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 07:41:07 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 9 -name r2cb_9 -include r2cb.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 32 FP additions, 24 FP multiplications, cannam@95: * (or, 8 additions, 0 multiplications, 24 fused multiply/add), cannam@95: * 40 stack variables, 12 constants, and 18 memory accesses cannam@95: */ cannam@95: #include "r2cb.h" cannam@95: cannam@95: static void r2cb_9(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_326827896, +1.326827896337876792410842639271782594433726619); cannam@95: DK(KP1_705737063, +1.705737063904886419256501927880148143872040591); cannam@95: DK(KP766044443, +0.766044443118978035202392650555416673935832457); cannam@95: DK(KP1_532088886, +1.532088886237956070404785301110833347871664914); cannam@95: DK(KP984807753, +0.984807753012208059366743024589523013670643252); cannam@95: DK(KP1_969615506, +1.969615506024416118733486049179046027341286503); cannam@95: DK(KP839099631, +0.839099631177280011763127298123181364687434283); cannam@95: DK(KP176326980, +0.176326980708464973471090386868618986121633062); cannam@95: DK(KP866025403, +0.866025403784438646763723170752936183471402627); cannam@95: DK(KP500000000, +0.500000000000000000000000000000000000000000000); cannam@95: DK(KP1_732050807, +1.732050807568877293527446341505872366942805254); 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(36, rs), MAKE_VOLATILE_STRIDE(36, csr), MAKE_VOLATILE_STRIDE(36, csi)) { cannam@95: E T4, Th, T3, Tb, Tp, Tk, T7, Tf, Ti, Ta, T1, T2; cannam@95: Ta = Ci[WS(csi, 3)]; cannam@95: T1 = Cr[0]; cannam@95: T2 = Cr[WS(csr, 3)]; cannam@95: T4 = Cr[WS(csr, 1)]; cannam@95: Th = Ci[WS(csi, 1)]; cannam@95: { cannam@95: E T5, T9, T6, Td, Te; cannam@95: T5 = Cr[WS(csr, 4)]; cannam@95: T9 = T1 - T2; cannam@95: T3 = FMA(KP2_000000000, T2, T1); cannam@95: T6 = Cr[WS(csr, 2)]; cannam@95: Td = Ci[WS(csi, 4)]; cannam@95: Te = Ci[WS(csi, 2)]; cannam@95: Tb = FNMS(KP1_732050807, Ta, T9); cannam@95: Tp = FMA(KP1_732050807, Ta, T9); cannam@95: Tk = T6 - T5; cannam@95: T7 = T5 + T6; cannam@95: Tf = Td + Te; cannam@95: Ti = Td - Te; cannam@95: } cannam@95: { cannam@95: E Tu, To, Tt, Tn, Tc, T8; cannam@95: Tc = FNMS(KP500000000, T7, T4); cannam@95: T8 = T4 + T7; cannam@95: { cannam@95: E Tw, Tj, Tr, Tg, Tv; cannam@95: Tw = Ti + Th; cannam@95: Tj = FNMS(KP500000000, Ti, Th); cannam@95: Tr = FMA(KP866025403, Tf, Tc); cannam@95: Tg = FNMS(KP866025403, Tf, Tc); cannam@95: Tv = T3 - T8; cannam@95: R0[0] = FMA(KP2_000000000, T8, T3); cannam@95: { cannam@95: E Tq, Tl, Ts, Tm; cannam@95: Tq = FMA(KP866025403, Tk, Tj); cannam@95: Tl = FNMS(KP866025403, Tk, Tj); cannam@95: R0[WS(rs, 3)] = FMA(KP1_732050807, Tw, Tv); cannam@95: R1[WS(rs, 1)] = FNMS(KP1_732050807, Tw, Tv); cannam@95: Ts = FNMS(KP176326980, Tr, Tq); cannam@95: Tu = FMA(KP176326980, Tq, Tr); cannam@95: Tm = FNMS(KP839099631, Tl, Tg); cannam@95: To = FMA(KP839099631, Tg, Tl); cannam@95: R0[WS(rs, 1)] = FNMS(KP1_969615506, Ts, Tp); cannam@95: Tt = FMA(KP984807753, Ts, Tp); cannam@95: R1[0] = FMA(KP1_532088886, Tm, Tb); cannam@95: Tn = FNMS(KP766044443, Tm, Tb); cannam@95: } cannam@95: } cannam@95: R1[WS(rs, 2)] = FNMS(KP1_705737063, Tu, Tt); cannam@95: R0[WS(rs, 4)] = FMA(KP1_705737063, Tu, Tt); cannam@95: R0[WS(rs, 2)] = FNMS(KP1_326827896, To, Tn); cannam@95: R1[WS(rs, 3)] = FMA(KP1_326827896, To, Tn); cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: cannam@95: static const kr2c_desc desc = { 9, "r2cb_9", {8, 0, 24, 0}, &GENUS }; cannam@95: cannam@95: void X(codelet_r2cb_9) (planner *p) { cannam@95: X(kr2c_register) (p, r2cb_9, &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 9 -name r2cb_9 -include r2cb.