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 07:40:15 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_r2cf.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 15 -name r2cfII_15 -dft-II -include r2cfII.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 72 FP additions, 41 FP multiplications, cannam@95: * (or, 38 additions, 7 multiplications, 34 fused multiply/add), cannam@95: * 57 stack variables, 12 constants, and 30 memory accesses cannam@95: */ cannam@95: #include "r2cfII.h" cannam@95: cannam@95: static void r2cfII_15(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs) cannam@95: { cannam@95: DK(KP823639103, +0.823639103546331925877420039278190003029660514); cannam@95: DK(KP910592997, +0.910592997310029334643087372129977886038870291); cannam@95: DK(KP951056516, +0.951056516295153572116439333379382143405698634); cannam@95: DK(KP866025403, +0.866025403784438646763723170752936183471402627); cannam@95: DK(KP500000000, +0.500000000000000000000000000000000000000000000); cannam@95: DK(KP559016994, +0.559016994374947424102293417182819058860154590); cannam@95: DK(KP690983005, +0.690983005625052575897706582817180941139845410); cannam@95: DK(KP552786404, +0.552786404500042060718165266253744752911876328); cannam@95: DK(KP447213595, +0.447213595499957939281834733746255247088123672); cannam@95: DK(KP809016994, +0.809016994374947424102293417182819058860154590); cannam@95: DK(KP618033988, +0.618033988749894848204586834365638117720309180); cannam@95: DK(KP250000000, +0.250000000000000000000000000000000000000000000); cannam@95: { cannam@95: INT i; cannam@95: 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)) { cannam@95: E T9, TQ, TV, TW, Tw, TJ; cannam@95: { cannam@95: E Ta, Tl, Tg, T8, T7, TF, TX, TT, Tm, Th, TM, TZ, Tr, Tn, Tj; cannam@95: E Tz, To, TN, TH, Tp, TO; cannam@95: Ta = R0[WS(rs, 5)]; cannam@95: Tl = R1[WS(rs, 2)]; cannam@95: { cannam@95: E T1, T2, T5, T3, T4; cannam@95: T1 = R0[0]; cannam@95: T2 = R0[WS(rs, 3)]; cannam@95: T5 = R1[WS(rs, 4)]; cannam@95: T3 = R0[WS(rs, 6)]; cannam@95: T4 = R1[WS(rs, 1)]; cannam@95: { cannam@95: E Tb, TL, Te, TK, TR, Tf, Ti, Ty; cannam@95: Tb = R1[0]; cannam@95: TR = T2 + T5; cannam@95: Tg = R0[WS(rs, 2)]; cannam@95: { cannam@95: E T6, TS, Tc, Td; cannam@95: T6 = T2 + T3 - T4 - T5; cannam@95: T8 = (T3 + T5 - T2) - T4; cannam@95: TS = T3 + T4; cannam@95: Tc = R1[WS(rs, 3)]; cannam@95: Td = R1[WS(rs, 6)]; cannam@95: T7 = FNMS(KP250000000, T6, T1); cannam@95: TF = T1 + T6; cannam@95: TX = FNMS(KP618033988, TR, TS); cannam@95: TT = FMA(KP618033988, TS, TR); cannam@95: TL = Tc - Td; cannam@95: Te = Tc + Td; cannam@95: } cannam@95: TK = Tg + Tb; cannam@95: Tm = R0[WS(rs, 7)]; cannam@95: Tf = Tb - Te; cannam@95: Th = Tb + Te; cannam@95: TM = FMA(KP618033988, TL, TK); cannam@95: TZ = FNMS(KP618033988, TK, TL); cannam@95: Ti = FMA(KP809016994, Th, Tg); cannam@95: Ty = FMA(KP447213595, Th, Tf); cannam@95: Tr = R1[WS(rs, 5)]; cannam@95: Tn = R0[WS(rs, 1)]; cannam@95: Tj = FNMS(KP552786404, Ti, Tf); cannam@95: Tz = FNMS(KP690983005, Ty, Tg); cannam@95: To = R0[WS(rs, 4)]; cannam@95: TN = Tr + Tm; cannam@95: } cannam@95: } cannam@95: TH = Ta + Tg - Th; cannam@95: Tp = Tn + To; cannam@95: TO = To - Tn; cannam@95: { cannam@95: E Tx, TA, TP, T14, T11, Tu, TD; cannam@95: { cannam@95: E T10, TI, TC, TY; cannam@95: T9 = FNMS(KP559016994, T8, T7); cannam@95: Tx = FMA(KP559016994, T8, T7); cannam@95: TA = FNMS(KP809016994, Tz, Ta); cannam@95: TP = FMA(KP618033988, TO, TN); cannam@95: TY = FNMS(KP618033988, TN, TO); cannam@95: { cannam@95: E Tq, Ts, TG, Tt, TB; cannam@95: Tq = Tm - Tp; cannam@95: Ts = Tm + Tp; cannam@95: T14 = TZ - TY; cannam@95: T10 = TY + TZ; cannam@95: TG = Ts - Tr - Tl; cannam@95: Tt = FMA(KP809016994, Ts, Tr); cannam@95: TB = FMA(KP447213595, Ts, Tq); cannam@95: T11 = FMA(KP500000000, T10, TX); cannam@95: Ci[WS(csi, 2)] = KP866025403 * (TH - TG); cannam@95: TI = TG + TH; cannam@95: Tu = FNMS(KP552786404, Tt, Tq); cannam@95: TC = FNMS(KP690983005, TB, Tr); cannam@95: } cannam@95: Ci[WS(csi, 1)] = KP951056516 * (T10 - TX); cannam@95: Cr[WS(csr, 7)] = TF + TI; cannam@95: Cr[WS(csr, 2)] = FNMS(KP500000000, TI, TF); cannam@95: TD = FNMS(KP809016994, TC, Tl); cannam@95: } cannam@95: { cannam@95: E TU, Tk, T13, Tv, T12, TE; cannam@95: TQ = TM - TP; cannam@95: TU = TP + TM; cannam@95: T12 = TD + TA; cannam@95: TE = TA - TD; cannam@95: Tk = FNMS(KP559016994, Tj, Ta); cannam@95: TV = FMA(KP500000000, TU, TT); cannam@95: Ci[WS(csi, 6)] = -(KP951056516 * (FMA(KP910592997, T12, T11))); cannam@95: Ci[WS(csi, 3)] = KP951056516 * (FNMS(KP910592997, T12, T11)); cannam@95: T13 = FNMS(KP500000000, TE, Tx); cannam@95: Cr[WS(csr, 1)] = Tx + TE; cannam@95: Tv = FNMS(KP559016994, Tu, Tl); cannam@95: Ci[WS(csi, 4)] = KP951056516 * (TT - TU); cannam@95: Cr[WS(csr, 6)] = FMA(KP823639103, T14, T13); cannam@95: Cr[WS(csr, 3)] = FNMS(KP823639103, T14, T13); cannam@95: TW = Tv + Tk; cannam@95: Tw = Tk - Tv; cannam@95: } cannam@95: } cannam@95: } cannam@95: Ci[WS(csi, 5)] = -(KP951056516 * (FNMS(KP910592997, TW, TV))); cannam@95: Ci[0] = -(KP951056516 * (FMA(KP910592997, TW, TV))); cannam@95: TJ = FNMS(KP500000000, Tw, T9); cannam@95: Cr[WS(csr, 4)] = T9 + Tw; cannam@95: Cr[0] = FMA(KP823639103, TQ, TJ); cannam@95: Cr[WS(csr, 5)] = FNMS(KP823639103, TQ, TJ); cannam@95: } cannam@95: } cannam@95: } cannam@95: cannam@95: static const kr2c_desc desc = { 15, "r2cfII_15", {38, 7, 34, 0}, &GENUS }; cannam@95: cannam@95: void X(codelet_r2cfII_15) (planner *p) { cannam@95: X(kr2c_register) (p, r2cfII_15, &desc); cannam@95: } cannam@95: cannam@95: #else /* HAVE_FMA */ cannam@95: cannam@95: /* Generated by: ../../../genfft/gen_r2cf.native -compact -variables 4 -pipeline-latency 4 -n 15 -name r2cfII_15 -dft-II -include r2cfII.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 72 FP additions, 33 FP multiplications, cannam@95: * (or, 54 additions, 15 multiplications, 18 fused multiply/add), cannam@95: * 37 stack variables, 8 constants, and 30 memory accesses cannam@95: */ cannam@95: #include "r2cfII.h" cannam@95: cannam@95: static void r2cfII_15(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs) cannam@95: { cannam@95: DK(KP500000000, +0.500000000000000000000000000000000000000000000); cannam@95: DK(KP866025403, +0.866025403784438646763723170752936183471402627); cannam@95: DK(KP809016994, +0.809016994374947424102293417182819058860154590); cannam@95: DK(KP309016994, +0.309016994374947424102293417182819058860154590); cannam@95: DK(KP250000000, +0.250000000000000000000000000000000000000000000); cannam@95: DK(KP559016994, +0.559016994374947424102293417182819058860154590); cannam@95: DK(KP587785252, +0.587785252292473129168705954639072768597652438); cannam@95: DK(KP951056516, +0.951056516295153572116439333379382143405698634); cannam@95: { cannam@95: INT i; cannam@95: 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)) { cannam@95: E