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 07:39:45 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 8 -name r2cf_8 -include r2cf.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 20 FP additions, 4 FP multiplications, Chris@10: * (or, 16 additions, 0 multiplications, 4 fused multiply/add), Chris@10: * 18 stack variables, 1 constants, and 16 memory accesses Chris@10: */ Chris@10: #include "r2cf.h" Chris@10: Chris@10: static void r2cf_8(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(KP707106781, +0.707106781186547524400844362104849039284835938); 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(32, rs), MAKE_VOLATILE_STRIDE(32, csr), MAKE_VOLATILE_STRIDE(32, csi)) { Chris@10: E T4, T7, T3, Tj, Td, T5, T8, T9; Chris@10: { Chris@10: E T1, T2, Tb, Tc; Chris@10: T1 = R0[0]; Chris@10: T2 = R0[WS(rs, 2)]; Chris@10: Tb = R1[WS(rs, 3)]; Chris@10: Tc = R1[WS(rs, 1)]; Chris@10: T4 = R0[WS(rs, 1)]; Chris@10: T7 = T1 - T2; Chris@10: T3 = T1 + T2; Chris@10: Tj = Tb + Tc; Chris@10: Td = Tb - Tc; Chris@10: T5 = R0[WS(rs, 3)]; Chris@10: T8 = R1[0]; Chris@10: T9 = R1[WS(rs, 2)]; Chris@10: } Chris@10: { Chris@10: E T6, Tf, Ta, Ti; Chris@10: T6 = T4 + T5; Chris@10: Tf = T4 - T5; Chris@10: Ta = T8 - T9; Chris@10: Ti = T8 + T9; Chris@10: { Chris@10: E Th, Tk, Te, Tg; Chris@10: Th = T3 + T6; Chris@10: Cr[WS(csr, 2)] = T3 - T6; Chris@10: Tk = Ti + Tj; Chris@10: Ci[WS(csi, 2)] = Tj - Ti; Chris@10: Te = Ta + Td; Chris@10: Tg = Td - Ta; Chris@10: Cr[0] = Th + Tk; Chris@10: Cr[WS(csr, 4)] = Th - Tk; Chris@10: Ci[WS(csi, 3)] = FMA(KP707106781, Tg, Tf); Chris@10: Ci[WS(csi, 1)] = FMS(KP707106781, Tg, Tf); Chris@10: Cr[WS(csr, 1)] = FMA(KP707106781, Te, T7); Chris@10: Cr[WS(csr, 3)] = FNMS(KP707106781, Te, T7); Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kr2c_desc desc = { 8, "r2cf_8", {16, 0, 4, 0}, &GENUS }; Chris@10: Chris@10: void X(codelet_r2cf_8) (planner *p) { Chris@10: X(kr2c_register) (p, r2cf_8, &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 8 -name r2cf_8 -include r2cf.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 20 FP additions, 2 FP multiplications, Chris@10: * (or, 20 additions, 2 multiplications, 0 fused multiply/add), Chris@10: * 14 stack variables, 1 constants, and 16 memory accesses Chris@10: */ Chris@10: #include "r2cf.h" Chris@10: Chris@10: static void r2cf_8(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(KP707106781, +0.707106781186547524400844362104849039284835938); 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(32, rs), MAKE_VOLATILE_STRIDE(32, csr), MAKE_VOLATILE_STRIDE(32, csi)) { Chris@10: E T3, T7, Td, Tj, T6, Tg, Ta, Ti; Chris@10: { Chris@10: E T1, T2, Tb, Tc; Chris@10: T1 = R0[0]; Chris@10: T2 = R0[WS(rs, 2)]; Chris@10: T3 = T1 + T2; Chris@10: T7 = T1 - T2; Chris@10: Tb = R1[WS(rs, 3)]; Chris@10: Tc = R1[WS(rs, 1)]; Chris@10: Td = Tb - Tc; Chris@10: Tj = Tb + Tc; Chris@10: } Chris@10: { Chris@10: E T4, T5, T8, T9; Chris@10: T4 = R0[WS(rs, 1)]; Chris@10: T5 = R0[WS(rs, 3)]; Chris@10: T6 = T4 + T5; Chris@10: Tg = T4 - T5; Chris@10: T8 = R1[0]; Chris@10: T9 = R1[WS(rs, 2)]; Chris@10: Ta = T8 - T9; Chris@10: Ti = T8 + T9; Chris@10: } Chris@10: Cr[WS(csr, 2)] = T3 - T6; Chris@10: Ci[WS(csi, 2)] = Tj - Ti; Chris@10: { Chris@10: E Te, Tf, Th, Tk; Chris@10: Te = KP707106781 * (Ta + Td); Chris@10: Cr[WS(csr, 3)] = T7 - Te; Chris@10: Cr[WS(csr, 1)] = T7 + Te; Chris@10: Tf = KP707106781 * (Td - Ta); Chris@10: Ci[WS(csi, 1)] = Tf - Tg; Chris@10: Ci[WS(csi, 3)] = Tg + Tf; Chris@10: Th = T3 + T6; Chris@10: Tk = Ti + Tj; Chris@10: Cr[WS(csr, 4)] = Th - Tk; Chris@10: Cr[0] = Th + Tk; Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kr2c_desc desc = { 8, "r2cf_8", {20, 2, 0, 0}, &GENUS }; Chris@10: Chris@10: void X(codelet_r2cf_8) (planner *p) { Chris@10: X(kr2c_register) (p, r2cf_8, &desc); Chris@10: } Chris@10: Chris@10: #endif /* HAVE_FMA */