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 07:41:36 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_r2cb.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -sign 1 -n 16 -name r2cbIII_16 -dft-III -include r2cbIII.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 66 FP additions, 36 FP multiplications, cannam@95: * (or, 46 additions, 16 multiplications, 20 fused multiply/add), cannam@95: * 55 stack variables, 9 constants, and 32 memory accesses cannam@95: */ cannam@95: #include "r2cbIII.h" cannam@95: cannam@95: static void r2cbIII_16(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(KP668178637, +0.668178637919298919997757686523080761552472251); cannam@95: DK(KP1_662939224, +1.662939224605090474157576755235811513477121624); cannam@95: DK(KP198912367, +0.198912367379658006911597622644676228597850501); cannam@95: DK(KP1_961570560, +1.961570560806460898252364472268478073947867462); cannam@95: DK(KP707106781, +0.707106781186547524400844362104849039284835938); cannam@95: DK(KP1_414213562, +1.414213562373095048801688724209698078569671875); cannam@95: DK(KP414213562, +0.414213562373095048801688724209698078569671875); cannam@95: DK(KP1_847759065, +1.847759065022573512256366378793576573644833252); cannam@95: DK(KP2_000000000, +2.000000000000000000000000000000000000000000000); cannam@95: { cannam@95: INT i; cannam@95: for (i = v; i > 0; i = i - 1, R0 = R0 + ovs, R1 = R1 + ovs, Cr = Cr + ivs, Ci = Ci + ivs, MAKE_VOLATILE_STRIDE(64, rs), MAKE_VOLATILE_STRIDE(64, csr), MAKE_VOLATILE_STRIDE(64, csi)) { cannam@95: E TA, TD, Tv, TG, TE, TF; cannam@95: { cannam@95: E TK, TP, T7, T13, TW, TH, Tj, TC, To, Te, TX, TS, T12, Tt, TB; cannam@95: { cannam@95: E T4, Tf, T3, TU, Tz, T5, Tg, Th; cannam@95: { cannam@95: E T1, T2, Tx, Ty; cannam@95: T1 = Cr[0]; cannam@95: T2 = Cr[WS(csr, 7)]; cannam@95: Tx = Ci[0]; cannam@95: Ty = Ci[WS(csi, 7)]; cannam@95: T4 = Cr[WS(csr, 4)]; cannam@95: Tf = T1 - T2; cannam@95: T3 = T1 + T2; cannam@95: TU = Ty - Tx; cannam@95: Tz = Tx + Ty; cannam@95: T5 = Cr[WS(csr, 3)]; cannam@95: Tg = Ci[WS(csi, 4)]; cannam@95: Th = Ci[WS(csi, 3)]; cannam@95: } cannam@95: { cannam@95: E Tb, Tk, Ta, TR, Tn, Tc, Tq, Tr; cannam@95: { cannam@95: E T8, T9, Tl, Tm; cannam@95: T8 = Cr[WS(csr, 2)]; cannam@95: { cannam@95: E Tw, T6, TV, Ti; cannam@95: Tw = T4 - T5; cannam@95: T6 = T4 + T5; cannam@95: TV = Th - Tg; cannam@95: Ti = Tg + Th; cannam@95: TK = Tw - Tz; cannam@95: TA = Tw + Tz; cannam@95: TP = T3 - T6; cannam@95: T7 = T3 + T6; cannam@95: T13 = TV + TU; cannam@95: TW = TU - TV; cannam@95: TH = Tf + Ti; cannam@95: Tj = Tf - Ti; cannam@95: T9 = Cr[WS(csr, 5)]; cannam@95: } cannam@95: Tl = Ci[WS(csi, 2)]; cannam@95: Tm = Ci[WS(csi, 