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 07:41:08 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 14 -name r2cb_14 -include r2cb.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 62 FP additions, 44 FP multiplications, cannam@95: * (or, 18 additions, 0 multiplications, 44 fused multiply/add), cannam@95: * 58 stack variables, 7 constants, and 28 memory accesses cannam@95: */ cannam@95: #include "r2cb.h" cannam@95: cannam@95: static void r2cb_14(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_949855824, +1.949855824363647214036263365987862434465571601); cannam@95: DK(KP1_801937735, +1.801937735804838252472204639014890102331838324); cannam@95: DK(KP692021471, +0.692021471630095869627814897002069140197260599); cannam@95: DK(KP2_000000000, +2.000000000000000000000000000000000000000000000); cannam@95: DK(KP356895867, +0.356895867892209443894399510021300583399127187); cannam@95: DK(KP801937735, +0.801937735804838252472204639014890102331838324); cannam@95: DK(KP554958132, +0.554958132087371191422194871006410481067288862); 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(56, rs), MAKE_VOLATILE_STRIDE(56, csr), MAKE_VOLATILE_STRIDE(56, csi)) { cannam@95: E Te, TO, TT, TG, TJ, TD, TR, TE; cannam@95: { cannam@95: E T3, TK, To, TM, Tu, TL, Tr, TS, TA, TN, TX, TF, Tv, T7, Tf; cannam@95: E T6, Th, Tc, T8, T1, T2; cannam@95: T1 = Cr[0]; cannam@95: T2 = Cr[WS(csr, 7)]; cannam@95: { cannam@95: E Ts, Tt, Tp, Tq, Tm, Tn; cannam@95: Tm = Ci[WS(csi, 4)]; cannam@95: Tn = Ci[WS(csi, 3)]; cannam@95: Ts = Ci[WS(csi, 6)]; cannam@95: Te = T1 + T2; cannam@95: T3 = T1 - T2; cannam@95: TK = Tm + Tn; cannam@95: To = Tm - Tn; cannam@95: Tt = Ci[WS(csi, 1)]; cannam@95: Tp = Ci[WS(csi, 2)]; cannam@95: Tq = Ci[WS(csi, 5)]; cannam@95: { cannam@95: E T4, T5, Ta, Tb; cannam@95: T4 = Cr[WS(csr, 2)]; cannam@95: TM = Ts + Tt; cannam@95: Tu = Ts - Tt; cannam@95: TL = Tp + Tq; cannam@95: Tr = Tp - Tq; cannam@95: TS = FMA(KP554958132, TK, TM); cannam@95: TA = FMA(KP554958132, To, Tu); cannam@95: TN = FMA(KP554958132, TM, TL); cannam@95: TX = FNMS(KP554958132, TL, TK); cannam@95: TF = FNMS(KP554958132, Tr, To); cannam@95: Tv = FMA(KP554958132, Tu, Tr); cannam@95: T5 = Cr[WS(csr, 5)]; cannam@95: Ta = Cr[WS(csr, 6)]; cannam@95: Tb = Cr[WS(csr, 1)]; cannam@95: T7 = Cr[WS(csr, 4)]; cannam@95: Tf = T4 + T5; cannam@95: T6 = T4 - T5; cannam@95: Th = Ta + Tb; cannam@95: Tc = Ta - Tb; cannam@95: T8 = Cr[WS(csr, 3)]; cannam@95: } cannam@95: } cannam@95: { cannam@95: E Tw, Tx, TP, Tg, T9, TY, TC, TI, TQ; cannam@95: Tw = FMA(KP801937735, Tv, To); cannam@95: Tx = FNMS(KP356895867, Tf, Th); cannam@95: TP = FNMS(KP356895867, T6, Tc); cannam@95: Tg = T7 + T8; cannam@95: T9 = T7 - T8; cannam@95: TY = FNMS(KP801937735, TX, TM); cannam@95: { cannam@95: E TB, TH, TV, Ty, Tl, Ti, TW, Tz; cannam@95: TB = FNMS(KP801937735, TA, Tr); cannam@95: Ti = Tf + Tg + Th; cannam@95: TC = FNMS(KP356895867, Th, Tg); cannam@95: { cannam@95: E Tj, Td, TU, Tk; cannam@95: Tj = FNMS(KP356895867, Tg, Tf); cannam@95: Td = T6 + T9 + Tc; cannam@95: TH = FNMS(KP356895867, T9, T6); cannam@95: TU = FNMS(KP356895867, Tc, T9); cannam@95: R0[0] = FMA(KP2_000000000, Ti, Te); cannam@95: Tk = FNMS(KP692021471, Tj, Th); cannam@95: R1[WS(rs, 3)] = FMA(KP2_000000000, Td, T3); cannam@95: TV = FNMS(KP692021471, TU, T6); cannam@95: Ty = FNMS(KP692021471, Tx, Tg); cannam@95: Tl = FNMS(KP1_801937735, Tk, Te); cannam@95: } cannam@95: TO = FMA(KP801937735, TN, TK); cannam@95: TW = FNMS(KP1_801937735, TV, T3); cannam@95: Tz = FNMS(KP1_801937735, Ty, Te); cannam@95: R0[WS(rs, 3)] = FMA(KP1_949855824, Tw, Tl); cannam@95: R0[WS(rs, 4)] = FNMS(KP1_949855824, Tw, Tl); cannam@95: R1[WS(rs, 5)] = FMA(KP1_949855824, TY, TW); cannam@95: R1[WS(rs, 1)] = FNMS(KP1_949855824, TY, TW); cannam@95: R0[WS(rs, 6)] = FMA(KP1_949855824, TB, Tz); cannam@95: R0[WS(rs, 1)] = FNMS(KP1_949855824, TB, Tz); cannam@95: TI = FNMS(KP692021471, TH, Tc); cannam@95: } cannam@95: TT = FNMS(KP801937735, TS, TL); cannam@95: TQ = FNMS(KP692021471, TP, T9); cannam@95: TG = FNMS(KP801937735, TF, Tu); cannam@95: TJ = FNMS(KP1_801937735, TI, T3); cannam@95: TD = FNMS(KP692021471, TC, Tf); cannam@95: TR = FNMS(KP1_801937735, TQ, T3); cannam@95: } cannam@95: } cannam@95: R1[WS(rs, 6)] = FMA(KP1_949855824, TO, TJ); cannam@95: R1[0] = FNMS(KP1_949855824, TO, TJ); cannam@95: TE = FNMS(KP1_801937735, TD, Te); cannam@95: R1[WS(rs, 2)] = FMA(KP1_949855824, TT, TR); cannam@95: R1[WS(rs, 4)] = FNMS(KP1_949855824, TT, TR); cannam@95: R0[WS(rs, 2)] = FMA(KP1_949855824, TG, TE); cannam@95: R0[WS(rs, 5)] = FNMS(KP1_949855824, TG, TE); cannam@95: } cannam@95: } cannam@95: } cannam@95: cannam@95: static const kr2c_desc desc = { 14, "r2cb_14", {18, 0, 44, 0}, &GENUS }; cannam@95: cannam@95: void X(codelet_r2cb_14) (planner *p) { cannam@95: X(kr2c_register) (p, r2cb_14, &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 14 -name r2cb_14 -include r2cb.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 62 FP additions, 38 FP multiplications, cannam@95: * (or, 36 additions, 12 multiplications, 26 fused multiply/add), cannam@95: * 28 stack variables, 7 constants, and 28 memory accesses cannam@95: */ cannam@95: #include "r2cb.h" cannam@95: cannam@95: static void r2cb_14(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_801937735, +1.801937735804838252472204639014890102331838324); cannam@95: DK(KP445041867, +0.445041867912628808577805128993589518932711138); cannam@95: DK(KP1_246979603, +1.246979603717467061050009768008479621264549462); cannam@95: DK(KP867767478, +0.867767478235116240951536665696717509219981456); cannam@95: DK(KP1_949855824, +1.949855824363647214036263365987862434465571601); cannam@95: DK(KP1_563662964, +1.563662964936059617416889053348115500464669037); 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(56, rs), MAKE_VOLATILE_STRIDE(56, csr), MAKE_VOLATILE_STRIDE(56, csi)) { cannam@95: E T3, Td, T6, Te, Tq, Tz, Tn, Ty, Tc, Tg, Tk, Tx, T9, Tf, T1; cannam@95: