cannam@127: /* cannam@127: * Copyright (c) 2003, 2007-14 Matteo Frigo cannam@127: * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology cannam@127: * cannam@127: * This program is free software; you can redistribute it and/or modify cannam@127: * it under the terms of the GNU General Public License as published by cannam@127: * the Free Software Foundation; either version 2 of the License, or cannam@127: * (at your option) any later version. cannam@127: * cannam@127: * This program is distributed in the hope that it will be useful, cannam@127: * but WITHOUT ANY WARRANTY; without even the implied warranty of cannam@127: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the cannam@127: * GNU General Public License for more details. cannam@127: * cannam@127: * You should have received a copy of the GNU General Public License cannam@127: * along with this program; if not, write to the Free Software cannam@127: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA cannam@127: * cannam@127: */ cannam@127: cannam@127: /* This file was automatically generated --- DO NOT EDIT */ cannam@127: /* Generated on Sat Jul 30 16:47:46 EDT 2016 */ cannam@127: cannam@127: #include "codelet-rdft.h" cannam@127: cannam@127: #ifdef HAVE_FMA cannam@127: cannam@127: /* Generated by: ../../../genfft/gen_r2cf.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 20 -name r2cfII_20 -dft-II -include r2cfII.h */ cannam@127: cannam@127: /* cannam@127: * This function contains 102 FP additions, 63 FP multiplications, cannam@127: * (or, 39 additions, 0 multiplications, 63 fused multiply/add), cannam@127: * 67 stack variables, 10 constants, and 40 memory accesses cannam@127: */ cannam@127: #include "r2cfII.h" cannam@127: cannam@127: static void r2cfII_20(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs) cannam@127: { cannam@127: DK(KP707106781, +0.707106781186547524400844362104849039284835938); cannam@127: DK(KP951056516, +0.951056516295153572116439333379382143405698634); cannam@127: DK(KP559016994, +0.559016994374947424102293417182819058860154590); cannam@127: DK(KP690983005, +0.690983005625052575897706582817180941139845410); cannam@127: DK(KP552786404, +0.552786404500042060718165266253744752911876328); cannam@127: DK(KP447213595, +0.447213595499957939281834733746255247088123672); cannam@127: DK(KP809016994, +0.809016994374947424102293417182819058860154590); cannam@127: DK(KP250000000, +0.250000000000000000000000000000000000000000000); cannam@127: DK(KP618033988, +0.618033988749894848204586834365638117720309180); cannam@127: DK(KP381966011, +0.381966011250105151795413165634361882279690820); cannam@127: { cannam@127: INT i; cannam@127: for (i = v; i > 0; i = i - 1, R0 = R0 + ivs, R1 = R1 + ivs, Cr = Cr + ovs, Ci = Ci + ovs, MAKE_VOLATILE_STRIDE(80, rs), MAKE_VOLATILE_STRIDE(80, csr), MAKE_VOLATILE_STRIDE(80, csi)) { cannam@127: E Tv, TK, TN, Th, T1l, T1n, Ts, TH; cannam@127: { cannam@127: E Ti, T1d, T1f, T1e, T1g, T1p, TS, Tg, To, T8, T7, T19, T1r, T1k, Tx; cannam@127: E Tp, TX, Ty, TF, Tr, TV, Tz, TA, TI; cannam@127: { cannam@127: E Ta, Tb, Td, Te; cannam@127: Ti = R1[WS(rs, 