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 07:41:32 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_r2cb.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -sign 1 -n 2 -name r2cbIII_2 -dft-III -include r2cbIII.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 0 FP additions, 2 FP multiplications, Chris@10: * (or, 0 additions, 2 multiplications, 0 fused multiply/add), Chris@10: * 4 stack variables, 1 constants, and 4 memory accesses Chris@10: */ Chris@10: #include "r2cbIII.h" Chris@10: Chris@10: static void r2cbIII_2(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(KP2_000000000, +2.000000000000000000000000000000000000000000000); Chris@10: { Chris@10: INT i; Chris@10: for (i = v; i > 0; i = i - 1, R0 = R0 + ovs, R1 = R1 + ovs, Cr = Cr + ivs, Ci = Ci + ivs, MAKE_VOLATILE_STRIDE(8, rs), MAKE_VOLATILE_STRIDE(8, csr), MAKE_VOLATILE_STRIDE(8, csi)) { Chris@10: E T1, T2; Chris@10: T1 = Cr[0]; Chris@10: T2 = Ci[0]; Chris@10: R0[0] = KP2_000000000 * T1; Chris@10: R1[0] = -(KP2_000000000 * T2); Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kr2c_desc desc = { 2, "r2cbIII_2", {0, 2, 0, 0}, &GENUS }; Chris@10: Chris@10: void X(codelet_r2cbIII_2) (planner *p) { Chris@10: X(kr2c_register) (p, r2cbIII_2, &desc); Chris@10: } Chris@10: Chris@10: #else /* HAVE_FMA */ Chris@10: Chris@10: /* Generated by: ../../../genfft/gen_r2cb.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 2 -name r2cbIII_2 -dft-III -include r2cbIII.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 0 FP additions, 2 FP multiplications, Chris@10: * (or, 0 additions, 2 multiplications, 0 fused multiply/add), Chris@10: * 4 stack variables, 1 constants, and 4 memory accesses Chris@10: */ Chris@10: #include "r2cbIII.h" Chris@10: Chris@10: static void r2cbIII_2(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(KP2_000000000, +2.000000000000000000000000000000000000000000000); Chris@10: { Chris@10: INT i; Chris@10: for (i = v; i > 0; i = i - 1, R0 = R0 + ovs, R1 = R1 + ovs, Cr = Cr + ivs, Ci = Ci + ivs, MAKE_VOLATILE_STRIDE(8, rs), MAKE_VOLATILE_STRIDE(8, csr), MAKE_VOLATILE_STRIDE(8, csi)) { Chris@10: E T1, T2; Chris@10: T1 = Cr[0]; Chris@10: T2 = Ci[0]; Chris@10: R0[0] = KP2_000000000 * T1; Chris@10: R1[0] = -(KP2_000000000 * T2); Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kr2c_desc desc = { 2, "r2cbIII_2", {0, 2, 0, 0}, &GENUS }; Chris@10: Chris@10: void X(codelet_r2cbIII_2) (planner *p) { Chris@10: X(kr2c_register) (p, r2cbIII_2, &desc); Chris@10: } Chris@10: Chris@10: #endif /* HAVE_FMA */