Chris@82: /* Chris@82: * Copyright (c) 2003, 2007-14 Matteo Frigo Chris@82: * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology Chris@82: * Chris@82: * This program is free software; you can redistribute it and/or modify Chris@82: * it under the terms of the GNU General Public License as published by Chris@82: * the Free Software Foundation; either version 2 of the License, or Chris@82: * (at your option) any later version. Chris@82: * Chris@82: * This program is distributed in the hope that it will be useful, Chris@82: * but WITHOUT ANY WARRANTY; without even the implied warranty of Chris@82: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the Chris@82: * GNU General Public License for more details. Chris@82: * Chris@82: * You should have received a copy of the GNU General Public License Chris@82: * along with this program; if not, write to the Free Software Chris@82: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA Chris@82: * Chris@82: */ Chris@82: Chris@82: /* This file was automatically generated --- DO NOT EDIT */ Chris@82: /* Generated on Thu May 24 08:06:26 EDT 2018 */ Chris@82: Chris@82: #include "rdft/codelet-rdft.h" Chris@82: Chris@82: #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA) Chris@82: Chris@82: /* Generated by: ../../../genfft/gen_r2cf.native -fma -compact -variables 4 -pipeline-latency 4 -n 8 -name r2cf_8 -include rdft/scalar/r2cf.h */ Chris@82: Chris@82: /* Chris@82: * This function contains 20 FP additions, 4 FP multiplications, Chris@82: * (or, 16 additions, 0 multiplications, 4 fused multiply/add), Chris@82: * 14 stack variables, 1 constants, and 16 memory accesses Chris@82: */ Chris@82: #include "rdft/scalar/r2cf.h" Chris@82: Chris@82: 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@82: { Chris@82: DK(KP707106781, +0.707106781186547524400844362104849039284835938); Chris@82: { Chris@82: INT i; Chris@82: 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@82: E T3, T7, Td, Tj, T6, Tf, Ta, Ti; Chris@82: { Chris@82: E T1, T2, Tb, Tc; Chris@82: T1 = R0[0]; Chris@82: T2 = R0[WS(rs, 2)]; Chris@82: T3 = T1 + T2; Chris@82: T7 = T1 - T2; Chris@82: Tb = R1[WS(rs, 3)]; Chris@82: Tc = R1[WS(rs, 1)]; Chris@82: Td = Tb - Tc; Chris@82: Tj = Tb + Tc; Chris@82: } Chris@82: { Chris@82: E T4, T5, T8, T9; Chris@82: T4 = R0[WS(rs, 1)]; Chris@82: T5 = R0[WS(rs, 3)]; Chris@82: T6 = T4 + T5; Chris@82: Tf = T4 - T5; Chris@82: T8 = R1[0]; Chris@82: T9 = R1[WS(rs, 2)]; Chris@82: Ta = T8 - T9; Chris@82: Ti = T8 + T9; Chris@82: } Chris@82: Cr[WS(csr, 2)] = T3 - T6; Chris@82: Ci[WS(csi, 2)] = Tj - Ti; Chris@82: { Chris@82: E Te, Tg, Th, Tk; Chris@82: Te = Ta + Td; Chris@82: Cr[WS(csr, 3)] = FNMS(KP707106781, Te, T7); Chris@82: Cr[WS(csr, 1)] = FMA(KP707106781, Te, T7); Chris@82: Tg = Td - Ta; Chris@82: Ci[WS(csi, 1)] = FMS(KP707106781, Tg, Tf); Chris@82: Ci[WS(csi, 3)] = FMA(KP707106781, Tg, Tf); Chris@82: Th = T3 + T6; Chris@82: Tk = Ti + Tj; Chris@82: Cr[WS(csr, 4)] = Th - Tk; Chris@82: Cr[0] = Th + Tk; Chris@82: } Chris@82: } Chris@82: } Chris@82: } Chris@82: Chris@82: static const kr2c_desc desc = { 8, "r2cf_8", {16, 0, 4, 0}, &GENUS }; Chris@82: Chris@82: void X(codelet_r2cf_8) (planner *p) { Chris@82: X(kr2c_register) (p, r2cf_8, &desc); Chris@82: } Chris@82: Chris@82: #else Chris@82: Chris@82: /* Generated by: ../../../genfft/gen_r2cf.native -compact -variables 4 -pipeline-latency 4 -n 8 -name r2cf_8 -include rdft/scalar/r2cf.h */ Chris@82: Chris@82: /* Chris@82: * This function contains 20 FP additions, 2 FP multiplications, Chris@82: * (or, 20 additions, 2 multiplications, 0 fused multiply/add), Chris@82: * 14 stack variables, 1 constants, and 16 memory accesses Chris@82: */ Chris@82: #include "rdft/scalar/r2cf.h" Chris@82: Chris@82: 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@82: { Chris@82: DK(KP707106781, +0.707106781186547524400844362104849039284835938); Chris@82: { Chris@82: INT i; Chris@82: 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@82: E T3, T7, Td, Tj, T6, Tg, Ta, Ti; Chris@82: { Chris@82: E T1, T2, Tb, Tc; Chris@82: T1 = R0[0]; Chris@82: T2 = R0[WS(rs, 2)]; Chris@82: T3 = T1 + T2; Chris@82: T7 = T1 - T2; Chris@82: Tb = R1[WS(rs, 3)]; Chris@82: Tc = R1[WS(rs, 1)]; Chris@82: Td = Tb - Tc; Chris@82: Tj = Tb + Tc; Chris@82: } Chris@82: { Chris@82: E T4, T5, T8, T9; Chris@82: T4 = R0[WS(rs, 1)]; Chris@82: T5 = R0[WS(rs, 3)]; Chris@82: T6 = T4 + T5; Chris@82: Tg = T4 - T5; Chris@82: T8 = R1[0]; Chris@82: T9 = R1[WS(rs, 2)]; Chris@82: Ta = T8 - T9; Chris@82: Ti = T8 + T9; Chris@82: } Chris@82: Cr[WS(csr, 2)] = T3 - T6; Chris@82: Ci[WS(csi, 2)] = Tj - Ti; Chris@82: { Chris@82: E Te, Tf, Th, Tk; Chris@82: Te = KP707106781 * (Ta + Td); Chris@82: Cr[WS(csr, 3)] = T7 - Te; Chris@82: Cr[WS(csr, 1)] = T7 + Te; Chris@82: Tf = KP707106781 * (Td - Ta); Chris@82: Ci[WS(csi, 1)] = Tf - Tg; Chris@82: Ci[WS(csi, 3)] = Tg + Tf; Chris@82: Th = T3 + T6; Chris@82: Tk = Ti + Tj; Chris@82: Cr[WS(csr, 4)] = Th - Tk; Chris@82: Cr[0] = Th + Tk; Chris@82: } Chris@82: } Chris@82: } Chris@82: } Chris@82: Chris@82: static const kr2c_desc desc = { 8, "r2cf_8", {20, 2, 0, 0}, &GENUS }; Chris@82: Chris@82: void X(codelet_r2cf_8) (planner *p) { Chris@82: X(kr2c_register) (p, r2cf_8, &desc); Chris@82: } Chris@82: Chris@82: #endif