Mercurial > hg > sv-dependency-builds
diff src/fftw-3.3.3/rdft/scalar/r2cf/hc2cfdft_4.c @ 10:37bf6b4a2645
Add FFTW3
| author | Chris Cannam |
|---|---|
| date | Wed, 20 Mar 2013 15:35:50 +0000 |
| parents | |
| children |
line wrap: on
line diff
--- /dev/null Thu Jan 01 00:00:00 1970 +0000 +++ b/src/fftw-3.3.3/rdft/scalar/r2cf/hc2cfdft_4.c Wed Mar 20 15:35:50 2013 +0000 @@ -0,0 +1,217 @@ +/* + * Copyright (c) 2003, 2007-11 Matteo Frigo + * Copyright (c) 2003, 2007-11 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., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + * + */ + +/* This file was automatically generated --- DO NOT EDIT */ +/* Generated on Sun Nov 25 07:40:44 EST 2012 */ + +#include "codelet-rdft.h" + +#ifdef HAVE_FMA + +/* Generated by: ../../../genfft/gen_hc2cdft.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 4 -dit -name hc2cfdft_4 -include hc2cf.h */ + +/* + * This function contains 30 FP additions, 20 FP multiplications, + * (or, 24 additions, 14 multiplications, 6 fused multiply/add), + * 32 stack variables, 1 constants, and 16 memory accesses + */ +#include "hc2cf.h" + +static void hc2cfdft_4(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms) +{ + DK(KP500000000, +0.500000000000000000000000000000000000000000000); + { + INT m; + for (m = mb, W = W + ((mb - 1) * 6); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 6, MAKE_VOLATILE_STRIDE(16, rs)) { + E Td, Tu, Tr, T4, Tm, To, T9, T5, TA, Tp, Tv, TD, T6, Tq; + { + E Tk, Tl, Tf, TC, Tj, T7, T8, T1, Tn, Tb, Tc; + Tb = Ip[0]; + Tc = Im[0]; + { + E Ti, Tg, Th, T2, T3; + Tg = Rm[0]; + Th = Rp[0]; + Tk = W[1]; + Tl = Tb + Tc; + Td = Tb - Tc; + Tu = Th + Tg; + Ti = Tg - Th; + Tf = W[0]; + T2 = Ip[WS(rs, 1)]; + T3 = Im[WS(rs, 1)]; + TC = Tk * Ti; + Tj = Tf * Ti; + T7 = Rp[WS(rs, 1)]; + Tr = T2 + T3; + T4 = T2 - T3; + T8 = Rm[WS(rs, 1)]; + T1 = W[2]; + Tn = W[4]; + } + Tm = FNMS(Tk, Tl, Tj); + To = T7 - T8; + T9 = T7 + T8; + T5 = T1 * T4; + TA = Tn * Tr; + Tp = Tn * To; + Tv = T1 * T9; + TD = FMA(Tf, Tl, TC); + T6 = W[3]; + Tq = W[5]; + } + { + E Tw, Ta, TB, Ts; + Tw = FMA(T6, T4, Tv); + Ta = FNMS(T6, T9, T5); + TB = FNMS(Tq, To, TA); + Ts = FMA(Tq, Tr, Tp); + { + E TF, Tx, Te, Tz; + TF = Tu + Tw; + Tx = Tu - Tw; + Te = Ta + Td; + Tz = Td - Ta; + { + E TG, TE, Tt, Ty; + TG = TB + TD; + TE = TB - TD; + Tt = Tm - Ts; + Ty = Ts + Tm; + Im[0] = KP500000000 * (TE - Tz); + Ip[WS(rs, 1)] = KP500000000 * (Tz + TE); + Rp[0] = KP500000000 * (TF + TG); + Rm[WS(rs, 1)] = KP500000000 * (TF - TG); + Rp[WS(rs, 1)] = KP500000000 * (Tx + Ty); + Rm[0] = KP500000000 * (Tx - Ty); + Im[WS(rs, 1)] = KP500000000 * (Tt - Te); + Ip[0] = KP500000000 * (Te + Tt); + } + } + } + } + } +} + +static const tw_instr twinstr[] = { + {TW_FULL, 1, 4}, + {TW_NEXT, 1, 0} +}; + +static const hc2c_desc desc = { 4, "hc2cfdft_4", twinstr, &GENUS, {24, 14, 6, 0} }; + +void X(codelet_hc2cfdft_4) (planner *p) { + X(khc2c_register) (p, hc2cfdft_4, &desc, HC2C_VIA_DFT); +} +#else /* HAVE_FMA */ + +/* Generated by: ../../../genfft/gen_hc2cdft.native -compact -variables 4 -pipeline-latency 4 -n 4 -dit -name hc2cfdft_4 -include hc2cf.h */ + +/* + * This function contains 30 FP additions, 20 FP multiplications, + * (or, 24 additions, 14 multiplications, 6 fused multiply/add), + * 18 stack variables, 1 constants, and 16 memory accesses + */ +#include "hc2cf.h" + +static void hc2cfdft_4(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms) +{ + DK(KP500000000, +0.500000000000000000000000000000000000000000000); + { + INT m; + for (m = mb, W = W + ((mb - 1) * 6); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 6, MAKE_VOLATILE_STRIDE(16, rs)) { + E Tc, Tr, Tk, Tx, T9, Ts, Tp, Tw; + { + E Ta, Tb, Tj, Tf, Tg, Th, Te, Ti; + Ta = Ip[0]; + Tb = Im[0]; + Tj = Ta + Tb; + Tf = Rm[0]; + Tg = Rp[0]; + Th = Tf - Tg; + Tc = Ta - Tb; + Tr = Tg + Tf; + Te = W[0]; + Ti = W[1]; + Tk = FNMS(Ti, Tj, Te * Th); + Tx = FMA(Ti, Th, Te * Tj); + } + { + E T4, To, T8, Tm; + { + E T2, T3, T6, T7; + T2 = Ip[WS(rs, 1)]; + T3 = Im[WS(rs, 1)]; + T4 = T2 - T3; + To = T2 + T3; + T6 = Rp[WS(rs, 1)]; + T7 = Rm[WS(rs, 1)]; + T8 = T6 + T7; + Tm = T6 - T7; + } + { + E T1, T5, Tl, Tn; + T1 = W[2]; + T5 = W[3]; + T9 = FNMS(T5, T8, T1 * T4); + Ts = FMA(T1, T8, T5 * T4); + Tl = W[4]; + Tn = W[5]; + Tp = FMA(Tl, Tm, Tn * To); + Tw = FNMS(Tn, Tm, Tl * To); + } + } + { + E Td, Tq, Tz, TA; + Td = T9 + Tc; + Tq = Tk - Tp; + Ip[0] = KP500000000 * (Td + Tq); + Im[WS(rs, 1)] = KP500000000 * (Tq - Td); + Tz = Tr + Ts; + TA = Tw + Tx; + Rm[WS(rs, 1)] = KP500000000 * (Tz - TA); + Rp[0] = KP500000000 * (Tz + TA); + } + { + E Tt, Tu, Tv, Ty; + Tt = Tr - Ts; + Tu = Tp + Tk; + Rm[0] = KP500000000 * (Tt - Tu); + Rp[WS(rs, 1)] = KP500000000 * (Tt + Tu); + Tv = Tc - T9; + Ty = Tw - Tx; + Ip[WS(rs, 1)] = KP500000000 * (Tv + Ty); + Im[0] = KP500000000 * (Ty - Tv); + } + } + } +} + +static const tw_instr twinstr[] = { + {TW_FULL, 1, 4}, + {TW_NEXT, 1, 0} +}; + +static const hc2c_desc desc = { 4, "hc2cfdft_4", twinstr, &GENUS, {24, 14, 6, 0} }; + +void X(codelet_hc2cfdft_4) (planner *p) { + X(khc2c_register) (p, hc2cfdft_4, &desc, HC2C_VIA_DFT); +} +#endif /* HAVE_FMA */