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 07:42:27 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_r2r.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -redft01 -n 8 -name e01_8 -include r2r.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 26 FP additions, 24 FP multiplications, Chris@10: * (or, 2 additions, 0 multiplications, 24 fused multiply/add), Chris@10: * 27 stack variables, 8 constants, and 16 memory accesses Chris@10: */ Chris@10: #include "r2r.h" Chris@10: Chris@10: static void e01_8(const R *I, R *O, stride is, stride os, INT v, INT ivs, INT ovs) Chris@10: { Chris@10: DK(KP1_961570560, +1.961570560806460898252364472268478073947867462); Chris@10: DK(KP1_662939224, +1.662939224605090474157576755235811513477121624); Chris@10: DK(KP198912367, +0.198912367379658006911597622644676228597850501); Chris@10: DK(KP668178637, +0.668178637919298919997757686523080761552472251); Chris@10: DK(KP707106781, +0.707106781186547524400844362104849039284835938); Chris@10: DK(KP1_847759065, +1.847759065022573512256366378793576573644833252); Chris@10: DK(KP414213562, +0.414213562373095048801688724209698078569671875); Chris@10: DK(KP1_414213562, +1.414213562373095048801688724209698078569671875); Chris@10: { Chris@10: INT i; Chris@10: for (i = v; i > 0; i = i - 1, I = I + ivs, O = O + ovs, MAKE_VOLATILE_STRIDE(16, is), MAKE_VOLATILE_STRIDE(16, os)) { Chris@10: E T8, Td, Th, T7, Tp, Tl, Te, Tb; Chris@10: { Chris@10: E Tj, T3, Tk, T6, T9, Ta; Chris@10: { Chris@10: E T1, T2, T4, T5; Chris@10: T1 = I[0]; Chris@10: T2 = I[WS(is, 4)]; Chris@10: T4 = I[WS(is, 2)]; Chris@10: T5 = I[WS(is, 6)]; Chris@10: T8 = I[WS(is, 1)]; Chris@10: Tj = FNMS(KP1_414213562, T2, T1); Chris@10: T3 = FMA(KP1_414213562, T2, T1); Chris@10: Tk = FMS(KP414213562, T4, T5); Chris@10: T6 = FMA(KP414213562, T5, T4); Chris@10: Td = I[WS(is, 7)]; Chris@10: T9 = I[WS(is, 5)]; Chris@10: Ta = I[WS(is, 3)]; Chris@10: } Chris@10: Th = FNMS(KP1_847759065, T6, T3); Chris@10: T7 = FMA(KP1_847759065, T6, T3); Chris@10: Tp = FNMS(KP1_847759065, Tk, Tj); Chris@10: Tl = FMA(KP1_847759065, Tk, Tj); Chris@10: Te = Ta - T9; Chris@10: Tb = T9 + Ta; Chris@10: } Chris@10: { Chris@10: E Tn, Tf, Tc, Tm; Chris@10: Tn = FNMS(KP707106781, Te, Td); Chris@10: Tf = FMA(KP707106781, Te, Td); Chris@10: Tc = FMA(KP707106781, Tb, T8); Chris@10: Tm = FNMS(KP707106781, Tb, T8); Chris@10: { Chris@10: E Tq, To, Tg, Ti; Chris@10: Tq = FMA(KP668178637, Tm, Tn); Chris@10: To = FNMS(KP668178637, Tn, Tm); Chris@10: Tg = FMA(KP198912367, Tf, Tc); Chris@10: Ti = FNMS(KP198912367, Tc, Tf); Chris@10: O[WS(os, 1)] = FMA(KP1_662939224, To, Tl); Chris@10: O[WS(os, 6)] = FNMS(KP1_662939224, To, Tl); Chris@10: O[WS(os, 2)] = FMA(KP1_662939224, Tq, Tp); Chris@10: O[WS(os, 5)] = FNMS(KP1_662939224, Tq, Tp); Chris@10: O[WS(os, 4)] = FMA(KP1_961570560, Ti, Th); Chris@10: O[WS(os, 3)] = FNMS(KP1_961570560, Ti, Th); Chris@10: O[0] = FMA(KP1_961570560, Tg, T7); Chris@10: O[WS(os, 7)] = FNMS(KP1_961570560, Tg, T7); Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kr2r_desc desc = { 8, "e01_8", {2, 0, 24, 0}, &GENUS, REDFT01 }; Chris@10: Chris@10: void X(codelet_e01_8) (planner *p) { Chris@10: X(kr2r_register) (p, e01_8, &desc); Chris@10: } Chris@10: Chris@10: #else /* HAVE_FMA */ Chris@10: Chris@10: /* Generated by: ../../../genfft/gen_r2r.native -compact -variables 4 -pipeline-latency 4 -redft01 -n 8 -name e01_8 -include r2r.