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 16:36:05 EDT 2016 */ Chris@42: Chris@42: #include "codelet-dft.h" Chris@42: Chris@42: #ifdef HAVE_FMA Chris@42: Chris@42: /* Generated by: ../../../genfft/gen_twiddle.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 3 -name t1_3 -include t.h */ Chris@42: Chris@42: /* Chris@42: * This function contains 16 FP additions, 14 FP multiplications, Chris@42: * (or, 6 additions, 4 multiplications, 10 fused multiply/add), Chris@42: * 21 stack variables, 2 constants, and 12 memory accesses Chris@42: */ Chris@42: #include "t.h" Chris@42: Chris@42: static void t1_3(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) Chris@42: { Chris@42: DK(KP866025403, +0.866025403784438646763723170752936183471402627); Chris@42: DK(KP500000000, +0.500000000000000000000000000000000000000000000); Chris@42: { Chris@42: INT m; Chris@42: for (m = mb, W = W + (mb * 4); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 4, MAKE_VOLATILE_STRIDE(6, rs)) { Chris@42: E T1, Tm, T9, Tc, Tb, Th, T7, Ti, Ta, Tj, Td; Chris@42: T1 = ri[0]; Chris@42: Tm = ii[0]; Chris@42: { Chris@42: E T3, T6, T2, T5, Tg, T4, T8; Chris@42: T3 = ri[WS(rs, 1)]; Chris@42: T6 = ii[WS(rs, 1)]; Chris@42: T2 = W[0]; Chris@42: T5 = W[1]; Chris@42: T9 = ri[WS(rs, 2)]; Chris@42: Tc = ii[WS(rs, 2)]; Chris@42: Tg = T2 * T6; Chris@42: T4 = T2 * T3; Chris@42: T8 = W[2]; Chris@42: Tb = W[3]; Chris@42: Th = FNMS(T5, T3, Tg); Chris@42: T7 = FMA(T5, T6, T4); Chris@42: Ti = T8 * Tc; Chris@42: Ta = T8 * T9; Chris@42: } Chris@42: Tj = FNMS(Tb, T9, Ti); Chris@42: Td = FMA(Tb, Tc, Ta); Chris@42: { Chris@42: E Tk, Te, To, Tn, Tl, Tf; Chris@42: Tk = Th - Tj; Chris@42: Tl = Th + Tj; Chris@42: Te = T7 + Td; Chris@42: To = Td - T7; Chris@42: ii[0] = Tl + Tm; Chris@42: Tn = FNMS(KP500000000, Tl, Tm); Chris@42: ri[0] = T1 + Te; Chris@42: Tf = FNMS(KP500000000, Te, T1); Chris@42: ii[WS(rs, 1)] = FMA(KP866025403, To, Tn); Chris@42: ii[WS(rs, 2)] = FNMS(KP866025403, To, Tn); Chris@42: ri[WS(rs, 2)] = FNMS(KP866025403, Tk, Tf); Chris@42: ri[WS(rs, 1)] = FMA(KP866025403, Tk, Tf); Chris@42: } Chris@42: } Chris@42: } Chris@42: } Chris@42: Chris@42: static const tw_instr twinstr[] = { Chris@42: {TW_FULL, 0, 3}, Chris@42: {TW_NEXT, 1, 0} Chris@42: }; Chris@42: Chris@42: static const ct_desc desc = { 3, "t1_3", twinstr, &GENUS, {6, 4, 10, 0}, 0, 0, 0 }; Chris@42: Chris@42: void X(codelet_t1_3) (planner *p) { Chris@42: X(kdft_dit_register) (p, t1_3, &desc); Chris@42: } Chris@42: #else /* HAVE_FMA */ Chris@42: Chris@42: /* Generated by: ../../../genfft/gen_twiddle.native -compact -variables 4 -pipeline-latency 4 -n 3 -name t1_3 -include t.h */ Chris@42: Chris@42: /* Chris@42: * This function contains 16 FP additions, 12 FP multiplications, Chris@42: * (or, 10 additions, 6 multiplications, 6 fused multiply/add), Chris@42: * 15 stack variables, 2 constants, and 12 memory accesses Chris@42: */ Chris@42: #include "t.h" Chris@42: Chris@42: static void t1_3(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) Chris@42: { Chris@42: DK(KP866025403, +0.866025403784438646763723170752936183471402627); Chris@42: DK(KP500000000, +0.500000000000000000000000000000000000000000000); Chris@42: { Chris@42: INT m; Chris@42: for (m = mb, W = W + (mb * 4); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 4, MAKE_VOLATILE_STRIDE(6, rs)) { Chris@42: E T1, Ti, T6, Te, Tb, Tf, Tc, Th; Chris@42: T1 = ri[0]; Chris@42: Ti = ii[0]; Chris@42: { Chris@42: E T3, T5, T2, T4; Chris@42: T3 = ri[WS(rs, 1)]; Chris@42: T5 = ii[WS(rs, 1)]; Chris@42: T2 = W[0]; Chris@42: T4 = W[1]; Chris@42: T6 = FMA(T2, T3, T4 * T5); Chris@42: Te = FNMS(T4, T3, T2 * T5); Chris@42: } Chris@42: { Chris@42: E T8, Ta, T7, T9; Chris@42: T8 = ri[WS(rs, 2)]; Chris@42: Ta = ii[WS(rs, 2)]; Chris@42: T7 = W[2]; Chris@42: T9 = W[3]; Chris@42: Tb = FMA(T7, T8, T9 * Ta); Chris@42: Tf = FNMS(T9, T8, T7 * Ta); Chris@42: } Chris@42: Tc = T6 + Tb; Chris@42: Th = Te + Tf; Chris@42: ri[0] = T1 + Tc; Chris@42: ii[0] = Th + Ti; Chris@42: { Chris@42: E Td, Tg, Tj, Tk; Chris@42: Td = FNMS(KP500000000, Tc, T1); Chris@42: Tg = KP866025403 * (Te - Tf); Chris@42: ri[WS(rs, 2)] = Td - Tg; Chris@42: ri[WS(rs, 1)] = Td + Tg; Chris@42: Tj = KP866025403 * (Tb - T6); Chris@42: Tk = FNMS(KP500000000, Th, Ti); Chris@42: ii[WS(rs, 1)] = Tj + Tk; Chris@42: ii[WS(rs, 2)] = Tk - Tj; Chris@42: } Chris@42: } Chris@42: } Chris@42: } Chris@42: Chris@42: static const tw_instr twinstr[] = { Chris@42: {TW_FULL, 0, 3}, Chris@42: {TW_NEXT, 1, 0} Chris@42: }; Chris@42: Chris@42: static const ct_desc desc = { 3, "t1_3", twinstr, &GENUS, {10, 6, 6, 0}, 0, 0, 0 }; Chris@42: Chris@42: void X(codelet_t1_3) (planner *p) { Chris@42: X(kdft_dit_register) (p, t1_3, &desc); Chris@42: } Chris@42: #endif /* HAVE_FMA */