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 12 -name r2cb_12 -include r2cb.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 38 FP additions, 16 FP multiplications, cannam@95: * (or, 22 additions, 0 multiplications, 16 fused multiply/add), cannam@95: * 31 stack variables, 2 constants, and 24 memory accesses cannam@95: */ cannam@95: #include "r2cb.h" cannam@95: cannam@95: static void r2cb_12(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_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(48, rs), MAKE_VOLATILE_STRIDE(48, csr), MAKE_VOLATILE_STRIDE(48, csi)) { cannam@95: E Ts, Tr; cannam@95: { cannam@95: E Tz, Te, Tn, Tk, Tc, Tw, Ty, Th, T4, T3, Td, T5; cannam@95: { cannam@95: E T8, Tu, Tl, Tm, Tb, T9, Ta, T1, T2, Tv; cannam@95: T8 = Cr[WS(csr, 3)]; cannam@95: T9 = Cr[WS(csr, 5)]; cannam@95: Ta = Cr[WS(csr, 1)]; cannam@95: Tu = Ci[WS(csi, 3)]; cannam@95: Tl = Ci[WS(csi, 5)]; cannam@95: Tm = Ci[WS(csi, 1)]; cannam@95: Tb = T9 + Ta; cannam@95: Tz = T9 - Ta; cannam@95: Te = Ci[WS(csi, 4)]; cannam@95: Tn = Tl - Tm; cannam@95: Tv = Tl + Tm; cannam@95: Tk = FNMS(KP2_000000000, T8, Tb); cannam@95: Tc = T8 + Tb; cannam@95: T1 = Cr[0]; cannam@95: T2 = Cr[WS(csr, 4)]; cannam@95: Tw = Tu - Tv; cannam@95: Ty = FMA(KP2_000000000, Tu, Tv); cannam@95: Th = Ci[WS(csi, 2)]; cannam@95: T4 = Cr[WS(csr, 6)]; cannam@95: T3 = FMA(KP2_000000000, T2, T1); cannam@95: Td = T1 - T2; cannam@95: T5 = Cr[WS(csr, 2)]; cannam@95: } cannam@95: { cannam@95: E To, Tp, Tf, Tg, T6, TA, TC; cannam@95: To = FMA(KP1_732050807, Tn, Tk); cannam@95: Ts = FNMS(KP1_732050807, Tn, Tk); cannam@95: Tp = FNMS(KP1_732050807, Te, Td); cannam@95: Tf = FMA(KP1_732050807, Te, Td); cannam@95: Tg = T4 - T5; cannam@95: T6 = FMA(KP2_000000000, T5, T4); cannam@95: TA = FMA(KP1_732050807, Tz, Ty); cannam@95: TC = FNMS(KP1_732050807, Tz, Ty); cannam@95: { cannam@95: E Tt, T7, Ti, Tq, Tj, TB, Tx; cannam@95: Tt = T3 - T6; cannam@95: T7 = T3 + T6; cannam@95: Ti = FNMS(KP1_732050807, Th, Tg); cannam@95: Tq = FMA(KP1_732050807, Th, Tg); cannam@95: R0[0] = FMA(KP2_000000000, Tc, T7); cannam@95: R0[WS(rs, 3)] = FNMS(KP2_000000000, Tc, T7); cannam@95: Tj = Tf + Ti; cannam@95: TB = Tf - Ti; cannam@95: Tr = Tp + Tq; cannam@95: Tx = Tp - Tq; cannam@95: R1[WS(rs, 5)] = TB + TC; cannam@95: R1[WS(rs, 2)] = TB - TC; cannam@95: R0[WS(rs, 4)] = Tj - To; cannam@95: R0[WS(rs, 1)] = Tj + To; cannam@95: R1[WS(rs, 3)] = Tx + TA; cannam@95: R1[0] = Tx - TA; cannam@95: R1[WS(rs, 4)] = FNMS(KP2_000000000, Tw, Tt); cannam@95: R1[WS(rs, 1)] = FMA(KP2_000000000, Tw, Tt); cannam@95: } cannam@95: } cannam@95: } cannam@95: R0[WS(rs, 2)] = Tr - Ts; cannam@95: R0[WS(rs, 5)] = Tr + Ts; cannam@95: } cannam@95: } cannam@95: } cannam@95: cannam@95: static const kr2c_desc desc = { 12, "r2cb_12", {22, 0, 16, 0}, &GENUS }; cannam@95: cannam@95: void X(codelet_r2cb_12) (planner *p) { cannam@95: X(kr2c_register) (p, r2cb_12, &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 12 -name r2cb_12 -include r2cb.