cannam@95: /* cannam@95: * Copyright (c) 2003, 2007-11 Matteo Frigo cannam@95: * Copyright (c) 2003, 2007-11 Massachusetts Institute of Technology cannam@95: * cannam@95: * This program is free software; you can redistribute it and/or modify cannam@95: * it under the terms of the GNU General Public License as published by cannam@95: * the Free Software Foundation; either version 2 of the License, or cannam@95: * (at your option) any later version. cannam@95: * cannam@95: * This program is distributed in the hope that it will be useful, cannam@95: * but WITHOUT ANY WARRANTY; without even the implied warranty of cannam@95: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the cannam@95: * GNU General Public License for more details. cannam@95: * cannam@95: * You should have received a copy of the GNU General Public License cannam@95: * along with this program; if not, write to the Free Software cannam@95: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA cannam@95: * cannam@95: */ cannam@95: cannam@95: /* This file was automatically generated --- DO NOT EDIT */ cannam@95: /* Generated on Sun Nov 25 07:39:45 EST 2012 */ cannam@95: cannam@95: #include "codelet-rdft.h" cannam@95: cannam@95: #ifdef HAVE_FMA cannam@95: cannam@95: /* Generated by: ../../../genfft/gen_r2cf.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 10 -name r2cf_10 -include r2cf.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 34 FP additions, 14 FP multiplications, cannam@95: * (or, 24 additions, 4 multiplications, 10 fused multiply/add), cannam@95: * 29 stack variables, 4 constants, and 20 memory accesses cannam@95: */ cannam@95: #include "r2cf.h" cannam@95: cannam@95: static void r2cf_10(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs) cannam@95: { cannam@95: DK(KP559016994, +0.559016994374947424102293417182819058860154590); cannam@95: DK(KP250000000, +0.250000000000000000000000000000000000000000000); cannam@95: DK(KP618033988, +0.618033988749894848204586834365638117720309180); cannam@95: DK(KP951056516, +0.951056516295153572116439333379382143405698634); cannam@95: { cannam@95: INT i; cannam@95: 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)) { cannam@95: E Tt, T3, T7, Tq, T6, Tv, Tp, Tm, Th, T8, T1, T2, T9, Tr; cannam@95: T1 = R0[0]; cannam@95: T2 = R1[WS(rs, 2)]; cannam@95: { cannam@95: E Te, Tn, Td, Tf, Tb, Tc; cannam@95: Tb = R0[WS(rs, 2)]; cannam@95: Tc = R1[WS(rs, 4)]; cannam@95: Te = R0[WS(rs, 3)]; cannam@95: Tt = T1 + T2; cannam@95: T3 = T1 - T2; cannam@95: Tn = Tb + Tc; cannam@95: Td = Tb - Tc; cannam@95: Tf = R1[0]; cannam@95: { cannam@95: E T4, T5, To, Tg; cannam@95: T4 = R0[WS(rs, 1)]; cannam@95: T5 = R1[WS(rs, 3)]; cannam@95: T7 = R0[WS(rs, 4)]; cannam@95: To = Te + Tf; cannam@95: Tg = Te - Tf; cannam@95: Tq = T4 + T5; cannam@95: T6 = T4 - T5; cannam@95: Tv = Tn + To; cannam@95: Tp = Tn - To; cannam@95: Tm = Tg - Td; cannam@95: Th = Td + Tg; cannam@95: T8 = R1[WS(rs, 1)]; cannam@95: } cannam@95: } cannam@95: T9 = T7 - T8; cannam@95: Tr = T7 + T8; cannam@95: { cannam@95: E Ty, Tk, Tx, Tj, Tu, Ts; cannam@95: Tu = Tq + Tr; cannam@95: Ts = Tq - Tr; cannam@95: { cannam@95: E Ta, Tl, Tw, Ti; cannam@95: Ta = T6 + T9; cannam@95: Tl = T6 - T9; cannam@95: Ci[WS(csi, 4)] = KP951056516 * (FMA(KP618033988, Tp, Ts)); cannam@95: Ci[WS(csi, 2)] = KP951056516 * (FNMS(KP618033988, Ts, Tp)); cannam@95: Ty = Tu - Tv; cannam@95: Tw = Tu + Tv; cannam@95: Ci[WS(csi, 3)] = KP951056516 * (FMA(KP618033988, Tl, Tm)); cannam@95: Ci[WS(csi, 1)] = -(KP951056516 * (FNMS(KP618033988, Tm, Tl))); cannam@95: Tk = Ta - Th; cannam@95: Ti = Ta + Th; cannam@95: Cr[0] = Tt + Tw; cannam@95: Tx = FNMS(KP250000000, Tw, Tt); cannam@95: Cr[WS(csr, 5)] = T3 + Ti; cannam@95: Tj = FNMS(KP250000000, Ti, T3); cannam@95: } cannam@95: Cr[WS(csr, 4)] = FMA(KP559016994, Ty, Tx); cannam@95: Cr[WS(csr, 2)] = FNMS(KP559016994, Ty, Tx); cannam@95: Cr[WS(csr, 3)] = FNMS(KP559016994, Tk, Tj); cannam@95: Cr[WS(csr, 1)] = FMA(KP559016994, Tk, Tj); cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: cannam@95: static const kr2c_desc desc = { 10, "r2cf_10", {24, 4, 10, 0}, &GENUS }; cannam@95: cannam@95: void X(codelet_r2cf_10) (planner *p) { cannam@95: X(kr2c_register) (p, r2cf_10, &desc); cannam@95: } cannam@95: cannam@95: #else /* HAVE_FMA */ cannam@95: cannam@95: /* Generated by: ../../../genfft/gen_r2cf.native -compact -variables 4 -pipeline-latency 4 -n 10 -name r2cf_10 -include r2cf.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 34 FP additions, 12 FP multiplications, cannam@95: * (or, 28 additions, 6 multiplications, 6 fused multiply/add), cannam@95: * 26 stack variables, 4 constants, and 20 memory accesses cannam@95: */ cannam@95: #include "r2cf.h" cannam@95: cannam@95: static void r2cf_10(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs) cannam@95: { cannam@95: DK(KP250000000, +0.250000000000000000000000000000000000000000000); cannam@95: DK(KP559016994, +0.559016994374947424102293417182819058860154590); cannam@95: DK(KP951056516, +0.951056516295153572116439333379382143405698634); cannam@95: DK(KP587785252, +0.587785252292473129168705954639072768597652438); cannam@95: { cannam@95: INT i; cannam@95: 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)) { cannam@95: E Ti, Tt, Ta, Tn, Td, To, Te, Tv, T3, Tq, T6, Tr, T7, Tu, Tg; cannam@95: E Th; cannam@95: Tg = R0[0]; cannam@95: Th = R1[WS(rs, 2)]; cannam@95: Ti = Tg - Th; cannam@95: Tt = Tg + Th; cannam@95: { cannam@95: E T8, T9, Tb, Tc; cannam@95: T8 = R0[WS(rs, 2)]; cannam@95: T9 = R1[WS(rs, 4)]; cannam@95: Ta = T8 - T9; cannam@95: Tn = T8 + T9; cannam@95: Tb = R0[WS(rs, 3)]; cannam@95: Tc = R1[0]; cannam@95: Td = Tb - Tc; cannam@95: To = Tb + Tc; cannam@95: } cannam@95: Te = Ta + Td; cannam@95: Tv = Tn + To; cannam@95: { cannam@95: E T1, T2, T4, T5; cannam@95: T1 = R0[WS(rs, 1)]; cannam@95: T2 = R1[WS(rs, 3)]; cannam@95: T3 = T1 - T2; cannam@95: Tq = T1 + T2; cannam@95: T4 = R0[WS(rs, 4)]; cannam@95: T5 = R1[WS(rs, 1)]; cannam@95: T6 = T4 - T5; cannam@95: Tr = T4 + T5; cannam@95: } cannam@95: T7 = T3 + T6; cannam@95: Tu = Tq + Tr; cannam@95: { cannam@95: E Tl, Tm, Tf, Tj, Tk; cannam@95: Tl = Td - Ta; cannam@95: Tm = T3 - T6; cannam@95: Ci[WS(csi, 1)] = FNMS(KP951056516, Tm, KP587785252 * Tl); cannam@95: Ci[WS(csi, 3)] = FMA(KP587785252, Tm, KP951056516 * Tl); cannam@95: Tf = KP559016994 * (T7 - Te); cannam@95: Tj = T7 + Te; cannam@95: Tk = FNMS(KP250000000, Tj, Ti); cannam@95: Cr[WS(csr, 1)] = Tf + Tk; cannam@95: Cr[WS(csr, 5)] = Ti + Tj; cannam@95: Cr[WS(csr, 3)] = Tk - Tf; cannam@95: } cannam@95: { cannam@95: E Tp, Ts, Ty, Tw, Tx; cannam@95: Tp = Tn - To; cannam@95: Ts = Tq - Tr; cannam@95: Ci[WS(csi, 2)] = FNMS(KP587785252, Ts, KP951056516 * Tp); cannam@95: Ci[WS(csi, 4)] = FMA(KP951056516, Ts, KP587785252 * Tp); cannam@95: Ty = KP559016994 * (Tu - Tv); cannam@95: Tw = Tu + Tv; cannam@95: Tx = FNMS(KP250000000, Tw, Tt); cannam@95: Cr[WS(csr, 2)] = Tx - Ty; cannam@95: Cr[0] = Tt + Tw; cannam@95: Cr[WS(csr, 4)] = Ty + Tx; cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: cannam@95: static const kr2c_desc desc = { 10, "r2cf_10", {28, 6, 6, 0}, &GENUS }; cannam@95: cannam@95: void X(codelet_r2cf_10) (planner *p) { cannam@95: X(kr2c_register) (p, r2cf_10, &desc); cannam@95: } cannam@95: cannam@95: #endif /* HAVE_FMA */