5)]; cannam@95: Tb = Cr[WS(csr, 1)]; cannam@95: Tk = T8 - T9; cannam@95: Ta = T8 + T9; cannam@95: TR = Tl - Tm; cannam@95: Tn = Tl + Tm; cannam@95: Tc = Cr[WS(csr, 6)]; cannam@95: Tq = Ci[WS(csi, 1)]; cannam@95: Tr = Ci[WS(csi, 6)]; cannam@95: } cannam@95: TC = Tk + Tn; cannam@95: To = Tk - Tn; cannam@95: { cannam@95: E Tp, Td, TQ, Ts; cannam@95: Tp = Tb - Tc; cannam@95: Td = Tb + Tc; cannam@95: TQ = Tr - Tq; cannam@95: Ts = Tq + Tr; cannam@95: Te = Ta + Td; cannam@95: TX = Ta - Td; cannam@95: TS = TQ - TR; cannam@95: T12 = TR + TQ; cannam@95: Tt = Tp - Ts; cannam@95: TB = Tp + Ts; cannam@95: } cannam@95: } cannam@95: } cannam@95: { cannam@95: E T10, TT, TY, TZ; cannam@95: R0[0] = KP2_000000000 * (T7 + Te); cannam@95: R0[WS(rs, 4)] = KP2_000000000 * (T13 - T12); cannam@95: T10 = TP - TS; cannam@95: TT = TP + TS; cannam@95: TY = TW - TX; cannam@95: TZ = TX + TW; cannam@95: { cannam@95: E T11, T14, TI, TL, Tu; cannam@95: T11 = T7 - Te; cannam@95: T14 = T12 + T13; cannam@95: R0[WS(rs, 5)] = KP1_847759065 * (FNMS(KP414213562, TT, TY)); cannam@95: R0[WS(rs, 1)] = KP1_847759065 * (FMA(KP414213562, TY, TT)); cannam@95: R0[WS(rs, 6)] = KP1_414213562 * (T14 - T11); cannam@95: R0[WS(rs, 2)] = KP1_414213562 * (T11 + T14); cannam@95: TD = TB - TC; cannam@95: TI = TC + TB; cannam@95: TL = To - Tt; cannam@95: Tu = To + Tt; cannam@95: { cannam@95: E TO, TJ, TN, TM; cannam@95: R0[WS(rs, 7)] = -(KP1_847759065 * (FNMS(KP414213562, TZ, T10))); cannam@95: R0[WS(rs, 3)] = KP1_847759065 * (FMA(KP414213562, T10, TZ)); cannam@95: TO = FMA(KP707106781, TI, TH); cannam@95: TJ = FNMS(KP707106781, TI, TH); cannam@95: TN = FMA(KP707106781, TL, TK); cannam@95: TM = FNMS(KP707106781, TL, TK); cannam@95: Tv = FMA(KP707106781, Tu, Tj); cannam@95: TG = FNMS(KP707106781, Tu, Tj); cannam@95: R1[WS(rs, 3)] = KP1_961570560 * (FMA(KP198912367, TO, TN)); cannam@95: R1[WS(rs, 7)] = -(KP1_961570560 * (FNMS(KP198912367, TN, TO))); cannam@95: R1[WS(rs, 5)] = KP1_662939224 * (FNMS(KP668178637, TJ, TM)); cannam@95: R1[WS(rs, 1)] = KP1_662939224 * (FMA(KP668178637, TM, TJ)); cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: TE = FNMS(KP707106781, TD, TA); cannam@95: TF = FMA(KP707106781, TD, TA); cannam@95: R1[WS(rs, 2)] = -(KP1_662939224 * (FNMS(KP668178637, TG, TF))); cannam@95: R1[WS(rs, 6)] = -(KP1_662939224 * (FMA(KP668178637, TF, TG))); cannam@95: R1[WS(rs, 4)] = -(KP1_961570560 * (FMA(KP198912367, Tv, TE))); cannam@95: R1[0] = KP1_961570560 * (FNMS(KP198912367, TE, Tv)); cannam@95: } cannam@95: } cannam@95: } cannam@95: cannam@95: static const kr2c_desc desc = { 16, "r2cbIII_16", {46, 16, 20, 0}, &GENUS }; cannam@95: cannam@95: void X(codelet_r2cbIII_16) (planner *p) { cannam@95: X(kr2c_register) (p, r2cbIII_16, &desc); cannam@95: } cannam@95: cannam@95: #else /* HAVE_FMA */ cannam@95: cannam@95: /* Generated by: ../../../genfft/gen_r2cb.