cannam@95: /* cannam@95: * Copyright (c) 2003, 2007-11 Matteo Frigo cannam@95: * Copyright (c) 2003, 2007-11 Massachusetts Institute of Technology cannam@95: * cannam@95: * This program is free software; you can redistribute it and/or modify cannam@95: * it under the terms of the GNU General Public License as published by cannam@95: * the Free Software Foundation; either version 2 of the License, or cannam@95: * (at your option) any later version. cannam@95: * cannam@95: * This program is distributed in the hope that it will be useful, cannam@95: * but WITHOUT ANY WARRANTY; without even the implied warranty of cannam@95: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the cannam@95: * GNU General Public License for more details. cannam@95: * cannam@95: * You should have received a copy of the GNU General Public License cannam@95: * along with this program; if not, write to the Free Software cannam@95: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA cannam@95: * cannam@95: */ cannam@95: cannam@95: /* This file was automatically generated --- DO NOT EDIT */ cannam@95: /* Generated on Sun Nov 25 07:40:15 EST 2012 */ cannam@95: cannam@95: #include "codelet-rdft.h" cannam@95: cannam@95: #ifdef HAVE_FMA cannam@95: cannam@95: /* Generated by: ../../../genfft/gen_r2cf.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 12 -name r2cfII_12 -dft-II -include r2cfII.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 45 FP additions, 24 FP multiplications, cannam@95: * (or, 21 additions, 0 multiplications, 24 fused multiply/add), cannam@95: * 37 stack variables, 3 constants, and 24 memory accesses cannam@95: */ cannam@95: #include "r2cfII.h" cannam@95: cannam@95: static void r2cfII_12(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs) cannam@95: { cannam@95: DK(KP707106781, +0.707106781186547524400844362104849039284835938); cannam@95: DK(KP866025403, +0.866025403784438646763723170752936183471402627); cannam@95: DK(KP500000000, +0.500000000000000000000000000000000000000000000); cannam@95: { cannam@95: INT i; cannam@95: 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)) { cannam@95: E TD, TB, Tp, T9, Tq, Tr, TE, To, Ts, TC; cannam@95: { cannam@95: E T8, T1, Tv, Tm, TF, Tz, Tl, Ta, Tb, Tt, TA, T4, Tc; cannam@95: { cannam@95: E Tx, Th, Ti, Tj, Ty, T6, T7, T2, T3, Tk; cannam@95: Tx = R0[WS(rs, 3)]; cannam@95: T6 = R0[WS(rs, 5)]; cannam@95: T7 = R0[WS(rs, 1)]; cannam@95: Th = R1[WS(rs, 4)]; cannam@95: Ti = R1[WS(rs, 2)]; cannam@95: Tj = R1[0]; cannam@95: Ty = T6 + T7; cannam@95: T8 = T6 - T7; cannam@95: T1 = R0[0]; cannam@95: Tv = Ti - Tj - Th; cannam@95: Tk = Ti - Tj; cannam@95: Tm = Ti + Tj; cannam@95: TF = Tx - Ty; cannam@95: Tz = FMA(KP500000000, Ty, Tx); cannam@95: T2 = R0[WS(rs, 2)]; cannam@95: T3 = R0[WS(rs, 4)]; cannam@95: Tl = FMA(KP500000000, Tk, Th); cannam@95: Ta = R1[WS(rs, 1)]; cannam@95: Tb = R1[WS(rs, 3)]; cannam@95: Tt = T1 + T3 - T2; cannam@95: TA = T3 + T2; cannam@95: T4 = T2 - T3; cannam@95: Tc = R1[WS(rs, 5)]; cannam@95: } cannam@95: { cannam@95: E Tn, Tg, T5, Tu; cannam@95: TD = FNMS(KP866025403, TA, Tz); cannam@95: TB = FMA(KP866025403, TA, Tz); cannam@95: T5 = FMA(KP500000000, T4, T1); cannam@95: Tu = Ta + Tc - Tb; cannam@95: { cannam@95: E Td, Tf, TG, Tw, Te; cannam@95: Td = Tb - Tc; cannam@95: Tf = Tc + Tb; cannam@95: Tp = FMA(KP866025403, T8, T5); cannam@95: T9 = FNMS(KP866025403, T8, T5); cannam@95: TG = Tv - Tu; cannam@95: Tw = Tu + Tv; cannam@95: Te = FMA(KP500000000, Td, Ta); cannam@95: Tq = FMA(KP866025403, Tm, Tl); cannam@95: Tn = FNMS(KP866025403, Tm, Tl); cannam@95: Ci[WS(csi, 1)] = FMA(KP707106781, TG, TF); cannam@95: Ci[WS(csi, 4)] = FMS(KP707106781, TG, TF); cannam@95: Cr[WS(csr, 4)] = FMA(KP707106781, Tw, Tt); cannam@95: Cr[WS(csr, 1)] = FNMS(KP707106781, Tw, Tt); cannam@95: Tg = FNMS(KP866025403, Tf, Te); cannam@95: Tr = FMA(KP866025403, Tf, Te); cannam@95: } cannam@95: TE = Tg + Tn; cannam@95: To = Tg - Tn; cannam@95: } cannam@95: } cannam@95: Ci[WS(csi, 2)] = FMS(KP707106781, TE, TD); cannam@95: Ci[WS(csi, 3)] = FMA(KP707106781, TE, TD); cannam@95: