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 07:40:15 EST 2012 */ Chris@10: Chris@10: #include "codelet-rdft.h" Chris@10: Chris@10: #ifdef HAVE_FMA Chris@10: Chris@10: /* Generated by: ../../../genfft/gen_r2cf.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 10 -name r2cfII_10 -dft-II -include r2cfII.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 32 FP additions, 18 FP multiplications, Chris@10: * (or, 14 additions, 0 multiplications, 18 fused multiply/add), Chris@10: * 37 stack variables, 4 constants, and 20 memory accesses Chris@10: */ Chris@10: #include "r2cfII.h" Chris@10: Chris@10: static void r2cfII_10(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs) 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 i; Chris@10: for (i = v; i > 0; i = i - 1, R0 = R0 + ivs, R1 = R1 + ivs, Cr = Cr + ovs, Ci = Ci + ovs, MAKE_VOLATILE_STRIDE(40, rs), MAKE_VOLATILE_STRIDE(40, csr), MAKE_VOLATILE_STRIDE(40, csi)) { Chris@10: E Tq, Ti, Tk, Tu, Tw, Tp, Tb, Tj, Tr, Tv; Chris@10: { Chris@10: E T1, To, Ts, Tt, T8, Ta, Te, Tm, Tl, Th, Tn, T9; Chris@10: T1 = R0[0]; Chris@10: To = R1[WS(rs, 2)]; Chris@10: { Chris@10: E T2, T3, T5, T6; Chris@10: T2 = R0[WS(rs, 2)]; Chris@10: T3 = R0[WS(rs, 3)]; Chris@10: T5 = R0[WS(rs, 4)]; Chris@10: T6 = R0[WS(rs, 1)]; Chris@10: { Chris@10: E Tc, T4, T7, Td, Tf, Tg; Chris@10: Tc = R1[0]; Chris@10: Ts = T2 + T3; Chris@10: T4 = T2 - T3; Chris@10: Tt = T5 + T6; Chris@10: T7 = T5 - T6; Chris@10: Td = R1[WS(rs, 4)]; Chris@10: Tf = R1[WS(rs, 1)]; Chris@10: Tg = R1[WS(rs, 3)]; Chris@10: T8 = T4 + T7; Chris@10: Ta = T4 - T7; Chris@10: Te = Tc - Td; Chris@10: Tm = Tc + Td; Chris@10: Tl = Tf + Tg; Chris@10: Th = Tf - Tg; Chris@10: } Chris@10: } Chris@10: Cr[WS(csr, 2)] = T1 + T8; Chris@10: Tn = Tl - Tm; Chris@10: Tq = Tm + Tl; Chris@10: Ti = FMA(KP618033988, Th, Te); Chris@10: Tk = FNMS(KP618033988, Te, Th); Chris@10: Ci[WS(csi, 2)] = Tn - To; Chris@10: T9 = FNMS(KP250000000, T8, T1); Chris@10: Tu = FMA(KP618033988, Tt, Ts); Chris@10: Tw = FNMS(KP618033988, Ts, Tt); Chris@10: Tp = FMA(KP250000000, Tn, To); Chris@10: Tb = FMA(KP559016994, Ta, T9); Chris@10: Tj = FNMS(KP559016994, Ta, T9); Chris@10: } Chris@10: Tr = FMA(KP559016994, Tq, Tp); Chris@10: Tv = FNMS(KP559016994, Tq, Tp); Chris@10: Cr[WS(csr, 1)] = FNMS(KP951056516, Tk, Tj); Chris@10: Cr[WS(csr, 3)] = FMA(KP951056516, Tk, Tj); Chris@10: Cr[0] = FMA(KP951056516, Ti, Tb); Chris@10: Cr[WS(csr, 4)] = FNMS(KP951056516, Ti, Tb); Chris@10: Ci[WS(csi, 1)] = FNMS(KP951056516, Tw, Tv); Chris@10: Ci[WS(csi, 3)] = FMA(KP951056516, Tw, Tv); Chris@10: Ci[WS(csi, 4)] = FMS(KP951056516, Tu, Tr); Chris@10: Ci[0] = -(FMA(KP951056516, Tu, Tr)); Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kr2c_desc desc = { 10, "r2cfII_10", {14, 0, 18, 0}, &GENUS }; Chris@10: Chris@10: void X(codelet_r2cfII_10) (planner *p) { Chris@10: X(kr2c_register) (p, r2cfII_10, &desc); Chris@10: } Chris@10: Chris@10: #else /* HAVE_FMA */ Chris@10: Chris@10: /* Generated by: ../../../genfft/gen_r2cf.native -compact -variables 4 -pipeline-latency 4 -n 10 -name r2cfII_10 -dft-II -include r2cfII.