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 26 FP additions, 15 FP multiplications, Chris@10: * (or, 20 additions, 9 multiplications, 6 fused multiply/add), Chris@10: * 28 stack variables, 8 constants, and 16 memory accesses Chris@10: */ Chris@10: #include "r2r.h" Chris@10: Chris@10: static void e01_8(const R *I, R *O, stride is, stride os, INT v, INT ivs, INT ovs) Chris@10: { Chris@10: DK(KP1_662939224, +1.662939224605090474157576755235811513477121624); Chris@10: DK(KP1_111140466, +1.111140466039204449485661627897065748749874382); Chris@10: DK(KP390180644, +0.390180644032256535696569736954044481855383236); Chris@10: DK(KP1_961570560, +1.961570560806460898252364472268478073947867462); Chris@10: DK(KP707106781, +0.707106781186547524400844362104849039284835938); Chris@10: DK(KP1_414213562, +1.414213562373095048801688724209698078569671875); Chris@10: DK(KP765366864, +0.765366864730179543456919968060797733522689125); Chris@10: DK(KP1_847759065, +1.847759065022573512256366378793576573644833252); Chris@10: { Chris@10: INT i; Chris@10: for (i = v; i > 0; i = i - 1, I = I + ivs, O = O + ovs, MAKE_VOLATILE_STRIDE(16, is), MAKE_VOLATILE_STRIDE(16, os)) { Chris@10: E T7, Tl, T4, Tk, Td, To, Tg, Tn; Chris@10: { Chris@10: E T5, T6, T1, T3, T2; Chris@10: T5 = I[WS(is, 2)]; Chris@10: T6 = I[WS(is, 6)]; Chris@10: T7 = FMA(KP1_847759065, T5, KP765366864 * T6); Chris@10: Tl = FNMS(KP1_847759065, T6, KP765366864 * T5); Chris@10: T1 = I[0]; Chris@10: T2 = I[WS(is, 4)]; Chris@10: T3 = KP1_414213562 * T2; Chris@10: T4 = T1 + T3; Chris@10: Tk = T1 - T3; Chris@10: { Chris@10: E T9, Tf, Tc, Te, Ta, Tb; Chris@10: T9 = I[WS(is, 1)]; Chris@10: Tf = I[WS(is, 7)]; Chris@10: Ta = I[WS(is, 5)]; Chris@10: Tb = I[WS(is, 3)]; Chris@10: Tc = KP707106781 * (Ta + Tb); Chris@10: Te = KP707106781 * (Ta - Tb); Chris@10: Td = T9 + Tc; Chris@10: To = Te + Tf; Chris@10: Tg = Te - Tf; Chris@10: Tn = T9 - Tc; Chris@10: } Chris@10: } Chris@10: { Chris@10: E T8, Th, Tq, Tr; Chris@10: T8 = T4 + T7; Chris@10: Th = FNMS(KP390180644, Tg, KP1_961570560 * Td); Chris@10: O[WS(os, 7)] = T8 - Th; Chris@10: O[0] = T8 + Th; Chris@10: Tq = Tk - Tl; Chris@10: Tr = FMA(KP1_111140466, Tn, KP1_662939224 * To); Chris@10: O[WS(os, 5)] = Tq - Tr; Chris@10: O[WS(os, 2)] = Tq + Tr; Chris@10: } Chris@10: { Chris@10: E Ti, Tj, Tm, Tp; Chris@10: Ti = T4 - T7; Chris@10: Tj = FMA(KP390180644, Td, KP1_961570560 * Tg); Chris@10: O[WS(os, 4)] = Ti - Tj; Chris@10: O[WS(os, 3)] = Ti + Tj; Chris@10: Tm = Tk + Tl; Chris@10: Tp = FNMS(KP1_111140466, To, KP1_662939224 * Tn); Chris@10: O[WS(os, 6)] = Tm - Tp; Chris@10: O[WS(os, 1)] = Tm + Tp; Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kr2r_desc desc = { 8, "e01_8", {20, 9, 6, 0}, &GENUS, REDFT01 }; Chris@10: Chris@10: void X(codelet_e01_8) (planner *p) { Chris@10: X(kr2r_register) (p, e01_8, &desc); Chris@10: } Chris@10: Chris@10: #endif /* HAVE_FMA */