T23; T29 = FMA(KP559016994, T28, T27); T2d = FNMS(KP559016994, T28, T27); T2c = FMA(KP618033988, T2b, T2a); T2e = FNMS(KP618033988, T2a, T2b); } } } } ci[WS(rs, 7)] = FMA(KP951056516, T2e, T2d); cr[WS(rs, 8)] = FMS(KP951056516, T2e, T2d); ci[WS(rs, 5)] = FMA(KP951056516, T2c, T29); cr[WS(rs, 6)] = FMS(KP951056516, T2c, T29); } } static const tw_instr twinstr[] = { {TW_FULL, 1, 10}, {TW_NEXT, 1, 0} }; static const hc2hc_desc desc = { 10, "hf_10", twinstr, &GENUS, {48, 18, 54, 0} }; void X(codelet_hf_10) (planner *p) { X(khc2hc_register) (p, hf_10, &desc); } #else /* HAVE_FMA */ /* Generated by: ../../../genfft/gen_hc2hc -compact -variables 4 -pipeline-latency 4 -n 10 -dit -name hf_10 -include hf.h */ /* * This function contains 102 FP additions, 60 FP multiplications, * (or, 72 additions, 30 multiplications, 30 fused multiply/add), * 45 stack variables, 4 constants, and 40 memory accesses */ #include "hf.h" static void hf_10(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms) { DK(KP587785252, +0.587785252292473129168705954639072768597652438); DK(KP951056516, +0.951056516295153572116439333379382143405698634); DK(KP250000000, +0.250000000000000000000000000000000000000000000); DK(KP559016994, +0.559016994374947424102293417182819058860154590); INT m; for (m = mb, W = W + ((mb - 1) * 18); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 18, MAKE_VOLATILE_STRIDE(rs)) { E T7, T1R, TT, T1C, TF, TQ, TR, T1o, T1p, T1P, TX, TY, TZ, T1d, T1g; E T1x, Ti, Tt, Tu, T1r, T1s, T1O, TU, TV, TW, T16, T19, T1y; { E T1, T1A, T6, T1B; T1 = cr[0]; T1A = ci[0]; { E T3, T5, T2, T4; T3 = cr[WS(rs, 5)]; T5 = ci[WS(rs, 5)]; T2 = W[8]; T4 = W[9]; T6 = FMA(T2, T3, T4 * T5); T1B = FNMS(T4, T3, T2 * T5); } T7 = T1 - T6; T1R = T1B + T1A; TT = T1 + T6; T1C = T1A - T1B; } { E Tz, T1b, TP, T1e, TE, T1c, TK, T1f; { E Tw, Ty, Tv, Tx; Tw = cr[WS(rs, 4)]; Ty = ci[WS(rs, 4)]; Tv = W[6]; Tx = W[7]; Tz = FMA(Tv, Tw, Tx * Ty); T1b = FNMS(Tx, Tw, Tv * Ty); } { E TM, TO, TL, TN; TM = cr[WS(rs, 1)]; TO = ci[WS(rs, 1)]; TL = W[0]; TN = W[1]; TP = FMA(TL, TM, TN * TO); T1e = FNMS(TN, TM, TL * TO); } { E TB, TD, TA, TC; TB = cr[WS(rs, 9)]; TD = ci[WS(rs, 9)]; TA = W[16]; TC = W[17]; TE = FMA(TA, TB, TC * TD); T1c = FNMS(TC, TB, TA * TD); } { E TH, TJ, TG, TI; TH = cr[WS(rs, 6)]; TJ = ci[WS(rs, 6)]; TG = W[10]; TI = W[11]; TK = FMA(TG, TH, TI * TJ); T1f = FNMS(TI, TH, TG * TJ); } TF = Tz - TE; TQ = TK - TP; TR = TF + TQ; T1o = T1b + T1c; T1p = T1f + T1e; T1P = T1o + T1p; TX = Tz + TE; TY = TK + TP; TZ = TX + TY; T1d = T1b - T1c; T1g = T1e - T1f; T1x = T1g - T1d; } { E Tc, T14, Ts, T18, Th, T15, Tn, T17; { E T9, Tb, T8, Ta; T9 = cr[WS(rs, 2)]; Tb = ci[WS(rs, 