T1, T2, Tx, TR, TE, T7, TD, Th, Tm, Tr, TQ, TA, TB, Tf, Te; cannam@95: E Tu, TS, Td, TH, TO; cannam@95: T1 = R0[WS(rs, 5)]; cannam@95: { cannam@95: E T3, Tv, T6, Tw, T4, T5; cannam@95: T2 = R0[WS(rs, 2)]; cannam@95: T3 = R1[0]; cannam@95: Tv = T2 + T3; cannam@95: T4 = R1[WS(rs, 3)]; cannam@95: T5 = R1[WS(rs, 6)]; cannam@95: T6 = T4 + T5; cannam@95: Tw = T4 - T5; cannam@95: Tx = FMA(KP951056516, Tv, KP587785252 * Tw); cannam@95: TR = FNMS(KP587785252, Tv, KP951056516 * Tw); cannam@95: TE = KP559016994 * (T3 - T6); cannam@95: T7 = T3 + T6; cannam@95: TD = KP250000000 * T7; cannam@95: } cannam@95: { cannam@95: E Ti, Tl, Tj, Tk, Tp, Tq; cannam@95: Th = R0[0]; cannam@95: Ti = R1[WS(rs, 4)]; cannam@95: Tl = R0[WS(rs, 6)]; cannam@95: Tj = R1[WS(rs, 1)]; cannam@95: Tk = R0[WS(rs, 3)]; cannam@95: Tp = Tk + Ti; cannam@95: Tq = Tl + Tj; cannam@95: Tm = Ti + Tj - (Tk + Tl); cannam@95: Tr = FMA(KP951056516, Tp, KP587785252 * Tq); cannam@95: TQ = FNMS(KP951056516, Tq, KP587785252 * Tp); cannam@95: TA = FMA(KP250000000, Tm, Th); cannam@95: TB = KP559016994 * (Tl + Ti - (Tk + Tj)); cannam@95: } cannam@95: { cannam@95: E T9, Tt, Tc, Ts, Ta, Tb, TG; cannam@95: Tf = R1[WS(rs, 2)]; cannam@95: T9 = R0[WS(rs, 7)]; cannam@95: Te = R1[WS(rs, 5)]; cannam@95: Tt = T9 + Te; cannam@95: Ta = R0[WS(rs, 1)]; cannam@95: Tb = R0[WS(rs, 4)]; cannam@95: Tc = Ta + Tb; cannam@95: Ts = Ta - Tb; cannam@95: Tu = FNMS(KP951056516, Tt, KP587785252 * Ts); cannam@95: TS = FMA(KP951056516, Ts, KP587785252 * Tt); cannam@95: Td = T9 + Tc; cannam@95: TG = KP559016994 * (T9 - Tc); cannam@95: TH = FNMS(KP309016994, Te, TG) + FNMA(KP250000000, Td, Tf); cannam@95: TO = FMS(KP809016994, Te, Tf) + FNMA(KP250000000, Td, TG); cannam@95: } cannam@95: { cannam@95: E Tn, T8, Tg, To; cannam@95: Tn = Th - Tm; cannam@95: T8 = T1 + T2 - T7; cannam@95: Tg = Td - Te - Tf; cannam@95: To = T8 + Tg; cannam@95: Ci[WS(csi, 2)] = KP866025403 * (T8 - Tg); cannam@95: Cr[WS(csr, 2)] = FNMS(KP500000000, To, Tn); cannam@95: Cr[WS(csr, 7)] = Tn + To; cannam@95: } cannam@95: { cannam@95: E TM, TX, TT, TV, TP, TU, TN, TW; cannam@95: TM = TB + TA; cannam@95: TX = KP866025403 * (TR + TS); cannam@95: TT = TR - TS; cannam@95: TV = FMS(KP500000000, TT, TQ); cannam@95: TN = T1 + TE + FNMS(KP809016994, T2, TD); cannam@95: TP = TN + TO; cannam@95: TU = KP866025403 * (TO - TN); cannam@95: Cr[WS(csr, 1)] = TM + TP; cannam@95: Ci[WS(csi, 1)] = TQ + TT; cannam@95: Ci[WS(csi, 6)] = TU - TV; cannam@95: Ci[WS(csi, 3)] = TU + TV; cannam@95: TW = FNMS(KP500000000, TP, TM); cannam@95: Cr[WS(csr, 3)] = TW - TX; cannam@95: Cr[WS(csr, 6)] = TW + TX; cannam@95: } cannam@95: { cannam@95: E Tz, TC, Ty, TK, TI, TL, TF, TJ; cannam@95: Tz = KP866025403 * (Tx + Tu); cannam@95: TC = TA - TB; cannam@95: Ty = Tu - Tx; cannam@95: TK = FMS(KP500000000, Ty, Tr); cannam@95: TF = FMA(KP309016994, T2, T1) + TD - TE; cannam@95: TI = TF + TH; cannam@95: TL = KP866025403 * (TH - TF); cannam@95: Ci[WS(csi, 4)] = Tr + Ty; cannam@95: Cr[WS(csr, 4)] = TC + TI; cannam@95: Ci[WS(csi, 5)] = TK - TL; cannam@95: Ci[0] = TK + TL; cannam@95: TJ = FNMS(KP500000000, TI, TC); cannam@95: Cr[0] = Tz + TJ; cannam@95: Cr[WS(csr, 5)] = TJ - Tz; cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: cannam@95: static const kr2c_desc desc = { 15, "r2cfII_15", {54, 15, 18, 0}, &GENUS }; cannam@95: cannam@95: void X(codelet_r2cfII_15) (planner *p) { cannam@95: X(kr2c_register) (p, r2cfII_15, &desc); cannam@95: } cannam@95: cannam@95: #endif /* HAVE_FMA */