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 38 FP additions, 10 FP multiplications, cannam@95: * (or, 34 additions, 6 multiplications, 4 fused multiply/add), cannam@95: * 25 stack variables, 2 constants, and 24 memory accesses cannam@95: */ cannam@95: #include "r2cb.h" cannam@95: cannam@95: static void r2cb_12(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_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(48, rs), MAKE_VOLATILE_STRIDE(48, csr), MAKE_VOLATILE_STRIDE(48, csi)) { cannam@95: E T8, Tb, Tm, TA, Tw, Tx, Tp, TB, T3, Tr, Tg, T6, Ts, Tk; cannam@95: { cannam@95: E T9, Ta, Tn, To; cannam@95: T8 = Cr[WS(csr, 3)]; cannam@95: T9 = Cr[WS(csr, 5)]; cannam@95: Ta = Cr[WS(csr, 1)]; cannam@95: Tb = T9 + Ta; cannam@95: Tm = FMS(KP2_000000000, T8, Tb); cannam@95: TA = KP1_732050807 * (T9 - Ta); cannam@95: Tw = Ci[WS(csi, 3)]; cannam@95: Tn = Ci[WS(csi, 5)]; cannam@95: To = Ci[WS(csi, 1)]; cannam@95: Tx = Tn + To; cannam@95: Tp = KP1_732050807 * (Tn - To); cannam@95: TB = FMA(KP2_000000000, Tw, Tx); cannam@95: } cannam@95: { cannam@95: E Tf, T1, T2, Td, Te; cannam@95: Te = Ci[WS(csi, 4)]; cannam@95: Tf = KP1_732050807 * Te; cannam@95: T1 = Cr[0]; cannam@95: T2 = Cr[WS(csr, 4)]; cannam@95: Td = T1 - T2; cannam@95: T3 = FMA(KP2_000000000, T2, T1); cannam@95: Tr = Td - Tf; cannam@95: Tg = Td + Tf; cannam@95: } cannam@95: { cannam@95: E Tj, T4, T5, Th, Ti; cannam@95: Ti = Ci[WS(csi, 2)]; cannam@95: Tj = KP1_732050807 * Ti; cannam@95: T4 = Cr[WS(csr, 6)]; cannam@95: T5 = Cr[WS(csr, 2)]; cannam@95: Th = T4 - T5; cannam@95: T6 = FMA(KP2_000000000, T5, T4); cannam@95: Ts = Th + Tj; cannam@95: Tk = Th - Tj; cannam@95: } cannam@95: { cannam@95: E T7, Tc, Tz, TC; cannam@95: T7 = T3 + T6; cannam@95: Tc = KP2_000000000 * (T8 + Tb); cannam@95: R0[WS(rs, 3)] = T7 - Tc; cannam@95: R0[0] = T7 + Tc; cannam@95: { cannam@95: E Tl, Tq, TD, TE; cannam@95: Tl = Tg + Tk; cannam@95: Tq = Tm - Tp; cannam@95: R0[WS(rs, 1)] = Tl - Tq; cannam@95: R0[WS(rs, 4)] = Tl + Tq; cannam@95: TD = Tg - Tk; cannam@95: TE = TB - TA; cannam@95: R1[WS(rs, 2)] = TD - TE; cannam@95: R1[WS(rs, 5)] = TD + TE; cannam@95: } cannam@95: Tz = Tr - Ts; cannam@95: TC = TA + TB; cannam@95: R1[0] = Tz - TC; cannam@95: R1[WS(rs, 3)] = Tz + TC; cannam@95: { cannam@95: E Tv, Ty, Tt, Tu; cannam@95: Tv = T3 - T6; cannam@95: Ty = KP2_000000000 * (Tw - Tx); cannam@95: R1[WS(rs, 4)] = Tv - Ty; cannam@95: R1[WS(rs, 1)] = Tv + Ty; cannam@95: Tt = Tr + Ts; cannam@95: Tu = Tm + Tp; cannam@95: R0[WS(rs, 5)] = Tt - Tu; cannam@95: R0[WS(rs, 2)] = Tt + Tu; cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: cannam@95: static const kr2c_desc desc = { 12, "r2cb_12", {34, 6, 4, 0}, &GENUS }; cannam@95: cannam@95: void X(codelet_r2cb_12) (planner *p) { cannam@95: X(kr2c_register) (p, r2cb_12, &desc); cannam@95: } cannam@95: cannam@95: #endif /* HAVE_FMA */