2)]; cannam@127: T1d = R0[WS(rs, 5)]; cannam@127: Ta = R0[WS(rs, 9)]; cannam@127: Tb = R0[WS(rs, 1)]; cannam@127: Td = R0[WS(rs, 3)]; cannam@127: Te = R0[WS(rs, 7)]; cannam@127: { cannam@127: E T1, T2, T5, T3, T4, T1i, Tc, Tf; cannam@127: T1 = R0[0]; cannam@127: T1f = Ta + Tb; cannam@127: Tc = Ta - Tb; cannam@127: T1e = Td + Te; cannam@127: Tf = Td - Te; cannam@127: T2 = R0[WS(rs, 4)]; cannam@127: T5 = R0[WS(rs, 6)]; cannam@127: T1g = FMA(KP381966011, T1f, T1e); cannam@127: T1p = FMA(KP381966011, T1e, T1f); cannam@127: TS = FMA(KP618033988, Tc, Tf); cannam@127: Tg = FNMS(KP618033988, Tf, Tc); cannam@127: T3 = R0[WS(rs, 8)]; cannam@127: T4 = R0[WS(rs, 2)]; cannam@127: T1i = T2 + T5; cannam@127: { cannam@127: E Tj, Tu, Tm, Tt, Tn, Tq, TU; cannam@127: Tj = R1[WS(rs, 8)]; cannam@127: To = R1[WS(rs, 6)]; cannam@127: { cannam@127: E T6, T1j, Tk, Tl; cannam@127: T6 = T2 + T3 - T4 - T5; cannam@127: T8 = (T3 + T5 - T2) - T4; cannam@127: T1j = T3 + T4; cannam@127: Tk = R1[0]; cannam@127: Tl = R1[WS(rs, 4)]; cannam@127: T7 = FNMS(KP250000000, T6, T1); cannam@127: T19 = T1 + T6; cannam@127: T1r = FNMS(KP618033988, T1i, T1j); cannam@127: T1k = FMA(KP618033988, T1j, T1i); cannam@127: Tu = Tk - Tl; cannam@127: Tm = Tk + Tl; cannam@127: } cannam@127: Tt = To + Tj; cannam@127: Tx = R1[WS(rs, 7)]; cannam@127: Tn = Tj - Tm; cannam@127: Tp = Tj + Tm; cannam@127: Tv = FNMS(KP618033988, Tu, Tt); cannam@127: TX = FMA(KP618033988, Tt, Tu); cannam@127: Tq = FMA(KP809016994, Tp, To); cannam@127: TU = FMA(KP447213595, Tp, Tn); cannam@127: Ty = R1[WS(rs, 1)]; cannam@127: TF = R1[WS(rs, 3)]; cannam@127: Tr = FNMS(KP552786404, Tq, Tn); cannam@127: TV = FNMS(KP690983005, TU, To); cannam@127: Tz = R1[WS(rs, 5)]; cannam@127: TA = R1[WS(rs, 9)]; cannam@127: TI = TF + Ty; cannam@127: } cannam@127: } cannam@127: } cannam@127: { cannam@127: E T1w, TJ, TB, T1a; cannam@127: T1w = T1f + T1d - T1e; cannam@127: TJ = Tz - TA; cannam@127: TB = Tz + TA; cannam@127: T1a = Ti + To - Tp; cannam@127: { cannam@127: E T9, T12, TT, T15, TG, TD, T1s, T1u, TW, T11, T10, T1h; cannam@127: { cannam@127: E TE, TC, TR, T1b; cannam@127: T9 = FNMS(KP559016994, T8, T7); cannam@127: TR = FMA(KP559016994, T8, T7); cannam@127: TK = FMA(KP618033988, TJ, TI); cannam@127: T12 = FNMS(KP618033988, TI, TJ); cannam@127: TE = Ty - TB; cannam@127: TC = Ty + TB; cannam@127: TT = FMA(KP951056516, TS, TR); cannam@127: T15 = FNMS(KP951056516, TS, TR); cannam@127: TG = FNMS(KP552786404, TF, TE); cannam@127: T1b = TC - TF - Tx; cannam@127: { cannam@127: E TZ, T1q, T1c, T1x; cannam@127: TZ = FMA(KP447213595, TC, TE); cannam@127: TD = FMA(KP250000000, TC, Tx); cannam@127: T1q = FNMS(KP809016994, T1p, T1d); cannam@127: T1c = T1a + T1b; cannam@127: T1x = T1a - T1b; cannam@127: T10 = FNMS(KP690983005, TZ, TF); cannam@127: T1s = FNMS(KP951056516, T1r, T1q); cannam@127: T1u = FMA(KP951056516, T1r, T1q); cannam@127: Ci[WS(csi, 7)] = FMA(KP707106781, T1x, T1w); cannam@127: Ci[WS(csi, 2)] = FMS(KP707106781, T1x, T1w); cannam@127: Cr[WS(csr, 7)] = FMA(KP707106781, T1c, T19); cannam@127: Cr[WS(csr, 2)] = FNMS(KP707106781, T1c, T19); cannam@127: } cannam@127: } cannam@127: TW = FNMS(KP809016994, TV, Ti); cannam@127: T11 = FNMS(KP809016994, T10, Tx); cannam@127: T1h = FMA(KP809016994, T1g, T1d); cannam@127: { cannam@127: E T17, TY, T16, T13; cannam@127: T17 = FNMS(KP951056516, TX, TW); cannam@127: TY = FMA(KP951056516, TX, TW); cannam@127: T16 = FMA(KP951056516, T12, T11); cannam@127: T13 = FNMS(KP951056516, T12, T11); cannam@127: TN = FMA(KP951056516, Tg, T9); cannam@127: Th = FNMS(KP951056516, Tg, T9); cannam@127: { cannam@127: E T18, T1v, T1t, T14; cannam@127: T18 = T16 - T17; cannam@127: T1v = T17 + T16; cannam@127: T1t = TY + T13; cannam@127: T14 = TY - T13; cannam@127: Cr[WS(csr, 1)] = FMA(KP707106781, T18, T15); cannam@127: Cr[WS(csr, 8)] = FNMS(KP707106781, T18, T15); cannam@127: Ci[WS(csi, 3)] = FMA(KP707106781, T1v, T1u); cannam@127: Ci[WS(csi, 6)] = FMS(KP707106781, T1v, T1u); cannam@127: Ci[WS(csi, 1)] = FNMS(KP707106781, T1t, T1s); cannam@127: Ci[WS(csi, 8)] = -(FMA(KP707106781, T1t, T1s)); cannam@127: Cr[WS(csr, 3)] = FMA(KP707106781, T14, TT); cannam@127: Cr[WS(csr, 6)] = FNMS(KP707106781, T14, TT); cannam@127: T1l = FMA(KP951056516, T1k, T1h); cannam@127: T1n = FNMS(KP951056516, T1k, T1h); cannam@127: } cannam@127: } cannam@127: Ts = FNMS(KP559016994, Tr, Ti); cannam@127: TH = FNMS(KP559016994, TG, TD); cannam@127: } cannam@127: } cannam@127: } cannam@127: { cannam@127: E TO, Tw, TP, TL; cannam@127: TO = FMA(KP951056516, Tv, Ts); cannam@127: Tw = FNMS(KP951056516, Tv, Ts); cannam@127: TP = FMA(KP951056516, TK, TH); cannam@127: TL = FNMS(KP951056516, TK, TH); cannam@127: { cannam@127: E TQ, T1m, T1o, TM; cannam@127: TQ = TO - TP; cannam@127: T1m = TO + TP; cannam@127: T1o = Tw + TL; cannam@127: TM = Tw - TL; cannam@127: Cr[WS(csr, 4)] = FMA(KP707106781, TQ, TN); cannam@127: Cr[WS(csr, 5)] = FNMS(KP707106781, TQ, TN); cannam@127: Ci[WS(csi, 9)] = FNMS(KP707106781, T1m, T1l); cannam@127: Ci[0] = -(FMA(KP707106781, T1m, T1l)); cannam@127: Ci[WS(csi, 5)] = FNMS(KP707106781, T1o, T1n); cannam@127: Ci[WS(csi, 4)] = -(FMA(KP707106781, T1o, T1n)); cannam@127: Cr[0] = FMA(KP707106781, TM, Th); cannam@127: Cr[WS(csr, 9)] = FNMS(KP707106781, TM, Th); cannam@127: } cannam@127: } cannam@127: } cannam@127: } cannam@127: } cannam@127: cannam@127: static const kr2c_desc desc = { 20, "r2cfII_20", {39, 0, 63, 0}, &GENUS }; cannam@127: cannam@127: void X(codelet_r2cfII_20) (planner *p) { cannam@127: X(kr2c_register) (p, r2cfII_20, &desc); cannam@127: } cannam@127: cannam@127: #else /* HAVE_FMA */ cannam@127: cannam@127: /* Generated by: ../../../genfft/gen_r2cf.native -compact -variables 4 -pipeline-latency 4 -n 20 -name r2cfII_20 -dft-II -include r2cfII.h */ cannam@127: cannam@127: /* cannam@127: * This function contains 102 FP additions, 34 FP multiplications, cannam@127: * (or, 86 additions, 18 multiplications, 16 fused multiply/add), cannam@127: * 60 stack variables, 13 constants, and 40 memory accesses cannam@127: */ cannam@127: #include "r2cfII.h" cannam@127: cannam@127: static void r2cfII_20(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs) cannam@127: { cannam@127: