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 07:41:35 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_r2cb.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -sign 1 -n 10 -name r2cbIII_10 -dft-III -include r2cbIII.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 32 FP additions, 28 FP multiplications, cannam@95: * (or, 14 additions, 10 multiplications, 18 fused multiply/add), cannam@95: * 38 stack variables, 5 constants, and 20 memory accesses cannam@95: */ cannam@95: #include "r2cbIII.h" cannam@95: cannam@95: static void r2cbIII_10(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(KP951056516, +0.951056516295153572116439333379382143405698634); cannam@95: DK(KP559016994, +0.559016994374947424102293417182819058860154590); cannam@95: DK(KP250000000, +0.250000000000000000000000000000000000000000000); cannam@95: DK(KP618033988, +0.618033988749894848204586834365638117720309180); cannam@95: DK(KP2_000000000, +2.000000000000000000000000000000000000000000000); cannam@95: { cannam@95: INT i; cannam@95: for (i = v; i > 0; i = i - 1, R0 = R0 + ovs, R1 = R1 + ovs, Cr = Cr + ivs, Ci = Ci + ivs, MAKE_VOLATILE_STRIDE(40, rs), MAKE_VOLATILE_STRIDE(40, csr), MAKE_VOLATILE_STRIDE(40, csi)) { cannam@95: E Tq, Ti, Tk, Tu, Tw, Tp, Tb, Tj, Tr, Tv; cannam@95: { cannam@95: E T1, To, Ts, Tt, T8, Ta, Te, Tl, Tm, Th, Tn, T9; cannam@95: T1 = Cr[WS(csr, 2)]; cannam@95: To = Ci[WS(csi, 2)]; cannam@95: { cannam@95: E T2, T3, T5, T6; cannam@95: T2 = Cr[WS(csr, 4)]; cannam@95: T3 = Cr[0]; cannam@95: T5 = Cr[WS(csr, 3)]; cannam@95: T6 = Cr[WS(csr, 1)]; cannam@95: { cannam@95: E Tc, T4, T7, Td, Tf, Tg; cannam@95: Tc = Ci[WS(csi, 3)]; cannam@95: Ts = T2 - T3; cannam@95: T4 = T2 + T3; cannam@95: Tt = T5 - T6; cannam@95: T7 = T5 + T6; cannam@95: Td = Ci[WS(csi, 1)]; cannam@95: Tf = Ci[WS(csi, 4)]; cannam@95: Tg = Ci[0]; cannam@95: T8 = T4 + T7; cannam@95: Ta = T7 - T4; cannam@95: Te = Tc - Td; cannam@95: Tl = Tc + Td; cannam@95: Tm = Tf + Tg; cannam@95: Th = Tf - Tg; cannam@95: } cannam@95: } cannam@95: R0[0] = KP2_000000000 * (T1 + T8); cannam@95: Tn = Tl - Tm; cannam@95: Tq = Tl + Tm; cannam@95: Ti = FMA(KP618033988, Th, Te); cannam@95: Tk = FNMS(KP618033988, Te, Th); cannam@95: R1[WS(rs, 2)] = KP2_000000000 * (Tn - To); cannam@95: T9 = FMS(KP250000000, T8, T1); cannam@95: Tu = FMA(KP618033988, Tt, Ts); cannam@95: Tw = FNMS(KP618033988, Ts, Tt); cannam@95: Tp = FMA(KP250000000, Tn, To); cannam@95: Tb = FNMS(KP559016994, Ta, T9); cannam@95: Tj = FMA(KP559016994, Ta, T9); cannam@95: } cannam@95: Tr = FMA(KP559016994, Tq, Tp); cannam@95: Tv = FNMS(KP559016994, Tq, Tp); cannam@95: R0[WS(rs, 2)] = -(KP2_000000000 * (FNMS(KP951056516, Tk, Tj))); cannam@95: R0[WS(rs, 3)] = KP2_000000000 * (FMA(KP951056516, Tk, Tj)); cannam@95: R0[WS(rs, 4)] = -(KP2_000000000 * (FNMS(KP951056516, Ti, Tb))); cannam@95: R0[WS(rs, 1)] = KP2_000000000 * (FMA(KP951056516, Ti, Tb)); cannam@95: R1[WS(rs, 1)] = KP2_000000000 * (FMA(KP951056516, Tw, Tv)); cannam@95: R1[WS(rs, 3)] = KP2_000000000 * (FNMS(KP951056516, Tw, Tv)); cannam@95: R1[WS(rs, 4)] = -(KP2_000000000 * (FNMS(KP951056516, Tu, Tr))); cannam@95: R1[0] = -(KP2_000000000 * (FMA(KP951056516, Tu, Tr))); cannam@95: } cannam@95: } cannam@95: } cannam@95: cannam@95: static const kr2c_desc desc = { 10, "r2cbIII_10", {14, 10, 18, 0}, &GENUS }; cannam@95: cannam@95: void X(codelet_r2cbIII_10) (planner *p) { cannam@95: X(kr2c_register) (p, r2cbIII_10, &desc); cannam@95: } cannam@95: cannam@95: #else /* HAVE_FMA */ cannam@95: cannam@95: /* Generated by: ../../../genfft/gen_r2cb.