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 32 FP additions, 12 FP multiplications, Chris@10: * (or, 26 additions, 6 multiplications, 6 fused multiply/add), Chris@10: * 21 stack variables, 4 constants, and 20 memory accesses Chris@10: */ Chris@10: #include "r2cfII.h" Chris@10: Chris@10: static void r2cfII_10(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs) Chris@10: { Chris@10: DK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@10: DK(KP587785252, +0.587785252292473129168705954639072768597652438); Chris@10: DK(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@10: DK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@10: { Chris@10: INT i; Chris@10: for (i = v; i > 0; i = i - 1, R0 = R0 + ivs, R1 = R1 + ivs, Cr = Cr + ovs, Ci = Ci + ovs, MAKE_VOLATILE_STRIDE(40, rs), MAKE_VOLATILE_STRIDE(40, csr), MAKE_VOLATILE_STRIDE(40, csi)) { Chris@10: E T1, To, T8, Tq, T9, Tp, Te, Ts, Th, Tn; Chris@10: T1 = R0[0]; Chris@10: To = R1[WS(rs, 2)]; Chris@10: { Chris@10: E T2, T3, T4, T5, T6, T7; Chris@10: T2 = R0[WS(rs, 2)]; Chris@10: T3 = R0[WS(rs, 3)]; Chris@10: T4 = T2 - T3; Chris@10: T5 = R0[WS(rs, 4)]; Chris@10: T6 = R0[WS(rs, 1)]; Chris@10: T7 = T5 - T6; Chris@10: T8 = T4 + T7; Chris@10: Tq = T5 + T6; Chris@10: T9 = KP559016994 * (T4 - T7); Chris@10: Tp = T2 + T3; Chris@10: } Chris@10: { Chris@10: E Tc, Td, Tm, Tf, Tg, Tl; Chris@10: Tc = R1[0]; Chris@10: Td = R1[WS(rs, 4)]; Chris@10: Tm = Tc + Td; Chris@10: Tf = R1[WS(rs, 1)]; Chris@10: Tg = R1[WS(rs, 3)]; Chris@10: Tl = Tf + Tg; Chris@10: Te = Tc - Td; Chris@10: Ts = KP559016994 * (Tm + Tl); Chris@10: Th = Tf - Tg; Chris@10: Tn = Tl - Tm; Chris@10: } Chris@10: Cr[WS(csr, 2)] = T1 + T8; Chris@10: Ci[WS(csi, 2)] = Tn - To; Chris@10: { Chris@10: E Ti, Tk, Tb, Tj, Ta; Chris@10: Ti = FMA(KP951056516, Te, KP587785252 * Th); Chris@10: Tk = FNMS(KP587785252, Te, KP951056516 * Th); Chris@10: Ta = FNMS(KP250000000, T8, T1); Chris@10: Tb = T9 + Ta; Chris@10: Tj = Ta - T9; Chris@10: Cr[WS(csr, 4)] = Tb - Ti; Chris@10: Cr[WS(csr, 3)] = Tj + Tk; Chris@10: Cr[0] = Tb + Ti; Chris@10: Cr[WS(csr, 1)] = Tj - Tk; Chris@10: } Chris@10: { Chris@10: E Tr, Tw, Tu, Tv, Tt; Chris@10: Tr = FMA(KP951056516, Tp, KP587785252 * Tq); Chris@10: Tw = FNMS(KP587785252, Tp, KP951056516 * Tq); Chris@10: Tt = FMA(KP250000000, Tn, To); Chris@10: Tu = Ts + Tt; Chris@10: Tv = Tt - Ts; Chris@10: Ci[0] = -(Tr + Tu); Chris@10: Ci[WS(csi, 3)] = Tw + Tv; Chris@10: Ci[WS(csi, 4)] = Tr - Tu; Chris@10: Ci[WS(csi, 1)] = Tv - Tw; Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kr2c_desc desc = { 10, "r2cfII_10", {26, 6, 6, 0}, &GENUS }; Chris@10: Chris@10: void X(codelet_r2cfII_10) (planner *p) { Chris@10: X(kr2c_register) (p, r2cfII_10, &desc); Chris@10: } Chris@10: Chris@10: #endif /* HAVE_FMA */