DK(KP572061402, +0.572061402817684297600072783580302076536153377); cannam@127: DK(KP218508012, +0.218508012224410535399650602527877556893735408); cannam@127: DK(KP309016994, +0.309016994374947424102293417182819058860154590); cannam@127: DK(KP809016994, +0.809016994374947424102293417182819058860154590); cannam@127: DK(KP559016994, +0.559016994374947424102293417182819058860154590); cannam@127: DK(KP951056516, +0.951056516295153572116439333379382143405698634); cannam@127: DK(KP587785252, +0.587785252292473129168705954639072768597652438); cannam@127: DK(KP250000000, +0.250000000000000000000000000000000000000000000); cannam@127: DK(KP176776695, +0.176776695296636881100211090526212259821208984); cannam@127: DK(KP395284707, +0.395284707521047416499861693054089816714944392); cannam@127: DK(KP672498511, +0.672498511963957326960058968885748755876783111); cannam@127: DK(KP415626937, +0.415626937777453428589967464113135184222253485); cannam@127: DK(KP707106781, +0.707106781186547524400844362104849039284835938); cannam@127: { cannam@127: INT i; cannam@127: for (i = v; i > 0; i = i - 1, R0 = R0 + ivs, R1 = R1 + ivs, Cr = Cr + ovs, Ci = Ci + ovs, MAKE_VOLATILE_STRIDE(80, rs), MAKE_VOLATILE_STRIDE(80, csr), MAKE_VOLATILE_STRIDE(80, csi)) { cannam@127: E T8, TD, Tm, TN, T9, TC, TY, TE, Te, TF, Tl, TK, T12, TL, Tk; cannam@127: E TM, T1, T6, Tq, T1l, T1c, Tp, T1f, T1e, T1d, Ty, TW, T1g, T1m, Tx; cannam@127: E Tu; cannam@127: T8 = R1[WS(rs, 2)]; cannam@127: TD = KP707106781 * T8; cannam@127: Tm = R1[WS(rs, 7)]; cannam@127: TN = KP707106781 * Tm; cannam@127: { cannam@127: E Ta, TA, Td, TB, Tb, Tc; cannam@127: T9 = R1[WS(rs, 6)]; cannam@127: Ta = R1[WS(rs, 8)]; cannam@127: TA = T9 + Ta; cannam@127: Tb = R1[0]; cannam@127: Tc = R1[WS(rs, 4)]; cannam@127: Td = Tb + Tc; cannam@127: TB = Tb - Tc; cannam@127: TC = FMA(KP415626937, TA, KP672498511 * TB); cannam@127: TY = FNMS(KP415626937, TB, KP672498511 * TA); cannam@127: TE = KP395284707 * (Ta - Td); cannam@127: Te = Ta + Td; cannam@127: TF = KP176776695 * Te; cannam@127: } cannam@127: { cannam@127: E Tg, TJ, Tj, TI, Th, Ti; cannam@127: Tg = R1[WS(rs, 1)]; cannam@127: Tl = R1[WS(rs, 3)]; cannam@127: TJ = Tg + Tl; cannam@127: Th = R1[WS(rs, 5)]; cannam@127: Ti = R1[WS(rs, 9)]; cannam@127: Tj = Th + Ti; cannam@127: TI = Th - Ti; cannam@127: TK = FNMS(KP415626937, TJ, KP672498511 * TI); cannam@127: T12 = FMA(KP415626937, TI, KP672498511 * TJ); cannam@127: TL = KP395284707 * (Tg - Tj); cannam@127: Tk = Tg + Tj; cannam@127: TM = KP176776695 * Tk; cannam@127: } cannam@127: { cannam@127: E T2, T5, T3, T4, T1a, T1b; cannam@127: T1 = R0[0]; cannam@127: T2 = R0[WS(rs, 6)]; cannam@127: T5 = R0[WS(rs, 8)]; cannam@127: T3 = R0[WS(rs, 2)]; cannam@127: T4 = R0[WS(rs, 4)]; cannam@127: T1a = T4 + T2; cannam@127: T1b = T5 + T3; cannam@127: T6 = T2 + T3 - (T4 + T5); cannam@127: Tq = FMA(KP250000000, T6, T1); cannam@127: T1l = FNMS(KP951056516, T1b, KP587785252 * T1a); cannam@127: T1c = FMA(KP951056516, T1a, KP587785252 * T1b); cannam@127: Tp = KP559016994 * (T5 + T2 - (T4 + T3)); cannam@127: } cannam@127: T1f = R0[WS(rs, 5)]; cannam@127: { cannam@127: E