Cr[0] = FMA(KP707106781, To, T9); cannam@95: Cr[WS(csr, 5)] = FNMS(KP707106781, To, T9); cannam@95: Ts = Tq - Tr; cannam@95: TC = Tr + Tq; cannam@95: Ci[0] = -(FMA(KP707106781, TC, TB)); cannam@95: Ci[WS(csi, 5)] = FNMS(KP707106781, TC, TB); cannam@95: Cr[WS(csr, 2)] = FMA(KP707106781, Ts, Tp); cannam@95: Cr[WS(csr, 3)] = FNMS(KP707106781, Ts, Tp); cannam@95: } cannam@95: } cannam@95: } cannam@95: cannam@95: static const kr2c_desc desc = { 12, "r2cfII_12", {21, 0, 24, 0}, &GENUS }; cannam@95: cannam@95: void X(codelet_r2cfII_12) (planner *p) { cannam@95: X(kr2c_register) (p, r2cfII_12, &desc); cannam@95: } cannam@95: cannam@95: #else /* HAVE_FMA */ cannam@95: cannam@95: /* Generated by: ../../../genfft/gen_r2cf.native -compact -variables 4 -pipeline-latency 4 -n 12 -name r2cfII_12 -dft-II -include r2cfII.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 43 FP additions, 12 FP multiplications, cannam@95: * (or, 39 additions, 8 multiplications, 4 fused multiply/add), cannam@95: * 28 stack variables, 5 constants, and 24 memory accesses cannam@95: */ cannam@95: #include "r2cfII.h" cannam@95: cannam@95: static void r2cfII_12(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs) cannam@95: { cannam@95: DK(KP353553390, +0.353553390593273762200422181052424519642417969); cannam@95: DK(KP707106781, +0.707106781186547524400844362104849039284835938); cannam@95: DK(KP612372435, +0.612372435695794524549321018676472847991486870); cannam@95: DK(KP500000000, +0.500000000000000000000000000000000000000000000); cannam@95: DK(KP866025403, +0.866025403784438646763723170752936183471402627); cannam@95: { cannam@95: INT i; cannam@95: 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)) { cannam@95: E Tx, Tg, T4, Tz, Ty, Tj, TA, T9, Tm, Tl, Te, Tp, To, Tf, TE; cannam@95: E TF; cannam@95: { cannam@95: E T1, T3, T2, Th, Ti; cannam@95: T1 = R0[0]; cannam@95: T3 = R0[WS(rs, 2)]; cannam@95: T2 = R0[WS(rs, 4)]; cannam@95: Tx = KP866025403 * (T2 + T3); cannam@95: Tg = FMA(KP500000000, T3 - T2, T1); cannam@95: T4 = T1 + T2 - T3; cannam@95: Tz = R0[WS(rs, 3)]; cannam@95: Th = R0[WS(rs, 5)]; cannam@95: Ti = R0[WS(rs, 1)]; cannam@95: Ty = Th + Ti; cannam@95: Tj = KP866025403 * (Th - Ti); cannam@95: TA = FMA(KP500000000, Ty, Tz); cannam@95: } cannam@95: { cannam@95: E T5, T6, T7, T8; cannam@95: T5 = R1[WS(rs, 1)]; cannam@95: T6 = R1[WS(rs, 5)]; cannam@95: T7 = R1[WS(rs, 3)]; cannam@95: T8 = T6 - T7; cannam@95: T9 = T5 + T8; cannam@95: Tm = KP612372435 * (T6 + T7); cannam@95: Tl = FNMS(KP353553390, T8, KP707106781 * T5); cannam@95: } cannam@95: { cannam@95: E Td, Ta, Tb, Tc; cannam@95: Td = R1[WS(rs, 4)]; cannam@95: Ta = R1[WS(rs, 2)]; cannam@95: Tb = R1[0]; cannam@95: Tc = Ta - Tb; cannam@95: Te = Tc - Td; cannam@95: Tp = FMA(KP353553390, Tc, KP707106781 * Td); cannam@95: To = KP612372435 * (Ta + Tb); cannam@95: } cannam@95: Tf = KP707106781 * (T9 + Te); cannam@95: Cr[WS(csr, 1)] = T4 - Tf; cannam@95: Cr[WS(csr, 4)] = T4 + Tf; cannam@95: TE = KP707106781 * (Te - T9); cannam@95: TF = Tz - Ty; cannam@95: Ci[WS(csi, 4)] = TE - TF; cannam@95: Ci[WS(csi, 1)] = TE + TF; cannam@95: { cannam@95: E Tk, TB, Tr, Tw, Tn, Tq; cannam@95: Tk = Tg - Tj; cannam@95: TB = Tx - TA; cannam@95: Tn = Tl - Tm; cannam@95: Tq = To - Tp; cannam@95: Tr = Tn + Tq; cannam@95: Tw = Tn - Tq; cannam@95: Cr[WS(csr, 5)] = Tk - Tr; cannam@95: Ci[WS(csi, 2)] = Tw + TB; cannam@95: Cr[0] = Tk + Tr; cannam@95: Ci[WS(csi, 3)] = Tw - TB; cannam@95: } cannam@95: { cannam@95: E Ts, TD, Tv, TC, Tt, Tu; cannam@95: Ts = Tg + Tj; cannam@95: TD = Tx + TA; cannam@95: Tt = To + Tp; cannam@95: Tu = Tm + Tl; cannam@95: Tv = Tt - Tu; cannam@95: TC = Tu + Tt; cannam@95: Cr[WS(csr, 3)] = Ts - Tv; cannam@95: Ci[WS(csi, 5)] = TD - TC; cannam@95: Cr[WS(csr, 2)] = Ts + Tv; cannam@95: Ci[0] = -(TC + TD); cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: cannam@95: static const kr2c_desc desc = { 12, "r2cfII_12", {39, 8, 4, 0}, &GENUS }; cannam@95: cannam@95: void X(codelet_r2cfII_12) (planner *p) { cannam@95: X(kr2c_register) (p, r2cfII_12, &desc); cannam@95: } cannam@95: cannam@95: #endif /* HAVE_FMA */