Chris@42: /* Chris@42: * Copyright (c) 2003, 2007-14 Matteo Frigo Chris@42: * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology Chris@42: * Chris@42: * This program is free software; you can redistribute it and/or modify Chris@42: * it under the terms of the GNU General Public License as published by Chris@42: * the Free Software Foundation; either version 2 of the License, or Chris@42: * (at your option) any later version. Chris@42: * Chris@42: * This program is distributed in the hope that it will be useful, Chris@42: * but WITHOUT ANY WARRANTY; without even the implied warranty of Chris@42: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the Chris@42: * GNU General Public License for more details. Chris@42: * Chris@42: * You should have received a copy of the GNU General Public License Chris@42: * along with this program; if not, write to the Free Software Chris@42: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA Chris@42: * Chris@42: */ Chris@42: Chris@42: /* This file was automatically generated --- DO NOT EDIT */ Chris@42: /* Generated on Sat Jul 30 16:50:41 EDT 2016 */ Chris@42: Chris@42: #include "codelet-rdft.h" Chris@42: Chris@42: #ifdef HAVE_FMA Chris@42: Chris@42: /* 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 */ Chris@42: Chris@42: /* Chris@42: * This function contains 32 FP additions, 28 FP multiplications, Chris@42: * (or, 14 additions, 10 multiplications, 18 fused multiply/add), Chris@42: * 38 stack variables, 5 constants, and 20 memory accesses Chris@42: */ Chris@42: #include "r2cbIII.h" Chris@42: Chris@42: static void r2cbIII_10(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs) Chris@42: { Chris@42: DK(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@42: DK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@42: DK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@42: DK(KP618033988, +0.618033988749894848204586834365638117720309180); Chris@42: DK(KP2_000000000, +2.000000000000000000000000000000000000000000000); Chris@42: { Chris@42: INT i; Chris@42: 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)) { Chris@42: E Tq, Ti, Tk, Tu, Tw, Tp, Tb, Tj, Tr, Tv; Chris@42: { Chris@42: E T1, To, Ts, Tt, T8, Ta, Te, Tl, Tm, Th, Tn, T9; Chris@42: T1 = Cr[WS(csr, 2)]; Chris@42: To = Ci[WS(csi, 2)]; Chris@42: { Chris@42: E T2, T3, T5, T6; Chris@42: T2 = Cr[WS(csr, 4)]; Chris@42: T3 = Cr[0]; Chris@42: T5 = Cr[WS(csr, 3)]; Chris@42: T6 = Cr[WS(csr, 1)]; Chris@42: { Chris@42: E Tc, T4, T7, Td, Tf, Tg; Chris@42: Tc = Ci[WS(csi, 3)]; Chris@42: Ts = T2 - T3; Chris@42: T4 = T2 + T3; Chris@42: Tt = T5 - T6; Chris@42: T7 = T5 + T6; Chris@42: Td = Ci[WS(csi, 1)]; Chris@42: Tf = Ci[WS(csi, 4)]; Chris@42: Tg = Ci[0]; Chris@42: T8 = T4 + T7; Chris@42: Ta = T7 - T4; Chris@42: Te = Tc - Td; Chris@42: Tl = Tc + Td; Chris@42: Tm = Tf + Tg; Chris@42: Th = Tf - Tg; Chris@42: } Chris@42: } Chris@42: R0[0] = KP2_000000000 * (T1 + T8); Chris@42: Tn = Tl - Tm; Chris@42: Tq = Tl + Tm; Chris@42: Ti = FMA(KP618033988, Th, Te); Chris@42: Tk = FNMS(KP618033988, Te, Th); Chris@42: R1[WS(rs, 2)] = KP2_000000000 * (Tn - To); Chris@42: T9 = FMS(KP250000000, T8, T1); Chris@42: Tu = FMA(KP618033988, Tt, Ts); Chris@42: Tw = FNMS(KP618033988, Ts, Tt); Chris@42: Tp = FMA(KP250000000, Tn, To); Chris@42: Tb = FNMS(KP559016994, Ta, T9); Chris@42: Tj = FMA(KP559016994, Ta, T9); Chris@42: } Chris@42: Tr = FMA(KP559016994, Tq, Tp); Chris@42: Tv = FNMS(KP559016994, Tq, Tp); Chris@42: R0[WS(rs, 2)] = -(KP2_000000000 * (FNMS(KP951056516, Tk, Tj))); Chris@42: R0[WS(rs, 3)] = KP2_000000000 * (FMA(KP951056516, Tk, Tj)); Chris@42: R0[WS(rs, 4)] = -(KP2_000000000 * (FNMS(KP951056516, Ti, Tb))); Chris@42: R0[WS(rs, 