Tv, Tw, Ts, Tt; cannam@127: Tv = R0[WS(rs, 9)]; cannam@127: Tw = R0[WS(rs, 1)]; cannam@127: Tx = Tv - Tw; cannam@127: T1e = Tv + Tw; cannam@127: Ts = R0[WS(rs, 3)]; cannam@127: Tt = R0[WS(rs, 7)]; cannam@127: Tu = Ts - Tt; cannam@127: T1d = Ts + Tt; cannam@127: } cannam@127: Ty = FMA(KP951056516, Tu, KP587785252 * Tx); cannam@127: TW = FNMS(KP951056516, Tx, KP587785252 * Tu); cannam@127: T1g = FMA(KP809016994, T1d, KP309016994 * T1e) + T1f; cannam@127: T1m = FNMS(KP809016994, T1e, T1f) - (KP309016994 * T1d); cannam@127: { cannam@127: E T7, T1r, To, T1q, Tf, Tn; cannam@127: T7 = T1 - T6; cannam@127: T1r = T1e + T1f - T1d; cannam@127: Tf = T8 + (T9 - Te); cannam@127: Tn = (Tk - Tl) - Tm; cannam@127: To = KP707106781 * (Tf + Tn); cannam@127: T1q = KP707106781 * (Tf - Tn); cannam@127: Cr[WS(csr, 2)] = T7 - To; cannam@127: Ci[WS(csi, 2)] = T1q - T1r; cannam@127: Cr[WS(csr, 7)] = T7 + To; cannam@127: Ci[WS(csi, 7)] = T1q + T1r; cannam@127: } cannam@127: { cannam@127: E T1h, T1j, TX, T15, T10, T16, T13, T17, TV, TZ, T11; cannam@127: T1h = T1c - T1g; cannam@127: T1j = T1c + T1g; cannam@127: TV = Tq - Tp; cannam@127: TX = TV - TW; cannam@127: T15 = TV + TW; cannam@127: TZ = FMA(KP218508012, T9, TD) + TF - TE; cannam@127: T10 = TY + TZ; cannam@127: T16 = TZ - TY; cannam@127: T11 = FNMS(KP218508012, Tl, TL) - (TM + TN); cannam@127: T13 = T11 - T12; cannam@127: T17 = T11 + T12; cannam@127: { cannam@127: E T14, T19, T18, T1i; cannam@127: T14 = T10 + T13; cannam@127: Cr[WS(csr, 5)] = TX - T14; cannam@127: Cr[WS(csr, 4)] = TX + T14; cannam@127: T19 = T17 - T16; cannam@127: Ci[WS(csi, 5)] = T19 - T1h; cannam@127: Ci[WS(csi, 4)] = T19 + T1h; cannam@127: T18 = T16 + T17; cannam@127: Cr[WS(csr, 9)] = T15 - T18; cannam@127: Cr[0] = T15 + T18; cannam@127: T1i = T13 - T10; cannam@127: Ci[0] = T1i - T1j; cannam@127: Ci[WS(csi, 9)] = T1i + T1j; cannam@127: } cannam@127: } cannam@127: { cannam@127: E T1n, T1p, Tz, TR, TH, TS, TP, TT, Tr, TG, TO; cannam@127: T1n = T1l + T1m; cannam@127: T1p = T1m - T1l; cannam@127: Tr = Tp + Tq; cannam@127: Tz = Tr + Ty; cannam@127: TR = Tr - Ty; cannam@127: TG = TD + TE + FNMS(KP572061402, T9, TF); cannam@127: TH = TC + TG; cannam@127: TS = TC - TG; cannam@127: TO = TL + TM + FNMS(KP572061402, Tl, TN); cannam@127: TP = TK - TO; cannam@127: TT = TK + TO; cannam@127: { cannam@127: E TQ, T1o, TU, T1k; cannam@127: TQ = TH + TP; cannam@127: Cr[WS(csr, 6)] = Tz - TQ; cannam@127: Cr[WS(csr, 3)] = Tz + TQ; cannam@127: T1o = TT - TS; cannam@127: Ci[WS(csi, 6)] = T1o - T1p; cannam@127: Ci[WS(csi, 3)] = T1o + T1p; cannam@127: TU = TS + TT; cannam@127: Cr[WS(csr, 8)] = TR - TU; cannam@127: Cr[WS(csr, 1)] = TR + TU; cannam@127: T1k = TP - TH; cannam@127: Ci[WS(csi, 8)] = T1k - T1n; cannam@127: Ci[WS(csi, 1)] = T1k + T1n; cannam@127: } cannam@127: } cannam@127: } cannam@127: } cannam@127: } cannam@127: cannam@127: static const kr2c_desc desc = { 20, "r2cfII_20", {86, 18, 16, 0}, &GENUS }; cannam@127: cannam@127: void X(codelet_r2cfII_20) (planner *p) { cannam@127: X(kr2c_register) (p, r2cfII_20, &desc); cannam@127: } cannam@127: cannam@127: #endif /* HAVE_FMA */