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 16 -name r2cbIII_16 -dft-III -include r2cbIII.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 66 FP additions, 32 FP multiplications, cannam@95: * (or, 54 additions, 20 multiplications, 12 fused multiply/add), cannam@95: * 40 stack variables, 9 constants, and 32 memory accesses cannam@95: */ cannam@95: #include "r2cbIII.h" cannam@95: cannam@95: static void r2cbIII_16(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(KP1_961570560, +1.961570560806460898252364472268478073947867462); cannam@95: DK(KP390180644, +0.390180644032256535696569736954044481855383236); cannam@95: DK(KP1_111140466, +1.111140466039204449485661627897065748749874382); cannam@95: DK(KP1_662939224, +1.662939224605090474157576755235811513477121624); cannam@95: DK(KP707106781, +0.707106781186547524400844362104849039284835938); cannam@95: DK(KP1_414213562, +1.414213562373095048801688724209698078569671875); cannam@95: DK(KP765366864, +0.765366864730179543456919968060797733522689125); cannam@95: DK(KP1_847759065, +1.847759065022573512256366378793576573644833252); cannam@95: DK(KP2_000000000, +2.000000000000000000000000000000000000000000000); cannam@95: { cannam@95: INT i; cannam@95: for (i = v; i > 0; i = i - 1, R0 = R0 + ovs, R1 = R1 + ovs, Cr = Cr + ivs, Ci = Ci + ivs, MAKE_VOLATILE_STRIDE(64, rs), MAKE_VOLATILE_STRIDE(64, csr), MAKE_VOLATILE_STRIDE(64, csi)) { cannam@95: E T7, TW, T13, Tj, TD, TK, TP, TH, Te, TX, T12, To, Tt, Tx, TS; cannam@95: E Tw, TT, TY; cannam@95: { cannam@95: E T3, Tf, TC, TV, T6, Tz, Ti, TU; cannam@95: { cannam@95: E T1, T2, TA, TB; cannam@95: T1 = Cr[0]; cannam@95: T2 = Cr[WS(csr, 7)]; cannam@95: T3 = T1 + T2; cannam@95: Tf = T1 - T2; cannam@95: TA = Ci[0]; cannam@95: TB = Ci[WS(csi, 7)]; cannam@95: TC = TA + TB; cannam@95: TV = TB - TA; cannam@95: } cannam@95: { cannam@95: E T4, T5, Tg, Th; cannam@95: T4 = Cr[WS(csr, 4)]; cannam@95: T5 = Cr[WS(csr, 3)]; cannam@95: T6 = T4 + T5; cannam@95: Tz = T4 - T5; cannam@95: Tg = Ci[WS(csi, 4)]; cannam@95: Th = Ci[WS(csi, 3)]; cannam@95: Ti = Tg + Th; cannam@95: TU = Tg - Th; cannam@95: } cannam@95: T7 = T3 + T6; cannam@95: TW = TU + TV; cannam@95: T13 = TV - TU; cannam@95: Tj = Tf - Ti; cannam@95: TD = Tz + TC; cannam@95: TK = Tz - TC; cannam@95: TP = T3 - T6; cannam@95: TH = Tf + Ti; cannam@95: } cannam@95: { cannam@95: E Ta, Tk, Tn, TR, Td, Tp, Ts, TQ; cannam@95: { cannam@95: E T8, T9, Tl, Tm; cannam@95: T8 = Cr[WS(csr, 2)]; cannam@95: T9 = Cr[WS(csr, 5)]; cannam@95: Ta = T8 + T9; cannam@95: Tk = T8 - T9; cannam@95: Tl = Ci[WS(csi, 2)]; cannam@95: Tm = Ci[WS(csi, 5)]; cannam@95: Tn = Tl + Tm; cannam@95: TR = Tl - Tm; cannam@95: } cannam@95: { cannam@95: E Tb, Tc, Tq, Tr; cannam@95: Tb = Cr[WS(csr, 1)]; cannam@95: Tc = Cr[WS(csr, 6)]; cannam@95: Td = Tb + Tc; cannam@95: Tp = Tb - Tc; cannam@95: Tq = Ci[WS(csi, 1)]; cannam@95: Tr = Ci[WS(csi, 6)]; cannam@95: Ts = Tq + Tr; cannam@95: TQ = Tr - Tq; cannam@95: } cannam@95: Te = Ta + Td; cannam@95: TX = Ta - Td; cannam@95: T12 = TR + TQ; cannam@95: To = Tk - Tn; cannam@95: Tt = Tp - Ts; cannam@95: Tx = Tp + Ts; cannam@95: TS = TQ - TR; cannam@95: Tw = Tk + Tn; cannam@95: } cannam@95: R0[0] = KP2_000000000 * (T7 + Te); cannam@95: R0[WS(rs, 4)] = KP2_000000000 * (T13 - T12); cannam@95: TT = TP + TS; cannam@95: TY = TW - TX; cannam@95: R0[WS(rs, 1)] = FMA(KP1_847759065, TT, KP765366864 * TY); cannam@95: R0[WS(rs, 5)] = FNMS(KP765366864, TT, KP1_847759065 * TY); cannam@95: { cannam@95: E T11, T14, TZ, T10; cannam@95: T11 = T7 - Te; cannam@95: T14 = T12 + T13; cannam@95: R0[WS(rs, 2)] = KP1_414213562 * (T11 + T14); cannam@95: R0[WS(rs, 6)] = KP1_414213562 * (T14 - T11); cannam@95: TZ = TP - TS; cannam@95: T10 = TX + TW; cannam@95: R0[WS(rs, 3)] = FMA(KP765366864, TZ, KP1_847759065 * T10); cannam@95: R0[WS(rs, 7)] = FNMS(KP1_847759065, TZ, KP765366864 * T10); cannam@95: } cannam@95: { cannam@95: E TJ, TN, TM, TO, TI, TL; cannam@95: TI = KP707106781 * (Tw + Tx); cannam@95: TJ = TH - TI; cannam@95: TN = TH + TI; cannam@95: TL = KP707106781 * (To - Tt); cannam@95: TM = TK - TL; cannam@95: TO = TL + TK; cannam@95: R1[WS(rs, 1)] = FMA(KP1_662939224, TJ, KP1_111140466 * TM); cannam@95: R1[WS(rs, 7)] = FNMS(KP1_961570560, TN, KP390180644 * TO); cannam@95: R1[WS(rs, 5)] = FNMS(KP1_111140466, TJ, KP1_662939224 * TM); cannam@95: R1[WS(rs, 3)] = FMA(KP390180644, TN, KP1_961570560 * TO); cannam@95: } cannam@95: { cannam@95: E Tv, TF, TE, TG, Tu, Ty; cannam@95: Tu = KP707106781 * (To + Tt); cannam@95: Tv = Tj + Tu; cannam@95: TF = Tj - Tu; cannam@95: Ty = KP707106781 * (Tw - Tx); cannam@95: TE = Ty + TD; cannam@95: TG = Ty - TD; cannam@95: R1[0] = FNMS(KP390180644, TE, KP1_961570560 * Tv); cannam@95: R1[WS(rs, 6)] = FNMS(KP1_662939224, TF, KP1_111140466 * TG); cannam@95: R1[WS(rs, 4)] = -(FMA(KP390180644, Tv, KP1_961570560 * TE)); cannam@95: R1[WS(rs, 2)] = FMA(KP1_111140466, TF, KP1_662939224 * TG); cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: cannam@95: static const kr2c_desc desc = { 16, "r2cbIII_16", {54, 20, 12, 0}, &GENUS }; cannam@95: cannam@95: void X(codelet_r2cbIII_16) (planner *p) { cannam@95: X(kr2c_register) (p, r2cbIII_16, &desc); cannam@95: } cannam@95: cannam@95: #endif /* HAVE_FMA */