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 10 -name r2cbIII_10 -dft-III -include r2cbIII.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 32 FP additions, 16 FP multiplications, cannam@95: * (or, 26 additions, 10 multiplications, 6 fused multiply/add), cannam@95: * 22 stack variables, 5 constants, and 20 memory accesses cannam@95: */ cannam@95: #include "r2cbIII.h" cannam@95: cannam@95: static void r2cbIII_10(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(KP500000000, +0.500000000000000000000000000000000000000000000); cannam@95: DK(KP1_902113032, +1.902113032590307144232878666758764286811397268); cannam@95: DK(KP1_175570504, +1.175570504584946258337411909278145537195304875); cannam@95: DK(KP2_000000000, +2.000000000000000000000000000000000000000000000); cannam@95: DK(KP1_118033988, +1.118033988749894848204586834365638117720309180); cannam@95: { cannam@95: INT i; cannam@95: for (i = v; i > 0; i = i - 1, R0 = R0 + ovs, R1 = R1 + ovs, Cr = Cr + ivs, Ci = Ci + ivs, MAKE_VOLATILE_STRIDE(40, rs), MAKE_VOLATILE_STRIDE(40, csr), MAKE_VOLATILE_STRIDE(40, csi)) { cannam@95: E T1, To, T8, Tq, Ta, Tp, Te, Ts, Th, Tn; cannam@95: T1 = Cr[WS(csr, 2)]; cannam@95: To = Ci[WS(csi, 2)]; cannam@95: { cannam@95: E T2, T3, T4, T5, T6, T7; cannam@95: T2 = Cr[WS(csr, 4)]; cannam@95: T3 = Cr[0]; cannam@95: T4 = T2 + T3; cannam@95: T5 = Cr[WS(csr, 3)]; cannam@95: T6 = Cr[WS(csr, 1)]; cannam@95: T7 = T5 + T6; cannam@95: T8 = T4 + T7; cannam@95: Tq = T5 - T6; cannam@95: Ta = KP1_118033988 * (T7 - T4); cannam@95: Tp = T2 - T3; cannam@95: } cannam@95: { cannam@95: E Tc, Td, Tm, Tf, Tg, Tl; cannam@95: Tc = Ci[WS(csi, 4)]; cannam@95: Td = Ci[0]; cannam@95: Tm = Tc + Td; cannam@95: Tf = Ci[WS(csi, 1)]; cannam@95: Tg = Ci[WS(csi, 3)]; cannam@95: Tl = Tg + Tf; cannam@95: Te = Tc - Td; cannam@95: Ts = KP1_118033988 * (Tl + Tm); cannam@95: Th = Tf - Tg; cannam@95: Tn = Tl - Tm; cannam@95: } cannam@95: R0[0] = KP2_000000000 * (T1 + T8); cannam@95: R1[WS(rs, 2)] = KP2_000000000 * (Tn - To); cannam@95: { cannam@95: E Ti, Tj, Tb, Tk, T9; cannam@95: Ti = FNMS(KP1_902113032, Th, KP1_175570504 * Te); cannam@95: Tj = FMA(KP1_175570504, Th, KP1_902113032 * Te); cannam@95: T9 = FNMS(KP2_000000000, T1, KP500000000 * T8); cannam@95: Tb = T9 - Ta; cannam@95: Tk = T9 + Ta; cannam@95: R0[WS(rs, 1)] = Tb + Ti; cannam@95: R0[WS(rs, 3)] = Tk + Tj; cannam@95: R0[WS(rs, 4)] = Ti - Tb; cannam@95: R0[WS(rs, 2)] = Tj - Tk; cannam@95: } cannam@95: { cannam@95: E Tr, Tv, Tu, Tw, Tt; cannam@95: Tr = FMA(KP1_902113032, Tp, KP1_175570504 * Tq); cannam@95: Tv = FNMS(KP1_175570504, Tp, KP1_902113032 * Tq); cannam@95: Tt = FMA(KP500000000, Tn, KP2_000000000 * To); cannam@95: Tu = Ts + Tt; cannam@95: Tw = Tt - Ts; cannam@95: R1[0] = -(Tr + Tu); cannam@95: R1[WS(rs, 3)] = Tw - Tv; cannam@95: R1[WS(rs, 4)] = Tr - Tu; cannam@95: R1[WS(rs, 1)] = Tv + Tw; cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: cannam@95: static const kr2c_desc desc = { 10, "r2cbIII_10", {26, 10, 6, 0}, &GENUS }; cannam@95: cannam@95: void X(codelet_r2cbIII_10) (planner *p) { cannam@95: X(kr2c_register) (p, r2cbIII_10, &desc); cannam@95: } cannam@95: cannam@95: #endif /* HAVE_FMA */