Chris@19: /* Chris@19: * Copyright (c) 2003, 2007-14 Matteo Frigo Chris@19: * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology Chris@19: * Chris@19: * This program is free software; you can redistribute it and/or modify Chris@19: * it under the terms of the GNU General Public License as published by Chris@19: * the Free Software Foundation; either version 2 of the License, or Chris@19: * (at your option) any later version. Chris@19: * Chris@19: * This program is distributed in the hope that it will be useful, Chris@19: * but WITHOUT ANY WARRANTY; without even the implied warranty of Chris@19: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the Chris@19: * GNU General Public License for more details. Chris@19: * Chris@19: * You should have received a copy of the GNU General Public License Chris@19: * along with this program; if not, write to the Free Software Chris@19: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA Chris@19: * Chris@19: */ Chris@19: Chris@19: /* This file was automatically generated --- DO NOT EDIT */ Chris@19: /* Generated on Tue Mar 4 13:50:25 EST 2014 */ Chris@19: Chris@19: #include "codelet-rdft.h" Chris@19: Chris@19: #ifdef HAVE_FMA Chris@19: Chris@19: /* Generated by: ../../../genfft/gen_hc2hc.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -sign 1 -n 5 -dif -name hb_5 -include hb.h */ Chris@19: Chris@19: /* Chris@19: * This function contains 40 FP additions, 34 FP multiplications, Chris@19: * (or, 14 additions, 8 multiplications, 26 fused multiply/add), Chris@19: * 42 stack variables, 4 constants, and 20 memory accesses Chris@19: */ Chris@19: #include "hb.h" Chris@19: Chris@19: static void hb_5(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms) Chris@19: { Chris@19: DK(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@19: DK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@19: DK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@19: DK(KP618033988, +0.618033988749894848204586834365638117720309180); Chris@19: { Chris@19: INT m; Chris@19: for (m = mb, W = W + ((mb - 1) * 8); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 8, MAKE_VOLATILE_STRIDE(10, rs)) { Chris@19: E TQ, TP, TT, TR, TS, TU; Chris@19: { Chris@19: E T1, Tn, TM, Tw, Tb, T8, To, Tf, Ta, Tg, Th; Chris@19: { Chris@19: E T2, T3, T5, T6, T4, Tu; Chris@19: T1 = cr[0]; Chris@19: T2 = cr[WS(rs, 1)]; Chris@19: T3 = ci[0]; Chris@19: T5 = cr[WS(rs, 2)]; Chris@19: T6 = ci[WS(rs, 1)]; Chris@19: Tn = ci[WS(rs, 4)]; Chris@19: T4 = T2 + T3; Chris@19: Tu = T2 - T3; Chris@19: { Chris@19: E T7, Tv, Td, Te; Chris@19: T7 = T5 + T6; Chris@19: Tv = T5 - T6; Chris@19: Td = ci[WS(rs, 3)]; Chris@19: Te = cr[WS(rs, 4)]; Chris@19: TM = FNMS(KP618033988, Tu, Tv); Chris@19: Tw = FMA(KP618033988, Tv, Tu); Chris@19: Tb = T4 - T7; Chris@19: T8 = T4 + T7; Chris@19: To = Td - Te; Chris@19: Tf = Td + Te; Chris@19: Ta = FNMS(KP250000000, T8, T1); Chris@19: Tg = ci[WS(rs, 2)]; Chris@19: Th = cr[WS(rs, 3)]; Chris@19: } Chris@19: } Chris@19: cr[0] = T1 + T8; Chris@19: { Chris@19: E TG, T9, Tm, Tz, TH, TC, TA, Tk, Tt, TL, Tc, Ti, Tp, TI; Chris@19: TG = FNMS(KP559016994, Tb, Ta); Chris@19: Tc = FMA(KP559016994, Tb, Ta); Chris@19: T9 = W[0]; Chris@19: Ti = Tg + Th; Chris@19: Tp = Tg - Th; Chris@19: Tm = W[1]; Chris@19: { Chris@19: E Ts, Tj, Tr, Tq; Chris@19: Tz = W[6]; Chris@19: Ts = To - Tp; Chris@19: Tq = To + Tp; Chris@19: Tj = FMA(KP618033988, Ti, Tf); Chris@19: TH = FNMS(KP618033988, Tf, Ti); Chris@19: ci[0] = Tn + Tq; Chris@19: Tr = FNMS(KP250000000, Tq, Tn); Chris@19: TC = W[7]; Chris@19: TA = FMA(KP951056516, Tj, Tc); Chris@19: Tk = FNMS(KP951056516, Tj, Tc); Chris@19: Tt = FMA(KP559016994, Ts, Tr); Chris@19: TL = FNMS(KP559016994, Ts, Tr); Chris@19: } Chris@19: { Chris@19: E TE, TB, Ty, Tl, TD, Tx; Chris@19: TE = TC * TA; Chris@19: TB = Tz * TA; Chris@19: Ty = Tm * Tk; Chris@19: Tl = T9 * Tk; Chris@19: TD = FNMS(KP951056516, Tw, Tt); Chris@19: Tx = FMA(KP951056516, Tw, Tt); Chris@19: TI = FMA(KP951056516, TH, TG); Chris@19: TQ = FNMS(KP951056516, TH, TG); Chris@19: ci[WS(rs, 4)] = FMA(Tz, TD, TE); Chris@19: cr[WS(rs, 4)] = FNMS(TC, TD, TB); Chris@19: ci[WS(rs, 1)] = FMA(T9, Tx, Ty); Chris@19: cr[WS(rs, 1)] = FNMS(Tm, Tx, Tl); Chris@19: } Chris@19: { Chris@19: E TF, TK, TN, TJ, TO; Chris@19: TF = W[2]; Chris@19: TK = W[3]; Chris@19: TP = W[4]; Chris@19: TT = FMA(KP951056516, TM, TL); Chris@19: TN = FNMS(KP951056516, TM, TL); Chris@19: TJ = TF * TI; Chris@19: TO = TK * TI; Chris@19: TR = TP * TQ; Chris@19: TS = W[5]; Chris@19: cr[WS(rs, 2)] = FNMS(TK, TN, TJ); Chris@19: ci[WS(rs, 2)] = FMA(TF, TN, TO); Chris@19: } Chris@19: } Chris@19: } Chris@19: cr[WS(rs, 3)] = FNMS(TS, TT, TR); Chris@19: TU = TS * TQ; Chris@19: ci[WS(rs, 3)] = FMA(TP, TT, TU); Chris@19: } Chris@19: } Chris@19: } Chris@19: Chris@19: static const tw_instr twinstr[] = { Chris@19: {TW_FULL, 1, 5}, Chris@19: {TW_NEXT, 1, 0} Chris@19: }; Chris@19: Chris@19: static const hc2hc_desc desc = { 5, "hb_5", twinstr, &GENUS, {14, 8, 26, 0} }; Chris@19: Chris@19: void X(codelet_hb_5) (planner *p) { Chris@19: X(khc2hc_register) (p, hb_5, &desc); Chris@19: } Chris@19: #else /* HAVE_FMA */ Chris@19: Chris@19: /* Generated by: ../../../genfft/gen_hc2hc.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 5 -dif -name hb_5 -include hb.h */ Chris@19: Chris@19: /* Chris@19: * This function contains 40 FP additions, 28 FP multiplications, Chris@19: * (or, 26 additions, 14 multiplications, 14 fused multiply/add), Chris@19: * 27 stack variables, 4 constants, and 20 memory accesses Chris@19: */ Chris@19: #include "hb.h" Chris@19: Chris@19: static void hb_5(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms) Chris@19: { Chris@19: DK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@19: DK(KP587785252, +0.587785252292473129168705954639072768597652438); Chris@19: DK(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@19: