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 07:40:15 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_r2cf.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 15 -name r2cfII_15 -dft-II -include r2cfII.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 72 FP additions, 41 FP multiplications, Chris@10: * (or, 38 additions, 7 multiplications, 34 fused multiply/add), Chris@10: * 57 stack variables, 12 constants, and 30 memory accesses Chris@10: */ Chris@10: #include "r2cfII.h" Chris@10: Chris@10: static void r2cfII_15(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs) Chris@10: { Chris@10: DK(KP823639103, +0.823639103546331925877420039278190003029660514); Chris@10: DK(KP910592997, +0.910592997310029334643087372129977886038870291); Chris@10: DK(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@10: DK(KP866025403, +0.866025403784438646763723170752936183471402627); Chris@10: DK(KP500000000, +0.500000000000000000000000000000000000000000000); Chris@10: DK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@10: DK(KP690983005, +0.690983005625052575897706582817180941139845410); Chris@10: DK(KP552786404, +0.552786404500042060718165266253744752911876328); Chris@10: DK(KP447213595, +0.447213595499957939281834733746255247088123672); Chris@10: DK(KP809016994, +0.809016994374947424102293417182819058860154590); Chris@10: DK(KP618033988, +0.618033988749894848204586834365638117720309180); Chris@10: DK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@10: { Chris@10: INT i; Chris@10: for (i = v; i > 0; i = i - 1, R0 = R0 + ivs, R1 = R1 + ivs, Cr = Cr + ovs, Ci = Ci + ovs, MAKE_VOLATILE_STRIDE(60, rs), MAKE_VOLATILE_STRIDE(60, csr), MAKE_VOLATILE_STRIDE(60, csi)) { Chris@10: E T9, TQ, TV, TW, Tw, TJ; Chris@10: { Chris@10: E Ta, Tl, Tg, T8, T7, TF, TX, TT, Tm, Th, TM, TZ, Tr, Tn, Tj; Chris@10: E Tz, To, TN, TH, Tp, TO; Chris@10: Ta = R0[WS(rs, 5)]; Chris@10: Tl = R1[WS(rs, 2)]; Chris@10: { Chris@10: E T1, T2, T5, T3, T4; Chris@10: T1 = R0[0]; Chris@10: T2 = R0[WS(rs, 3)]; Chris@10: T5 = R1[WS(rs, 4)]; Chris@10: T3 = R0[WS(rs, 6)]; Chris@10: T4 = R1[WS(rs, 1)]; Chris@10: { Chris@10: E Tb, TL, Te, TK, TR, Tf, Ti, Ty; Chris@10: Tb = R1[0]; Chris@10: TR = T2 + T5; Chris@10: Tg = R0[WS(rs, 2)]; Chris@10: { Chris@10: E T6, TS, Tc, Td; Chris@10: T6 = T2 + T3 - T4 - T5; Chris@10: T8 = (T3 + T5 - T2) - T4; Chris@10: TS = T3 + T4; Chris@10: Tc = R1[WS(rs, 3)]; Chris@10: Td = R1[WS(rs, 6)]; Chris@10: T7 = FNMS(KP250000000, T6, T1); Chris@10: TF = T1 + T6; Chris@10: TX = FNMS(KP618033988, TR, TS); Chris@10: TT = FMA(KP618033988, TS, TR); Chris@10: TL = Tc - Td; Chris@10: Te = Tc + Td; Chris@10: } Chris@10: TK = Tg + Tb; Chris@10: Tm = R0[WS(rs, 7)]; Chris@10: Tf = Tb - Te; Chris@10: Th = Tb + Te; Chris@10: TM = FMA(KP618033988, TL, TK); Chris@10: TZ = FNMS(KP618033988, TK, TL); Chris@10: Ti = FMA(KP809016994, Th, Tg); Chris@10: Ty = FMA(KP447213595, Th, Tf); Chris@10: Tr = R1[WS(rs, 5)]; Chris@10: Tn = R0[WS(rs, 1)]; Chris@10: Tj = FNMS(KP552786404, Ti, Tf); Chris@10: Tz = FNMS(KP690983005, Ty, Tg); Chris@10: To = R0[WS(rs, 4)]; Chris@10: TN = Tr + Tm; Chris@10: } Chris@10: } Chris@10: TH = Ta + Tg - Th; Chris@10: Tp = Tn + To; Chris@10: TO = To - Tn; Chris@10: { Chris@10: E Tx, TA, TP, T14, T11, Tu, TD; Chris@10: { Chris@10: E T10, TI, TC, TY; Chris@10: T9 = FNMS(KP559016994, T8, T7); Chris@10: Tx = FMA(KP559016994, T8, T7); Chris@10: TA = FNMS(KP809016994, Tz, Ta); Chris@10: TP = FMA(KP618033988, TO, TN); Chris@10: TY = FNMS(KP618033988, TN, TO); Chris@10: { Chris@10: E Tq, Ts, TG, Tt, TB; Chris@10: Tq = Tm - Tp; Chris@10: Ts = Tm + Tp; Chris@10: T14 = TZ - TY; Chris@10: T10 = TY + TZ; Chris@10: TG = Ts - Tr - Tl; Chris@10: Tt = FMA(KP809016994, Ts, Tr); Chris@10: TB = FMA(KP447213595, Ts, Tq); Chris@10: T11 = FMA(KP500000000, T10, TX); Chris@10: Ci[WS(csi, 2)] = KP866025403 * (TH - TG); Chris@10: TI = TG + TH; Chris@10: Tu = FNMS(KP552786404, Tt, Tq); Chris@10: TC = FNMS(KP690983005, TB, Tr); Chris@10: } Chris@10: Ci[WS(csi, 1)] = KP951056516 * (T10 - TX); Chris@10: Cr[WS(csr, 7)] = TF + TI; Chris@10: Cr[WS(csr, 2)] = FNMS(KP500000000, TI, TF); Chris@10: TD = FNMS(KP809016994, TC, Tl); Chris@10: } Chris@10: { Chris@10: E TU, Tk, T13, Tv, T12, TE; Chris@10: TQ = TM - TP; Chris@10: TU = TP + TM; Chris@10: T12 = TD + TA; Chris@10: TE = TA - TD; Chris@10: Tk = FNMS(KP559016994, Tj, Ta); Chris@10: TV = FMA(KP500000000, TU, TT); Chris@10: Ci[WS(csi, 6)] = -(KP951056516 * (FMA(KP910592997, T12, T11))); Chris@10: Ci[WS(csi, 3)] = KP951056516 * (FNMS(KP910592997, T12, T11)); Chris@10: T13 = FNMS(KP500000000, TE, Tx); Chris@10: Cr[WS(csr, 1)] = Tx + TE; Chris@10: Tv = FNMS(KP559016994, Tu, Tl); Chris@10: Ci[WS(csi, 4)] = KP951056516 * (TT - TU); Chris@10: Cr[WS(csr, 6)] = FMA(KP823639103, T14, T13); Chris@10: Cr[WS(csr, 3)] = FNMS(KP823639103, T14, T13); Chris@10: TW = Tv + Tk; Chris@10: Tw = Tk - Tv; Chris@10: } Chris@10: } Chris@10: } Chris@10: Ci[WS(csi, 5)] = -(KP951056516 * (FNMS(KP910592997, TW, TV))); Chris@10: Ci[0] = -(KP951056516 * (FMA(KP910592997, TW, TV))); Chris@10: TJ = FNMS(KP500000000, Tw, T9); Chris@10: Cr[WS(csr, 4)] = T9 + Tw; Chris@10: Cr[0] = FMA(KP823639103, TQ, TJ); Chris@10: Cr[WS(csr, 5)] = FNMS(KP823639103, TQ, TJ); Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kr2c_desc desc = { 15, "r2cfII_15", {38, 7, 34, 0}, &GENUS }; Chris@10: Chris@10: void X(codelet_r2cfII_15) (planner *p) { Chris@10: X(kr2c_register) (p, r2cfII_15, &desc); Chris@10: } Chris@10: Chris@10: #else /* HAVE_FMA */ Chris@10: Chris@10: /* Generated by: ../../../genfft/gen_r2cf.native -compact -variables 4 -pipeline-latency 4 -n 15 -name r2cfII_15 -dft-II -include r2cfII.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 72 FP additions, 33 FP multiplications, Chris@10: * (or, 54 additions, 15 multiplications, 18 fused multiply/add), Chris@10: * 37 stack variables, 8 constants, and 30 memory accesses Chris@10: */ Chris@10: #include "r2cfII.h" Chris@10: Chris@10: static void r2cfII_15(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs) Chris@10: { Chris@10: DK(KP500000000, +0.500000000000000000000000000000000000000000000); Chris@10: DK(KP866025403, +0.866025403784438646763723170752936183471402627); Chris@10: DK(KP809016994, +0.809016994374947424102293417182819058860154590); Chris@10: DK(KP309016994, +0.309016994374947424102293417182819058860154590); Chris@10: DK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@10: DK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@10: DK(KP587785252, +0.587785252292473129168705954639072768597652438); Chris@10: DK(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@10: { Chris@10: INT i; Chris@10: for (i = v; i > 0; i = i - 1, R0 = R0 + ivs, R1 = R1 + ivs, Cr = Cr + ovs, Ci = Ci + ovs, MAKE_VOLATILE_STRIDE(60, rs), MAKE_VOLATILE_STRIDE(60, csr), MAKE_VOLATILE_STRIDE(60, csi)) { Chris@10: E T1, T2, Tx, TR, TE, T7, TD, Th, Tm, Tr, TQ, TA, TB, Tf, Te; Chris@10: E Tu, TS, Td, TH, TO; Chris@10: T1 = R0[WS(rs, 5)]; Chris@10: { Chris@10: E T3, Tv, T6, Tw, T4, T5; Chris@10: T2 = R0[WS(rs, 2)]; Chris@10: T3 = R1[0]; Chris@10: Tv = T2 + T3; Chris@10: T4 = R1[WS(rs, 3)]; Chris@10: T5 = R1[WS(rs, 6)]; Chris@10: T6 = T4 + T5; Chris@10: Tw = T4 - T5; Chris@10: Tx = FMA(KP951056516, Tv, KP587785252 * Tw); Chris@10: TR = FNMS(KP587785252, Tv, KP951056516 * Tw); Chris@10: TE = KP559016994 * (T3 - T6); Chris@10: T7 = T3 + T6; Chris@10: TD = KP250000000 * T7; Chris@10: } Chris@10: { Chris@10: E Ti, Tl, Tj, Tk, Tp, Tq; Chris@10: Th = R0[0]; Chris@10: Ti = R1[WS(rs, 4)]; Chris@10: Tl = R0[WS(rs, 6)]; Chris@10: Tj = R1[WS(rs, 1)]; Chris@10: Tk = R0[WS(rs, 3)]; Chris@10: Tp = Tk + Ti; Chris@10: Tq = Tl + Tj; Chris@10: Tm = Ti + Tj - (Tk + Tl); Chris@10: Tr = FMA(KP951056516, Tp, KP587785252 * Tq); Chris@10: TQ = FNMS(KP951056516, Tq, KP587785252 * Tp); Chris@10: TA = FMA(KP250000000, Tm, Th); Chris@10: TB = KP559016994 * (Tl + Ti - (Tk + Tj)); Chris@10: } Chris@10: { Chris@10: E T9, Tt, Tc, Ts, Ta, Tb, TG; Chris@10: Tf = R1[WS(rs, 2)]; Chris@10: T9 = R0[WS(rs, 7)]; Chris@10: Te = R1[WS(rs, 5)]; Chris@10: Tt = T9 + Te; Chris@10: Ta = R0[WS(rs, 1)]; Chris@10: Tb = R0[WS(rs, 4)]; Chris@10: Tc = Ta + Tb; Chris@10: Ts = Ta - Tb; Chris@10: Tu = FNMS(KP951056516, Tt, KP587785252 * Ts); Chris@10: TS = FMA(KP951056516, Ts, KP587785252 * Tt); Chris@10: Td = T9 + Tc; Chris@10: TG = KP559016994 * (T9 - Tc); Chris@10: TH = FNMS(KP309016994, Te, TG) + FNMA(KP250000000, Td, Tf); Chris@10: TO = FMS(KP809016994, Te, Tf) + FNMA(KP250000000, Td, TG); Chris@10: } Chris@10: { Chris@10: E Tn, T8, Tg, To; Chris@10: Tn = Th - Tm; Chris@10: T8 = T1 + T2 - T7; Chris@10: Tg = Td - Te - Tf; Chris@10: To = T8 + Tg; Chris@10: Ci[WS(csi, 2)] = KP866025403 * (T8 - Tg); Chris@10: Cr[WS(csr, 2)] = FNMS(KP500000000, To, Tn); Chris@10: Cr[WS(csr, 7)] = Tn + To; Chris@10: } Chris@10: { Chris@10: E TM, TX, TT, TV, TP, TU, TN, TW; Chris@10: TM = TB + TA; Chris@10: TX = KP866025403 * (TR + TS); Chris@10: TT = TR - TS; Chris@10: TV = FMS(KP500000000, TT, TQ); Chris@10: TN = T1 + TE + FNMS(KP809016994, T2, TD); Chris@10: TP = TN + TO; Chris@10: TU = KP866025403 * (TO - TN); Chris@10: Cr[WS(csr, 1)] = TM + TP; Chris@10: Ci[WS(csi, 1)] = TQ + TT; Chris@10: Ci[WS(csi, 6)] = TU - TV; Chris@10: Ci[WS(csi, 3)] = TU + TV; Chris@10: TW = FNMS(KP500000000, TP, TM); Chris@10: Cr[WS(csr, 3)] = TW - TX; Chris@10: Cr[WS(csr, 6)] = TW + TX; Chris@10: } Chris@10: { Chris@10: E Tz, TC, Ty, TK, TI, TL, TF, TJ; Chris@10: Tz = KP866025403 * (Tx + Tu); Chris@10: TC = TA - TB; Chris@10: Ty = Tu - Tx; Chris@10: TK = FMS(KP500000000, Ty, Tr); Chris@10: TF = FMA(KP309016994, T2, T1) + TD - TE; Chris@10: TI = TF + TH; Chris@10: TL = KP866025403 * (TH - TF); Chris@10: Ci[WS(csi, 4)] = Tr + Ty; Chris@10: Cr[WS(csr, 4)] = TC + TI; Chris@10: Ci[WS(csi, 5)] = TK - TL; Chris@10: Ci[0] = TK + TL; Chris@10: TJ = FNMS(KP500000000, TI, TC); Chris@10: Cr[0] = Tz + TJ; Chris@10: Cr[WS(csr, 5)] = TJ - Tz; Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kr2c_desc desc = { 15, "r2cfII_15", {54, 15, 18, 0}, &GENUS }; Chris@10: Chris@10: void X(codelet_r2cfII_15) (planner *p) { Chris@10: X(kr2c_register) (p, r2cfII_15, &desc); Chris@10: } Chris@10: Chris@10: #endif /* HAVE_FMA */