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 10 -name r2cf_10 -include r2cf.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 34 FP additions, 14 FP multiplications, Chris@10: * (or, 24 additions, 4 multiplications, 10 fused multiply/add), Chris@10: * 29 stack variables, 4 constants, and 20 memory accesses Chris@10: */ Chris@10: #include "r2cf.h" Chris@10: Chris@10: static void r2cf_10(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(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@10: DK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@10: DK(KP618033988, +0.618033988749894848204586834365638117720309180); Chris@10: DK(KP951056516, +0.951056516295153572116439333379382143405698634); 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(40, rs), MAKE_VOLATILE_STRIDE(40, csr), MAKE_VOLATILE_STRIDE(40, csi)) { Chris@10: E Tt, T3, T7, Tq, T6, Tv, Tp, Tm, Th, T8, T1, T2, T9, Tr; Chris@10: T1 = R0[0]; Chris@10: T2 = R1[WS(rs, 2)]; Chris@10: { Chris@10: E Te, Tn, Td, Tf, Tb, Tc; Chris@10: Tb = R0[WS(rs, 2)]; Chris@10: Tc = R1[WS(rs, 4)]; Chris@10: Te = R0[WS(rs, 3)]; Chris@10: Tt = T1 + T2; Chris@10: T3 = T1 - T2; Chris@10: Tn = Tb + Tc; Chris@10: Td = Tb - Tc; Chris@10: Tf = R1[0]; Chris@10: { Chris@10: E T4, T5, To, Tg; Chris@10: T4 = R0[WS(rs, 1)]; Chris@10: T5 = R1[WS(rs, 3)]; Chris@10: T7 = R0[WS(rs, 4)]; Chris@10: To = Te + Tf; Chris@10: Tg = Te - Tf; Chris@10: Tq = T4 + T5; Chris@10: T6 = T4 - T5; Chris@10: Tv = Tn + To; Chris@10: Tp = Tn - To; Chris@10: Tm = Tg - Td; Chris@10: Th = Td + Tg; Chris@10: T8 = R1[WS(rs, 1)]; Chris@10: } Chris@10: } Chris@10: T9 = T7 - T8; Chris@10: Tr = T7 + T8; Chris@10: { Chris@10: E Ty, Tk, Tx, Tj, Tu, Ts; Chris@10: Tu = Tq + Tr; Chris@10: Ts = Tq - Tr; Chris@10: { Chris@10: E Ta, Tl, Tw, Ti; Chris@10: Ta = T6 + T9; Chris@10: Tl = T6 - T9; Chris@10: Ci[WS(csi, 4)] = KP951056516 * (FMA(KP618033988, Tp, Ts)); Chris@10: Ci[WS(csi, 2)] = KP951056516 * (FNMS(KP618033988, Ts, Tp)); Chris@10: Ty = Tu - Tv; Chris@10: Tw = Tu + Tv; Chris@10: Ci[WS(csi, 3)] = KP951056516 * (FMA(KP618033988, Tl, Tm)); Chris@10: Ci[WS(csi, 1)] = -(KP951056516 * (FNMS(KP618033988, Tm, Tl))); Chris@10: Tk = Ta - Th; Chris@10: Ti = Ta + Th; Chris@10: Cr[0] = Tt + Tw; Chris@10: Tx = FNMS(KP250000000, Tw, Tt); Chris@10: Cr[WS(csr, 5)] = T3 + Ti; Chris@10: Tj = FNMS(KP250000000, Ti, T3); Chris@10: } Chris@10: Cr[WS(csr, 4)] = FMA(KP559016994, Ty, Tx); Chris@10: Cr[WS(csr, 2)] = FNMS(KP559016994, Ty, Tx); Chris@10: Cr[WS(csr, 3)] = FNMS(KP559016994, Tk, Tj); Chris@10: Cr[WS(csr, 1)] = FMA(KP559016994, Tk, Tj); Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kr2c_desc desc = { 10, "r2cf_10", {24, 4, 10, 0}, &GENUS }; Chris@10: Chris@10: void X(codelet_r2cf_10) (planner *p) { Chris@10: X(kr2c_register) (p, r2cf_10, &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 10 -name r2cf_10 -include r2cf.