DK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@19: { Chris@19: INT m; Chris@19: for (m = mb, W = W + ((mb - 1) * 8); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 8, MAKE_VOLATILE_STRIDE(10, rs)) { Chris@19: E T1, Tj, TG, Ts, T8, Ti, T9, Tn, TD, Tu, Tg, Tt; Chris@19: { Chris@19: E T4, Tq, T7, Tr; Chris@19: T1 = cr[0]; Chris@19: { Chris@19: E T2, T3, T5, T6; Chris@19: T2 = cr[WS(rs, 1)]; Chris@19: T3 = ci[0]; Chris@19: T4 = T2 + T3; Chris@19: Tq = T2 - T3; Chris@19: T5 = cr[WS(rs, 2)]; Chris@19: T6 = ci[WS(rs, 1)]; Chris@19: T7 = T5 + T6; Chris@19: Tr = T5 - T6; Chris@19: } Chris@19: Tj = KP559016994 * (T4 - T7); Chris@19: TG = FMA(KP951056516, Tq, KP587785252 * Tr); Chris@19: Ts = FNMS(KP951056516, Tr, KP587785252 * Tq); Chris@19: T8 = T4 + T7; Chris@19: Ti = FNMS(KP250000000, T8, T1); Chris@19: } Chris@19: { Chris@19: E Tc, Tl, Tf, Tm; Chris@19: T9 = ci[WS(rs, 4)]; Chris@19: { Chris@19: E Ta, Tb, Td, Te; Chris@19: Ta = ci[WS(rs, 3)]; Chris@19: Tb = cr[WS(rs, 4)]; Chris@19: Tc = Ta - Tb; Chris@19: Tl = Ta + Tb; Chris@19: Td = ci[WS(rs, 2)]; Chris@19: Te = cr[WS(rs, 3)]; Chris@19: Tf = Td - Te; Chris@19: Tm = Td + Te; Chris@19: } Chris@19: Tn = FNMS(KP951056516, Tm, KP587785252 * Tl); Chris@19: TD = FMA(KP951056516, Tl, KP587785252 * Tm); Chris@19: Tu = KP559016994 * (Tc - Tf); Chris@19: Tg = Tc + Tf; Chris@19: Tt = FNMS(KP250000000, Tg, T9); Chris@19: } Chris@19: cr[0] = T1 + T8; Chris@19: ci[0] = T9 + Tg; Chris@19: { Chris@19: E To, Ty, Tw, TA, Tk, Tv; Chris@19: Tk = Ti - Tj; Chris@19: To = Tk - Tn; Chris@19: Ty = Tk + Tn; Chris@19: Tv = Tt - Tu; Chris@19: Tw = Ts + Tv; Chris@19: TA = Tv - Ts; Chris@19: { Chris@19: E Th, Tp, Tx, Tz; Chris@19: Th = W[2]; Chris@19: Tp = W[3]; Chris@19: cr[WS(rs, 2)] = FNMS(Tp, Tw, Th * To); Chris@19: ci[WS(rs, 2)] = FMA(Th, Tw, Tp * To); Chris@19: Tx = W[4]; Chris@19: Tz = W[5]; Chris@19: cr[WS(rs, 3)] = FNMS(Tz, TA, Tx * Ty); Chris@19: ci[WS(rs, 3)] = FMA(Tx, TA, Tz * Ty); Chris@19: } Chris@19: } Chris@19: { Chris@19: E TE, TK, TI, TM, TC, TH; Chris@19: TC = Tj + Ti; Chris@19: TE = TC - TD; Chris@19: TK = TC + TD; Chris@19: TH = Tu + Tt; Chris@19: TI = TG + TH; Chris@19: TM = TH - TG; Chris@19: { Chris@19: E TB, TF, TJ, TL; Chris@19: TB = W[0]; Chris@19: TF = W[1]; Chris@19: cr[WS(rs, 1)] = FNMS(TF, TI, TB * TE); Chris@19: ci[WS(rs, 1)] = FMA(TB, TI, TF * TE); Chris@19: TJ = W[6]; Chris@19: TL = W[7]; Chris@19: cr[WS(rs, 4)] = FNMS(TL, TM, TJ * TK); Chris@19: ci[WS(rs, 4)] = FMA(TJ, TM, TL * TK); Chris@19: } Chris@19: } Chris@19: } Chris@19: } Chris@19: } Chris@19: Chris@19: static const tw_instr twinstr[] = { Chris@19: {TW_FULL, 1, 5}, Chris@19: {TW_NEXT, 1, 0} Chris@19: }; Chris@19: Chris@19: static const hc2hc_desc desc = { 5, "hb_5", twinstr, &GENUS, {26, 14, 14, 0} }; Chris@19: Chris@19: void X(codelet_hb_5) (planner *p) { Chris@19: X(khc2hc_register) (p, hb_5, &desc); Chris@19: } Chris@19: #endif /* HAVE_FMA */