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 07:41:07 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 8 -name r2cb_8 -include r2cb.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 20 FP additions, 12 FP multiplications, Chris@10: * (or, 8 additions, 0 multiplications, 12 fused multiply/add), Chris@10: * 19 stack variables, 2 constants, and 16 memory accesses Chris@10: */ Chris@10: #include "r2cb.h" Chris@10: Chris@10: static void r2cb_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(KP1_414213562, +1.414213562373095048801688724209698078569671875); 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(32, rs), MAKE_VOLATILE_STRIDE(32, csr), MAKE_VOLATILE_STRIDE(32, csi)) { Chris@10: E Th, Tb, Tg, Ti; Chris@10: { Chris@10: E T4, Ta, Td, T9, T3, Tc, T8, Te; Chris@10: T4 = Cr[WS(csr, 2)]; Chris@10: Ta = Ci[WS(csi, 2)]; Chris@10: { Chris@10: E T1, T2, T6, T7; Chris@10: T1 = Cr[0]; Chris@10: T2 = Cr[WS(csr, 4)]; Chris@10: T6 = Cr[WS(csr, 1)]; Chris@10: T7 = Cr[WS(csr, 3)]; Chris@10: Td = Ci[WS(csi, 1)]; Chris@10: T9 = T1 - T2; Chris@10: T3 = T1 + T2; Chris@10: Tc = T6 - T7; Chris@10: T8 = T6 + T7; Chris@10: Te = Ci[WS(csi, 3)]; Chris@10: } Chris@10: { Chris@10: E Tj, T5, Tk, Tf; Chris@10: Tj = FNMS(KP2_000000000, T4, T3); Chris@10: T5 = FMA(KP2_000000000, T4, T3); Chris@10: Th = FMA(KP2_000000000, Ta, T9); Chris@10: Tb = FNMS(KP2_000000000, Ta, T9); Chris@10: Tk = Td - Te; Chris@10: Tf = Td + Te; Chris@10: R0[0] = FMA(KP2_000000000, T8, T5); Chris@10: R0[WS(rs, 2)] = FNMS(KP2_000000000, T8, T5); Chris@10: R0[WS(rs, 3)] = FMA(KP2_000000000, Tk, Tj); Chris@10: R0[WS(rs, 1)] = FNMS(KP2_000000000, Tk, Tj); Chris@10: Tg = Tc - Tf; Chris@10: Ti = Tc + Tf; Chris@10: } Chris@10: } Chris@10: R1[0] = FMA(KP1_414213562, Tg, Tb); Chris@10: R1[WS(rs, 2)] = FNMS(KP1_414213562, Tg, Tb); Chris@10: R1[WS(rs, 3)] = FMA(KP1_414213562, Ti, Th); Chris@10: R1[WS(rs, 1)] = FNMS(KP1_414213562, Ti, Th); Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kr2c_desc desc = { 8, "r2cb_8", {8, 0, 12, 0}, &GENUS }; Chris@10: Chris@10: void X(codelet_r2cb_8) (planner *p) { Chris@10: X(kr2c_register) (p, r2cb_8, &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 8 -name r2cb_8 -include r2cb.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 20 FP additions, 6 FP multiplications, Chris@10: * (or, 20 additions, 6 multiplications, 0 fused multiply/add), Chris@10: * 21 stack variables, 2 constants, and 16 memory accesses Chris@10: */ Chris@10: #include "r2cb.h" Chris@10: Chris@10: static void r2cb_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(KP1_414213562, +1.414213562373095048801688724209698078569671875); 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(32, rs), MAKE_VOLATILE_STRIDE(32, csr), MAKE_VOLATILE_STRIDE(32, csi)) { Chris@10: E T5, Tg, T3, Te, T9, Ti, Td, Tj, T6, Ta; Chris@10: { Chris@10: E T4, Tf, T1, T2; Chris@10: T4 = Cr[WS(csr, 2)]; Chris@10: T5 = KP2_000000000 * T4; Chris@10: Tf = Ci[WS(csi, 2)]; Chris@10: Tg = KP2_000000000 * Tf; Chris@10: T1 = Cr[0]; Chris@10: T2 = Cr[WS(csr, 4)]; Chris@10: T3 = T1 + T2; Chris@10: Te = T1 - T2; Chris@10: { Chris@10: E T7, T8, Tb, Tc; Chris@10: T7 = Cr[WS(csr, 1)]; Chris@10: T8 = Cr[WS(csr, 3)]; Chris@10: T9 = KP2_000000000 * (T7 + T8); Chris@10: Ti = T7 - T8; Chris@10: Tb = Ci[WS(csi, 1)]; Chris@10: Tc = Ci[WS(csi, 3)]; Chris@10: Td = KP2_000000000 * (Tb - Tc); Chris@10: Tj = Tb + Tc; Chris@10: } Chris@10: } Chris@10: T6 = T3 + T5; Chris@10: R0[WS(rs, 2)] = T6 - T9; Chris@10: R0[0] = T6 + T9; Chris@10: Ta = T3 - T5; Chris@10: R0[WS(rs, 1)] = Ta - Td; Chris@10: R0[WS(rs, 3)] = Ta + Td; Chris@10: { Chris@10: E Th, Tk, Tl, Tm; Chris@10: Th = Te - Tg; Chris@10: Tk = KP1_414213562 * (Ti - Tj); Chris@10: R1[WS(rs, 2)] = Th - Tk; Chris@10: R1[0] = Th + Tk; Chris@10: Tl = Te + Tg; Chris@10: Tm = KP1_414213562 * (Ti + Tj); Chris@10: R1[WS(rs, 1)] = Tl - Tm; Chris@10: R1[WS(rs, 3)] = Tl + Tm; Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kr2c_desc desc = { 8, "r2cb_8", {20, 6, 0, 0}, &GENUS }; Chris@10: Chris@10: void X(codelet_r2cb_8) (planner *p) { Chris@10: X(kr2c_register) (p, r2cb_8, &desc); Chris@10: } Chris@10: Chris@10: #endif /* HAVE_FMA */