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 34 FP additions, 12 FP multiplications, Chris@10: * (or, 28 additions, 6 multiplications, 6 fused multiply/add), Chris@10: * 26 stack variables, 4 constants, and 20 memory accesses Chris@10: */ Chris@10: #include "r2cf.h" Chris@10: Chris@10: static void r2cf_10(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(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@10: DK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@10: DK(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@10: DK(KP587785252, +0.587785252292473129168705954639072768597652438); 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(40, rs), MAKE_VOLATILE_STRIDE(40, csr), MAKE_VOLATILE_STRIDE(40, csi)) { Chris@10: E Ti, Tt, Ta, Tn, Td, To, Te, Tv, T3, Tq, T6, Tr, T7, Tu, Tg; Chris@10: E Th; Chris@10: Tg = R0[0]; Chris@10: Th = R1[WS(rs, 2)]; Chris@10: Ti = Tg - Th; Chris@10: Tt = Tg + Th; Chris@10: { Chris@10: E T8, T9, Tb, Tc; Chris@10: T8 = R0[WS(rs, 2)]; Chris@10: T9 = R1[WS(rs, 4)]; Chris@10: Ta = T8 - T9; Chris@10: Tn = T8 + T9; Chris@10: Tb = R0[WS(rs, 3)]; Chris@10: Tc = R1[0]; Chris@10: Td = Tb - Tc; Chris@10: To = Tb + Tc; Chris@10: } Chris@10: Te = Ta + Td; Chris@10: Tv = Tn + To; Chris@10: { Chris@10: E T1, T2, T4, T5; Chris@10: T1 = R0[WS(rs, 1)]; Chris@10: T2 = R1[WS(rs, 3)]; Chris@10: T3 = T1 - T2; Chris@10: Tq = T1 + T2; Chris@10: T4 = R0[WS(rs, 4)]; Chris@10: T5 = R1[WS(rs, 1)]; Chris@10: T6 = T4 - T5; Chris@10: Tr = T4 + T5; Chris@10: } Chris@10: T7 = T3 + T6; Chris@10: Tu = Tq + Tr; Chris@10: { Chris@10: E Tl, Tm, Tf, Tj, Tk; Chris@10: Tl = Td - Ta; Chris@10: Tm = T3 - T6; Chris@10: Ci[WS(csi, 1)] = FNMS(KP951056516, Tm, KP587785252 * Tl); Chris@10: Ci[WS(csi, 3)] = FMA(KP587785252, Tm, KP951056516 * Tl); Chris@10: Tf = KP559016994 * (T7 - Te); Chris@10: Tj = T7 + Te; Chris@10: Tk = FNMS(KP250000000, Tj, Ti); Chris@10: Cr[WS(csr, 1)] = Tf + Tk; Chris@10: Cr[WS(csr, 5)] = Ti + Tj; Chris@10: Cr[WS(csr, 3)] = Tk - Tf; Chris@10: } Chris@10: { Chris@10: E Tp, Ts, Ty, Tw, Tx; Chris@10: Tp = Tn - To; Chris@10: Ts = Tq - Tr; Chris@10: Ci[WS(csi, 2)] = FNMS(KP587785252, Ts, KP951056516 * Tp); Chris@10: Ci[WS(csi, 4)] = FMA(KP951056516, Ts, KP587785252 * Tp); Chris@10: Ty = KP559016994 * (Tu - Tv); Chris@10: Tw = Tu + Tv; Chris@10: Tx = FNMS(KP250000000, Tw, Tt); Chris@10: Cr[WS(csr, 2)] = Tx - Ty; Chris@10: Cr[0] = Tt + Tw; Chris@10: Cr[WS(csr, 4)] = Ty + Tx; Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kr2c_desc desc = { 10, "r2cf_10", {28, 6, 6, 0}, &GENUS }; Chris@10: Chris@10: void X(codelet_r2cf_10) (planner *p) { Chris@10: X(kr2c_register) (p, r2cf_10, &desc); Chris@10: } Chris@10: Chris@10: #endif /* HAVE_FMA */