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 32 FP additions, 18 FP multiplications, cannam@95: * (or, 22 additions, 8 multiplications, 10 fused multiply/add), cannam@95: * 35 stack variables, 12 constants, and 18 memory accesses cannam@95: */ cannam@95: #include "r2cb.h" cannam@95: cannam@95: static void r2cb_9(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(KP984807753, +0.984807753012208059366743024589523013670643252); cannam@95: DK(KP173648177, +0.173648177666930348851716626769314796000375677); cannam@95: DK(KP300767466, +0.300767466360870593278543795225003852144476517); cannam@95: DK(KP1_705737063, +1.705737063904886419256501927880148143872040591); cannam@95: DK(KP642787609, +0.642787609686539326322643409907263432907559884); cannam@95: DK(KP766044443, +0.766044443118978035202392650555416673935832457); cannam@95: DK(KP1_326827896, +1.326827896337876792410842639271782594433726619); cannam@95: DK(KP1_113340798, +1.113340798452838732905825904094046265936583811); cannam@95: DK(KP500000000, +0.500000000000000000000000000000000000000000000); cannam@95: DK(KP866025403, +0.866025403784438646763723170752936183471402627); cannam@95: DK(KP2_000000000, +2.000000000000000000000000000000000000000000000); cannam@95: DK(KP1_732050807, +1.732050807568877293527446341505872366942805254); 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(36, rs), MAKE_VOLATILE_STRIDE(36, csr), MAKE_VOLATILE_STRIDE(36, csi)) { cannam@95: E T3, Tq, Tc, Tk, Tj, T8, Tm, Ts, Th, Tr, Tw, Tx; cannam@95: { cannam@95: E Tb, T1, T2, T9, Ta; cannam@95: Ta = Ci[WS(csi, 3)]; cannam@95: Tb = KP1_732050807 * Ta; cannam@95: T1 = Cr[0]; cannam@95: T2 = Cr[WS(csr, 3)]; cannam@95: T9 = T1 - T2; cannam@95: T3 = FMA(KP2_000000000, T2, T1); cannam@95: Tq = T9 + Tb; cannam@95: Tc = T9 - Tb; cannam@95: } cannam@95: { cannam@95: E T4, T7, Ti, Tg, Tl, Td; cannam@95: T4 = Cr[WS(csr, 1)]; cannam@95: Tk = Ci[WS(csi, 1)]; cannam@95: { cannam@95: E T5, T6, Te, Tf; cannam@95: T5 = Cr[WS(csr, 4)]; cannam@95: T6 = Cr[WS(csr, 2)]; cannam@95: T7 = T5 + T6; cannam@95: Ti = KP866025403 * (T5 - T6); cannam@95: Te = Ci[WS(csi, 4)]; cannam@95: Tf = Ci[WS(csi, 2)]; cannam@95: Tg = KP866025403 * (Te + Tf); cannam@95: Tj = Tf - Te; cannam@95: } cannam@95: T8 = T4 + T7; cannam@95: Tl = FMA(KP500000000, Tj, Tk); cannam@95: Tm = Ti + Tl; cannam@95: Ts = Tl - Ti; cannam@95: Td = FNMS(KP500000000, T7, T4); cannam@95: Th = Td - Tg; cannam@95: Tr = Td + Tg; cannam@95: } cannam@95: R0[0] = FMA(KP2_000000000, T8, T3); cannam@95: Tw = T3 - T8; cannam@95: Tx = KP1_732050807 * (Tk - Tj); cannam@95: R1[WS(rs, 1)] = Tw - Tx; cannam@95: R0[WS(rs, 3)] = Tw + Tx; cannam@95: { cannam@95: E Tp, Tn, To, Tv, Tt, Tu; cannam@95: Tp = FMA(KP1_113340798, Th, KP1_326827896 * Tm); cannam@95: Tn = FNMS(KP642787609, Tm, KP766044443 * Th); cannam@95: To = Tc - Tn; cannam@95: R1[0] = FMA(KP2_000000000, Tn, Tc); cannam@95: R1[WS(rs, 3)] = To + Tp; cannam@95: R0[WS(rs, 2)] = To - Tp; cannam@95: Tv = FMA(KP1_705737063, Tr, KP300767466 * Ts); cannam@95: Tt = FNMS(KP984807753, Ts, KP173648177 * Tr); cannam@95: Tu = Tq - Tt; cannam@95: R0[WS(rs, 1)] = FMA(KP2_000000000, Tt, Tq); cannam@95: R0[WS(rs, 4)] = Tu + Tv; cannam@95: R1[WS(rs, 2)] = Tu - Tv; cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: cannam@95: static const kr2c_desc desc = { 9, "r2cb_9", {22, 8, 10, 0}, &GENUS }; cannam@95: cannam@95: void X(codelet_r2cb_9) (planner *p) { cannam@95: X(kr2c_register) (p, r2cb_9, &desc); cannam@95: } cannam@95: cannam@95: #endif /* HAVE_FMA */