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 07:42:27 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_r2r.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -redft10 -n 8 -name e10_8 -include r2r.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 26 FP additions, 18 FP multiplications, cannam@95: * (or, 16 additions, 8 multiplications, 10 fused multiply/add), cannam@95: * 28 stack variables, 9 constants, and 16 memory accesses cannam@95: */ cannam@95: #include "r2r.h" cannam@95: cannam@95: static void e10_8(const R *I, R *O, stride is, stride os, INT v, INT ivs, INT ovs) cannam@95: { cannam@95: DK(KP668178637, +0.668178637919298919997757686523080761552472251); cannam@95: DK(KP1_662939224, +1.662939224605090474157576755235811513477121624); cannam@95: DK(KP198912367, +0.198912367379658006911597622644676228597850501); cannam@95: DK(KP1_961570560, +1.961570560806460898252364472268478073947867462); cannam@95: DK(KP707106781, +0.707106781186547524400844362104849039284835938); cannam@95: DK(KP1_414213562, +1.414213562373095048801688724209698078569671875); cannam@95: DK(KP2_000000000, +2.000000000000000000000000000000000000000000000); cannam@95: DK(KP414213562, +0.414213562373095048801688724209698078569671875); cannam@95: DK(KP1_847759065, +1.847759065022573512256366378793576573644833252); cannam@95: { cannam@95: INT i; cannam@95: for (i = v; i > 0; i = i - 1, I = I + ivs, O = O + ovs, MAKE_VOLATILE_STRIDE(16, is), MAKE_VOLATILE_STRIDE(16, os)) { cannam@95: E T3, Te, Tl, Tp, Tm, T6, Tn, T9; cannam@95: { cannam@95: E T4, Tj, Tk, T5, T7, T8; cannam@95: { cannam@95: E T1, T2, Tc, Td; cannam@95: T1 = I[0]; cannam@95: T2 = I[WS(is, 7)]; cannam@95: Tc = I[WS(is, 4)]; cannam@95: Td = I[WS(is, 3)]; cannam@95: T4 = I[WS(is, 2)]; cannam@95: Tj = T1 + T2; cannam@95: T3 = T1 - T2; cannam@95: Tk = Tc + Td; cannam@95: Te = Tc - Td; cannam@95: T5 = I[WS(is, 5)]; cannam@95: T7 = I[WS(is, 1)]; cannam@95: T8 = I[WS(is, 6)]; cannam@95: } cannam@95: Tl = Tj - Tk; cannam@95: Tp = Tj + Tk; cannam@95: Tm = T4 + T5; cannam@95: T6 = T4 - T5; cannam@95: Tn = T7 + T8; cannam@95: T9 = T7 - T8; cannam@95: } cannam@95: { cannam@95: E Tg, Ti, Tb, Th; cannam@95: { cannam@95: E Tq, To, Ta, Tf; cannam@95: Tq = Tm + Tn; cannam@95: To = Tm - Tn; cannam@95: Ta = T6 + T9; cannam@95: Tf = T6 - T9; cannam@95: O[WS(os, 6)] = KP1_847759065 * (FMA(KP414213562, Tl, To)); cannam@95: O[WS(os, 2)] = KP1_847759065 * (FNMS(KP414213562, To, Tl)); cannam@95: O[0] = KP2_000000000 * (Tp + Tq); cannam@95: O[WS(os, 4)] = KP1_414213562 * (Tp - Tq); cannam@95: Tg = FNMS(KP707106781, Tf, Te); cannam@95: Ti = FMA(KP707106781, Tf, Te); cannam@95: Tb = FNMS(KP707106781, Ta, T3); cannam@95: Th = FMA(KP707106781, Ta, T3); cannam@95: } cannam@95: O[WS(os, 7)] = KP1_961570560 * (FMA(KP198912367, Th, Ti)); cannam@95: O[WS(os, 1)] = KP1_961570560 * (FNMS(KP198912367, Ti, Th)); cannam@95: O[WS(os, 5)] = -(KP1_662939224 * (FNMS(KP668178637, Tb, Tg))); cannam@95: O[WS(os, 3)] = KP1_662939224 * (FMA(KP668178637, Tg, Tb)); cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: cannam@95: static const kr2r_desc desc = { 8, "e10_8", {16, 8, 10, 0}, &GENUS, REDFT10 }; cannam@95: cannam@95: void X(codelet_e10_8) (planner *p) { cannam@95: X(kr2r_register) (p, e10_8, &desc); cannam@95: } cannam@95: cannam@95: #else /* HAVE_FMA */ cannam@95: cannam@95: /* Generated by: ../../../genfft/gen_r2r.native -compact -variables 4 -pipeline-latency 4 -redft10 -n 8 -name e10_8 -include r2r.