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 07:39:46 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 12 -name r2cf_12 -include r2cf.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 38 FP additions, 10 FP multiplications, Chris@10: * (or, 30 additions, 2 multiplications, 8 fused multiply/add), Chris@10: * 31 stack variables, 2 constants, and 24 memory accesses Chris@10: */ Chris@10: #include "r2cf.h" Chris@10: Chris@10: static void r2cf_12(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(KP866025403, +0.866025403784438646763723170752936183471402627); Chris@10: DK(KP500000000, +0.500000000000000000000000000000000000000000000); 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(48, rs), MAKE_VOLATILE_STRIDE(48, csr), MAKE_VOLATILE_STRIDE(48, csi)) { Chris@10: E Tm, T6, Ty, Tp, T5, Tk, Tt, Tb, Tc, Td, T9, Tn; Chris@10: { Chris@10: E T1, Tg, Th, Ti, T4, T2, T3, T7, T8, Tj; Chris@10: T1 = R0[0]; Chris@10: T2 = R0[WS(rs, 2)]; Chris@10: T3 = R0[WS(rs, 4)]; Chris@10: Tg = R1[WS(rs, 1)]; Chris@10: Th = R1[WS(rs, 3)]; Chris@10: Ti = R1[WS(rs, 5)]; Chris@10: T4 = T2 + T3; Chris@10: Tm = T3 - T2; Chris@10: T6 = R0[WS(rs, 3)]; Chris@10: Ty = Ti - Th; Chris@10: Tj = Th + Ti; Chris@10: Tp = FNMS(KP500000000, T4, T1); Chris@10: T5 = T1 + T4; Chris@10: T7 = R0[WS(rs, 5)]; Chris@10: Tk = FNMS(KP500000000, Tj, Tg); Chris@10: Tt = Tg + Tj; Chris@10: T8 = R0[WS(rs, 1)]; Chris@10: Tb = R1[WS(rs, 4)]; Chris@10: Tc = R1[0]; Chris@10: Td = R1[WS(rs, 2)]; Chris@10: T9 = T7 + T8; Chris@10: Tn = T8 - T7; Chris@10: } Chris@10: { Chris@10: E Te, Tz, To, TC; Chris@10: Te = Tc + Td; Chris@10: Tz = Td - Tc; Chris@10: To = Tm - Tn; Chris@10: TC = Tm + Tn; Chris@10: { Chris@10: E Ta, Tq, TA, TB; Chris@10: Ta = T6 + T9; Chris@10: Tq = FNMS(KP500000000, T9, T6); Chris@10: TA = Ty - Tz; Chris@10: TB = Ty + Tz; Chris@10: { Chris@10: E Tf, Tu, Tx, Tr; Chris@10: Tf = FNMS(KP500000000, Te, Tb); Chris@10: Tu = Tb + Te; Chris@10: Tx = Tp - Tq; Chris@10: Tr = Tp + Tq; Chris@10: { Chris@10: E Tv, Tw, Tl, Ts; Chris@10: Tv = T5 + Ta; Chris@10: Cr[WS(csr, 3)] = T5 - Ta; Chris@10: Ci[WS(csi, 4)] = KP866025403 * (TC + TB); Chris@10: Ci[WS(csi, 2)] = KP866025403 * (TB - TC); Chris@10: Tw = Tt + Tu; Chris@10: Ci[WS(csi, 3)] = Tt - Tu; Chris@10: Tl = Tf - Tk; Chris@10: Ts = Tk + Tf; Chris@10: Cr[WS(csr, 1)] = FMA(KP866025403, TA, Tx); Chris@10: Cr[WS(csr, 5)] = FNMS(KP866025403, TA, Tx); Chris@10: Cr[0] = Tv + Tw; Chris@10: Cr[WS(csr, 6)] = Tv - Tw; Chris@10: Cr[WS(csr, 4)] = Tr + Ts; Chris@10: Cr[WS(csr, 2)] = Tr - Ts; Chris@10: Ci[WS(csi, 5)] = FNMS(KP866025403, To, Tl); Chris@10: Ci[WS(csi, 1)] = FMA(KP866025403, To, Tl); Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kr2c_desc desc = { 12, "r2cf_12", {30, 2, 8, 0}, &GENUS }; Chris@10: Chris@10: void X(codelet_r2cf_12) (planner *p) { Chris@10: X(kr2c_register) (p, r2cf_12, &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 12 -name r2cf_12 -include r2cf.