1)] = KP2_000000000 * (FMA(KP951056516, Ti, Tb)); Chris@42: R1[WS(rs, 1)] = KP2_000000000 * (FMA(KP951056516, Tw, Tv)); Chris@42: R1[WS(rs, 3)] = KP2_000000000 * (FNMS(KP951056516, Tw, Tv)); Chris@42: R1[WS(rs, 4)] = -(KP2_000000000 * (FNMS(KP951056516, Tu, Tr))); Chris@42: R1[0] = -(KP2_000000000 * (FMA(KP951056516, Tu, Tr))); Chris@42: } Chris@42: } Chris@42: } Chris@42: Chris@42: static const kr2c_desc desc = { 10, "r2cbIII_10", {14, 10, 18, 0}, &GENUS }; Chris@42: Chris@42: void X(codelet_r2cbIII_10) (planner *p) { Chris@42: X(kr2c_register) (p, r2cbIII_10, &desc); Chris@42: } Chris@42: Chris@42: #else /* HAVE_FMA */ Chris@42: Chris@42: /* Generated by: ../../../genfft/gen_r2cb.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 10 -name r2cbIII_10 -dft-III -include r2cbIII.h */ Chris@42: Chris@42: /* Chris@42: * This function contains 32 FP additions, 16 FP multiplications, Chris@42: * (or, 26 additions, 10 multiplications, 6 fused multiply/add), Chris@42: * 22 stack variables, 5 constants, and 20 memory accesses Chris@42: */ Chris@42: #include "r2cbIII.h" Chris@42: Chris@42: static void r2cbIII_10(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs) Chris@42: { Chris@42: DK(KP500000000, +0.500000000000000000000000000000000000000000000); Chris@42: DK(KP1_902113032, +1.902113032590307144232878666758764286811397268); Chris@42: DK(KP1_175570504, +1.175570504584946258337411909278145537195304875); Chris@42: DK(KP2_000000000, +2.000000000000000000000000000000000000000000000); Chris@42: DK(KP1_118033988, +1.118033988749894848204586834365638117720309180); Chris@42: { Chris@42: INT i; Chris@42: 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)) { Chris@42: E T1, To, T8, Tq, Ta, Tp, Te, Ts, Th, Tn; Chris@42: T1 = Cr[WS(csr, 2)]; Chris@42: To = Ci[WS(csi, 2)]; Chris@42: { Chris@42: E T2, T3, T4, T5, T6, T7; Chris@42: T2 = Cr[WS(csr, 4)]; Chris@42: T3 = Cr[0]; Chris@42: T4 = T2 + T3; Chris@42: T5 = Cr[WS(csr, 3)]; Chris@42: T6 = Cr[WS(csr, 1)]; Chris@42: T7 = T5 + T6; Chris@42: T8 = T4 + T7; Chris@42: Tq = T5 - T6; Chris@42: Ta = KP1_118033988 * (T7 - T4); Chris@42: Tp = T2 - T3; Chris@42: } Chris@42: { Chris@42: E Tc, Td, Tm, Tf, Tg, Tl; Chris@42: Tc = Ci[WS(csi, 4)]; Chris@42: Td = Ci[0]; Chris@42: Tm = Tc + Td; Chris@42: Tf = Ci[WS(csi, 1)]; Chris@42: Tg = Ci[WS(csi, 3)]; Chris@42: Tl = Tg + Tf; Chris@42: Te = Tc - Td; Chris@42: Ts = KP1_118033988 * (Tl + Tm); Chris@42: Th = Tf - Tg; Chris@42: Tn = Tl - Tm; Chris@42: } Chris@42: R0[0] = KP2_000000000 * (T1 + T8); Chris@42: R1[WS(rs, 2)] = KP2_000000000 * (Tn - To); Chris@42: { Chris@42: E Ti, Tj, Tb, Tk, T9; Chris@42: Ti = FNMS(KP1_902113032, Th, KP1_175570504 * Te); Chris@42: Tj = FMA(KP1_175570504, Th, KP1_902113032 * Te); Chris@42: T9 = FNMS(KP2_000000000, T1, KP500000000 * T8); Chris@42: Tb = T9 - Ta; Chris@42: Tk = T9 + Ta; Chris@42: R0[WS(rs, 1)] = Tb + Ti; Chris@42: R0[WS(rs, 3)] = Tk + Tj; Chris@42: R0[WS(rs, 4)] = Ti - Tb; Chris@42: R0[WS(rs, 2)] = Tj - Tk; Chris@42: } Chris@42: { Chris@42: E Tr, Tv, Tu, Tw, Tt; Chris@42: Tr = FMA(KP1_902113032, Tp, KP1_175570504 * Tq); Chris@42: Tv = FNMS(KP1_175570504, Tp, KP1_902113032 * Tq); Chris@42: Tt = FMA(KP500000000, Tn, KP2_000000000 * To); Chris@42: Tu = Ts + Tt; Chris@42: Tw = Tt - Ts; Chris@42: R1[0] = -(Tr + Tu); Chris@42: R1[WS(rs, 3)] = Tw - Tv; Chris@42: R1[WS(rs, 4)] = Tr - Tu; Chris@42: R1[WS(rs, 1)] = Tv + Tw; Chris@42: } Chris@42: } Chris@42: } Chris@42: } Chris@42: Chris@42: static const kr2c_desc desc = { 10, "r2cbIII_10", {26, 10, 6, 0}, &GENUS }; Chris@42: Chris@42: void X(codelet_r2cbIII_10) (planner *p) { Chris@42: X(kr2c_register) (p, r2cbIII_10, &desc); Chris@42: } Chris@42: Chris@42: #endif /* HAVE_FMA */