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 26 FP additions, 16 FP multiplications, cannam@95: * (or, 20 additions, 10 multiplications, 6 fused multiply/add), cannam@95: * 28 stack variables, 9 constants, and 16 memory accesses cannam@95: */ cannam@95: #include "r2r.h" cannam@95: cannam@95: static void e10_8(const R *I, R *O, stride is, stride os, INT v, INT ivs, INT ovs) cannam@95: { cannam@95: DK(KP765366864, +0.765366864730179543456919968060797733522689125); cannam@95: DK(KP1_847759065, +1.847759065022573512256366378793576573644833252); cannam@95: DK(KP390180644, +0.390180644032256535696569736954044481855383236); cannam@95: DK(KP1_961570560, +1.961570560806460898252364472268478073947867462); cannam@95: DK(KP2_000000000, +2.000000000000000000000000000000000000000000000); cannam@95: DK(KP1_414213562, +1.414213562373095048801688724209698078569671875); cannam@95: DK(KP1_111140466, +1.111140466039204449485661627897065748749874382); cannam@95: DK(KP1_662939224, +1.662939224605090474157576755235811513477121624); cannam@95: DK(KP707106781, +0.707106781186547524400844362104849039284835938); cannam@95: { cannam@95: INT i; cannam@95: for (i = v; i > 0; i = i - 1, I = I + ivs, O = O + ovs, MAKE_VOLATILE_STRIDE(16, is), MAKE_VOLATILE_STRIDE(16, os)) { cannam@95: E T3, Tj, Tf, Tk, Ta, Tn, Tc, Tm; cannam@95: { cannam@95: E T1, T2, Td, Te; cannam@95: T1 = I[0]; cannam@95: T2 = I[WS(is, 7)]; cannam@95: T3 = T1 - T2; cannam@95: Tj = T1 + T2; cannam@95: Td = I[WS(is, 4)]; cannam@95: Te = I[WS(is, 3)]; cannam@95: Tf = Td - Te; cannam@95: Tk = Td + Te; cannam@95: { cannam@95: E T4, T5, T6, T7, T8, T9; cannam@95: T4 = I[WS(is, 2)]; cannam@95: T5 = I[WS(is, 5)]; cannam@95: T6 = T4 - T5; cannam@95: T7 = I[WS(is, 1)]; cannam@95: T8 = I[WS(is, 6)]; cannam@95: T9 = T7 - T8; cannam@95: Ta = KP707106781 * (T6 + T9); cannam@95: Tn = T7 + T8; cannam@95: Tc = KP707106781 * (T6 - T9); cannam@95: Tm = T4 + T5; cannam@95: } cannam@95: } cannam@95: { cannam@95: E Tb, Tg, Tp, Tq; cannam@95: Tb = T3 - Ta; cannam@95: Tg = Tc - Tf; cannam@95: O[WS(os, 3)] = FNMS(KP1_111140466, Tg, KP1_662939224 * Tb); cannam@95: O[WS(os, 5)] = FMA(KP1_662939224, Tg, KP1_111140466 * Tb); cannam@95: Tp = Tj + Tk; cannam@95: Tq = Tm + Tn; cannam@95: O[WS(os, 4)] = KP1_414213562 * (Tp - Tq); cannam@95: O[0] = KP2_000000000 * (Tp + Tq); cannam@95: } cannam@95: { cannam@95: E Th, Ti, Tl, To; cannam@95: Th = T3 + Ta; cannam@95: Ti = Tf + Tc; cannam@95: O[WS(os, 1)] = FNMS(KP390180644, Ti, KP1_961570560 * Th); cannam@95: O[WS(os, 7)] = FMA(KP1_961570560, Ti, KP390180644 * Th); cannam@95: Tl = Tj - Tk; cannam@95: To = Tm - Tn; cannam@95: O[WS(os, 2)] = FNMS(KP765366864, To, KP1_847759065 * Tl); cannam@95: O[WS(os, 6)] = FMA(KP765366864, Tl, KP1_847759065 * To); cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: cannam@95: static const kr2r_desc desc = { 8, "e10_8", {20, 10, 6, 0}, &GENUS, REDFT10 }; cannam@95: cannam@95: void X(codelet_e10_8) (planner *p) { cannam@95: X(kr2r_register) (p, e10_8, &desc); cannam@95: } cannam@95: cannam@95: #endif /* HAVE_FMA */