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 38 FP additions, 8 FP multiplications, Chris@10: * (or, 34 additions, 4 multiplications, 4 fused multiply/add), Chris@10: * 21 stack variables, 2 constants, and 24 memory accesses Chris@10: */ Chris@10: #include "r2cf.h" Chris@10: Chris@10: static void r2cf_12(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(KP866025403, +0.866025403784438646763723170752936183471402627); Chris@10: DK(KP500000000, +0.500000000000000000000000000000000000000000000); 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(48, rs), MAKE_VOLATILE_STRIDE(48, csr), MAKE_VOLATILE_STRIDE(48, csi)) { Chris@10: E T5, Tp, Tb, Tn, Ty, Tt, Ta, Tq, Tc, Ti, Tz, Tu, Td, To; Chris@10: { Chris@10: E T1, T2, T3, T4; Chris@10: T1 = R0[0]; Chris@10: T2 = R0[WS(rs, 2)]; Chris@10: T3 = R0[WS(rs, 4)]; Chris@10: T4 = T2 + T3; Chris@10: T5 = T1 + T4; Chris@10: Tp = FNMS(KP500000000, T4, T1); Chris@10: Tb = T3 - T2; Chris@10: } Chris@10: { Chris@10: E Tj, Tk, Tl, Tm; Chris@10: Tj = R1[WS(rs, 1)]; Chris@10: Tk = R1[WS(rs, 3)]; Chris@10: Tl = R1[WS(rs, 5)]; Chris@10: Tm = Tk + Tl; Chris@10: Tn = FNMS(KP500000000, Tm, Tj); Chris@10: Ty = Tl - Tk; Chris@10: Tt = Tj + Tm; Chris@10: } Chris@10: { Chris@10: E T6, T7, T8, T9; Chris@10: T6 = R0[WS(rs, 3)]; Chris@10: T7 = R0[WS(rs, 5)]; Chris@10: T8 = R0[WS(rs, 1)]; Chris@10: T9 = T7 + T8; Chris@10: Ta = T6 + T9; Chris@10: Tq = FNMS(KP500000000, T9, T6); Chris@10: Tc = T8 - T7; Chris@10: } Chris@10: { Chris@10: E Te, Tf, Tg, Th; Chris@10: Te = R1[WS(rs, 4)]; Chris@10: Tf = R1[0]; Chris@10: Tg = R1[WS(rs, 2)]; Chris@10: Th = Tf + Tg; Chris@10: Ti = FNMS(KP500000000, Th, Te); Chris@10: Tz = Tg - Tf; Chris@10: Tu = Te + Th; Chris@10: } Chris@10: Cr[WS(csr, 3)] = T5 - Ta; Chris@10: Ci[WS(csi, 3)] = Tt - Tu; Chris@10: Td = KP866025403 * (Tb - Tc); Chris@10: To = Ti - Tn; Chris@10: Ci[WS(csi, 1)] = Td + To; Chris@10: Ci[WS(csi, 5)] = To - Td; Chris@10: { Chris@10: E Tx, TA, Tv, Tw; Chris@10: Tx = Tp - Tq; Chris@10: TA = KP866025403 * (Ty - Tz); Chris@10: Cr[WS(csr, 5)] = Tx - TA; Chris@10: Cr[WS(csr, 1)] = Tx + TA; Chris@10: Tv = T5 + Ta; Chris@10: Tw = Tt + Tu; Chris@10: Cr[WS(csr, 6)] = Tv - Tw; Chris@10: Cr[0] = Tv + Tw; Chris@10: } Chris@10: { Chris@10: E Tr, Ts, TB, TC; Chris@10: Tr = Tp + Tq; Chris@10: Ts = Tn + Ti; Chris@10: Cr[WS(csr, 2)] = Tr - Ts; Chris@10: Cr[WS(csr, 4)] = Tr + Ts; Chris@10: TB = Ty + Tz; Chris@10: TC = Tb + Tc; Chris@10: Ci[WS(csi, 2)] = KP866025403 * (TB - TC); Chris@10: Ci[WS(csi, 4)] = KP866025403 * (TC + TB); Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kr2c_desc desc = { 12, "r2cf_12", {34, 4, 4, 0}, &GENUS }; Chris@10: Chris@10: void X(codelet_r2cf_12) (planner *p) { Chris@10: X(kr2c_register) (p, r2cf_12, &desc); Chris@10: } Chris@10: Chris@10: #endif /* HAVE_FMA */