2)]; T8 = W[2]; Ta = W[3]; Tc = FMA(T8, T9, Ta * Tb); T14 = FNMS(Ta, T9, T8 * Tb); } { E Tp, Tr, To, Tq; Tp = cr[WS(rs, 3)]; Tr = ci[WS(rs, 3)]; To = W[4]; Tq = W[5]; Ts = FMA(To, Tp, Tq * Tr); T18 = FNMS(Tq, Tp, To * Tr); } { E Te, Tg, Td, Tf; Te = cr[WS(rs, 7)]; Tg = ci[WS(rs, 7)]; Td = W[12]; Tf = W[13]; Th = FMA(Td, Te, Tf * Tg); T15 = FNMS(Tf, Te, Td * Tg); } { E Tk, Tm, Tj, Tl; Tk = cr[WS(rs, 8)]; Tm = ci[WS(rs, 8)]; Tj = W[14]; Tl = W[15]; Tn = FMA(Tj, Tk, Tl * Tm); T17 = FNMS(Tl, Tk, Tj * Tm); } Ti = Tc - Th; Tt = Tn - Ts; Tu = Ti + Tt; T1r = T14 + T15; T1s = T17 + T18; T1O = T1r + T1s; TU = Tc + Th; TV = Tn + Ts; TW = TU + TV; T16 = T14 - T15; T19 = T17 - T18; T1y = T16 + T19; } { E T11, TS, T12, T1i, T1k, T1a, T1h, T1j, T13; T11 = KP559016994 * (Tu - TR); TS = Tu + TR; T12 = FNMS(KP250000000, TS, T7); T1a = T16 - T19; T1h = T1d + T1g; T1i = FMA(KP951056516, T1a, KP587785252 * T1h); T1k = FNMS(KP587785252, T1a, KP951056516 * T1h); ci[WS(rs, 4)] = T7 + TS; T1j = T12 - T11; ci[WS(rs, 2)] = T1j - T1k; cr[WS(rs, 3)] = T1j + T1k; T13 = T11 + T12; ci[0] = T13 - T1i; cr[WS(rs, 1)] = T13 + T1i; } { E T1m, T10, T1l, T1u, T1w, T1q, T1t, T1v, T1n; T1m = KP559016994 * (TW - TZ); T10 = TW + TZ; T1l = FNMS(KP250000000, T10, TT); T1q = T1o - T1p; T1t = T1r - T1s; T1u = FNMS(KP587785252, T1t, KP951056516 * T1q); T1w = FMA(KP951056516, T1t, KP587785252 * T1q); cr[0] = TT + T10; T1v = T1m + T1l; cr[WS(rs, 4)] = T1v - T1w; ci[WS(rs, 3)] = T1v + T1w; T1n = T1l - T1m; cr[WS(rs, 2)] = T1n - T1u; ci[WS(rs, 1)] = T1n + T1u; } { E T1H, T1z, T1G, T1F, T1J, T1D, T1E, T1K, T1I; T1H = KP559016994 * (T1y + T1x); T1z = T1x - T1y; T1G = FMA(KP250000000, T1z, T1C); T1D = Ti - Tt; T1E = TQ - TF; T1F = FMA(KP587785252, T1D, KP951056516 * T1E); T1J = FNMS(KP951056516, T1D, KP587785252 * T1E); cr[WS(rs, 5)] = T1z - T1C; T1K = T1H + T1G; cr[WS(rs, 9)] = T1J - T1K; ci[WS(rs, 8)] = T1J + T1K; T1I = T1G - T1H; cr[WS(rs, 7)] = T1F - T1I; ci[WS(rs, 6)] = T1F + T1I; } { E T1Q, T1S, T1T, T1N, T1V, T1L, T1M, T1W, T1U; T1Q = KP559016994 * (T1O - T1P); T1S = T1O + T1P; T1T = FNMS(KP250000000, T1S, T1R); T1L = TU - TV; T1M = TX - TY; T1N = FMA(KP951056516, T1L, KP587785252 * T1M); T1V = FNMS(KP587785252, T1L, KP951056516 * T1M); ci[WS(rs, 9)] = T1S + T1R; T1W = T1T - T1Q; cr[WS(rs, 8)] = T1V - T1W; ci[WS(rs, 7)] = T1V + T1W; T1U = T1Q + T1T; cr[WS(rs, 6)] = T1N - T1U; ci[WS(rs, 5)] = T1N + T1U; } } } static const tw_instr twinstr[] = { {TW_FULL, 1, 10}, {TW_NEXT, 1, 0} }; static const hc2hc_desc desc = { 10, "hf_10", twinstr, &GENUS, {72, 30, 30, 0} }; void X(codelet_hf_10) (planner *p) { X(khc2hc_register) (p, hf_10, &desc); } #endif /* HAVE_FMA */