E T2; cannam@95: T1 = Cr[0]; cannam@95: T2 = Cr[WS(csr, 7)]; cannam@95: T3 = T1 - T2; cannam@95: Td = T1 + T2; cannam@95: { cannam@95: E T4, T5, To, Tp; cannam@95: T4 = Cr[WS(csr, 2)]; cannam@95: T5 = Cr[WS(csr, 5)]; cannam@95: T6 = T4 - T5; cannam@95: Te = T4 + T5; cannam@95: To = Ci[WS(csi, 2)]; cannam@95: Tp = Ci[WS(csi, 5)]; cannam@95: Tq = To - Tp; cannam@95: Tz = To + Tp; cannam@95: } cannam@95: { cannam@95: E Tl, Tm, Ta, Tb; cannam@95: Tl = Ci[WS(csi, 6)]; cannam@95: Tm = Ci[WS(csi, 1)]; cannam@95: Tn = Tl - Tm; cannam@95: Ty = Tl + Tm; cannam@95: Ta = Cr[WS(csr, 6)]; cannam@95: Tb = Cr[WS(csr, 1)]; cannam@95: Tc = Ta - Tb; cannam@95: Tg = Ta + Tb; cannam@95: } cannam@95: { cannam@95: E Ti, Tj, T7, T8; cannam@95: Ti = Ci[WS(csi, 4)]; cannam@95: Tj = Ci[WS(csi, 3)]; cannam@95: Tk = Ti - Tj; cannam@95: Tx = Ti + Tj; cannam@95: T7 = Cr[WS(csr, 4)]; cannam@95: T8 = Cr[WS(csr, 3)]; cannam@95: T9 = T7 - T8; cannam@95: Tf = T7 + T8; cannam@95: } cannam@95: R1[WS(rs, 3)] = FMA(KP2_000000000, T6 + T9 + Tc, T3); cannam@95: R0[0] = FMA(KP2_000000000, Te + Tf + Tg, Td); cannam@95: { cannam@95: E Tr, Th, TE, TD; cannam@95: Tr = FNMS(KP1_949855824, Tn, KP1_563662964 * Tk) - (KP867767478 * Tq); cannam@95: Th = FMA(KP1_246979603, Tf, Td) + FNMA(KP445041867, Tg, KP1_801937735 * Te); cannam@95: R0[WS(rs, 2)] = Th - Tr; cannam@95: R0[WS(rs, 5)] = Th + Tr; cannam@95: TE = FMA(KP867767478, Tx, KP1_563662964 * Ty) - (KP1_949855824 * Tz); cannam@95: TD = FMA(KP1_246979603, Tc, T3) + FNMA(KP1_801937735, T9, KP445041867 * T6); cannam@95: R1[WS(rs, 2)] = TD - TE; cannam@95: R1[WS(rs, 4)] = TD + TE; cannam@95: } cannam@95: { cannam@95: E Tt, Ts, TA, Tw; cannam@95: Tt = FMA(KP867767478, Tk, KP1_563662964 * Tn) - (KP1_949855824 * Tq); cannam@95: Ts = FMA(KP1_246979603, Tg, Td) + FNMA(KP1_801937735, Tf, KP445041867 * Te); cannam@95: R0[WS(rs, 6)] = Ts - Tt; cannam@95: R0[WS(rs, 1)] = Ts + Tt; cannam@95: TA = FNMS(KP1_949855824, Ty, KP1_563662964 * Tx) - (KP867767478 * Tz); cannam@95: Tw = FMA(KP1_246979603, T9, T3) + FNMA(KP445041867, Tc, KP1_801937735 * T6); cannam@95: R1[WS(rs, 5)] = Tw - TA; cannam@95: R1[WS(rs, 1)] = Tw + TA; cannam@95: } cannam@95: { cannam@95: E TC, TB, Tv, Tu; cannam@95: TC = FMA(KP1_563662964, Tz, KP1_949855824 * Tx) + (KP867767478 * Ty); cannam@95: TB = FMA(KP1_246979603, T6, T3) + FNMA(KP1_801937735, Tc, KP445041867 * T9); cannam@95: R1[0] = TB - TC; cannam@95: R1[WS(rs, 6)] = TB + TC; cannam@95: Tv = FMA(KP1_563662964, Tq, KP1_949855824 * Tk) + (KP867767478 * Tn); cannam@95: Tu = FMA(KP1_246979603, Te, Td) + FNMA(KP1_801937735, Tg, KP445041867 * Tf); cannam@95: R0[WS(rs, 4)] = Tu - Tv; cannam@95: R0[WS(rs, 3)] = Tu + Tv; cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: cannam@95: static const kr2c_desc desc = { 14, "r2cb_14", {36, 12, 26, 0}, &GENUS }; cannam@95: cannam@95: void X(codelet_r2cb_14) (planner *p) { cannam@95: X(kr2c_register) (p, r2cb_14, &desc); cannam@95: } cannam@95: cannam@95: #endif /* HAVE_FMA */